Anda di halaman 1dari 328

Widespan Roof Structures

Compiled by Michael Barnes and Michael Dickson University of Bath

"l

ThomasTelford

Published by Thomas Telford Publishing, Thomas Telford Ltd, 1 Heron Quay, London E14 4JD. URL: http://www.thomastelford.com

Distributors for Thomas Telford books are USA: ASCE Press, 1801 Alexander Bell Drive, Reston, VA 20191-4400, USA Japan: Maruzen Co. Ltd, Book Department, 3-10 Nihonbashi 2-chome, Chuo-ku, Tokyo 103 Australia: DA Books and Journals, 648 Whitehorse Road, Mitcham 3132, Victoria

First published 2000

Also available from Thomas Telford Books Engineer's Contribution to Contemporary Architecture (series) Antony Hunt by A. Macdonald, 2000, ISBN 0 7277 2769 9 Heinz Isler by J. Chilton, 2000, ISBN 0 7277 2878 4 Peter Rice by A. Brown, 2000, ISBN 0 7277 2770 2 Eladio Dieste by R. Pedreschi, 2000, ISBN 0 7277 277 9 An introduction to cable roof structures, second edition, by H.A. Buchholdt, 1998, ISBN 0 7277 2624 2 Analysis of cable and catenary structures, by P. Broughton and P. Ndumbaro, 1994, ISBN 0 7277 2008 2 Windloading: a practical guide to BS 6399-2 by N. Cook, 1999, ISBN 0 7277 2755 9 The architecture of bridge design by D. Bennett, 1997, ISBN 0 7277 2529 7

A catalogue record for this book is available from the British Library

ISBN: 0 7277 2877 6 Michael Barnes and Michael Dickson, 2000 All rights, including translation, reserved. Except as permitted by the Copyright, Designs and Patents Act 1988, no part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying or otherwise, without the prior written permission of the Publishing Director, Thomas Telford Publishing, Thomas Telford Ltd, 1 Heron Quay, London E14 4JD. This book is published on the understanding that the authors are solely responsible for the statements made and opinions expressed in it and that its publication does not necessarily imply that such statements and/or opinions are or reflect the views or opinions of the publishers. While every effort has been made to ensure that the statements made and the opinions expressed in this publication provide a safe and accurate guide, no liability or responsibility can be accepted in this respect by the authors, editors or publishers. Typeset by the Department of Architecture and Civil Engineering, University of Bath Printed and bound in Great Britain by Polestar Scientifica, Exeter

DEDICATION

SIR EDMUND HAPPOLD


FRENG, FISTRUCTE, HON FRIBA, FCIOB

1930 - 1996

The Symposium on which this book is based was dedicated to the memory of Ted Happold. Many of the papers presented are by his personal friends and are a tribute to the esteem in which he is held.

FOREWORD
This book includes all of the papers presented at the International Symposium on Widespan Enclosures which was held at the University of Bath in April 2000. It also contains additional papers which expand on verbal presentations given at the Symposium and other invited contributions for completeness of the text.

The principal aim of the book is to bring together expert knowledge in the design and development of widespan roof structures and space enclosures at a time when so many innovative structures which mark the Millennium have come to fruition. As was the case for the original symposium, the book is intended to be multidisciplinary in its approach and to present a broad spectrum of contributions appropriate to the holistic design of widespan structures.

It is hoped that the book will provide a useful reference text of precedent studies of innovative work by leading architects and engineers for students and practitioners of the future.

Michael Barnes and Michael Dickson

ACKNOWLEDGMENTS
The original symposium on Widespan Enclosures was sponsored jointly by the Happold Trust, the DETR and the University of Bath. The authors are grateful for their help, and also for the work of Sue Fairhurst of the University of Bath who designed the layout and compiled the papers for this book.

CONTENTS
3 4 Dedication Foreword and A c k n o w l e d g e m e n t s

Michael

Barnes

and Michael

Dickson

: Preface and O v e r v i e w : The Evolution of L o n g s p a n Lightweight Structures

17 19 31 41 50

FORM A N D STRUCTURE Dickson Barnes : On Frei Otto's Philosophy of Widespan Lightweight Structures : Form and Stress M o d e l l i n g o f Tension Structures : The Definition of Curved Geometry for Widespan Structures Concepts and Reliability in the D e s i g n o f Widespan Structures

Michael Michael

Chris J K Williams Massimo Majowiecki:

62

Peter Irwin, Frank Hochstenbach

and Scott Gamble

: W i n d and S n o w Considerations for Widespan Enclosures

73 75 85 89 100 105 117 127 129 133 137 145 149 159 169 777 178 189 199 208 213

II F O R M A N D E N V I R O N M E N T Andrew Peter Whalley Thoday : The Eden Project, Glass H o u s e s , World Environments

: Planting Environment for the Eden Project : Civil and Structural D e s i g n for the Eden Project : Climatic E n v e l o p e s : The Environmental C o n s e q u e n c e s o f a Building with a W i d e Span : Controlling the Indoor Climate in Widespan Enclosures - 4 Case Studies

Alan C Jones Ben Morris

Max Fordham Nick Cullen

III M E Z O T E C T U R E A N D T H E M I L L E N N I U M D O M E Jennie Page Mike Davies : The M i l l e n n i u m D o m e , 'Introduction to Client Concept' : Mezotecture : Servicing the D o m e Environment

Tony McLaughlin Martin Kealy

: Fire Engineering Large Environmental Enclosures, T h e R o o f o f the M i l l e n n i u m D o m e : Principles o f Construction for Widespan Structures with e x a m p l e s from the M i l l e n n i u m D o m e

Ian Liddell: Peter Miller David

Trench : Construction of Widespan Enclosure

IV C O N S T R U C T I O N A N D M A T E R I A L S J org Schlaich Wolfgang : Lightweight Structures : Tensile Space Structures

Renner

Johan Sischka Robert Silman

: Engineering the Construction o f the Great Court R o o f for the British M u s e u m : A History o f Widespan Structures in the United States : Materials for the N e w M i l l e n n i u m

Nick Goldsmith

218 229 230 241 247 252 261

Horst Berger

: Engineering an Integrated Architecture for Widespan Enclosures

V SPORTS STADIA Mike Otlet: Michael The M i l l e n n i u m Stadium, Cardiff : D e v e l o p m e n t o f the N e w W e m b l e y Stadium R o o f

Barker

Steve Morley

: K e e p i n g the D o o r s Open: The O l y m p i c Stadium, S y d n e y

Alan Willby : Widespan Enclosures and Structures: Cost Considerations with Illustrated Case Studies Kazuo Ishii : Lightweight Enclosures in Japan, and Stadia for the 2 0 0 2 World Cup

275 276

VI

G L A S S A N D F A B R I C ATRIA

Spencer de Grey : The D e s i g n o f the R o o f s of the British M u s e u m Great Court and the Music Centre at Gateshead Stephen Brown : Engineering the British M u s e u m Great Court R o o f : From Schlumberger to the D y n a m i c Earth, a Sequence of Membrane Roofs

283 287 297 304 309 317

Bill Taylor and John Thornton Brian Forster Tim Macfarlane Bernard Rudiger

: Tensioned Braced Ribs in Architectural Projects and Damian Murphy : Three Widespan Space Enclosures in the U S A : A Philosophy of "Widespan" as applied to Smaller Enclosures

Vaudeville

and Mitsu Edwards Janner

Lutz and Michele

: Structural D e s i g n of T w o Large Span Buildings in Germany

PREFACE AND OVERVIEW THE EVOLUTION OF LONGSPAN LIGHTWEIGHT STRUCTURES


Michael Barnes, Professor of Civil/Structural Engineering, Department of Architecture and Civil Engineering, University of Bath; Michael Dickson, FREng, Visiting Professor of Engineering Design, University of Bath

INTRODUCTION AND HISTORICAL BACKGROUND


The historical forms of the sail, the tent and the vault have developed under the influence of craft, art, architecture, mathematical science and engineering science into the multiplicity of constructional forms available today. One of the early pioneers in this field was Frei Otto [1]. Circus tents and early nomadic tents led to experimentation with sail tents for the Lausanne Exhibition (Otto, Stromeyer), and the introduction of high strength pvc coated polyester fabrics led to wider span and more intricately patterned prestressed membranes [2]. With increased use of these tensile membranes, and the perceived benefits of large airsupported structures, teflon glass fabrics and a variety of special coatings were specifically developed to provide fire resistance, self cleansing and durability - the conical membranes at La Verne College (Shaver, Campbell and Bird) (Fig 1) were constructed nearly thirty years ago.
Fig 2 The Raleigh Arena

Fig 3 German Pavilion EXPO 67

Fig 1 La Verne College

Public and professional acceptance of tensile structures as permanent systems was enhanced by large and historically important prestressed cable-net structures such as the Raleigh Arena, N Carolina (Nowicki, Severud) (Fig 2), the German Pavilion at Montreal (Gutbrod, Otto, Stromeyer, Leonhardt & Andra, Linkwitz) (Fig3), the Olympic Stadium at Munich (Behnisch, Otto, Leonardt & Andra) (Fig 4), and the Skating Arena, Munich (Ackerman, Schlaich) [3].

Fig 4

Munich Olympic Stadium

In contrast to those forms needing to be anticlastic for stability, single curvature hanging forms have been largely stabilised by selfweight, damping and shell stiffness as for the Aircraft Hangar, Frankfurt (Finsterwalder), Conference Centre, Mecca (Gutbrod,

Otto, Arup), and Dulles International Airport (Saarinan, Severud). The Yale Hockey Rink (Saarinan, Severud), the wonderful National Gymnasiums for Tokyo Olympics (Tange, Tsuboi) (Fig 5), and Tsim Sha Tsui Cultural Centre, Hong Kong (Public Works, Happold), use the same engineering devices.

Fig 6a

Council of Ministers Building - Physical model

Fig 6b

Council of Ministers Building - Numerical model

Fig 5 National Gymnasium Tokyo Olympics

From the traditional brick and stone vaults of our cathedrals, mosques and temples come the synclastic shells of Isler and the hypar forms of Candela [4]. Indeed springing from Gaudi's proposals for the Segrada Familia in Barcelona, came the form studies on grid shells at the Institut fur Leichteflachentragwerk, Stuttgart (IL), culminating in the Timber Grid shell at Mannheim (Mutschler, Langner, Otto, Arup, Happold). A further development on this theme was the funicular lattice shell for the Council of Ministers Building in Riyadh (Gutbrod,Otto,Arup,Happold) (Figs 6(a) and (b). The vela for the Coliseum in Rome have transformed into the "Wandelbares Dach" at Bad Herschfelt (Frei Otto) (Fig 7), at Cannes (Taillibert, Otto, du Chateau), and indeed the proposal for the retractable Roof at the Montreal Olympic Stadium (Taillibert). Also progressing from the US Pavilion at Montreal 1967 (Buckminster Fuller) and the many large scale space dome structures of that date there are now a number of constructed projects which have large moving and sliding roofs. Notable amongst these are the Sky Dome,Toronto (Robbie, Allen) (Fig 23), the Ajax Stadium, Amsterdam (HBG) and the Millennium Stadium, Cardiff (HOK, Lobb, Atkins, Laing) - the latter presented in section V of this book..

Sail power for the Cutty Sark and early hot air balloons led to the patent by Lanchester for an air supported structure which was converted to reality by Walter Bird in air supported structures for Radomes and the US Pavilion at the Osaka Exhibition (Cambridge 5, Geiger, Bird), the Sports Centre, Riyadh University (Birdair), the 220 x 159m Pontiac Silver Dome (OHL, Geiger Berger, Birdair) (Fig 8), and the Tokyo Dome (Nikkei Sekkei, Takenaka). The above resume provides only an introduction to the development of widespan enclosures. Detailed reviews are given by Forster [5], Ishii [6], and others [3,4,7,10,12]. The overview in the following sections, discussing various types of lightweight widespan structures, is clearly linked to the authors' own involvements and is not intended as an objective historical record.

Fig 7

Retractable tent at Bad Herchfelt

CENTRAL QUESTIONS TOWARDS FUTURE DEVELOPMENT


The foregoing projects and many others are the role models for the next generation to develop further. They have been facilitated by the gradual purposeful scholastic evolution of building forms, techniques and material technologies, coupled to more sophisticated ways of evolving architectural form, defining engineering response and predicting complex geometry. This introduction prefaces the language and grammar of current widespan structures as defined by emerging materials technology, forms and building techniques for widespan enclosures. The most important questions for debate are: (1) How do we use engineering and materials science and art to produce better quality widespan enclosures offering better environmental and architectural quality for less resource thereby creating minimal impact on our planet (Life Cycle Assessment). What are the likely future design construction and operational developments which will allow the evolution of new building forms in response to tomorrows architectural and engineering challenges.

A BACKGROUND TO FORM D E V E L O P M E N T AND ANALYSIS


Many of the concepts for the formfinding, analysis and design of lightweight and longspan structures originated from the observation of structures in nature and from physical modelling techniques, the latter principally through work at the Institute for Lightweight Structures, (IL) Stuttgart University directed by Frei Otto |9]. The engineering science was largely developed in parallel by SFB64 (1976-84) directed by Leonhardt, Argyris, Linkwitz, Otto and others at the University of Stuttgart 110, 11]. SFB64 in effect pulled together all the technology, design method and material science that had been necessary to successfully design and construct the Olympic Stadia at Munich [12 1. Argyris et al [13], and Haug [ 14], outline the background to the development of matrix based numerical methods for the analysis of tension structures. Reference [15] reviews the Computer Aided Design of lightweight tensile structures, and explains particularly the development of systems based on the method of Dynamic Relaxation [16]. To simplify presentation, this introductory paper will discuss the development of different types of lightweight widespan enclosures in separate categories:

(2)

AIR S U P P O R T E D S T R U C T U R E S
In parallel with, and as a corresponding member of SFB64 at Stuttgart University, the Wolfson Air Supported Research Group at the University of Bath directed by the late Sir Ted Happold assembled all the pieces of the engineering jigsaw that were required for the design of lightweight tension structures and in particular air supported structures materials research and testing, forensic observation, design methodology, form-finding, analysis, economics, wind tunnel studies and environmental design. Above all this research group related its work to the practical constraints of real construction, and established a basic range of testing the physical and dimensional characteristics as well as resistance to degradation and abrasion of structural fabrics. The work was presented in two conferences and subsequent design guide publications sponsored by the Institution of Structural Engineers [17, 18]. This work coincided with the rapid developments being achieved by Geiger-Berger and Walter Bird of Birdair at the Pontiac Silver Dome, the Sports Centre in Santa Barbara and at the Riyadh University Sports Centre following the precedent of the US Pavilion at the Osaka Expo in 1970. At Osaka, Geiger's low profile pvc coated polyester membrane supported with 0.2 kPa internal air pressure was restrained by spiral cables on a diagonal grid. These were organised within a 140m x 82m oval bearing structure to minimise bending in the perimeter anchorage berms. The final structure created a naturally light internal space of exceptional economy. It was the

Clearly structural forms which carry their principal loads over long spans through direct compression or tension offer an ideal opportunity to approach a minimum of embodied energy (and therefore minimal depletion of scarce materials) by their construction [8]. Coincidentally a conscious choice of structural and architectural form can enhance natural ventilation patterns, so that orientation and choice of openings or roofing material will admit natural light and energy but exclude glare and summer heat. To achieve this requires definition of internal environmental performance for comfort and the incorporation of purpose designed mechanical, electrical and water systems, all controlled to reduce their demand for non renewable energy. Above all such major investments in built fabric must appeal to the sense of delight of those who look upon them and use their facility. For long span enclosures the potential is considerable.

Fig 8 Pontiac Silver Dome

10

principle innovation that launched many large air supported structures that now exist in the US, the Middle East, Japan and other countries. In temperate climates, where high snow drifting is likely, or high localised wind impact pressures may be dominant, additional care is required in design both of the form and the inflation system as is demonstrated by the interesting forensic paper on the Minnesota Metrodome [19]. The Wolfson work resulted in a design proposal for a 35 acre air supported structure to cover a 600 x 300m enclosure to house a 2,000 person village for 58 North in Alberta (Fullerton, Otto, Happold). This was orientated in the east/west direction to admit winter light through 20m high south facing glazing and would be supported by an air pressure of 0.4- 0.5 kPa to cater for adverse snow conditions. For reasons of social habitation the armadillo primary cable arrangement supported triple layer 5 x 3m inflated transparent ETFE foil cushions as the cladding module (instead of the conventional double ptfe translucent technology). Foil cushions had been developed by Stefan Leonhardt of Texlon Gmbh at Arnhem Zoo and for European Leisure Market. They have been subsequently used at the Westminster & Chelsea Hospital, London, the Hampshire Tennis Centre, Eastleigh, and Ferguslie Park, Paisley). Each in its small way is a forerunner to the Eden Dome, Cornwall (Grimshaw, Hunt), to be discussed in the second section of this book. At the time of his tragic death, Harald Mulberger of IPL was working on the design of a low rise 550 x 280m kidney shape air supported structure to cover industrial waste and minimise the treatment cost for contaminated water. The proposal was to use a very heavy grade pvc coated polyester fabricated in 20 x 20m panels with mechanical joints without a primary cable reinforcing grid in order to eliminate stress concentrations between membrane and reinforcing cables, and to provide a smoothe surface form. For flexible air-supported structures of these large spans, and even larger proposals such as "City in the Arctic" (Fig 9) (Otto, Arup Structures 3) [20], the structural function of the membrane becomes almost to separate an internal static mass of air from the external moving air mass. The implications of air / structure interaction are crucial in the context of avoiding divergence and the occurence of travelling surface waves under critical wind speeds, and equally to avoid the mobilisation of many thousands of tonnes of air through gross deformations of the structure. The implications are also potentially very important for the shape finding of both the surface structure and surrounding structures to minimise nonuniformity of the loading; so that form modelling as a whole might be regarded as a fluid / structure interaction problem [21, 22].

Fig 9 City in the Arctic

TENTS
The form and design of early modern tent structures was originally tied to a process rooted in physical modelling of readily replicable physical surfaces. They developed from the technology of the Stromeyer Circus Tent on the basis of minimal soap film surfaces and stretch fabric modelling, largely through the influence of Frei Otto [23]. These doubly curved, generally saddle shaped surfaces of varying curvature and often (with fabric models) widely varying ratios of biaxial prestress, were designed to resist different ranges of applied loading with a calculable and acceptable level of 'improving' deformation. The small wave form tents of cotton canvas for early aircraft hangars have developed through improved materials technology, numerical modelling and design techniques, into the dramatic yet inexpensive pvc coated Baltimore Harbour Lights Theatre (FTL) (Fig 10), and the simple arch supported entrance tent for Koln Garden Festival into the "Oleada" (Bird in Flight) entrance canopy for Expo 92 in Seville (IPL) (Fig 11). These more complex forms, using highly stressed heavier grade fabrics and rather intricate cable stressing and supporting systems, have become possible because of the development of interactive graphics and colour controlled stress evaluation which facilitate form-finding and stress optimisation [15, 24]. Detailing of connections for tent constructions has also become much more sophisticated [25, 26]. Taking a perspective of parallel development of membrane and fabric technology with the evolution of tent forms, one can see the gradual development from the geometric, flat patterns of the original circus and wedding tents into wave tents and the doubly curved, hump forms of 80 x 40m Dyce Tent (DRU, Otto, Arup) (Fig 12). Such designs used relatively weak cotton or polyester canvas in uniform 3ft widths, sewn with overlap chain stitched joints, with fields terminated on simple scallop boundaries of webbing (or cables). The

Fig 10 Baltimore Harbour Lights Theatre

analysis, coupled with advances in suitable materials have since led to structures such as the pvc coated polyester humped tent for the Imagination Building, London (Heron, Happold), the conical forms for the 420,000m2 Hajj Terminal in ptfe coated heavy grade glass fibre (SOM, Khan), and to the remarkable ptfe coated glass fibre enclosure for the Schlumberger research centre at Cambridge (Hopkins, Hunt, Arup) (Fig 14). The development of stiffer and more robust fabrics has also enabled the recent construction of forms with flatter surfaces than the traditional highly shaped saddle surfaces. Static and dynamic imposed loads are absorbed by elongation of biaxially prestressed panels, with deflections being controlled by pretension and increased material stiffness. Above all, this development, coupled to improved structural detailing and local patterning has led to the successful completion of a series of far flatter building forms - NEC Arena (Happold/Liddell) with a frame of steelwork lattice trusses at 12.5m centres

Fig 11 The Oleada Main Entrance, Seville

double curved surfaces were thus achieved by changing the angle between warp and weft fibres of the original fabric without elaborate patterns; made feasible because of the very low shear stiffness of the open weave fabrics. The 90m x 60m free form aviary at Munich (and similar structures at San Diego and Hong Kong) (Fig 13), formed from 3.2mm stainless steel wire mesh supported off tubular masts and curved concrete boundary walls, might be considered as a sophisticated development of the Dyce humped tent technique. By 1980, analytical methods had advanced to enable numerical erection of such forms from the original flat geometry to the double curved "cloud" and the production of stress patterns and detailed definitions within the zig zag pentograph at the mast tops - but still however relying on conceptual physical models, hard copy plots and numerical output. Developments in patterning, form determination and

I
I

Til
Fig 12 Dyce Tent

12

for static equilibrium. The scaling up to the real structure avoided tolerance problems by the inclusion of sufficient turnbuckles in the surface net. Its prototype, the IL building at Stuttgart University, remains an outstanding example of permanent cable net construction and has recently been restored. This technology [27] was extended to the Munich Olympic Stadium (Behnisch, Otto, Leonhardt & Andra) and more directly to the 120 x 90m Jeddah Sports Complex (Gutbrod, Otto, Happold, Linkwitz). Here the nomadic saddle shape was divided in 9 separate fields of cable net using galvanised and polyurethane dipped 12mm cables and galvanised drop forged clamps. For safety during construction the cable net is a 500mm grid (Fig 15a). Each field is connected via 3 8 0 galvanised steel wire ropes to groups of 2 or 4 3 8 0 composite ridge or stay cables continuous through supporting inertia saddles on the 4 southern and 4 northern tubular steel masts. Prestress was installed by use of sand jacks at the base of each mast. The lowly lit internal environment is created by 9500m2 of translucent external skin of ivory coloured specially fabricated pvc coated polyester grid

Fig 13 Munich Aviary

Fig 14

Schlumberger Research Centre, Cambridge

provides an internal clear span of 63m. The pvc coated polyester fabric in 23 x 12m sheets is prestressed in position to form 'stable' flat sides to the Arena. For summer use only, the 150 x 120m RSSB audience tent as a meeting place for 20,000 people is constructed from a primary system of internal masts and highly tensioned straight cables to which intermediate flat panels of pvc coated polyester fabric are attached - in its way a forerunner to the Millennium Dome at Greenwich (discussed in section 3 of this book).

Fig 15a

Cable net for King Abdul Aziz Stadium. Jeddah

CABLE STRUCTURES AND NETS


Nowicki and Severud's Saddle Surface Cable Net for the Raleigh Arena, discussed in the paper by Bob Silman, inspired the work of the young Frei Otto and led to a very fruitful collaboration with tent builder Peter Stromeyer. Early low strength cotton based membrane materials were limited to unreinforced spans of less than 10m, so their proposal for the sail structures at Lausanne Expo '64 was for cable net reinforced cotton roofs, designed through physical models and hand analyses. Even the subsequent German Pavilion for Montreal Expo '67 with its 50 x 50 centimetre cable net of 12mm cables, tubular support masts and underhanging light pvc coated membrane, was designed entirely on the basis of 'accurate' physical models followed by photogrammatic restitution of the boundary zones and simple calculations

Fig 15b The Clad Stadium

weave, supported on plates off the cable net and a lighter internal fabric hung off 'bretzel' supports as at Montreal (Fig 15b). The double layer membrane was designed to naturally vent the external heat through the intermediate cavity up to the mast - hence reducing internal cooling requirements [28].

13

The conceptual design of this structure was again based on physical modelling techniques ranging from simple chain link systems, through delicate soap film studies, to very accurate models using relatively inextensible silk screen material. This provided the architectural form and also the "target" for a form matching and analysis procedure based on Dynamic Relaxation, later to be developed into a complete CAD system, but in this case purely to provide the required precision for analysis and simulation of the erection and stressing out process (which entailed no surface turnbuckle adjustments). A more conceptual use of these numerical methods is in form-finding by physical and numerical modelling in parallel, with the simultaneous display of form and stress distributions guiding design decisions (see paper 2 of this book). The Jeddah Sports Centre required 8 kg/m2 of steel for the cables and clamps and a further 8 kg/m2 in the anchorages and masts in order to support the two membrane layers which weighed 1.5 and 0.8 kg/m2. In contrast the 84m x 48m clear span tubular lattice roof to the Olympic Swimming Pool, Kowloon (Walker, Kwan, Buro Happold), although for significantly heavier live loadings, required 65 kg/m2 of primary tubular steelwork and 21 kg/m2 of secondary steel in order to support its ridge and furrow roof of metal decking and 40% single glazing. This illustrates from the point of view of materials consumption, the primary difference between "flexible" and "stiff" structures in respect of their requirement for embodied structural materials. A similar contrast could be made between the Millennium Dome at Greenwich and the Millennium Stadium at Cardiff, although of course with their differences of purpose and expectation. As analytical techniques and knowledge improve so designers can evolve practical constructions further. Following the precedent of the IL building, the Diplomatic Quarters Club in Riyadh has two tiled and insulated shallow conexes supported on orthogonal cable nets [29]. More recent and sharply curved orthogonal nets have been used for the Aviary in Victoria Park, Hong Kong (Arup) and for the undulating wall for the German Pavilion at Expo '92, Seville (IPL) with its standardised glazed shingle cladding (Fig 16). In the US, Buckminster Fuller's 1954 patent for tensegrity structures has led to a number of interesting cable domes. That at St Petersburg, Florida (HOK, Geiger) has a diameter of 230m and more recently the oval dome for 70,000 spectators by Weidlinger at Georgia, Atlanta spans a full 250 x 180m (Fig 17). The largest and most recent cable dome structure is the Millennium Dome at Greenwich with a diameter of 320m. In contrast to the multiple flying struts of the foregoing "tensegrity cable domes", the masts at Greenwich extend from the ground through the structure to support a purely tensile radial net and membrane surface. This resulted in a very economical structure discussed in detail in later papers by Mike Davies, Ian Liddell and others.

Fig 16 The German Pavilion, Seville

Fig 17 The Georgia Stadium Cable Dome Plan

GRID SHELLS AND COMPRESSION VAULTS


The elegance of the grid shell at Mannheim (1975) was that it was formed from a double layer grid of continuous finger jointed 50 x 50mm hemlock laths at 500mm centres (Fig 18). These were laid out flat, hoisted up and bent into the precalculated funicular form required to support permanent self weight loads in direct compression [30]. Thus the effects of creep and shrinkage on the stability of the resulting free form shells were largely avoided while the cross grain moisture movements were allowed for by the use of cupped shaped washers on the nodal bolt detail. When doubled in layer and diagonally stiffened the resulting free form structures of a grid shell possess some of the load carrying characteristics of a pure shell without the accompanying cost of construction. More recently, as a demonstration of environmental sustainability, timber gridshells were felt to be suitable for the Earth Centre in Doncaster (1998). These use a single layer of green oak lathes 50 x 16mm spliced together into suitable lengths. These were laid out flat on the ground in a 400m square mat and bolted together, trimmed to a pre-determined shape established by the combination of physical modelling and computer generated hanging forms. It was then coerced on site into the correct form by adjustable props and cable 'Tirfors' and fixed to specified boundary positions.

Fig 18 Timber Grid Shell, Mannheim Fig 20 Al Faisaliah Banqueting Hall

As a result of the development of numerical methods to assess buckling resistance of splined structural systems and to determine their precise initially strained forms [31], two gridshell projects are being designed at present. The 45m x 15m amphora shaped grid shell for the Weald & Downland Museum, Sussex (Cullinan, Buro Happold, Green Oak Carpentry) (fig 19) is formed by a double layer of 50 x 35 oak timbers fingerjointed into continuous lengths and stiffened diagonally by a top layer of lathes (instead of the alternative of 6mm of cables). The other is a grid shell of 12Omm0 paper tubes 75m x 35m at 1.0m centres on sand box foundations, stiffened by undertied timber ladders and end walls for the Japanese Pavilion, Hannover 2000. The building, complete with 'paper' enclosing membrane and honeycomb end walls (Ban/Otto/BuroHappold/ Takenaka), is designed to touch the earth lightly and to be 100% recyclable. In Riyadh beneath the landscaped plaza to the hotel, office tower and apartments complex of the Al Faisaliah Complex is a 63m x 81m clear span exhibition and banqueting hall (Foster/Happold) [32]. Pairs of arches of varying cross section and depth (maximum 1300mm) wishbone shaped in plan, spring from buttresses 81 m apart (Fig 20). The void within each pair of arches formed between the upper and lower slabs provides the air supply ducts for air diffusers to the hall.

The benefits of Computer Aided Design (CAD) coupled to Integrated Manufacture (CIM), which have facilitated the construction of many elegant glass and steel gridshells, now make possible the reintroduction of massive vaulted forms where the effects of structural mass acting as a thermal flywheel can be accurately modelled using Computational Fluid Dynamics.

RETRACTABLE MEMBRANES AND SLIDING ROOFS


The earliest modern example of a demountable roof is really the pneumatic covering to the swimming pool part of the Sports Complex at Santa Barbara, California. In contrast to the adjacent permanent air supported sports hall of cable restrained ptfe coated glass fibre membrane, the seasonal cover to the swimming pool can be readily taken away for summer use on account of the great ease of foldability and handling of its lighter pvc coated translucent membrane. Of course deployable toldo and vela have been in use for centuries as enclosures for the courtyards of Islamic houses at Granada or in the streets of Seville for protection against the summer sun. Indeed this technique was extended by El Wakil and Rasch to cover the 36 x 54m courtyard within the new all brick dome structures of the Quba Mosque in Medina. A more recent version was proposed for the deployable membrane to be hung from cables from a central spine arch for the proposed King Fahd Sports Centre in Jeddah (HOK, Buro Happold). The retractable roof for the Theatre at Bad Herschfeld Castle in Germany (Otto, Romberg/ Leonhardt & Andra) evolved a design for deploying a membrane surface from a single off-centre central mast. This design was achieved largely by physical modelling. The dropper

Fig 19

Weald and Downland Museum Grid Shell

15

support points in their deployed position had to be carefully located beneath the radiating support cables so that the membrane could be deployed using individual crawler tractors and geometrical control systems. The relative height and position of the mast support point in relation to the geometry of the boundary supports, required to give the deployed membrane surface positive curvature against wind uplift and to prevent water/snow ponding, was difficult to achieve even at this relatively small scale. The subsequent scheme for Regensburg, Germany (Fig 21), indicates the height of the primary mast support that is needed to achieve adequate structural conditioning. Additionally the larger scale permanent uses for such a structure requires a significantly heavier grade of enclosure membrane which in turn makes "folding" more difficult to predict and control. Similar difficulties were encountered in the original design for the 20,000m retractable roof at Montreal.
2

In consequence of the innovations and advances to date, briefly described above, retractable systems that use tensile membranes as the primary envelope represent a structural form that could have enormous potential for future implementation provided good geometrical conditioning of the resulting membrane forms is achieved. This is especially so since forms of deployment can now be examined and adjusted using computer simulation systems. Retractable roofs offer space, largely for sports and leisure where in good summer conditions spectators can experience the open air. However they can also be

Fig 22

Retractable Umbrellas, Grand Mosque, Medina

Fig 21

Retractable Tent, Regensburg

Indeed designers need to spend a substantial part of their time studying the nature and arrangement of fold patterns for such schemes - to date very much a physical rather than a numerical art. Even this restriction is about to change given that Haug and others have developed iterative numerical techniques to "observe" the instant deployment of parachutes and air bags installed for driver security in cars. [33]. As materials develop, solutions develop. Nine 17m x 18m retractable umbrellas enclose the courtyard of the Grand Mosque at Medina (Rasch, Happold). They use an open weave teflon (ptfe) 'Tenara' fabric deployed in inverted form beneath high strength folding steelwork arms, supported from a single central cantilever column. The 'Tenara' fabric is resistant to fire and to intense ultraviolet radiation and is protected when the parasol is furled by a shell similar to that which protects a folded insect wing. Integrated systems of hydraulic actuators are built into the main body of each structure alongside supply ducts for cold air circulation, loudspeakers for public announcements and lighting (Fig 22).

offered protection in the advent of adverse weather during the staging of high profile events, exhibitions, pop concerts etc at such venues. Many of these large scale structures have been constructed in the US and Japan. However opening roofs have not yet provided the answer to the health of grass playing surfaces. At the 20,000 person Gelredome, Arnhem (HBG) not only does the roof close but also the tray containing the grass playing surface is capable of sliding out beneath the south stand on the teflon bearings so as to enjoy the full photosynthetic effect of the sun. Because retractable domes are very large heavy 'truss spanning' type structures, their action when closed as well as during deployment needs special consideration. Their design and technology has been reviewed by Prof Ishii [34]. One of the first such complex projects is the 51,500 seat Sky Dome, Toronto for basketball and other uses. The C$532 building was completed in 1989, includes hotel and conference facilities and has a maximum span of 674 ft (Fig 23). The fully retractable roof weighs 11,000 tons, covers 340,000 ft (8 acres) and consists of four panels, 3 of which are moveable. Panels 2 and 3 slide over each other along straight rails. Panel 1 rotates 180 degrees moving along a curved rail to the space between panels 3 & 4. 54 drive mechanisms
2

16

operate the roof with a series of 10 horsepower motors (750Hp total). The roof moves at 72 ft/min and takes 20 minutes to open. The successors to this roof and others are discussed in Section V of this book.

8.

9. 10.

11. 12. 13.

14. 15.

16. 17. 18. Fig 23 Toronto Sky Dome 19.

SUMMARY
What is an enclosure? It results from physical construction which is deployed in order to moderate the interior climate to be comfortable for the uses to which it is put in all weather conditions. Since these spaces are widespan they need to be able to support a wide range of external wind and snow loadings and other superimposed loads as well as to incorporate within their design the effects, sometimes significant, of thermal expansion, contraction and relative settlement. Above all, designs of such enclosures need to respect their urban location and express to the user the sense of delight. We hope that the papers for this symposium will convey that same sense. They are all invited papers submitted by internationally acclaimed architects and engineers which contain valuable precedents for the future.

20. 21. 22.

23. 24.

25. 26. 27.

28. 29.

REFERENCES
1. 2. 3. Frei Otto, "Das Hangende Dach", 1954 Ullstein A G Berlin, reprinted 1990 Verlag der Kunst Dresden ISBN 3.364 001987 Philip Drew, "Frei Otto: Form & Structure", Granada/Hatje 1976 ISBN 0258 970537 A Holgate, "The Art of the Structural Engineer - the work of Jorg Schlaich and his team", Axel Menges, 1997, ISBN 30930698-67-6 J Joedicke, "Shell Architecture" , Karl Kramer/Tiranti, London 1963 B Forster, "Cable and Membrane Roofs - A Historical Survey", Structural Eng Review Vol 6, n 3-4 pp 145-174 (1994) Elevier, ISSN 0952-5807/94 Kazuo Ishii, "Membrane Structures in Japan" , SPS Publishing Co 1995 ISBN 4-90540-05-06 Bethlehem Steel: "Cable Roof Structures" , (No 5 Roofs), November 1968

30.

31.

32.

4. 5.

33.

6. 7.

34.

M Dickson, P Westbury , "Intelligent use of Longspan Structures for Tomorrow's Environment" . Developments in materials, forms and techniques - Engineering a New Architecture, Aarhus, June 1998 Crosby, Burkhardt, Kendel et al: "Frei Otto at Work". Architectural Design, Vol XL I, March 1971 Mick Eckhart, "Frei Otto and Munich Olympic Games - From Measuring Experimental Models to Computer Determination of the Pattern", Zodiac 21 ed Maria Bottero, Olivetti 1974 Happold, Dickson, "The Story of Munich" . Zodiac 21 Architectural Design 6/1974 SFB65 International Symposia. Stuttgart. No 1 Vol 1-3 (1976). No. 2 Vol 1-2 (1979), No. 3 Vol 1-2 (1985) H Argyris, T Angelopoulos, B Bichet, "A general method for shape finding of lightweight structures". International Conference on Tension Structures, London 1974 E Haug, G H Pavell, "Finite element analysis of non linear membrane structures", I ASS Symposium, Tokyo 1971 Michael Barnes, "Form and Stress Engineering of Tension Structures", Structural Engineering Review, Vol 6. n 3-4 Nov 1994 ISSN 0952 5807. A S Day, "An introduction to Dynamic Relaxation" , The Engineer 219 (1965) The Institution of Structural Engineers: "Air Supported Structures: the State of the Art" June 1980 The Institution of Structural Engineers: "The Design of Air Supported Structures", Bristol 1984 Ian Liddell, "Minnesota Metrodome - A study in the behaviour of air supported roofs under environmental loads". Structural Engineering Review Vol 6 n 3-4, November 1994, ISSN 0952 580794 IL2 : "Stadt in Der Arktic - City in the Arctic", Institut for Leichte Flachentrawerke, University of Stuttgart, 1971 C J K Williams, "The Design and Analysis of Surface Structures", PhD thesis. University of Bath, 1991 M Barnes and C J K Williams. "Design and analytical aspects of very wide span flexible structures", in Developments in Computer aided Design for Structural Engineering, Civil-Comp Press, 1995 Stromeyer, "Textiles Bauen", (einBegriff fiir leichte Flachentragwerke) - L Stromeyer GmbH (c 1973) M Barnes, "Form Finding and Analysis of Tension Structures by Dynamic Relaxation", Int. Journal of Space Structures, Vol 14 n 2 , 1999 Torvald Faegre Tents, "Architecture of the Nomads", John Murray (Publishers) 1979 Bubner E , "Membrane Construction - Connection Details" , publ. Wehlemann GmbH Essen, ISBN 3-9805790.7.7 Fritz Leonhardt, Frei Otto: "Prestressed Cable Net Constructions - the Munich Olympic Roof", SFB64, 1976, (translated from German) Pugh E, "King Abdul Aziz University Sports Hall, Jeddah", in Patterns 5 published by Buro Happold, May 1989 Happold, Ealey, Liddell, Pugh, Webster , "The design and construction of the Diplomatic Club, Riyadh", The Structural Engineer Vol 65 A , January 1987 E Happold, W 1 Liddell, "Timber Lattice Roof for Mannheim Bundesgarten", The Structural Engineer Vol 53 No. 3 , March 1975 S. Adriaenssens, M. Barnes and C. Williams, " A n e w analytical and numerical basis for the form-finding and analysis of spline and grid-shell structures", in Computing Developments in Civil and Structural Engineering. Civil-Comp Press, 1999 Dickson, "Developments in Structural Form to Minimise Environmental Impact", in IABSE Symposium, Rio de Janeiro - Structures for the Future - The Search for Quality Haug E, Dubois J, et al: "Transport Vehicle Crash - safety and manufacturing simulation in the perspective of high performance computing and networking". Future Generation Computer Systems, V10 n2 - 3, June 1994 Kazuo Ishii , "Retractable Roof Structures in Japan". LSA International Congress 1998

SECTION I Form and structure

On Frei Otto's Philosophy for Widespan Lightweight Structures

Form and Stress Modelling of Tension Structures

The Definition of Curved Geometry for Widespan Structures

Concepts and Reliability in the Design of Widespan Structures

Wind and Snow considerations for Widespan Structures

19

ON FREI OTTO'S PHILOSOPHY OF WIDESPAN LIGHTWEIGHT STRUCTURES


Michael Dickson

This is a text prepared in collaboration with Professor Frei Otto and based on a presentation given on his behalf at the inaugural session of the Bath University Symposium on Widespan Enclosures in April 2000. Frei Otto's long career in lightweight structures includes the development of stressed tensile sails for the Lausanne EXP064, the distinguished German Pavilion membrane and pre-stressed cable nets for EXP067 in Montreal, the Munich Olympic Roofs in 1972, and the Gridshell at Mannheim in 1975. The conceptual design studies for these and many other projects were carried out at the Institut fur LeichteFlachentragwerke (I.L.) which Frei founded at the University of Stuttgart. Between 1967 and 1995 he worked often with Ted Happold, a friend and fellow spirit, on such projects as the 120m x 90m cable net structure for Jeddah University and the Diplomatic Club, Riyadh, (Aga Khan prize for architecture 1998). Professor Otto's current work includes consultancy on the Venezuelan and Japanese pavilions for the EXPO at Hannover 2000, and conceptual design for the new railway station for Stuttgart 21. He is one of the leading innovators in the development of lightweight structures and has recently received the following international prizes: 0 Honda prize for ecological technology 1990, Tokyo 77ie* principle prize of the German Institution of Architects and Engineers in Berlin, 1996 The Wolf Prize for Arts, Jerusalem, 1997

material usage are the fundamentals in securing this aim. Yet in the solution of this theme the use of large spans is not just a game to make the Guinness Book of Records but a search for real solutions for mankind. To know about large spans also opens opportunities for advances for shorter spans. In the absence of scale factors on short spans it is possible to use material less effectively. Conversely for the larger spans, it is necessary to seek out fundamental 'absolutes' of performance and to recognise the significance of 'scale' and the problems of enclosure. Part of this search is the recognition of optimal performance and benefits of different structural forms in ascending order of opportunity - so this paper tackles the fundamentals of performance of successive structural types - bending structures for smaller spans, lattice structures, gridshells and compression vaults, tension structures and finally the opportunities for pneumatic structures. These structural systems are discussed and illustrated principally through a wide variety of projects.

BENDING AND LATTICE STRUCTURES


The scaffolding lattice system for roofs was devised to avoid the volume of material that would have been required of a 'bending' structure. Small diameter compression or tension tubes in a three dimensional lattice transfer the roof loads back to a few columns for the 102m x 52m membrane covered Interbau Buildings Berlin 1961. This was a 'first' for Mero and in a way the precursor to the many lattice space frame structures by Kenzo Tange at Osaka 1970 (fig 1).

He is an Honorary Fellow of the Institution of Structural Engineers and of the Royal Institute of British Architects, and Dr of Science honoris causa at the University of Bath.

INTRODUCTION
In both the developed and the developing world, widespan enclosures are increasingly required to house and facilitate many of the collective activities of society. Such enclosures need to do this without drawing down excessive quantities of scarce construction material or drawing upon unnecessary quantities of energy in their operation. To ensure such aims requires an efficacy of construction, a delight in their occupation and appropriateness to their location. Beauty of architecture, efficiency of structural form and appropriateness of
Fig 1

20

Prefabricated standardised galvanised Delta units and 'bolted' cross nodes made up the 42 cm deep intermediate viewing platforms for the German Pavilion at Montreal (1967). Engineered by Leonhardt and Andra the cruciform head units positioned diagonally across the grid concentrate loading from the floor grid onto the column top, each element as in the human body prepared for its particular duty (fig 2).

Fig 3

Fig 2

DIRECT FORCE STRUCTURES


Inescapably, the most efficient way to transfer load back to foundations is by a 'direct' way - an inclined straight spar. Early investigations with the mushroom support 'spars' to the 'humped' tent at Koln 1957 led to studies for radiating 'fan' systems for the Transrapide Maglev viaduct system (figure 4a). It should be noted in passing that the alternate form to the nose of the model capsule

In the 24m high Bell Tower for Berlin (fig 3), the architectural form of the virendeel truss is retained while the plate thickness is aggregated from 10mm at the top to 50mm at the bottom in order to restrain the drift of the Tower to 16mm under the ringing action of the 3 bells function following form:

21

on the bridge is itself a holistic proposal to reduce wind resistance at speed, hence the required magnet power and therefore also the weight to be supported by the bridge itself. The purpose of the viaduct design for the Maglev was to reduce the impact of the linear induction Transrapide support system on the countryside of Northern Germany from Hamburg to Berlin. The Transrapide Maglev is a light multicar system capable of travelling at up to 450 km/hr using the technology of aircraft systems. Breaking loads from emergency deceleration are more critical to the support structures than vertical loading. Structural continuity and close accuracy of construction, allowing also for thermal distortions, is essential for ride comfort - hence the structural concept of a minimal triangulated tubular network casting little shade on the ground below and supporting loads onto simple foundations.

Fig 6

But the aesthetic of a design in its surrounding is also of great importance - the tree fountain on its well in Warmbronn drips its water carefully into the well (fig 7). Sometimes a symbol will be sufficient. For the one day meeting of the Evangelic Church at the Berlin Olympic Stadium (1961), only a single 40m high guyed cross structure was needed to express the enclosure (fig 8).

Fig 4b

A further development of these thoughts has led to the fan pedestrian bridge system for Gelsenkirchen 1999 (fig 4b). In line with earlier studies of bamboo structures at the I.L., the various spans of this radiating system are to be made from kit form of solid 70mm galvanised bars and 4-bolt cup-clamp systems. Individual buckling lengths are to be reduced by an internal criss cross of stabilising bars - also 70mm 0 . The 1960 study at Yale for a thin roof did so by dividing the individual spars to form a "triangulated" network of stable compression elements (fig 5). For the Council of Ministers project in Riyadh (1978) the loading from the 3'D' hexagonal grid shell for the seating bowl is gathered by irregular triangulated configurations of tubes of successively increasing diameter. These match the buckling length restrictions to the requirements of increasing load back onto a single composite column of 3 individual braced tubes (fig 6).

Fig 7

fey

L 1 iA ^ 1
''ii'-tnnMi''

v,

'

Fig 5

Fig 8

22

FUNDAMENTALS OF MATERIAL AND FORM


In terms of the 'absolutes' of measurement, illustrations on a logarithmic scale relate the basic forms of stability of everything from mountains to grasses and hairs - a i m high grass has a H/D of 100 or more (fig 9). Of consideration too, for all enclosure tension studies, are the fundamental rupture lengths of different materials under their self-weight - wood is the leader.

Other studies have also shed light on fundamentals of performance: What stable forms does sand create when allowed to run away? (fig 13). What are the laws of form for spine structures (fig 10) and for hanging vaults? (fig 11).

1
1
f QB'a
!

/ f
y

///
r \ f
| * 1

1' 1 i m
L.

,J f 7h ? ; i

s
* -r-

'

' j
i

'j

Fig 9

Fig 13

Based on the historic shells from Harran (fig 12) what crucial forms from local brickwork can resist the lateral forces of an earthquake? - as measured on the tipping table, the cone of make 0.3g (30).

Fig 11

Fig 14

23

The proposals for the naturally light and ventilated forms for Islamic University at Uzbekistan, constructed solely of bricks is the outcome of such studies (fig 14). To optimise bridging, theoretical studies show that you can bridge 10 miles so probably at least 1/10 of that can be achieved in reality. How do vaults really work? In the vousoir model, it is to be noted that the zig-zag string transmits the 'shear' for stability (fig 15).

Indeed the study of arch forms led directly to the form of the openings in the supporting walls of the Diplomatic Club, Riyadh [now CasaTuwaiq] (fig 18).

Fig 18

S H E L L S , G R I D - S H E L L S A N D VAULTS
In 1958, with the help of students at Washington University, a rubber membrane weighted with nails was used to investigate forms "without bending". Such forms were then stiffened with plaster and inverted into a shell form. There followed the single layer timber gridshells for Essen (1962) (fig 19) and that by students at Berkeley (1962) constructed out of steel rods and washers from the hardware store (fig 20).

Fig 15

On the tipping table, lower arch forms are more stable than high forms. Even arches can be curved in plan, (figures 16 & 17).

Fig 16

Fig 20

24

The bending free grid-shell form is really a low cost construction method for creating complex forms for public space. An early example is the auditorium of the German Pavilion, Montreal (1967) prefabricated in Germany and drawn out into its final form on site (fig 21). This was a forerunner to the minimal energy house designed for Ted Happold. Here oak lath gridshell, turf covering, south facing glass wall, pv cells and wind generation are all part of a holistic approach to design (fig 22).

end walls. In turn these were used to attach the paper membrane. All components including the "sand box" foundations were designed to be easily recyclable and so give an enclosure which specifically "touches both the 'planet' and ground lightly" (fig 25).

Fig 22

With the help of the computer, there are now forms which are difficult/impossible to model physically - the naturally light and ventilated workshop in Dorset for John Makepeace of roundwood spruce trees formed by green bending the tapered green debarked trees is one (fig 23). Another, the Japanese Pavilion at Hannover with Shigeru Ban is in reality only "findable" on the computer even though here physical modelling gets close to the final form (fig 24). An originally flat grid of 12cm diameter paper tubes banded together in a 1 metre grid is pushed up to form a bended amphora form subsequently stiffened by the cable - undertied timber ladders and diagonally braced cable formed honeycomb

Fig 25

The double layer gridshells for the Bundesgartenschau, Mannheim designed by Frei with Mutschler, Langer, Happold and Liddell were most courageous enclosures and extremely inexpensive (fig 26). So inexpensive that Kikutake followed them with a much larger 'shell' for the Japanese Silk Road Exhibition in 1988 (fig 27).

Fig 26

'Inversion' of the tension eye for Montreal and the I.L. (fig 28) led to the development of the compression forms for the new below-ground naturally light station beneath the Schlossgarten, Stuttgart with Ingenhoven and Buro Happold/LAP (fig 29). The inverted forms modelled in plaster span a grid of 60m x 30m using only a concrete vault 35cm thick at the crown thickened to 65 cms around the eye. Indeed each pier supports of the order of 35,OOOkN of loading from the landforms above. Recent form models for Stuttgart 21 envisage inexpensive construction techniques using propped timber grid shell forms (remember the bending free forms of Mannheim) to create the free vaulted form from the plasticity of wet reinforced concrete.

Fig 27

26

HANGING STRUCTURES AND DEAD WEIGHT FORMS


Simple hanging forms are able to exploit the effectiveness of the long rupture lengths of tension fibres - especially if they can be stabilised against disturbing loads by self weight, damping or enclosure. Early studies for a pagoda roof 1983 previewed the prototype house at Hooke Park with Richard Burton. The hanging roundwood spruce thinnings curved down under dead weight are opposed by A frame compression spans (fig 30). The elegant Wilkhahn factory with Gestering architects and Speik und Hinkes engineers for timber products in its agricultural landscape uses a similar philosophy but employs square sawn timber (fig 31). Really this was a focussed study in the use of minimum embodied energy and of minimum operational energy in the industrial context - built in the countryside.

Fig 32

More daringly, the wind tunnel at Teddington was used to investigate the stability of the proposed hanging roof for covering the Berlin Olympic Stadium. Solid steel rods supported on tension cables add sufficient weight to the plexiglass forms (1973) (fig 33).

Fig 33

PRESTRESSED TENSIONED ROOFS


At the heart any discussion of prestressed tensile roofs are the many studies that contrast tents with a central support point and radial cutting patterns to those with double curved saddle and sail surfaces into which eyes, rings or mushroom supports are introduced. The Riyadh Heart tent (1986) with its radial spider net of stainless steel cables supporting painted glass panels onto a central mast (fig 34) is diametrically different to the 40 x 30m Berlin humped tent of deformed cotton canvas tied down at the edges but supported on a series of mushroom supports (1957) (fig 35). The 55m radial patterned squares of the Hadj tents by Fasler Khan of S.O.M. supported on central cable-hung rings of ptfe glass fabric by Chemfab/Birdair are forerunners to the triple layer central supported tensile enclosure for Storek Furniture in Leonberg (2000) (fig 36).

Fig 31

Both are predated by the aluminium covered, heavily insulated auditorium for Mecca with Gutbrod/Arup/Happold. The 22mm 0 spiral cables hanging from the central steel portal are cross connected by double angles that support and distribute the loads of the insulation and cladding and contain the enclosed air volume (1968) (fig 32).

Fig 34

Fig 37

The 36m 0 wave tent for Dance Stage at Koln 1957 is now also a protected structure for its six exceptionally slim supporting batten masts, each externally guyed to separate foundations (fig 38). To save on foundations, the high points of the waveform for the Biennale at Venice (1996) use A-frame masts to transfer loads to foundations shared with the tie down (fig 39).

Fig 35

Fig 38

Fig 36

The first wave form system had 3 parallel spans of 15m and was patterned by overlapping cotton canvas to create the enclosure (fig 37) for a flying priest, Pater Schulte. Multipurpose, it doubled by day as a translucent place of worship and at night as a covering to his small aircraft!
Fig 39

28

Tension structures offer a huge opportunity for very longspan lightweight structures. The recently restored building of the Institut fur LeichteFlachentragwerke (IL) is now also a listed building (fig 40). Originally this was the prototype eye structure for the many masted freeform translucent white pvc enclosure for the German Pavilion at Montreal (1967). The patterned membrane was hung underneath the cable net of 12mm galvanised cables at 500mm centre on springy bretzels (fig 41). This remarkable cable net construction brought the skills of Gutbrod architect, Fritz Leonhardt engineer, and Peter Stromeyer, tent maker and manufacturer together with Otto to create a longspan building that brought with it a paradigm shift in cable net technology. This technology was then transferred to the 120m x 90m double membrane cable net enclosure on eight masts for the Sports Centre, King Abdul Aziz University, Jeddah (1978) (fig 42). Clamped anchorages and chizel point masts and 'teller'plate membrane supports were introduced here.

Fig 42

the mesh over an existing ash tree. The particular form for this Voliere was devised to facilitate flight and resting patterns for the ornithological occupants within a natural landscape. The architect for Miinchen Tierpark was Jorg Gribl.

Fig 43

Another protected structure is the Olympic Roof at Munich (1972), by Behnisch, Otto and Leonhardt. These structures with their first use of the flying mast laid the corner stone to the wide-ranging research (SFB64) on long span structures directed by Leonhardt, Otto and others in many departments of the University of Stuttgart (1975-1985) (fig 44)..

Fig 41

The plan of the multimasted Voliere at Munich (1980) is reminiscent of Montreal but the doubly-curved snow supporting stainless steel woven mesh gets its form from the earlier humped tents at Berlin and Dyce (Aberdeen) 1975 (fig 43). Here computer visualisation enabled development of the zigzag eye form required to support
Fig 44

29

DEMOUNTABLE STRUCTURES
Historical studies at the IL on Roman Vela and Islamic Toldo's resulted in the design and execution of a number of demountable or semi demountable enclosures. Early projects at Cannes and Paris led to the demountable covering over the ruined church at Bad Herschfeld (1968). A central mast supports radial primary cables down which crawler tractors deploy the unfurling pvc membrane. A large scale study for deployment over the swimming pool at Regensburg (1972) demonstrates the importance of obtaining a construction geometry that is capable of delivering a well resolved geometry especially under the extreme load of wind, snow and snow ponding (fig 47). At Regensburg this was made possible by having a mast which was sufficiently high.

Fig 45

Along the way, the entrance arch tent originally of polyurethane covered glass fabric stabilising a single 150mm diameter tubular arch was a preview of the Otto/Tange/Arup proposal for arch supported cable net forms for the Kuwait Sports City. The design proposed only a 1.0m diameter steel pipe stabilised by a cable net roof for the 250m long main Olympic Stadium (fig 45).. The 1961 proposal with Leonhardt for a 1800m x 550m covering to Bremen Harbour employed high masts and primary cables supporting a cable net under which was hung a thin membrane skin (as subsequently executed at Montreal in 1967). This is probably the alternative technique for spanning spaces as large as that of the Dome at Greenwich (1999). In a way, the 80m x 40m study for a pvc covering at Sullom Voe, Shetland Islands (1981) which is supported from a number of masts by arrays of single cables is a scaled down version of the design for Bremen (fig 46).

Fig 47

PNEUMATICS
Studies and calculations in 1958 indicated that a factory of 3 pneumatic bubbles, made out of aluminium sheet might be able to enclose spans up to 800m (fig 48). Many fundamental pneumatic forms were originally investigated during research using flexible rubber membranes whose forms were "captured" by plaster casts (1960) (fig 49).

Fig 46

Fig 48

30

Fig 49

The creation of stable pneumatic enclosures is concerned with the differences of pressure between the internal and external medium - air, water, helium etc. At one level, air supported structures enable economic and safe enclosure of very large public spaces - City in the Arctic (1971) (fig 50) - or the project for "58 North". At the smaller scale for the Academieschiff, Berlin, positive internal pressures generate the stable screen for projection of images onto the largely cylindrical internal surfaces (fig 51) while on an earlier project negative pressure produces a concave spherical surface for external projection (fig 52).

Fig 52

The ultimate pneumatic structures are those of man himself - Bone is made as a liquid filled tension structure - a baby, a composite pneumatic structure, is gestated within a protective water sack. As a concluding thought such illustrations show that if we are to break open the discussions between architect, engineer, user and constructor, we need to attend to some of the 'absolutes of performance'. Reference to the underlying form of the many natural structures around us will help address the problem of achieving similar efficiences - or at least coming closer.

Fig 50

Fig 51

31

FORM AND STRESS MODELLING OF TENSION STRUCTURES

Michael Barnes

SYNOPSIS
This article gives a brief overview of the development of form-finding applications for modern tensile roof structures, particularly CAD methods with stress or fabrication control of form. Selected projects are chosen to illustrate the application to cable networks, woven steel mesh systems, prestressed mechanisms, coated fabric membranes and air-supported structures. The purpose is to review the form-modelling of direct force structures in which the form must physically be a reflection of the prestress distribution - to provide a background description, principally for architects who will work with engineers in the design process.
Fig 1

CONCEPTUAL DESIGN
The origins of the design of modern tensile roof structures are based on physical modelling techniques, particularly for the conceptual form-finding process [1]. These physical procedures have ranged, in order of accuracy, from minimum surface soap films, through stretch fabric nylon models and the use of specially formed hexagonal weave models, to the use of silkscreen fabric (with very low stretch) or uniform mesh wire models. The latter procedures were sufficiently accurate for fabrication patterning of simple structures provided that, for prestressed membranes, the material used in the real structure had adequate flexibility, or that boundary turnbuckles were incorporated in cable networks to adjust tension distributions on site. The most sophisticated design to be carried through to the stage of fabrication patterning using such techniques, with precisely measured wire models, was probably the cable network for the German pavilion at EXPO 67 (Figs 1 & 2), designed by Frei Otto in collaboration with Gutbrod, Leonhardt & Andrea, and Linkwitz. Whilst other structures may have been technically more precise through the use of analytical descriptions for prestressed shapes, the EXPO 67 pavilion represented a truly freeform system; and it was perhaps this structure, more than any other, which captured the imagination of both the architectural profession and the gneral public and led to the popularity of tensile structures.

Fig 2

During the design of the Munich Olympic Games stadium, it became apparent that even quite large-scale and accurate wire models of the network system could not be sufficiently precise for design purposes; and since this time, numerical methods for form-finding, load analysis and fabrication patterning of both prestressed membranes and cable networks have played an essential role in the engineering design process and in the development of conceptual models [2,3,4].

32

simultaneously by specified stress distributions in the surface for a weightless prestress state. The neglect of self-weight in form-finding is important since it allows the shape to be determined purely on the basis of stress ratios in the principal weave directions (the warp and weft of the fabric), with subsequent scaling of stress levels to satisfy load state design conditions (including self weight). For fabrication patterning, the "compensations" that must be applied in the factory cutting of initially stress-free fabric panels are also governed by the choice of design prestress levels and the warp/weft stress ratios, with allowances made for inservice load distributions and relaxation of the prestress.
Fig 3

COMPUTER-AIDED DESIGN
The aim of CAD procedures might be stated as the replacement of physical modelling techniques by computer programs enabling realistic graphical display of form and stress levels and interactive control of these aspects together with the patterning and detail design features. An "expert systems" approach incorporated in these CAD procedures might further provide a guide to conceptual design, and coupled with CFD procedures may eventually lead to far better wind load definitions for complex shapes. In reality, simple stretch-fabric physical models (Fig 3) continue to be useful for the initial form studies of complex free-form systems. In addition to the tactile value of such models in conceptual design, they provide a means of communication between various members of the design team and the best learning process for new members of a team. They also yield an estimate of surface curvatures and hence stress distributions and, when stiffened by a reinforced resin coating, the basis for wind tunnel model tests. Models may, although now unusually, also provide a guide to suitable patterning arrangements and mesh generation topology for finiteelement type analyses. For cable network systems, this entails the determination of preferred mesh orientation, generally with a uniform mesh (or grid of equal length links), while for prestressed membranes a model clarifies the choice of panels for fabric patterning - generally employing "geodesic" seam lines. These geodesies describe lines that follow a minimum distance over the surface - equivalent to the trajectories that would be followed by finite width tapes without shearing distortion. The geodesic seam lines can thus define the edges of shear-free panels for fabrication purposes. In most cases now, only an imagined physical model is necessary for the definition of an initial computer model. From an initial for membranes precise cutting the overall definition of topology, CAD procedures control the seam geometry to provide patterns with optimum use of material, structural form being governed

For cable networks the CAD process will start from an initial coarse definition of topology with equal length links in each net direction (apart from end links intersecting with curved boundaries). Tension ratios in each net direction are also specified to cuntrol the surface curvatures - as with membranes. However, the process is then usually one of gradual refinement - doubling and then qudarupling the mesh density - until a net is found with no untensioned links and sufficient accuracy for fabrication patterning; "sufficient" meaning that patterms in a large surface net can be obtained to the nearest millimetre - steel cables are far less forgiving than coated fabrics. During all of this process the topology must be changing, with links disappearing from boundaries in some regions and new links appearing in others. It can be seen from the above description of the interrelations between form, patterning, design loads and stress distributions that the use of CAD procedures is an essential feature in both the conceptual and developed design stages for tension structures; and that, together with changes in support geometry and topology, the specification of required stresses, or forces in discrete components, is of prime importance in the control of form.

Fig 4

33

Only the third type of net can theoretically exactly fit the membrane surface - provided that the equivalent membrane stresses are uniform and equal in all directions (as a soap fdm model). However, for all three types, the procedure yields a good starting topology for the net analysis and its form adjustments. The procedure is closely analogous to the initial generation of nets from fabric models used for concept studies. Figure 4 shows a fabric study model (for Gatlinberg community centre) which was subsequently fibreglassed for wind tunnel testing (Fig 5). The physical modelling in this case progressed in parallel with the numerical modelling which is shown in figures 6 - 9 . In this coarse CAD model the cable traverse spacings represent four grids of the real uniform link net and the colours represent tension levels: light blue or yellow for the target or low but acceptable tensions, dark blue for high tensions and red for slack (or buckled) links. At an early stage in form-finding from an initially flat net the tensions are rather random with slack or highly stressed links concentrated in the boundary or mast support areas (Fig 6). The first true equilibrium state is shown in figure 7. In arriving at this state, as much as possible in the CAD process may be automatic - for example, in short traverses with shallow curvature the average tension can be checked and end link lengths adjusted automatically to meet target values. But for longer traverses with more complex curvatures, increased stressing at their ends in an attempt to eliminate large slack areas (see buckled band in fig 7) will merely over-stress other areas of the net, particularly around mast supports. In such cases, a substantial shearing of the net and/or adjustment of mast configurations is necessary in order to force the slack bands to follow greater distances over the surface - the adjusted prestress state, with no slack or over-stressed links, is shown in figure 8. After evaluating network tensions under wind and snow loadings further adjustments of the form may evidently be necessary; for example, under snow loading, the tensions around the hoop and radial support cable ring of the central mast were too high, and the effective bearing diameter of this ring was therefore increased (fig 9).

Fig 5

Fig 6

CABLE NET AND UNIFORM MESH STRUCTURES


The initial form investigations for cable networks and analysis checks for various load states may often be based on an equivalent membrane model with appropriate properties and with one of the cable traverse directions approximately parallel to the warp (or geodesic direction) lines in each surface region. A coarse grid cable-net model may subsequently be numerically assembled over the membrane surface by choosing one of the warp lines as a control traverse, from which the network can be set out link by link until intercepting with the boundaries. Networks numerically constructed in this way may be of three types: (a) uniform grid nets (the most commonly used in practice); (b) geodesic nets - since all warp lines in the membrane surface are geodesies, one set of cable traverses will be parallel to these lines throughout the surface and the other set can be assembled so that angles of incidence onto the first set are equal to their angles of departure; in such geodesic nets all cables have constant tension along their lengths in the prestress state. (c) hybrid nets in which one set of cables are geodesies over the surface with equal link lengths and the other set have constant tensions throughout their lengths.

Fig 7

34

Fig 10

Fig 9

The form-finding and analysis of large or finely spaced networks may be carried out using comparatively coarse grid models. For fabrication patterning, corrections to the grid lengths of the numerical idealization must subsequently be made to allow for local curvatures. This may be achieved by fitting splines thriough the grid traverses and shorteneing the slack grid lengths by the difference between surface arc and chord lengths. The final stage of patterning involves precise boundary and mast support zone analyses using a fine grid spacing, with geometry and tension distributions interpolated from the overall analysis to set the local zone analyses. A structure which perhaps more vividly illustrates the concept of net angle (or shearing) distortions to fit a surface shape is Munich Aviary (Fig 10). The continuous surface is formed from butt welded constant width rolls of crimped stainless steel fine mesh (Fig 11). The wires are held in place mechanically by the crimping, but allow in-plane shearing angles of up to 30. The feasibility of constructing such a system without wrinkling (and hence, in this case, parting) of the mesh depends on the following factors: (a) the number and height of masts and tie-down systems; (b) the number of convex indentations and concave curves forming the plan boundary geometry (in relation to the number of masts); (c) the freedom given, assuming a fixed plan, for adjustments in the heights of the mesh attachment along the boundary curve; and (d) the precise arrangement of the trellis supporting plates around each mast or tie-down (Figs 12 & 13).

Fig 11

Fig 12

Fig 13

Fig 14

35

4 *

A simple analogy is of laying an open weave cotton net over a sphere: at the top the net will be orthogonal, but as the net progresses down the sphere it must shear to hold to the surface. At a particular depth parts of it must depart outwards from the surface otherwise it will wrinkle. In contrast to the structures reviewed above, the concept of overlapping shingle plates for the Radolfzell concert sail dictated that as many plates as possible were standardized within a tolerance governed by the size of the clamp bolt holes (Figs 14 & 15). The 0.55m cable grid was thus arranged to be nearly orthogonal throughout the surface. To achieve this the surface was split into regions by means of ridge cables which also limited deflections and provided security for the main mast. A similar but more free-form structure was proposed for the Munich Zoo large cat enclosure (Fig 16). The orthogonal net orientations for these shingle structures are clearly dictated by the drainage requirements. Since the sides of plates are parallel with the net cables the ideal orientation is at 45 to the steepest gradient direction, though with a practical tolerance of 20. An alternative which could provide a stiffer structure, with the net traverses more closely following lines of principal curvature, would be to orientate the net at 45 to the plate edges, although perhaps at the expense of aesthetic appearance. Since the shingles are physically attached to the net only at opposite corners this also suggests the use of a hybrid geodesic type net (type C referred to earlier). A third alternative of a true principal curvature net is probably impractical from the point of view of standardization of the cladding panels.

Fig 16

36

The same glazed shingle system as in the Radolfzell concert sails was used for the undulating wall enclosure of the German pavilion at EXPO 92 in Seville (Figs 17 & 18). Again, a major objective was to employ the greatest number of standard panel units. In this case, however, there were to be no regional splitter cables; the entire surface was continuous. Instead of splitting the surface into regions, two geodesic bands (AA and BB in figure 19a) were employed to induce more equal tensions in the longer set of traverse cables (inclined at about 30 to the horizontal). To disguise these "tucks" in the otherwise uniform network the necessary shortening of the traverses was spread over three consecutive links in each geodesic band - figure 19b shows a region of the final network viewed normal to the plane 123. Whilst the shingle panels adjacent to the band AA are non-standard, the great majority of the surface panels are orthogonal. The main roof of the pavilion was a pneumatic lens membrane structure 90m x 50m, with an elliptic internal boundary truss, which was intended to give the impression of floating over the main assembly / functions area (Figures 20 and 21). The structure was supported by a single main mast which, because of its inclination, allowed the system to be stabilized only by slender cables (varying from 22 - 42mm 0 ) attached to the perimeter of the pneumatic lens (Fig 22). At the base of the mast the reactions in the self-weight state are components vertical and horizontal (parallel with the main axis of the ellipse); the vertical component is resisted by purely vertical cables around the perimeter, with greatest tensions in cables at the end nearest the mast in order to counter the overturning weight. The horizontal component is resisted by just two cables (shown thicker in the image), taken from either side of the lens structure to strong points at the top of the exhibition building. To accomodate variations in live load, particularly the lateral components due to wind, two further cables are taken from these strong points to a single point at the apex of the lens furthest from the mast; this arrangement of cables also resists overall torsion of the structure [5]. In effect the entire system is "docked", rather analogously to a ship which is moored to a dockside by fore, aft, and shear lines. A similar principle, though with more complex structuring, was used in the design of the Guthrie pavilion in Singapore (Fig 23).

Fig 21

Fig 22

Fig 23

Fig 25

Fig 24

P R E S T R E S S E D M E M B R A N E A N D AIR SUPPORTED STRUCTURES


As with cable systems, certain controls in CAD formfinding of membranes, such as controls on stress distributions, can be interactive. A very simple example is shown in figures 24 and 25, in which necking contraction of a conic is occurring because of insufficient warp stress (in the radial direction). The system can be restored to a desired form (Fig 26) by altering the single parameter of warp line tension to induce increased warp stress in the necked areas. Other adjustments, for example of panel and warp orientations, will entail some resetting of topology - yet this can be established by specifying only a few principal control lines in each surface, with automatic interpolation between them. The colour display of behaviour (Fig 25), form and patterning (Fig 30), and stress distributions (Figs 35 and 36), are clearly useful aids to guide the design process. The display and simple storage of stress contours is important from an engineering point of view for the comparison of different load states (and adjustments to form or material properties). The same graphics routines for stress plots are also used for contours of height (to examine possible ponding in shallow areas), or any other variable such as slope, discontinuity of slope and wind incidence - the latter to aid load definitions. The two main entrances at EXPO 92 both employed widespan cable and membrane shade structures [5]. The

Fig 26

Fig 27

"Diadema", shown in figure 27, employed a porous fabric partly in order to alleviate the high wind loads on the structure which had a maximum height of about 55m and span of 77m. The main surface structure was a wide grid cable net of equal spacing in one direction and equal traverse tensions in the other, with the lightly stressed fabric acting solely as shade covering. The "Oleada" entrance structure, shown in figures 28 - 31, employed a more highly stressed heavy-grade fabric, together with cable reinforcing in the surface whose principal function was to stabilize the central compression booms.

Fig 29 Fig 28

The design of the Oleada was intended as a continuation and reflection of La Barqueta bridge which joined the EXPO site to Old Seville. An initial functional aim of the structure was to constrain and then open out the view of the EXPO site. The sculptural form that emerged was enhanced (Fig 30) by adjusting surface stress ratios and the positioning and heights of the boundary mast points so that the surface reflected the motion of a bird in flight - "Oleada". The membranes span from perimeter guyed masts to two main central arches of 60 and 70m span, one inflected downwards and the other upwards. The first downward arch is suspended and pre-compressed by cables from 65m high V-form masts, with the compression force in the arch balanced by ties to the ground at its free ends. The second, upward arch is precompressed by light ties to ground level with the main free end thrust in the arch sustained by cables to the top of the V masts (Figs 29 & 31). In order to decouple the interactions between these two systems the masts were additionally guyed by independent cables to ground anchorages.

Fig 30

39

The original concept for the design was that the two main arches should be tensegrity type systems - in the sense that each was to consist of eight pin-jointed slender compression booms stabilized by spiral cable bracing for torsional and general stability (Fig 29), assisted by the mast stay cables or ground ties for stability in the vertical plane and by transverse cables in the membrane surfaces to enhance lateral stability. In fact, because of the complexity and high cost of connection details for the tensegrity system, and the difficulty of construction in terms of required tolerances, alternative segmental arches using thin walled large diameter tubes were eventually employed. In spite of their sizes (650 and 810mm diameter) they still had a slender appearance (Fig 31).

In a subsequent study [6] the use of fabricated segments was investigated. Each segment consists of a central compression boom and three tie rods to either end braced apart in the centre of the segment by a triangular yoke. The segments are then prestressed into a pin-jointed tensegrity arch system using only three continuous longitudinal chord cables attached to the apices of the triangular yokes (Fig 32). The advantage of this system is that only the chord cables (or in fact, for a circular arc, only one of the chords) need prestressing adjustments, and the problems of tolerances on site should be alleviated. However, although the resulting arch is stiff in bending it has no torsional stiffness - since, in contrast to the Oleada concept model, there are no spiral bracing cables. In fact, this is quite acceptable and even beneficial provided the membranes (or cables within the membrane surface) act compositely with the arch structure: Figure 33 illustrates the stability of the system under extreme transverse wind load which induce greatest torsional load. The yokes twist along the arch and allow the stresses in the membrane regions on either side to equilibrate eachother. A more recent project taken to full engineering design by IPL, though not built, is a very large air-supported structure for covering industrial waste, principally employed in order to minimize the cost of contaminated water treatment. This low-rise structure has a kidney shape (Fig 34) with main and minor axis spans of 500m x 280m and was to be fabricated in very heavy-grade PVC polyester in standardized 20x20m panels with mechanical joints. The high strength of the fabric eliminates the need for a reinforcing cable grid, which otherwise would have to sustain the main spanning tensions because of the disparity between membrane and cable stiffnesses (particularly with longer term effects). In addition to its strength the membrane has very high visco-elastic damping which is of benefit in terms of dynamic behaviour. However, this also entails substantial creep, particularly in the weft direction of the material. A major aspect of the design and associated material testing was therefore to account for both first time loading elastic stiffnesses and long-term creep effects in order to ensure a reasonable degree of load sharing between warp and weft fibres under various load states - figures 35 and 36 show contour plots of warp and weft stress for one wind loading direction.

Fig 32

W
Fig 33

Fig 34

ACKNOWLEDGEMENTS
Consulting Engineers for the Gatlinberg Centre project and for Munich Aviary were Buro Happold. The Architects for the former were Sprankle Lynd and Sprague and for the latter was Jorg Gribl. Frei Otto was Consultant Architect for both projects. Consulting Engineers and Architects for all other projects illustrated were IPL under the direction of the late Harald Muehlberger.

REFERENCES
1. 2. Frei Otto: Tensile Structures, MIT Press , 1971 Haug, E, Powell, G H: membrane structures, 3. Finite element analysis of non-linear IASS Symp. on Tension Structures and

Space Frames, Tokyo, 1971 Argyris, J H, Angelopoulos, T, Birchat, B: A general method for the shape finding of lightweight structures, Int. Conf. on Tension Strucrtures, London, 1974 4. Barnes, M R: Applications of dynamic relaxation to the design and analysis of cable, membrane and pneumatic structures. Second Int. Conf. on Space Structures, Guildford, 1975 5. Barnes, M R, Renner, W, Kiefer, M: of widespan EXPO structures, Structures, Stuttgart, 1996 6. Adriaenssens, S, Barnes, M R, Mollaert, M: free tensegrity spines, Fig 36 Aarhus, 1998 Deployable torsion Proc. Engineering a New Architecture, Case studies in the design Proc. Conceptual Design of

Another important consideration in the design of very wide span air-supported structures is the question of dynamic stability. High-rise structures with greater curvatures may have potentially lower membrane stresses but are likely to be subject to non-uniform, shape-deforming distributions of wind pressures and suctions. In contrast, low-rise systems are likely to have much more uniform wind suctions over the entire surface, but higher tensions will be induced because of the shallow curvature. Although these higher tensions incur penalties in terms of foundation loads they significantly improve large scale dynamic stability. A greater problem for these large scale air-supported structures (and other shallow membrane systems) is snow loading and the potential occurrence of ponding and consequent failure. This has been a difficulty with several of the largest built structures of this type and is perhaps the reason for the reduced use of air-supported systems in recent years. An alternative view is that control systems and the means of modelling wind and snow loadings and their interaction with structural systems have now substantially improved. These aspects are considered in the final two papers of this section on form and structure.

THE DEFINITION OF CURVED GEOMETRY FOR WIDESPAN STRUCTURES


Chris J K Williams Department of Architecture and Civil Engineering University of Bath UK

1 INTRODUCTION
If an enclosure is to be constructed of curved lines and surfaces rather than straight lines and flat planes, the questions arise as to how the geometry is first to be chosen and then how it can be defined with sufficient accuracy for the structure to be built and clad. There are clearly many ways that the geometry can be chosen and defined, but they fall into three broad categories and the methods used on any one project may fall into more than one of these categories. The categories are: Sculptural in which a model is sculpted by hand or a computer model is constructed that can be deformed interactively. Geometric in which the form is defined in terms of geometrical objects which might be simple spheres, cylinders or cones, or much more complicated objects which can only be visualised using computers. Physical in which the shape is controlled by some physical process such as a soap film or a hanging chain. The physical process may be modelled by an actual physical model or a mathematical model which may be analytic or numerical in a computer. An example of a mixed sculptural and physical approach would be bending a piece of wire by hand (the sculptural part) and then dipping it into soap solution and withdrawing it to form a soap film (the physical part). An example of a mixed geometric and physical approach would be forming a soap film between two parallel circular rings so that the rings are simple geometric entities. In this case the soap film forms a catenary of revolution so that one might say that it is a relatively simple geometric object formed by a physical process. The methods used for any one project will depend upon many factors. Perhaps the most important of these is the relative importance of structural, architectural and other constraints. Another is the experience of the design team in using various techniques, especially since the technology may have evolved in other disciplines such as sculpture, medicine or automobile, aerospace or ship design.

In the following I shall discuss some recent experience using a number of methods to try and illustrate the possibilities of the three approaches.

2 SCULPTURAL
Traditionally large sculptures or even car bodies were first made as small clay models or maquettes which were measured and enlarged. Now much of this work is done using computers employing software written for the automobile and aerospace industries. Frank O. Gehry & Associates use aerospace software, but the starting point <is still physical models. Such software is expensive and time is needed to learn how it can be used. Curved lines are divided up into a series of spline curves which fit together with an appropriate continuity of orientation and curvature. Curved surfaces are constructed from curved patches. These patches, developed for Computer Aided Design are very similar to the finite elements developed for the analysis of shell structures.

Fig 1 Body Zone sculpture Architect: Branson Coates Architecture Engineer: Buro Happold

Figures 1 to 5 show the Body Zone in the Millennium Dome. The shape was defined by the small physical model in the photographs in figure 1. A structural grid was drawn on the model and this was measured using a standard CAD package from scanned images of the photographs. This data was used to construct the computer model shown in figure 2.

Fig 4 Computer model from sections Architect: Branson Coates Architecture Engineer: Buro Happold

Following this work it was decided to concentrate on defining the figures in terms of parallel cross-sections. A copy of the physical model was sliced using a saw and the resulting cross-sections were scanned and 'traced' to produce figure 3. These cross-sections were joined to produce the three dimensional image shown in figures 4 and 5.
Fig 5 Rendered image Architect: Branson Coates Engineer: Buro Happold

Architecture

The software used to produce figures 2, 3 and 4 was specially written for the project.

43

Fig 6a

Computer image

Fig 6b

Le Corbusier's sketch

3 GEOMETRIC
The limit of what can be done using geometry is the mathematical knowledge and imagination of the individual. As an example let us consider Le Corbusier's Modulor ^ shown in figure 6b. Modulor is based upon the Fibonacci series, examples of which are 1, 1 , 2 , 3 , 5 , 8 , 13,21,34. and 1,3,4, 7, 11, 18, 2 9 , 4 7 , In both cases each term is the sum of the two preceding terms. Thus if % is the n term,
t n

i - ^. H As n gets larger, the ratio of successive terms becomes closer to <f> regardless of the starting two terms of the series, *o and . If , then a=* and x ^ f * , , which is the basis of Modulor. The computer program below was used to produce the dxf file that is plotted in figure 6a, replicating Modulor. where is the Golden Section and
2
X l 2

* = i^l

dxf files are text files which can be read by CAD programs such as AutoCAD, MicroStation or MiniCAD.

This equation has the general solution

(A0 -l)0 + ( A - 0 ) ^ -

#include <fstream.h> #include <iostream.h> #include <math.h> int i,j,k,m,nhalfycles,finish; float P I , x , y , z , U , V , W , a l p h a , b e t a , A,C,phi; ofstream Julia("Modulor.dxf"); int main(void) { PI=4.0*atan(1.0); Julia"0\nSECTION\n2\nENTITIES\n"; nhalfycles=18;m=20*nhalfycles; phi=(1.0+sqrt(5.0))12 for(j=l;j<=4;j+=l) { if(j==4)finish=90.0; else finish=60; for(k=0;k<=finish;k+=3) .0;C=216.0;

Figure 7 Golden section log spiral

Figure 8 Two spirals

The program will run on any computer (Macintosh, PC etc.) with a C++ compiler and very little work would be required to convert the program to Basic or Fortran. An unlimited variety of curves and surfaces can be produced by such programs. For example, figures 7, 8 and 9 were produced using the formulae
x=4>" coSan I Y = 0 " sificwr J x=(p" co&an + </T c o m I
a n c

{
A=1.2*cos((PI*k)/180.0); for(i=0;i<=m;i+=l)

{
beta=-(1.0*i*nhalfycles)/(2.0*m); alpha=beta;if(j<=2)alpha-=0.5; y=C*pow(phi,alpha); if(j==l||j==3) x=A*y*fabs(sin(beta*PI)); if(j==2||j==4) x=A*0.5*y*fabs(sin(2.0*beta*PI)); if(j<=2)x=-x; x=x/8.0; z=0.0; if(i!=0)

Y=<P' si&ait-Q~

sia/r

x = ( l + p W *

+(i-p)0""cosot|

y = (l-p)<f" sian

respectively.

{
Julia"0\nLINE\n8\n0\n"; Julia"10\n"U"\n2 0\n" V"\n30\n"W"\n"; Julia"ll\n"x"\n21\n" y"\n31\n"z"\n" ; } U=x;V=y;W=z; }

Fig 9

Spirals to lines

}
}

Julia"0\nENDSEC\n0\nEOF\n"; Julia.close(); c o u t " D X F file written\n"; return 0; }


Fig 10 Rendered image

In each case a is varied to draw a curve and in the case of figure 9, a different value of p is used for each curve. The surface in figure 10 was obtained from the curves in figure 9 by giving each curve a different value of c.

Fig 11 Bridge study

Fig 12 Bridge study

The bridge studies in figures 11, 12 and 13 were also produced by purely mathematical methods as was the shell study in figure 14. In each case the whole object is defined by the just one set of mathematical formula so that there is complete continuity of all derivatives, orientation, curvature, rate of change of curvature etc.

Fig 15 Millennium Dome Rest Zone - system geometry Architect: Richard Rogers Partnership Engineer: Buro Happold

Fig 13 Bridge study

Fig 16 Rest Zone Architect: Richard Rogers Engineer: Buro Happold

Partnership

Fig 14 Shell study for Stuttgart railway station Architect: Ingenhoven Overdiek Kahlen und Partner Consultant Architect: Professor Frei Otto Engineer: Buro Happold

Figures 15 and 16 show the Rest Zone in the Millennium Dome which was produced by deforming a torus. Again there is complete continuity of all derivatives.

46

4 PHYSICAL
In the membrane theory of shell structures the geometry of the structure and the loads are assumed to be known and the three membrane stresses - two tensile or compressive and one shear are unknown. There are three equations of equilibrium, one in the direction normal to the surface,
<T%,+ p=0 ,

and two in the plane of the surface, Now much of this work is done with numerical models, although physical models are indispensable for initial studies. Form finding a fabric structure with a soap film or minimal surface is done by setting the membrane stress components 0- =Tg
a/} a/J

The notation here is similar to that in Green and Zerna^. In these equations the geometry of the shell is determined by the components of the metric tensor, <3ap , and of the curvature tensor, fcfyj. The components of load are p and P and the unknown membrane stress components are a }
a 1

a W

2 a

=a

1 2

The fact that there are three equilibrium equations and three unknown membrane stress components means that shells are essentially statically determinate if the overall shape and boundary supports permit. An inappropriate shape or lack of support may mean that a shell is a mechanism. It is not at all obvious which shapes and support conditions lead to mechanisms and which do not. Spivak^ discusses this issue in purely geometric terms, for example the Cohn-Vossen theorem states that any complete convex surface with positive Gaussian curvature is not a mechanism if membrane strains are prevented. A cooling tower on the ground is not a mechanism, but a spherical shell with a hole in the top is. A shell may also be a mechanism if it is made of masonry, so that the principal membrane stresses must be compressive, or if it is made of fabric in which case they must be tensile. A mechanism can carry certain loads if it has the correct shape. In the case of a masonry structure, the dominant load is the dead load and in the case of a fabric structure it is prestress upon which wind and snow are added. Form finding is the process of establishing a structural geometry for a mechanism to carry a particular load. Gaudi used hanging models which, when inverted, defined the shape of masonry arches and vaults for the Colonia Guell and the Sagrada Familia. Professor Frei Otto pioneered the use of physical models for fabric structures, cable nets and grid shells.

where T is the surface tension. In addition a geodesic co ordinate system for generating the cutting the pattern is obtained by imposing the conditions g = o and
12

9 2 2 constant
=

An equal mesh net is produced by writing


O" =0
12

and

g =g
1 1

2 2

= constant;

if there is no elastic extension, otherwise srn and g 2 increase with tension. Equal mesh nets are more difficult to form find than fabric structures due to the adjustment of cable lengths at the boundary.
2

Fig 18 Mannheim Bundesgartenscb.au 1.5m grid Architect: Mutschler & Partners Consultant Architect: Atelier Warmbronn (Professor Frei Otto) Form finding: Buro Linkwitz Engineer: Ove Arup & Partners (Ted Happold and Ian Liddell)

Fig 19 Mannheim erection

47

Figure 17 shows one of the equal mesh nets of the Tree of the Future intended for the Central Show in the Millennium Dome. In this case a computer program was specially written by Lynne Mabon of Bath University which automatically generated the boundary data. Figure 18 shows the hanging chain model for the Mannheim Bundesgartenschau. This is a computer generated model by Buro Linkwitz, based upon Frei Otto's accurate physical model. Figures 19 and 20 show the erection and load testing of the shells.

5 THE BRITISH M U S E U M GREAT COURT ROOF


Figures 23-31 show images of the computer model of the British Museum Great Court Roof. It was generated by a mixed approach.

Ml Ml Ml Ml { b) { *j 1-5! l rb M l M re) M l [
i *)
Fig 20 Mannheim load test

where r

TJX

+ y

Figures 21 and 22 show computer generated models of the Weald and Downland Museum. In this case the mathematical model had to contain bending stiffness during form finding, otherwise compressive stresses produced wrinkling.

Fig 23 British Museum Great Court Roof first function Architect: Foster and Partners Engineer: Buro Happold The shape of the surface was defined analytically by weighting and summing functions based on those shown in figures 23, 24 and 25. The weightings also varied with position to satisfy architectural, planning, structural and clearance requirements.

Fig 21 Weald and Downland Museum Architect: Edward Cullinan Architects Engineer: Buro Happold

Fig 24 British Museum Great Court Roof second function

Fig 22 Weald and Downland Museum Architect: Edward Cullinan Architects Engineer: Buro Happold

The positions of nodes on the surface were obtained from the starting grid shown in figure 26. A displacement was calculated for each interior node of this grid to make its x, y and z co-ordinates the weighted average of the current co-ordinates of the four surrounding nodes. However, before moving a node, the component of displacement normal to the surface (see figure 27) was

48

removed so that the node remained on the surface. This relaxation procedure was repeated thousands of times for the whole structure until the geometry settled down to that in figure 28.

Fig 25 British Museum Great Court Roof third function The weighting of the surrounding nodes was varied at different points on the surface to control the distribution of the nodes, in particular in relation to the sizes of the glass panels.

Fig 26 British Museum Great Court Roof original grid The spiraling members were obtained by joining points in the form finding grid as shown in figure 29 to produce figures 30 and 31.

Fig 27 British Museum Great Court Roof surface normals

Fig 28 British Museum Great Court Roof relaxed grid

49

Fig 31 British Museum Great Court Roof steel members Architect: Foster and Partners Engineer: Buro Happold

6 CONCLUSION
This paper discusses some of the ways in which curved forms can be generated. It is not possible to say that any one method is the optimum, because there are so many possibilities and the architectural, structural and environmental constraints will never be the same on two projects.

Fig 29 British Museum Great Court Roof steel members on grid

REFERENCES
1 Le Corbusier, The Modulor, translated by Peter de Francia and Anna Bostock, Faber and Faber, London 1961. 2 Green, A.E. and Zerna, W., Theoretical Oxford University Press, 1968. 3 Spivak, Michael, A comprehensive geometry,
n

elasticity,

2 * edition,

n(

introduction

to

differential

volume 5, 2 ^ edition, Publish or Perish Inc.,

Delaware, 1975, 1979.

Fig 30 British Museum Great Court Roof steel members

50

CONCEPTS AND RELIABILITY IN THE DESIGN OF WIDESPAN STRUCTURES

M. Majowiecki Department of Civil Engineering, University of Bologna, ITALY

ABSTRACT
Wide span structures are today widely applied for sport, social, industrial, ecological and other activities. The experience collected in the last decades identified structural typologies as space structures, cable structures, membrane structures and new - under tension - efficient materials which combination deals with lightweight structural systems, as the state of art on long span structural design. In order to increase the reliability assessment of wide span structural systems a knowledge based synthetical conceptual design approach is recommended. Theoretical and experimental in scale analysis, combined with a monitoring control of the subsequent performance of the structural system, can calibrate mathematical modelling and evaluate long term sufficiency of design. Some special remarks concerning the influence on the reliability level of detail design, are given at the end of the paper.

The state of the art trend on widespan enclosures: the lightweight structures from compression to tension
According to the state of the art, the more frequently used typologies and materials for wide span enclosures are: Space structures single layer grids double and multi layer grids single and double curvature space frames Cable structures cable stayed roofs suspended roofs cable trusses singleand multilayer nets Membrane structures prestressed anticlastic membranes pneumatic membranes Hybrid structures tensegrity systems beam-cable systems Convertible roofs overlapping sliding system pivoted system folding system The historical trend in the design and construction process of wide span enclosures was and is the minimization of the dead weight of the structure and , consequently, the ratio between dead and live loads (DL/LL). From ancient massive structures ( D L / L L 1 ) to modern lightweight structures ( D L / L L 1 ) , the DL/LL ratio was reduced more than 100 times due to the most effective exploitation of the properties of special high-strength materials, in combination with structural systems where tensile stresses are dominant (Tension structures). Due to the inherent stability of tension, in contrast to compression, tension structures lead naturally to optimization of the system energy compared with structures which are subjected to bending moments or are stressed axially with the possibility of reversal from tension to compression, as is the case with grids and framed structures. Therefore, the actual trend on

INTRODUCTION
Long span structures are today widely applied for: Sport buildings Stadia Sport halls Olympic swimming pools Ice tracks and skating rinks Indoor athletics Social buildings Fair pavillions Congress halls Auditorium and theatres Open air activities Industrial buildings Hangars Warehouses Airport terminals Ecology buildings Waste material storage Pollution isolation

51

lightweight structural typologies is to combine, as far as possible, a dominant tension mechanical system and histrength materials. In Table 1, it is possible to observe the exceptionally high strength to weight ratio (K=o7y) in tension (Kt) of hi-tech composite materials, important components of lightweight structural concepts. The different mechanical behaviour of compression and tension structures can be illustrated by Fig. I where, starting from a thin parabolic arch under uniform distributed load , it is possible to observe, during incremental loading, the following phases of the load displacement curve: Phase A: unloaded structure. Phase AB: compression phase; geometric softening; decrease of tangential stiffness, reduction in the positive value of the secondary term of the total potential energy (8 it).
J

MATERIALS

<V
N/mrn' N/mm 3 85 37 5 30 520 1050 900
!

7k N/m> 10' 18 5 25 79.5 79.5 45 6.664 13.376 20.000 21.250 m

K ni

Bricks Wood Concrete Sleel 52 Steel 105 Titanium

0.166 9 375 1.200

Composite materials hi-tech


Unidir Carbon fibres Textile carbon fibres Unidir Aramidk fibres Textile aramklk: fibres (Kevlar) L'nidir Glass fibres Textile glav. fibres

1400 800 1600 750 1100 450

15.5 15.5 13 13 20 20

90.000 52.000 123.000 58.000 55.000 22.500

--

--

Table 1

Mechanical properties of construction materials

Phase BCE: unstable phase; dynamic displacement from B to E with liberation of kinetic energy (cross hatched area). Here, the secondary term of total potential energy is negative (5 7t < 0).
2

SPECIAL ASPECTS OF CONCEPTUAL D E S I G N D E C I S I O N S O N L O N G SPAN STRUCTURES


From the observations of the in service performance, damages and collapses of all or part of structural systems, we have received many informations and teachings regarding the design and verification under the action of ultimate and serviceability limit states. Limit state violation for engineered structures have lead to spectacular collapses as the Tay (1879) and Tacoma bridges (1940). Sometimes an apparently "unimaginable" phenomenon occurs to cause structural failure. The Tacoma Narrows Bridge previously cited was apparently one such a case. It was also a design which departed considerably from earlier suspension bridge design. Long span coverings were subjected to partial and global failures as that of the Hartford Colisseum (1978), the Pontiac Stadium (1982) and the Milan Sport Hall (1985) due to snow storms, the Montreal Olympic Stadium due to wind excitations of the membrane roof (1988) and under snow accumulation (1999), the Minnesota Metrodome (1983) air supported structure that deflated under water ponding, etc. Those cases are lessons to be learned from the structural failure mechanism in order to identify the design and construction uncertainties in reliability assessment. Many novel projects of long span structures attempt to extend the "state of the art". New forms of construction and design techniques generate phenomenological uncertainties about any aspect of the possible behavior of the structure under construction service and extreme conditions.

Phase DEF: tension phase; geometric hardening increase in the tangent stiffness, branch of stable equilibrium with increasing value of secondary term of the total potential energy (5 n). Phase DEF is characteristic of the behaviour of tension structures. The non-linear geometric hardening results in a less than proportional increase of stresses in relation to increase external loads. This provides an increased nominal safety factor evaluated at ultimate limit state (p safety index).
2

Fig 1 Mechanical behaviour from arch to cable

52

Fortunately, structures rarely fail in a serious manner, but when they do it is often due to causes not directly related to the predicted nominal loading or strength probability distributions. Other factors as human error, negligence, poor workmanship or neglected loadings are most often involved [1]. Uncertainties related to the design process are also identified in structural modelling which represents the ratio between the actual and the foreseen model's response. According to Pugsley (1973), the main factors which may affect "proneness to structural accidents" are: new or unusual materials; new or unusual methods of construction; new or unusual types of structure; experience and organization of design and construction teams; research and development background; financial climate; industrial climate; political climate.

Considering the statistical results of table 2, and the "scale effect" of long span structures several special design aspects arise as: the snow distribution and accumulations on large covering areas in function of statistically correlated wind direction and intensity; the wind pressure distribution on large areas considering theoretical and experimental correlated power spectral densities or time histories; rigid and aeroelastic response of large structures under the action of cross-correlated random wind action considering static, quasi-static and resonant contributions; the time dependent effect of coactive indirect actions as pre-stressing, short and long term creeping and temperature effects; the local and global structural instability; the non linear geometric and material behaviour; reliability and safety factors of new hi-tech composite materials; the necessity to avoid and short-circuit progressive collapse of the structural system due to local secondary structural element and detail accidental failure; the compatibility of internal and external restrains and detail design, with the modelling hypothesis and real structural system response; the parametric sensibility of the structural system depending on the type and degree of static indeterminacy and hybrid collaboration between hardening and softening behaviour of substructures; In the case of movable structures, the knowledge base concerns mainly the moving cranes and the related conceptual design process have to consider existing observations, tests and specifications regarding the behaviour of similar structural systems. In order to fill the gap, the I ASS working group n16 prepared a state of the art report on retractable roof structures [2] including recommendations for structural design based on observations of malfunction and failures.

All these factors fit very well in the field of long span structures involving oftenly something "unusual" and clearly have an influence affecting human interaction. In Table 2, the prime cause of failure gives 4 3 % probability (Walker, 1981) to inadequate appreciation of loading conditions or structural behaviour. Apart from ignorance and negligence, it is possible to observe that the underestimation of influence and insufficient knowledge are the most probable factors in observed failure cases (Matousek & Schneider, 1976). Performance and serviceability limit states violation are also directly related to structural reliability. Expertise in structural detail design, which is oftenly considered as a micro task in conventional design, have an important role in special long span structures: reducing the model and physical uncertainties and avoiding chain failures of the structural system.
Cause Inadequate appreciation of loading conditions or structural behaviour Mistakes in drawings or calculations Inadequate information in contract documents or instructions % 43

.-"TECHNOLOGY

7
/

-"

ART \

\ \ ----

^'EXPERIENCE \

8
f gf

S'S
Jjj! | | \ \
X

' NATURE

8
;g

pi j ; SYNTHESIS

Contravention of requirements in contract documents or instructions Inadequate execution of erection procedure Unforeseeable misuse, abuse and/or sabotage, catastrophe, deterioration (partly "unimaginable"?) Random variations in loading, structure, materials, workmanship, etc. Others

| | j | ./ / /

X . OBSERV,, SCIENCE RESEARCH

"X \ .

13 7

Fig 2

Holistic approach to structural design

10

Table 2

Prime causes of failure. Adapted from Walker (1981)

53

Building Code of Canada (1990), point A-4.2.4.1: "It is important that innovative designs be carried out by a person especially qualified in the specific method applied...". Eurocode no. 1 intends to guarantee the level of safety and performance by a quality assurance (QA) strategy (point 2) and control procedures of the design process (point 8) in order to minimize human errors. Formalized methods of QA considers the need to achieve, by the institution of a "safety plan" the requirements of structural safety, serviceability and durability. A real danger is that excessive formalization of QA , born for tangible manufactured articles and not suitable for intangible conceptual control procedures, could lead to unacceptable and self-defeating degeneration of the design process, in a certain kind of Kafkian bureaucratic engineering and management.Notice about this phenomena is given by Carper (1996)in (Construction Pathology in the United States) [4]: "many repetitive problems and accidents occur, not from a lack of technical information, but due to procedural errors and failure to communicate and use available information". Important contributions concerning the matter was given by the International Symposium on "Conceptual design of Structures" organized by I ASS [5]. Furthermore, it would be necessary to have adequate and systematic feedback on the response of the design by monitoring the subsequent performance of such structures so that the long term sufficiency of the design can be evaluated. To assure a required reliability level, in the field of special structures, the design process must be checked in the following three principal phases: the conceptual design, the analytical model, and the working design phases as shown in Fig.3.

Fig 3

Conceptual design and analysis of structures

KNOWLEDGE BASED CONCEPTUAL DESIGN AND RELIABILITY LEVEL


The conceptual design is knowledge based and, basically, property of individual experts. Their involvement in early stages of design is equivalent, from the reliability point of view, to a human intervention strategy of checking and inspection and, from a statistical point of view, to a "filtering" action which can remove a significant part of "human errors". According to the design requirements, the conceptual design is defined by a knowledged expert synthetical approach [3] based on the reliability intuition of the selected model which has to be confirmed by the results of the analysis phase. The conceptual design approach is holistic (see Fig.2 Srivastava [5]) and directly depends on the skills and abilities of the design team members. This concept is now included in some national building codes, which are normally addressed only to conventional structural systems. As far as innovative designs are concerned, as in the case of most of the realized long span structures, only few comments are dedicated as, for instance, in the National

COMPUTER AIDED DESIGN & ANALYSIS


Conceptual errors are very hard to remove in the subsequent phase of structural analysis. In this phase the human intervention strategies as education, work environment, complexity reduction, self-checking and external checking and inspections are today assisted by new interactive computer aided design and analysis techniques. Specially the interactive graphic language will be very effective in obviating the effects of gross human errors during the structural modelling. Hardware and software interfaces make generate a useful interactive design & (Fig. 4). The computer aided design simplifies complex tasks and increases it possible to analysis cycle methodology the reliability

54

level, if the ergonomy and the logical flow of the interactive computer assistance is organized as follows: the interactive design methodology is not substitutive but rather integrates the creative aspects of the traditional design process (conceptual design). By means of an interactive graphic language (pre processing and post-processing software), the electronic computer becomes the useful mental and operating extension of the designer, while considerably increasing his capacity, speed and decision-making abilities; design optimization is a logical consequence of the interactive methodology, allowing very fast data modification and evaluation of consequences (what if) according to the classical step by step iterative procedure of trial and error (and trial and success) based on the experience of the designer, who is able to synthesize a considerable mass of data that is difficult to express as a mathematical problem.

INTERACTIVE GRAPHIC LANGUAGE ADDRESSED TO THE STRUCTURAL DESIGN OF LIGHTWEIGHT S T R U C T U R E S : from architectural to mathematical modelling
The development of the lightweight structural concept is historically correlated with the research in CAD technology. From the the initial empiric research made by Frei Otto, the theoretical and experimental investigation in the world concerning cable and membrane structures started in early 70s. In the Department of Structural Engineering of the University of Bologna, for instance, a research concerning an integrated computer aided analysis and design of lightweight structures produced a first interactive computer-aided shape-finding program that ran, in 1973, on an IBM mainframe with a video Console 2250. Nowadays, the interactive programs, written in C++ are object-oriented (OLE) under Windows 98 and Windows NT platform [6].
LINKED INTERFACE SOFTWARE

COMPUTER ANALYSIS

COMPUTER INPUT

COMPUTER OUTPUT

NKEO INTERFACE HARDWARE

Fig 5 Hardware and software network system


DATA PRE PROCESSING BASIC SOFTWARE FOR INPUT DEVICE INPUT DEVICE RESULT POST PROCESSING BASIC SOFTWARE (FOR OUTPUT DEVICEl OUTPUT DEVICE

DESIGNER OUTPUT

DESIGNER INPUT

The interactive software for analysis and design of special structural systems, as normally involved in wide span enclosures requires, more than general purpose programs, software to assist on many aspects of theoretical analysis, such as: state ' 0 ' form-finding analysis, for the shape-finding of cable, membrane and pneumatic structures; non linear material analysis for elastic, anelastic and plasticity including short and long term creeping; non linear geometrical analysis; for the static and dynamic analysis under large displacements; stocastic dynamic analysis in frequency domain for the buffeting response under the random wind action assisted by the experimental identification, on scale rigid models, of cross-correlated power spectral densities (PSD) of the internal and external pressures on large enclosures; stocastic dynamic analysis in time domain for the control of the aerodynamic stability of wide and flexible structural systems under wind excitation, assisted by the experimental identification, on aeroelastic scale models, of the cross-correlated time histories, considering fluid interactions; application of the optimization techniques to the structural design [7]; parametric stocastic sensibility & reliability analysis.

DESIGNER SYNTHESIS KNOWLEDGE BASE

Fig 4 Interactive design process

At the present time it seems to be very difficult to introduce aspects of artificial intelligence by a semantic software language inside the process of design of lightweight structures (expert system 3rd level software). The structural design, analysis & monitoring phases are today functionally linked and logically integrated, with other design components (architectural, mechanical, project management, etc), with a common topological and geometrical 3-D identification model, through a hardware and software network (Figure 5).

55

The advantages offered by the informatic and automation has been very important in the field of structural design in general and particularly essential in the case of long span lightweight structural systems. It is possible to examine more rigorous theoretical models avoiding, on the one hand, excessive simplifications that deprive the theoretical model, as a schematic reduction of the reality, of all significance and, on the other, that exhausting calculations lead to the loss of facts with a true influence, with the consequent discouragement of the designer from making efforts towards trying out different structural solutions. Under those apparently favourable circumstances, many documented structural failures have been detected where mistakes in the inadequate appreciation of structural behaviour was caused by unreliable man-machine interaction and the illusion that the computers, as powerful instruments of analysis, could replace conceptual design. For this purpose, IABSE have set up a special commission for the control of automation in structural design [8] . Documented.FEM modelling errors are illustrated in the First International Conference on computational Structures Technology [9].

[10] to provide design snow according to FAE (Finite Area Element) method, representing a state of the art on the matter.

Fig 7

Montreal Olympic Stadium - A cable stayed roof solution

The FAE method uses a combination of wind tunnel tests on a scale model and computer simulation to provide the most accurate assessment possible to estimate 30 year snow loads.

S O M E W I D E SPAN E N C L O S U R E S
Due to the lack of space, only some design&analysis illustrations of wide span enclosures, where the author was directly involved, will be included in the present paper with the intention to transmit some experiences that today may be part of the knowledge base. Long span structures needs special investigations concerning the actual live load distribution and intensity on large covering surfaces. Building codes normally are addressed only to small-medium scale projects. The uncertainties relate to the random distribution of live loads on long span structures imply very careful loading analysis using special experimental analysis. From the author's direct experience in designing large coverings, the most important experimental investigation regarding live load distribution concerns the snow drift and accumulation factors and the dynamic action of wind loading.

SNOW LOADING EXPERIMENTAL ANALYSIS O N S C A L E M O D E L S O L Y M P I C S T A D I U M IN M O N T R E A L .


During the design of the new roof for the Montreal Olympic Stadium Figure 7 a special analysis of snow loading was made considering three roof geometries varying the sag of the roof from 10 m, 11.5 m and 13 m.,in order to find a minimization of snow accumulation. The experimental investigation was carried out by RWDI

Fig 9

Comparative analysis of snow loading distribution in function of roof shape (10-13m)

Snow loads depend on many cumulative factors such as, snowfall intensity, redistribution of snow by the wind (speed and direction), geometry of the building and all surroundings affecting wind flow patterns, absorption of rain in the snowpack, and depletion of snow due to melting and subsequent runoff. The current NBCC (National Building Code of Canada) provides minimum

56

design loads for roofs which are based primarily on field observations made on a variety of roofs and on a statistical analysis of ground snow load data. There are, however, numerous situations where the geometry of the roof being studied and the particulars of the site are not well covered by the general provisions of the code. In these situations, a special study, using analytical, computational and model test methods, can be very beneficial since it allows the specific building geometry, site particulars and local climatic factors to all be taken into account. The National Building Code allows these types of studies through its "equivalency" clause and various references to special studies in its commentary. The model of the three new roof shapes were each constructed at 1:400 scale for the wind tunnel tests. The three model roof designs were each instrumented with 90o directional surface wind velocity vector sensors covering the surface. On the console roof, an additional 90 sensors were installed. Measurements of the local wind speed and direction, at an equivalent full-scale height of 1 m above the roof surface, were taken for 16 wind directions. The wind speed measurements were then converted to ratios of wind speed at the roof surface to the reference wind speed measured at a height equivalent at full scale to 600 m.

a * rB*oe*sau (wsi-ts?3) OoVOP-OTTOWat UHAtCfXtMttS B= T R O CASUU WTMUE OW s I A taS7-<tft) S

fa

'utmnx pmioo (tca*!)"

Fig 11 Statistical investigation for the reference 50 years return period wind speed

Results of structural load cases and local peak loading, not to be considered as acting over the roof simultaneously are shown in Fig. 9-10. The shape of the roof with a sag of more than 12m. gives separation of the air flow and turbulence in the wake increasing considerably the possibility of snow accumulations. The order of magnitude of the leopardized accumulations in the roof are of 4-6 kN!; local overdimensioning was necessary in order to avoid progressive collapse of the structural system.

WIND LOADING-EXPERIMENTAL ANALYSIS O N SCALE MODELS: rigid structures-quasi static behaviour


The integration of the wind tunnel data into the design process presents significant problems for wide span subhorizzontal enclosures; in contrast to buildings (high rise buildings) where knowledge of the base moment provides a sound basis for preliminary design, there is not single simple measure for the roof. The study of the Turin and Rome stadiums [11-12-13] drew attention to the inability of the measuring system employed to provide data in a form that could readily be based as input to the sophisticated dynamic numerical model developed by the designer and led to discussion between the designer and the wind tunnel researchers to examine alternate techniques that might be used in future projects [12]. In that case preliminary estimates of the resonant response were obtained from the panel spectra using interpolation to estimate magnitudes and assuming no correlation (at the natural frequencies) between panels .

Fig 10 Sliding and wind snow accumulations step loads

The plot shown in Figure 8, obtained by interpolation of the data using the Fisher-Typett type I extreme value distribution method, predicts the 30 year ground snow load, including both snow and rain ( S S + Sr), to be 2.8 kPa, which is in agreement with the code value.

57

One possible approach would be to produce a set of load histories, Qj(t), such that: Qj(t) = fp(x,y,fyj(x,y)dA where: p(x,y,t) (l

nett load per unit area at position (x,y);

(f)j(x,y)

weighting function.

Fig 13 Typical spectra of panel loads

For a series of pressure taps of the approximation to (f)j(t) would be: N


Qj

These estimates proved to be significantly larger than those observed on the aeroelastic model due to significant aerodynamic damping effects not included in the prediction process (see Figures 13-14). The situation is further complicated by the inability of the instrumentation system to provide a complete description of the loading. Only seven of about 60 panels were instrumented and the data obtained must be interpolated to provide estimates of the overall loading, the interpolation required is concerned not only with the magnitude of the panel loads but also the spectra and cross-spectra. The discussions centered on the use of high speed pressure scanning systems capable of producing essentially simultaneous pressure measurements at some 500 points at rates of perhaps 200 Hz per point. With such a system it would be possible to cover in excess of

(0 - J
/1 =

Pi QCnYf, t)A>; (X/,7/)

(2)

area of ith panel; pneumatic average of pressure at the taps in the ith panel; Xj, yf geometric centre of the taps on the ith panel; N number of panels. The requirements of a system designed to produce the load histories, (j)j(t), is discussed in the following section.

Ai Pi

The Thessaloniki Olympic sport complex: measurement and use of load time histories In collaboration with the Boundary layer wind tunnel laboratory of the University of Western Ontario, a new very practical method to obtain the structural response under the random wind action and small displacements (linear response) has been applied under the name of the "orthogonal decomposition method". If the weighting functions,^*), are chosen as mode shapes then <f)j(t) is a modal load and its use in conjunction with a dynamic model is clear; either as a set of time histories or a set of modal force spectra and cross-spectra. In the initial stages of a design the roof shape is probably known with reasonable accuracy but mode shapes not so. In such cases it might be appropriate to choose a suitable set of0y from which modal loads corresponding to shapes y/j can be estimated when the design is more advanced. In such a case we can approximate (pj as: (3)

a OUpicemnt tt Location 1 RMS 180.0* 0.013 i

o Ptpl<;cn>e<u t toet8o 3

Velocity 25.1 m / s

10"

10"

t" o
Frequency (f)

?<f

10*

o Diplacmnt at Location 4

Fig 14 Selected spectra of roof deflections

200 panels and produce a complete description of the load. Such a system would produce roughly 1 to 2 x l 0 observations for a single wind direction and it is clear that some compression of the data would be required.
6

the values of can be evaluated by minimizing the discrepancy between y/j and (j)j, ie: a, 0, I dA = 0
y

(4)

If the functions are chosen as a set of orthogonal shapes ty$jdA = 0;i 4= j then the coefficients are given as dA (5)

58

For finite panel sizes the corresponding relationship is:


N

(6)

North

Fig 17

Predicted 50 year return period peak differential pressures

Ued2

Fig 15 Views of pressure model

The experiment would involve the recording of the local histories yj(t) from which the model time histories could be constructed and the analysis conduced in either the time or frequency domain Figures 15-18). For the type of structure under consideration resonant effects are small and the response is largely a quasi-static to a spatially varied load. The deflections induced are closely related to the imposed loads and their distribution differs significantly from the Gaussian form [12]. In such a case the time domain solution, which preserves the extreme value distribution, is to be preferred over a frequency domain approach.

Fig 18

Orthogonal decomposition: pressure mode shapes

, g , re turk nn 00 m - Th.l.tlonltl oombin.d 40 mi tO = i 0 - i I lonii1 combtnad O - Than lorn lit enmbintd 40 ml - 0 r * "TbatlllonlM com Mn*d 40 mi - iOO yc
1

For the seismic analysis a frequency domain approach was adopted. The Kanai-Tajimi PSD was used under the design response spectra prescribed by Eurocode 1; under strong-motion, an acceleration time history was artificially generated according to site and durability characteristics.

A i
m

Fig 16

Relative contribution of Azimuthal Direction to the exceedance probability of various return period wind speeds for Thermi, Thessaloniki, Greece

Fig 19

Aeroelastic model for Rome Olympic Stadium

59

WIND LOADING-EXPERIMENTAL ANALYSIS O N S C A L E M O D E L S : flexible structures-aerodynamic behaviour


The Olympic stadium in Rome, The wind induced response of the cable supported stadium roof was analysed by a non linear model and a field of multicorrelated artificial generated wind loading time histories [14-15].Wind tunnel tests have been carried out at the BLWT Lab. of UWO on a model of 1:200 Fig. 19 scale determining:
j-* is-* frtqmencf [Hz] to*

time histories of the local pressures for every 10 of incoming flow direction;the maximun,minimun and average values of the wind pressure have then been evaluated, as well as the root mean square of its fluctuating part; presssure coefficients (maxima,minima and average) for every 10 of incoming direction; auto and cross-spectra of the fluctuating pressure (averaged on every single panel).

Fig 21

Target (1), simulated (2) and Kaimal's (3) normalized spectra of wind velocity

-1.3 -1.4 ^ -|..5i

,
i i 1

;
i >

IS

20

30

40

50 time [sec]

60

70

80

90

Fig 22 Time History of the displacement (leeward side at tension ring, run #2)

A fluid-interaction non linear analysis in time domain, made for the checking of La Plata stadium design [16-17] shows a better agreement between theoretical model and experimental values.

Fig 20 1st modal shape (T= 1.78 s)

The aerodynamic behaviour shows a clear shedding phenomenon. The external border of the structure, constituted of the trussed compression ring with triangular section and tubular elements and by the roofing of the upper part of the stands, disturbs the incoming horizzontal flow in such a way so that vortex shedding is built up. This causes the roofing structure to be subjected to a set of vortices with a characteristic frequency. This is confirmed by the resulting Power Spectra Density Function of the fluctuating pressures, which shows a peak at about 0.15Hz even if the values rapidly decrease with increasing distance Fig. 21.

RELIABILITY ANALYSIS. E Q U I L I B R I U M A N D COMPATIBILITY. THE SENSIBILITY ANALYSIS REGARDING THE NEW CABLE STAYED R O O F O F T H E O L Y M P I C S T A D I U M IN M O N T R E A L
Many times, during the initial stages of the architectural design, structural engineering intervention is almost absent. In this situation the geometry, typologies and materials are defined in preliminary design under poor structural knowledge base. Oftenly happens that, under particular political and financial conditions, in a final design phase the inertia of the design process does not allow any changement of an original unsatisfactory conceptual design. When such a situation is created, structural engineers are forced to find solutions under very strict boundary conditions. A very typical case was that of the retractable membrane roof of

60

the Montreal Olympic Stadium [18-19], built in 1985 under design conditions determined in a first construction phase related to the preparations of the Olympic games in 1976. Several local roof failures due to aeroelastic instability under non extreme wind conditions appeared in 1988 and 1991.

Reliability analysis of the roof structural system. Cable strain parameric sensibility. The uncertainties on the elastic modulus of the cable, geometrical and elastic long term creeping, tolerances of erections, non linear behaviour, created a sensitive response on the rigid space frame hanging from a set of 28 stay cables. The sensibility analysis showed that the response is sensitive to the standard deviation of the cable strain (De) variations. The failure probability is given by the probability that an outcome of the random variables (De) belongs to the failure domain D. This probability is expressed by the following integral [20] Fig. 24:
N

_ _k
v .
2

ZA
k

<k

<y *
k

Fig 23 Montreal Stadium cable stayed rigid roof - numerical model

Fig 23 Montreal Stadium cable stayed rigid roof - numerical model

A new cable stayed spatial steel framework was designed in order to replace the damaged membrane roof. This design collected all the previous problems related to the difficult boundary conditions, determined by the existing structure. The new roof of around 20000 m2 will be suspended from the inclined tower by 28 stay cables. Due to the particular boundary conditions and low ratio between dead loads and live loads the structural system, appears to be sensitive to differential stiffness between space frame ring and cable supporting substructure.lt was necessary to proceed to an elastic and anelastic response of the roof system according to an oriented statistical simulations' sensibility analysis in order to determine the probability of failure and the b-index in function of parametric variations related to random uncertainties of the stiffness of cable stays [20].

and the most probable failure mechanism will involve primarily the longest cables (Figure 25). The sensibility analysis was, therefore, extremely important to detect the weak points of the structural system and permits proper local dimensioning to prevent chain failure, as illustrated with the failure simulation of same sensitive cable elements.

DETAIL DESIGN
Conceptual and working detail design plays a significant role in the real response of long span structures under the influence of direct or indirect actions. Avoiding chain collapse. Many modern buildings' codes introduced the concept of preventing to avoid chain collapses in large structural system. Under the conceptual design phase of the Olympic Stadium in Rome [14], special attention has been paid to the incremental failure analysis of part of the structure. The membrane elements are uncorrected with the cable structure in order to avoid fragile rupture propagation and facilitate eventual repair and/or substitutions of membrane panels. A chain failure simulation analysis shows that the structural system allows for complete stress re-distribution under partial collapse of more than 4 rope trusses. The tension inner ring is formed by 12 cables minimizing the effect of global collapse due to individual failure.
2E-04-, 1.5E-04-J 1E-0<M 5E-0S

4.00E-05

6.00E-05

8.00E-06

1.00E-04

1.20E-04

1.40&04

1.60E-04

1.80E-04

2.00&04

3 OE+00-5E-05-1E-04-1.5E-04 -2E-04-4 0' , 1 '

'

<

<

<

Standard Deviation Fig 24 Failure probability for load case 1

, 3

, 4

, S

, 6

, 7

, 8

, 9

, 10

, 11

, 12

, 13

1 14

Fig 25

Most Probable De at failure load case 1

61

The project of another complex structural system as the Stadium of the Alpes [3], includes a special detail design avoiding collaborations between column and cable and eliminating unnecessary and dangerous interaction. This was determined by the high sensibility of the column stability to parametric local bending . Structural details, able to isolate sub-structures under strong seismic actions have been designed for the Thessaloniki Olympic sport centre [21].

Industrial Aerodynamics, 1992, pp. 1447-1458, 13. B.J. Vickery: Wind loads on the Olympic Stadium - orthogonal decomposition and dynamic (resonant) effects. BLWT-SS28A, 1993. 14. M. Majowiecki: The new suspended roof for the Olympic Stadium in Rome, IASS Symposium, Copenhagen, 1991. 15. C. Borri, M. Majowiecki, P. Spinelli: "Wind response of a large tensile structure - the new roof of the Olympic stadium in Rome, Journal of Wind Enginerring and Industrial Aerodynamics, 1992, pp. 1435-1446. 16. M. Lazzari, M. Majowiecki, A. Saetta, R. Vitaliani: "Generazione artificiale dell'azione del vento: analisi comparativa degli algoritmi di simulazione nel dominio del tempo", 5 Convegno Nazionale di Ingegneria del vento, ANIV, Perugia, settembre

CONCLUSIONS
It has been noted the influence of knowledge base on conceptual design in removing gross human intervention errors from initial design statements. Computer assisted methods of design and analysis allow to generate an interactive design cycle which optimizes the user synthetical capacities and the analytical computer elaboration power. Sensibility analysis is an extremely powerful tool to determine the influence of parametric design uncertainties for unusual long span structural systems.

1998. 17. M. Lazzari, M. Majowiecki, A. Saetta, R. Vitaliani : "Analisi dinamica non lineare di sistemi strutturali leggeri sub-horizzontali soggetti all'azione del vento",5 Convegno Nazionale di Ingegneria del vento, ANIV; Perugia 1998. 18. L. Lainey: Montreal Olympic Stadium, Innovative large span structures, IASS Congress, Toronto, 1992. 19. M. Majowiecki: Snow and wind experimental analysis in the design of long span sub-horizontal structures, J. Wind Eng. Ind. Aerodynamics, 1998. 20. A.G. Puppo, R.D. Bertero, M. Majowiecki: Oriented simulation sensibility analysis of a long span structural system", internal report. * 2 1 . M. Majowiecki, F. Zoulas, J. Ermopoulos: "The new sport centre in Thermi Thessaloniki": conceptual design of the structural steel system, IASS Congress Madrid, September 1999.

REFERENCES
1. 2. R.E. Melchers: Structural reliability, Elley Horwood ltd. 1987. Structural Design Of Retractable Roof Structures, IASS working group n16, 1996 3. M. Majowiecki: Conceptual design of some long span sport structures", Innovative large span structures, IASS Congress, Toronto, 1992. 4. K. L. Carper: " Construction Pathology in the United States", Lessons from structural failures, Structural Engineering International. 1/96. 5. Conceptual design of structures, International Symposium, Stuttgart, Germany, October 1996. 6. M. Majowiecki, R. Trevisan: "A graphic interactive software for structural modelling analysis and design", Space Structures 4, Thomas Telford, London, 1993. 7. A. Samartin: "Application of optimization techniques to structural design, Lightweight structures in civil engineering, Warsaw, Poland, September, 1995. 8. Working Commission V: "Using computers in the design of structures" - IABSE proceedings. 9. First International Conference on Computational Structures Technology, Heriot-Watt University, Edinburgh, U.K., August, 1991. 10. RWDI: Roof snow loading study-roof re-design Olympic Stadium Montreal, Quebec. Report 93-187F-15, 1993. 11. M. Majowiecki: Observations on theoretical and experimental investigations on lightweight wide span coverings, International Association for Wind Engineering, ANIV, 1990. 12. B.J. Vickery, M. Majowiecki: Wind induced response of a cable supported stadium roof. Journal of Wind Enginerring and

62

WIND AND SNOW CONSIDERATIONS FOR WIDE SPAN ENCLOSURES


byPeter A. Irwin, Frank M. Hochstenbach and Scott L. Gamble Rowan Williams Davies and Irwin Inc. Guelph, Ontario, Canada 1.0 I N T R O D U C T I O N
The structures that are the subject of this book, wide span enclosures, are often sensitive to environmental loads such as snow and wind. The finite element analysis of these structures has become highly sophisticated and now allows the effects of a given load distribution from wind or snow, or a combination thereof, to be computed with great accuracy. This has encouraged designers to move ahead with increasingly imaginative and daring solutions to the wide span enclosure problem. However, the results of such sophisticated analysis are, in the end, only as good as the assumptions concerning what the snow and wind load distributions really are. This puts the onus on trying make the specification of these loads as realistic as possible. Simply using standard building code load provisions is hardly appropriate in these circumstances. Special studies using scale models and computer simulations of loads are typically resorted to. The Ford Field project features an enclosed playing field and seating area attached to an existing building, which will serve as mixed use development. The overall dimensions of the curved stadium roof are 630 ft by 540 ft (192 m by 165 m). The principal structural members consist of four super-trusses, two cantilevered and two simply supported, with clear spans ranging from 535 ft (163 m) to 626 ft (191 m) (Figure 2).

Figure 2 Ford Field Interior (Artist's Concept) To establish the wind loads on a widespan roof, the preferred method is to undertake tests on a scale model in a boundary layer wind tunnel. A number of papers have been written on the detailed experimental techniques involved, see for example Grieve et al (1992) and Xie at al (2000), and it is not intended to go into them in this paper. However, the special issues related to retractable roofs, internal pressures and aeroelastic effects have not received much attention in the literature and will be briefly discussed. The majority of the paper will be devoted to snow loads. Methods of establishing snow loads on wide span structures had, until the late 1980's, been far behind those for wind loads. One reason is that the problem of scaling down the processes of snow drifting on a small scale model using dynamic similarity principles is more difficult than that for wind loads, due to the greater number of dynamic similarity parameters that must be satisfied, Irwin and Williams (1983), Kind (1986). Both wind tunnels and water flumes have been used. Figure 3 illustrates a test in a water flume in RWDI's laboratories using fine silica sand to represent the snow while the water simulates the wind. These types of model tests can yield useful qualitative information provided they are interpreted

Figure 1 Ford Field (Artist's Concept) To illustrate some of the applications of scale model testing and computer simulations, recent studies conducted by Rowan Williams Davies & Irwin will be referenced. In addition to studies in Korea and on Safeco Field in Seattle, Washington, frequent references will be made to studies performed on the proposed Ford Field in Detroit, Michigan, USA (see Figure 1). When completed, this 65,000 seat stadium will be the home of the NFL Detroit Lions football team. The stadium is owned by the Detroit/Wayne County Stadium Authority, and the principal design team for the stadium includes the firms of Smith Group Incorporated, Rossetti Associates Architects, and Thornton-Tomasetti Engineers.

63

carefully based on experience in the field. However, to obtain detailed quantitative information on the snow loads is difficult by the scale model methods alone. This is due in part to the scaling uncertainties but is due also to the snow accumulations being the result of a complex series of events. During a storm the snow fall rate, wind direction and wind speed all vary as a function of time. This is difficult to simulate. In some areas of prolonged low temperatures the problem is made yet more complex since several successive snow storms may all contribute to the load, sometimes interspersed by periods of partial melting. Snow may also turn to rain which then is absorbed into the snow already on the structure, thus adding to the weight.

a wide span enclosure are those acting on the main structural system, those on secondary structural members supporting the cladding and finally the local peak loads on the cladding. Therefore wind tunnel techniques have evolved to supply this information. The tests must be in a boundary layer type wind tunnel so that the mean velocity profile and the turbulence of the natural wind are properly simulated in the scale model tests. Local pressures can be determined directly on the wind tunnel model by point pressure measurements. Where the roof underside is exposed to wind the net wind loads are measured by installing pressure taps on the underside as well as the top side. Wind loads on secondary structural members can frequently be measured by the area average method in which the instantaneous values of the fluctuating pressures at a number of taps covering the required tributary area are averaged and the peak value of this area average determined. The wind loads on the main structural system can in some cases also be evaluated by the area averaging technique but more frequently it is necessary to evaluate more than just a simple area average. The load effect at a particular point in the main roof structure, which could be a member force or stress for example, or a bending moment at some point, depends not only on the local wind load at that point but on the integrated effect of the loads over the entire roof. Writing the load effect in the i structural member as E and assuming at this stage of the discussion that the roof responds instantaneously to applied loads, the load effect is given by
th t

Figure 3 Ford Field Water Flume Test Clearly to make systematic and realistic predictions of snow loads requires a procedure that can properly track the effects of the multiple variables involved, such as wind speed, wind direction, snow fall rate, rainfall, temperature, solar radiation, and heat transfer through the structure. Computer simulation methods are particularly well suited to tracking cumulative effects and this has led to the idea of combining the strength of this approach with that of the scale model study. At present a scale model test in a boundary layer wind tunnel is still the most effective way of determining the wind velocity patterns over the structure, although this may change in future. In the mid1980's RWDI developed the approach of using the wind tunnel to determine the detailed wind velocity field over the structure for various wind directions, and using the computer simulation method for essentially solving the rest of the problem, Irwin and Gamble (1987, 1988). This approach has been named the Finite Area Element (FAE) method because the snow drifting part of the simulation is done using a grid system that divides the roof into a large number of finite areas. In this paper the method will outline the main principles of the FAE method and give examples of its application.

E (t) =
i

\\C (x,y)P(x,y,l)dxdy
i

=
A
i

jjC (x,y)P(x,y)dxdy
!

+jjC (x,y)p(x,y,t)dxdy
A

(!)

where P(x,y,t) is the instantaneous pressure difference between top and bottom sides, P (x, y) is the time average of P(x,y,t),p(x,y,t) is the fluctuating component of P(x,y,t), the double integrals are over the whole area A of the roof, E is the mean load effect in member i and e (t) is the fluctuating part of the load effect in member i. The coefficient C (x,y) is the influence coefficient for member i and gives the load effect due to a unit load applied at the point (x,y). Clx,y) can be thought of an influence "surface" varying with x and y. The shape of the influence surface will be fixed for a given structural component but will vary for different components. What is required for structural design purposes is the peak load effect, E , which is given by
t t t t

2.0 W I N D L O A D I N G

(2)
2.1 W i n d Tunnel Testing
where e is the peak of the fluctuating load effect.
{

The wind loads that are needed for the structural design of

64

What can be done in a wind tunnel test effectively is to evaluate the right hand side of Equation 1 for some representative forms of influence coefficient. Thus, rather than evaluating a simple area average pressure, a weighted area average is computed with the influence function as the weighting factor. On a cantilevered roof for example the bending moment at the root of the cantilever is important. Therefore one of the chosen weighting functions in this case would be would be the distance out from the root. For wide span roofs their dynamic response is frequently important because their natural frequencies are often low. The moving mass of the roof in its various modes of vibration results in inertial forces that are in addition to the direct wind forces. To evaluate these additional forces the so called modal load for each mode of vibration of importance needs to be evaluated. Modal loads are again a form of weighted area average load, with the weighting factor now being the modal deflection shape. Once the additional forces from modal excitation have been evaluated they then need to be combined appropriately, put into the form of equivalent static loads and added to the direct wind loads. Methods of making these combinations are discussed by Xie at al (2000). The end results of such testing and analyses are illustrated in Figure 4 where a wind load case for the Ford Field roof is illustrated.

unless it is impact resistant. A helpful factor is that usually wide span enclosures have large interior volumes and, unless an opening in the walls is very large, it takes a significant time for the internal pressure to respond to changes in pressure caused by exterior gusts. As a result the transmission of gust pressures to the interior is attenuated for large interior volumes. Note that modern wind tunnel test procedures now enable the correlations of exterior pressures at a number of openings to be measured and thence the effects on internal pressure computed. These issues are discussed in more detail by Conley et al (1999).

2.3 Aeroelastic Effects


A final topic concerning wind loading is the possibility of unstable oscillation such as that which destroyed Tacoma Narrows bridge. Flexible large span roofs can have natural frequencies as low as some long span bridges and the possibility of aerodynamic instability needs to be considered. To investigate dynamic instabilities an aeroelastic model is needed, one that simulates the flexibility and mass properties of the real structure. Figure 5 shows the aeroelastic model of the new Safeco Field baseball stadium in Seattle, Washington.

Figure 5 Safeco Field Aeroelastic Model These models are more complex to design and build than a rigid model but have been fairly frequently used on roofs with natural frequencies less than 1 Hz. A simple measure of aerodynamic stability is the aerodynamic damping. The damping of the model can be measured fairly simply by disturbing it from equilibrium, recording the decay of oscillations and then using the logarithmic decrement method to evaluate the damping ratio. The difference between such measurements with wind on and wind off gives the aerodynamic damping. Other more sophisticated methods such as the random decrement method may also be used. Figure 6 shows the results of such measurements in tests on a recent stadium in Korea. It can be seen that in this case the aerodynamic damping initially increases with wind speed but then changes its behaviour and starts to decrease. At high enough wind speed it eventually goes negative which indicates the onset of aerodynamic instability. Fortunately the wind speed where this happened was high enough not to be a concern but it is an

Figure 4 Example of Structural Wind Loads (psf)

2.2 Internal Pressures


One issue which is important for roofs over enclosed spaces is the internal pressure. Frequently the internal pressure is similar in magnitude to the exterior uplift, and thus can have a large impact on the structural design. The internal pressure in storm conditions is sensitive to the locations and sizes of openings in the walls of the structure and it becomes critical that the probability of there being openings is addressed. In hurricane regions glass is readily shattered by flying debris, thereby creating openings,

65

indication that aerodynamic instabilities do exist on large span roofs.


Turbulent Flow

wind before themselves landing and throwing up yet more particles. Although a small fraction of particles do fly higher and can cause visibility problems, most particle trajectories, particularly those of the larger particles that make up the vast majority of the moving snow mass, are not more than a few centimetres high and a few tens of centimetres long. Saltation drifting begins once the wind speed exceeds a certain value. This threshold speed, U , is typically about 4 m/s for fresh snow (as measured about 1 m above the surface). Old snow, especially after a melting episode, or if it has been rained upon, is much less able to drift and has a much higher threshold velocity.
th

Negative Darpping

0.5

1.5

Wind Speed Ratio, V/V(100 yr)

- B -

120 Deg -m- 286 Deg

Figure 6 Aerodynamic

D a m p i n g Ratios

2.4 Retractable Roofs For retractable roofs the roof geometry varies significantly during the different stages of the retraction and closing processes. The wind load distribution therefore also varies and it is typical to examine several stages during the wind tunnel tests. This is not only for the purposes of structural design. Wind loads often dominate the power requirements * for the drive system and can affect the braking system design also. Usually there will be a range of wind speeds, e.g. 0-15 m/s mean hourly at 10 m height in open terrain, in which the roof is expected to open and close within a specified time such as 10 or 20 minutes. Slower operation may be tolerated at higher wind speeds. For the evaluation of power requirements it is not so much the peak loading condition that is important as the average loading together with a knowledge of what fraction of the time the wind loads will be within certain ranges. The presentation of wind tunnel results needs therefore to be in this form. The combination of roof uplift and drag is also important in some cases, particularly where friction between bogey wheels and rail tracks is being relied upon as part of the braking action.

The mass flux of saltating particles is ultimately limited by the drag effect of the particles on the wind near the snow surface. The wind is slowed down enough that it can no longer sustain a higher load of airborne particles and an equilibrium mass flux is reached. As discussed by Gamble, Kochanski and Irwin (1991) the distance required to reach the equilibrium flux is estimated to be of order 5m or less. The equilibrium flux rate q can be expressed as

q=f(U,Uj

(3)

where U is the mean wind velocity at the 1 m height and U is the threshold value of U for the onset of drifting.
th

An example of this relationship is that of Dyunin (1963)


q = c U
2

( U - U j

(4)

where the wind speed is in m/s, U is taken as 4 m/s and the constant c = 3.34 x 10" kg/(s m ). Another, suitable for fresh snow, is that given by Kobayashi (1973)
th 5 2 4

q = c U

(5)

Where c = 3x 1 0 kg/(s m ). Relationships such as this one essentially describe saltation type drifting and do not include snow in suspension. The additional mass flux in suspension is very small at low wind velocities but becomes more significant at higher speeds approaching 14 m/s. Schmidt (1986) provides information on the division between saltation drifting and suspension drifting which can be used to assess the probable contribution of suspension drifting to total snow flux at higher wind speeds.

3.0 S N O W L O A D I N G 3.1 The Process of Drifting


Snow drifting is a complex process which has been studied by a number of authors, e.g. Bagnold (1941), Dyunin (1963), Mellor (1965), Isyumov (1971), Kobayashi (1973), Kind (1981), Iversen (1982), and Schmidt (1986). What is important to know, and what is not very obvious visually, is that at the most frequently occurring wind speeds, i.e. less than about 14 m/s near the snow surface, the bulk of the movement of snow particles occurs within a few centimetres of the surface through a process called saltation. Saltation consists of particles being ejected from the surface by the impact of landing particles. Once thrown up into the air stream they are rapidly accelerated by the

3.2 Drifting on Roofs


An important implication of saltation drifting is that you will not generally find snow being drifted in any significant quantities from ground level up onto a roof. The particles drifting along the ground simply do not fly high enough. Therefore, since snow can usually drift off the roof, but not up onto it, there is a general tendency for the overall amount of snow on the roof to be depleted by

66

wind action. Exceptions to this only occur in extremely high winds or where the ground shape forms a "ramp" up to the roof level. The wind velocity patterns on a roof depend on the roofs geometry and can be very complex. Some understanding of why it is that snow will accumulate in some areas and not in others can be obtained by considering a simple case. We consider the case of two dimensional conditions and examine the rate of change of snow mass M on the roof in a i m wide lane between an initial station 1 and a second station 2, a distance Ax downwind. The value of Ax is assumed to be larger than the distance required for equilibrium snow flux to become established. From Equation 3 we may write the snow flux rates at stations 1 and 2 as

scouring. In the typical real life situation the wind flow is not two dimensional but the rule still holds that snow tends to accumulate in zones of decelerating wind and to be scoured in zones of accelerating wind. The Finite Area Element method uses a similar control volume approach to that described here to compute the snow mass transfer and deposition on roofs, but extended to three dimensions.

3.3 Simulation of Drifting in the Finite Area Element (FAE) M e t h o d


The basic assumptions of the FAE method follow from the discussion in the previous sections. The fact that the particle path lengths are small compared with the building dimensions and that the snow drifting can be envisaged as almost entirely consisting of a thin carpet of particles moving over the surface with a mass flux dictated by the local mean wind velocity at a reference height of 1 m, say, above the surface makes it feasible to think in terms of computing the snow drifting process based on a knowledge of the mean wind velocity field at this reference height. This is the starting point for the FAE method.

q,=f(uuj
<l2=f(U ,Uj
2

(6)

We consider the rate of build up of snow mass M in the control volume bounded by the snow surface, the vertical plane at station 1, the vertical plane at station 2, and the top of the saltation layer. The rate of build up of snow mass is simply the difference between what enters at station 1 and what leaves at station 2 (flow out of the top surface of the control volume being assumed negligible). Therefore we may write dM

dt = f(U,,U )-f(U ,U )
th 2 th

(7)

which may, in turn, be written

dM
dt

df(U,Uj
dx

-Ax

(8)

Figure

Finite Area Element Grid

where x is the coordinate parallel to the wind. We may express Equation 8 in terms of the snow mass m per unit surface area and the derivative/'= df/dU.

dm _ dt

^,dU dx

(9)

From this expression, with the knowledge that/" is always positive, it is clear that if the wind is accelerating between stations 1 and 2, i.e. dU/dx is positive, there will be a decrease in the snow mass lying on the surface between stations 1 and 2 because dm/dt will then be negative. Similarly, if the wind is decelerating then there will be a build up of snow mass between stations 1 and 2 since dm/dt is then positive. What Equation 9 tells us is that it is the gradient in wind velocity, rather than the velocity itself, that is important in creating snow accumulations or

Figure

Finite Area Element

As depicted in Figure 7, the roof is divided into a large number of control volumes, or finite areas, by a grid system and it requires that, for a given approaching wind speed and direction, the mean velocity field be known (through wind tunnel tests or calculation) at the 1 m height above the snow at each grid node point. Then expressions

67

such as Equation 4 can be used to compute the snow mass fluxes over the boundaries of the area elements and thence to compute the change in snow mass in each element over a given time interval. Figure 8 illustrates the typical area element. The mass flux of snow that would exist provided the snow were in a driftable state, and that there was an adequate supply of snow in upwind elements, is called the 'potential' mass flux and at grid point i,j is denoted by q j. Using Equation 4, the potential mass flux at grid point i,j is
{

fluxes at the two ends of each boundary. Thus

9, = T ( - 9 r 9i
= J

-9r
+ l

i J + l

) )
( 1 2 )

r(~ lx j
i i

~9x

i + l i j + I

if 9 = T(9
4 X i i j

\ +9
X i + l i j

\ )

where the convention is used that fluxes into the area element are positive and fluxes outwards are negative. The change in mass Sm in the element after time St may now be computed from

q.^culju^-uj
The components of mass flux at grid point i,j are

do)

Xij

=q , ( i,j) / i
l J

C0S a

Sm =(q l ( )
7

I 1

+q l

2 2

+q l

3 3

+q l )
4 4

+M . + M . - M
jail rain melt

where a is the angle of the local velocity vector to the grid, see Figure 8.
Lj

Normally wind tunnel tests are needed to determine the values of U and a but for some simple geometries it is possible to use calculated values. It is recognized that the formation of drifts can alter the values of U and a since the local wind velocities are influenced by roof shape which is effectively changed by the growth of drifts. However, for large roofs the drift depths are typically relatively small compared to the overall roof dimensions and the changes in shape presented to the wind will not be sufficient to appreciably alter the wind velocity field. Therefore, for large roofs, except near local surface obstacles such as penthouses or steps in roof elevation, there is justification for using the approximation that the patterns of and a measured on a model of the bare roof will still provide a reasonable basis for computing drift patterns. At features such as a step changes in roof elevation the step can eventually fill in with snow in the classic triangular shaped drift to the point where no more snow can be held. Effectively the wind velocity at the snow surface in this instance changes from being virtually zero prior to the step filling in to being the same as the velocity at the top of the step once the step is filled. It is reasonable in this instance to incorporate a switch in the computational procedure to change the velocity on the lower level roof near the step to that at the top of the step once sufficient mass has accumulated to cause the step to be filled to capacity. A similar approximate approach, i.e. giving an area element a maximum storage capacity, can be used with other types of discontinuous change in roof geometry where local drifts can form. A knowledge of the maximum volume of snow that can be held is required in these situations and good guidance on this can usually be obtained from scale model tests with particles in a water flume or wind tunnel.
itj lp Uj u y

where / through l are the lengths of the four sides of the area element. In this expression we have introduced the three other sources of mass change. These are: the added mass M due to snowfall landing in the area element during the time interval; the added mass M due to rainfall which soaked into the snow; and the loss of mass M due to snow melting and liquid water running out of the snow.
4
fall rain melt

Not all snow is driftable. As it ages there is a tendency for the threshold velocity for the onset of drifting to increase (see Kind, 1981, for example) until only very strong winds can initiate drifting. If snow has undergone significant melting then it will tend to become fixed without further drifting even if the temperature subsequently drops and refreezing takes place. If snow has been rained upon or experienced freezing rain these events also tend to fix it in place. This is modelled in the FAE simulations by incorporating a switch that turns off drifting if any of these events take place. Only new snow landing on top of the fixed snow is subsequently allowed to drift.

3.4 Melting and Retention of Liquid W a t e r in the Snowpack


The process of melting is complex and is one of the more difficult parts of the FAE simulation technique. A detailed approach to computing melting is taken in FAE studies, including a heat balance calculation, taking into account the sources of heat input and output from the snow pack, such as transfer from the roof below, solar radiation from above and the latent heat of melting and freezing. The ability of the snow to hold meltwater or rain is also an important variable which forms part of the computations. There is not space here to describe the full procedure but it is described in detail in Irwin et al (1995). Because of the complexity of melting and water retention, the FAE method is usually calibrated by computing ground snow loads at an idealised site with no drifting and

The average fluxes q q , q and q normal to the four element boundaries can be computed as the average of the
h 2 2 4

68

comparing the results with recorded ground snow load data. Adjustments can be made to variables such as the assumed density of freshly fallen snow, or to the capacity of the snow to hold liquid water, until the FAE simulation predictions of extreme ground loads are generally consistent with those deduced from the long term records. Also roof snow loads predicted by the FAE method are typically expressed as a fraction (or as a multiple) of the selected return period ground snow load (30 yr, 50 yr, 100 yr) computed by the FAE method. This normalization process removes some of the uncertainties associated with the melting processes. Roof loads are then arrived at by multiplying the computed roof snow load factors by the generally accepted value for the selected return period ground load for the area in question, such as given in the applicable building code.

IIMi-<i?<k-| i 4 < i i P M M i w n . . . . , , .

. k . i . ^ ' / V ' . M ' *afc*vvvarj

. , . .

m * V J VaVa

** v v v v v . *

'* ** :~4 : * ; v v a w b r w i V i M N ; ; i k k "J i ViWsZZi 'A l l l l

a / j ' i ' i ' < "i *j U'VtVs

i i i i i l l l i i

Figure 10 Roof Surface Wind Flow Pattern The practical implementation of the FAE method is fairly complex but can be summarised as follows: read digitized meteorological records of wind speed and direction, snowfall, air temperature, rainfall, cloud cover etc. for a given hour; compute local wind speeds and directions at l m height above roof at grid node points using local velocity ratios and directions from wind tunnel tests; determine if snow is in a driftable state based on temperature history, rainfall records, and storage capacity of the finite area element; compute the quantity of snow drifting into and out of each area element during the hour; perform mass balance computation for each area element for the one hour time interval; compute melting quantities and the snow load in each area element at the end of the hour; and then move on to the next hour of data and repeat computations.

3.5 Practical Implementation of t h e F A E Method

Figure 9 Scale Model in Wind Tunnel From the above description it can be seen that an important part of the FAE method is determining the wind velocity vectors on the roof for a series of wind directions. In order to accurately model the wind flow patterns across the roof, wind tunnel tests on a scale model are undertaken in a boundary layer wind tunnel to obtain data on wind velocities near the roof surface (Figure 9) in a simulated natural wind for typically 16 wind directions at 22.5 intervals. The local velocities are measured using special surface velocity vector (SVV) sensors described by Gamble et al. (1991). The local velocities are converted to velocity ratios by dividing by the velocity at gradient height above the simulated boundary layer. The results of the wind tunnel test are then combined with the finite area element model to produce a flow-field across the entire roof for the 16 directions tested. Figure 10 shows this plot for winds from 112.5. The velocity ratios at each node may then be used to compute full scale velocities on the roof by multiplying by the appropriate full scale gradient wind velocity inferred from the meteorological records and knowledge of the wind exposure of the meteorological station (usually an airport).

This procedure is continued first for one winter and then for a series of successive winters for which records are available, typically 30 to 40 winters in total. For each hour of each winter a complete pattern of snow loads on the roof is thus generated.

3.6 D e v e l o p m e n t o f Design S n o w L o a d s
For structural design the snow loads of interest are typically those with a return period of 30 to 100 years. A great number of different snow load patterns can occur in this number of years but from a practical point of view it is desirable initially to identify certain simple load cases and to express the FAE simulation results in that form. For example, building codes normally have as one load

69

case the simple uniform load. To obtain an analogous load case from the FAE simulation for a large span roof, the results can be put into the form of an equivalent uniform load by dividing the total weight of snow on the roof by the roof area. The resulting 'uniform' load can then be tracked over each winter and the maximum for each winter identified. An extreme value analysis may then be undertaken of the 30 to 40 annual maxima so as to compute the 'uniform' load for the selected return period. For an arch roof a more important case is where there is an unbalanced load, with more snow load on one side of the arch than the other. For an arched roof therefore the FAE results are used to compute the degree of unbalance between the loads on the two sides of the arch. The annual maxima of the difference in loads on the two sides of the arch would be evaluated in this case and again an extreme value analysis undertaken to find the value for the selected return period. For roofs with multiple bays all to be designed the same, the maximum bay load would be of interest. In this case the maximum bay load for each winter would be identified and again an extreme value analysis would be undertaken on the annual maxima to arrive at the load for the selected return period.

snow climate. Figure 12 illustrates the FAE-simulated ground snow accumulations for Ottawa during the winter of 1977-78.

i */* --.

Jij.n

iL

i-.L

* - H - l WHtl * - * - ! aial U

Figure 12 Time History of Accumulation and Precipitation Ottawa, Ontario, 1977-78 In more moderate regions, the temperature cycles below and above freezing several times over the course of the winter. A significant snow storm will deposit snow on the roof, which will often melt before the next storm arrives. In this climate, the design snow load can be achieved by the occurrence of two or three significant storms during a 'cold period, often followed by a rainfall as the temperature edges above freezing. In this climate, an extended period of unusually cold weather, where the snow does not melt over the course of the winter, will produce significant (design-type) loads on the roof.

Return Period |Years) Linear log Log normal

Figure 11 Extreme Value Fit In carrying out an extreme value analysis, the annual maxima are fitted by an assumed mathematical form for the probability P that a given load S will not be exceeded in any one year. Various mathematical forms can be used, the one giving best fit depends on the local climate. Typical forms are the Fisher-Tippet Type I and the Log Normal fits (Figure 11). The snow climate for the area where the building under study is to be built plays a large role in the performance of the roof in a finite area element snow loading simulation. In more northerly regions, the temperature drops below freezing in late fall or early winter, and rarely rises above freezing for any significant amount of time. In this climate, snow that falls on the roof will remain there throughout the winter. The design snow load is achieved as a result of several successive significant storms over the course of the winter, usually peaking in March or April. Ottawa, Ontario, Canada is a good example of this type of

18-Dec Rainfall

...H

ji
22-Jan 8-Fet) Snowfall

i
15-Mar Load

M i

*-M-*\ H-jN*l

Mi

Figure 13 Time History of Accumulation and Precipitation Detroit, Michigan, 1981-82 Detroit, Michigan is a good example of the second type of snow climate. As illustrated by the FAE-simulated ground snow accumulations in Figure 13 during the winter of 1981-82, there were four distinct loading periods, separated by warm periods when most of the snow was removed by melting.

70

3.7 Examples of Application of the F A E Method


There were several areas where the Finite Area Element method was used, in whole or in part, to assist in the design of the roofs of the Ford Field stadium. These areas will be discussed in more detail below. In addition to the wind tunnel test described above (Figure 9), the scale model was tested in the water flume snow and wind simulator to assist in the setup and analysis of the FAE simulation (Figure 3).

consideration. In reality however, a secondary structural member with a smaller tributary area is more sensitive to the effect of a localized drift than would be a primary structural member with a much larger tributary area. Therefore, the area-averaged snow load for a smaller structural member would be higher than that for a main structural member. By defining load cases for the tributary areas of these secondary members, the FAE simulation can more accurately determine these higher loads that the secondary members would experience. Similarly, the roof decking material would be even more sensitive to localized drifting. A localized drift could form on the roof which, due to area-averaging, would not be a major concern for the structural member, but which could result in the "punch-out" of the deck panel underneath. In addition to the defined load cases, the FAE simulation calculates the highest "local" loads experienced in each individual FAE element, for use in the design of the roof decking system. As experience has demonstrated, unbalanced loads due to snow on wide span roofs, especially arched wide span roofs, can be as important as total snow load. By defining loadcases to monitor the greatest difference in snow loads between both sides of the arch, the FAE method can determine more efficiently what the design unbalanced load should be. Figure 15 illustrates the maximum unbalanced snow load experienced for one of Ford Field's main trusses, during the winter of 1984-85. In this figure, loads are significantly higher on the south half of the roof (north being at the top of the figure). If the trend for the curved roof is to consistently experience greater unbalanced loads on one side than in the other, this can be identified in the FAE method as well, by comparing yearly maximums and looking at the time history of the load. In addition, the FAE simulation identifies whether the greatest unbalanced load occurs when one side is heavily loaded and the other is partially loaded, or if one side is almost completely unloaded and the other is partially loaded. This distinction is of great importance for certain types of structural systems. In total, for a single arched structural member, up to four different unbalanced loading patterns are generally identified.

3.7.1 S n o w Loads o n the M a i n W i d e s p a n Roof


In order to determine the design uniform and unbalanced snow loads on each super-truss, the tributary area for each truss was defined in the FAE simulation as a load case. The maximum uniform and unbalanced loads in these tributary areas were found for each year, and an extreme value analysis performed to determine design loads. If this roof were to be designed using the relevant building code, the building code would predict that all four super trusses would experience the same uniform load over the entire roof. By using the FAE method, the directionality of the prevailing winds and the aerodynamic effect of the non-symmetry of the curve in the roof are taken into consideration. Each truss is examined independently. Figure 14 illustrates the maximum snow load experienced for one of Ford Field's main trusses during the winter of 1981 -82. In this figure, the lighter shades of grey indicate higher snow load than the darker shades of grey, and north is at the top of the figure. As this figure illustrates, the maximum load varies from element to element across the roof. The general trend is for loads to be higher on the south half than on the north half and higher on the west half than on the east half. This analysis provides uniform loads tailored to each super-truss, leading to a more efficient design.

Figure 14 Example

F A E S n o w Load Pattern

The building code approach also provides only one overall uniform load for a typical flat roof, regardless of the tributary area of the structural member under

Figure 15 Example

F A E Unbalanced

S n o w Loading Pattern

71

monitored simultaneously. The building code would assume that both roofs which contribute to the loading of the super trusses experience their peak snow loads at the same instant in time. The FAE simulation can determine if this is the case or not. In the case of Ford Field, load cases were defined to look at snow loads on the roofs both simultaneously and independently, to assess this effect. It was determined that the maximum simultaneous load was in the order of 5-30% less than the load when considered separately.

3.7.3 Loads o n Existing Roofs


When constructing new buildings adjacent to older ones, or adding new additions to existing buildings, account needs to be taken of how the new structure will impact snow loads on the existing roof. There have been several examples of roof collapses on roofs that have functioned normally for years, but have experienced dramatically increased snow loads after the construction of an adjacent, higher roof. This is another function capably handled by the FAE method, either during the design stage, or postcollapse as a tool for forensic engineering. Since the FAE analysis uses local meteorological data in it's simulation, modelling an existing building will determine the snow loads that building has actually * experienced during its lifetime. Then, by running a second FAE simulation on the existing roof including the new addition, the loads before and after construction can be compared to determine whether the new addition will have a positive or negative impact on snow loads on the existing roofs. If it is determined that the snow loads on the existing roofs will increase, the results of this comparison may be used to identify which areas are affected, and which areas are not, as a cost saving tool in the retrofit of the existing roof. In some cases, remedial measures can be employed on the new roof in order to alleviate the additional accumulation on the existing roof. In the case of Ford Field, the new football stadium will be built attached to an existing warehouse, which will be converted into a mixed use development as part of the new facility. Since the new structure is much larger and higher than the buildings that previously occupied the site, the snow loads on the roof of the existing warehouse could change dramatically. In this case, the existing warehouse was included in the FAE simulation to determine, after the addition, the snow loads that the warehouse roof would be required to support.

Figure 16 Example

o f Structural S n o w Loads (psf)

In the case of the Ford Field FAE study, separate design load patterns were provided for: Uniform loads on the primary structural members (Figure 16), Unbalanced loads on the primary structural members (four patterns), Uniform loads on secondary structural members, Uniform loads for the design of the roof decking system.

3.7.2 Step Loading and Simultaneous Loading


For the Ford Field stadium, a lower roof over the seating area surrounds the main curved roof. Traditionally, the change in roof elevation would produce a localized aerodynamic calm region adjacent to the step, where significant quantities of snow could potentially collect. The building code is designed to consider a roof-step in the order of approximately one storey. For roof steps of this height, snow drifting off the higher roof deposits directly adjacent to the step, generally producing a typical triangular drift. When the change in elevation is considerably higher than one storey, as in the Ford Field roof, the drifting snow tends to be transported further from the upper roof before depositing on the lower roof. For this situation, the typical calculations in the building code tend to be overly conservative immediately adjacent to the roof change, and non-conservative further out from the roof. By using the observed results from the water flume snow and wind simulation, as well as empirical formulae developed by RWDI, snow transport across the elevation change is custom-tailored to the roof step under consideration, resulting in more accurate deposition patterns. In the structural design of Ford Field, the outside supertrusses support portions of the lower side roofs as well as the upper curved roof. By defining loadcases combining the tributary areas on both roofs, loads on both roofs are

3.7.4 C o m b i n e d W i n d and S n o w Loading


For some roofs the combination of snow and wind loading can be a governing case (wind loads are not always upwards). A number of building codes specify load combination factors to cover the possibility of snow and wind loads being additive. These code combination factors tend to be arrived at somewhat arbitrarily because of lack of detailed research in this area. The FAE method

72

is well suited for examining combined snow and wind loads in a rational manner for specific projects. The basic method produces snow loads for every hour of 30 or more winters based on the detailed hourly meteorological records. When, in addition, wind pressure coefficients on the wide span enclosure are available from wind tunnel tests as a function of wind direction, the FAE analysis is readily modified so as to compute also the hour by hour wind loads. Simultaneous loading conditions can then be examined statistically and equivalent load combination factors deduced. This is best carried out specifically for each project, since, in some areas of the world, peak wind loads are generally produced in summer storms; in other areas the highest winds are produced in the winter in combination with snow fall and high amounts of drifting. Specifics of the response of the building to wind directionality are also critical in wind and snow loading combinations.

Loading on Large Roofs", 7 Int. Conf. on Wind Engineering, Aachen, Germany, Pre-prints vol. 3, pp. 171180. Irwin, P.A., and Gamble, S.L. (1988) "Prediction of Snow Loading on the Toronto SkyDome", Proc. First International Conference on Snow Engineering, Santa Barbara, publ. by CRREL, Hanover NH. Irwin, P.A., Gamble, S.L., Retzlaff, R.N., and Taylor, D.A. (1992) "Effects of Roof Size and Heat Transfer On Snow Loads on Flat Roofs", Proc. Second International Conference on Snow Engineering, Santa Barbara, publ. by CRREL, Hanover NH. Irwin, P.A., Gamble, S.L., Hunter, M.A., and Kochanski, W.W. (1993) "Parametric Studies of Snow Loads on Large Roofs", Rowan Williams Davies and Irwin Inc. Report 92118T-15, submitted to the Institute of Research in Construction, National Research Council of Canada, 25 March. Irwin, P.A., Gamble, S.L., and Taylor, D.A. (1995) "Effects of Roof Size and Heat Transfer on Snow Load: Studies for the 1995 NBC", Canadian Journal of Civil Engineering, August, 1995. Isyumov, N. (1971) "An Approach to the Prediction of Snow Loads", PhD Thesis, University of Western Ontario. Iversen, J.D. (1982) "Small-scale modelling of Snowdrift Phenomena", Proc. Int. Workshop on Wind Tunnel Modelling Criteria in Civ. Eng. Applications, Gaithersburg, Maryland, Cambridge Univ. Press, pp.522545. Kind, R.J. (1981) "Snow Drifting", Hand book of Snow, Principles, Processes, Management and Use, Chap. 8, ed. D.M. Gray & D.H. Male, Pergamon Press. Kind, R.J. (1986) "Snowdrifting: a Review of Modelling Methods", Cold Regions Science and Technology, 12, pp.217-228. Kobayashi, D. (1973) "Studies of Snow Transport in Low Level Drifting Snow", Inst, of Low Temp. Sci., Sapporo, Japan, Report No. A31, pp. 1-58. Mellor,M. (1965) "Blowing Snow", Cold Regions Science and Engineering, Part III, Section A3C, U.S. Army Cold Regions Research and Engineering Laboratory, Hanover, N.H.. Schmidt, R.A. (1986) "Transport Rate of Drifting Snow and the Mean Wind Speed Profile", Boundary-Layer Meteorology 34, pp. 213-241. Xie, J., Irwin, P.A., Kilpatrick, J., Conley, C , and Soligo, M. (2000) Determination of Wind Loads on Large Roofs and Equivalent Gust Factors, Int. Conf. On Wind Eng., Seoul, Korea.

th

4.0 C O N C L U S I O N
Present day computing methods for the structural analysis of large span structures are highly sophisticated and are capable of producing very efficient structural systems. However, the validity of their results are only as good as the assumed loading conditions used for the structural design. This makes it important to employ an equal level of sophistication in determining the design loads. This paper has described state-of-the-art methods for determining the design wind and snow loads and identified some additional issues of importance not adequately addressed using standard code-based methods for specifying wind and snow loads for wide-span roof structures.

REFERENCES
Bagnold, R.A. (1941) "The Physics of Blown Sand and Desert Dunes", Methuen, London. Dyunin, A.K. (1963) "Solid Flux of Snow Bearing Airflow", National Research Council of Canada, Ottawa, Technical Translation TT-1102, (from Russian). Gamble, S.L., Kochanski, W.K., and Irwin, P.A. (1991) "Finite Area Element Snow Loading Prediction Applications and Advancements", Eighth Int. Conf. on Wind Eng., London, Ontario, pub. Elsevier. Grieve, J.B., Stone, G.K., and Irwin, P.A. (1992) Prediction of Structural Wind Loads on Large Span Roofs, Innovative Large Span Structures, Vol. 1, pp 404-414, IASS-CSCE International Congress, Toronto. Irwin, P.A., and Williams, C J . (1983) "Application of Snow Simulation Model Tests to Planning and Design", Proc. Eastern Snow Conf. Vol. 28, 40 Annual Meeting, pp. 118-130.
th

Irwin, P.A., and Gamble, S.L. (1987) "Prediction of Snow

SECTION II Form and Environment

The Eden Project, Glass Houses and World Environments Planting Environment for the Eden Project Civil and Structural Design of the Eden Ptoject Foil Climatic Envelopes The Environmental Consequences of a Building with a Widespan Controlling the Indoor Climate in Widespan Enclosures

75

THE EDEN PROJECT GLASS HOUSES WORLD ENVIRONMENTS


Andrew Whalley B.Arch AA Dipl AIA RIBA Director Nicholas Grimshaw & Partners

THE GENESIS OF EDEN


Without plants there would be no life on earth. Plants are unique in their ability to convert the energy of the sun through photosynthesis and supply us with our lifeline support systems. Oxygen - food - fuel - medicines - clothes. Even our use of fossil fuels coal gas and oil are exploiting the results of photosynthesis from millions of years ago. Our increasing understanding exploration of the world and inexorably entwined with the civilisation and the potential of horticultural sciences, exploitation of plants is development of our of mankind.

known for his scientific interests and most ships'doctors in particular had great scientific interest in what they saw and collected on their travels. In parallel with the work of such famous botanists as Sir Joseph Banks was the emergence of a new architectural form to house and protect these new and delicate specimens, a built form that could transform the temperate climate of the British Isles into the humid tropics -the Glass House. The first half of the nineteenth century is the period of rapid change during which the demand for more exotic specimens in turn demanded much higher levels of natural light than was provided by masonry buildings with large windows. In 1811 Thomas Knight, head of the Royal Society, set out a challenge: Horticultural

" not a single building of this kind has yet been erected in which the greatest possible quantity of space has been obtained and of light and heat admitted - proportionate to the capital expended." Interestingly this statement was almost repeated verbatim by Peter Thoday at our first briefing session on the Eden project.
The Earth Intensive farming Sunflowers

THE GLASS HOUSE


The Eden Project endeavours to both recognise our country's great past heritage of plant exploration and at the same time look to the future. Helping to define, through research, a world where mankind can live and develop within sustainable parameters. To design the built form of a glasshouse, and to represent these ideals, requires both an understanding of the architectural heritage that made our past botanical achievements possible, and ideas for a more sustainable future. As a nation of adventurers we have developed navigational and seafaring skills that have allowed us to explore the globe. The ports of Cornwall were the first landfalls encountered by many returning eighteenth and nineteenth century sea captains, many of whom collected plant specimens on their travels. Captain Cook was well

Sir Joseph Paxton is synonymous with the emergence of this new design approach. He introduced industrialised and prefabricated techniques that produced great economies and speed of construction. However it is interesting to note that all of his work including the Great Stove at Chats worth (1841) and indeed the Great Exhibition of 1852 kept to the relatively safe and known technologies of timber construction. A bigger influence on the new architectural form was through the exploitation of new technology.

The Great Stove, Chatsworth, 1841

Timber glazing systems, Crystal Palace, 1852

76

In 1816 John Claudius Loudon patented a technique that exploited the malleable qualities of wrought iron, drawing and curving it into a structural glazing sash bar. Immediately he saw that this technology offered possibilities for a new type of Architecture. In his own words: " it may be beautiful without exhibiting any of the orders of Grecian or of Gothic may not therefore glass roofs be rendered expressive of ideas of a higher and more appropriate kind, than those which are suggested by mere sheds or a glazed arcade."

Thomas Knight by harnessing the best of the current technological understanding and generating new architectural solutions that were truly great expressions of the era.

Palm House, Kew

Palm House. Kew

Kibble Palace, Glasgow

Glass during this period was still taxed, and this was applied relative to the size of the panes. Glass was also still an expensive luxury material produced by spinning plates or cutting it out of cylinders. Consequently it was highly desirable to construct these new plant houses with their huge expanses of glass using many small sheets rather than large sheets. Traditional timber glazing bars with small sheets of glass were relatively inefficient in comparison to the slender wrought iron glazing bar; this produced the optimum transparent skin. Loudon developed his system with the contractor W and D Bailey using the glass as part of the structural system. This gave rise to completely new expressions of architectural form. Many of these glass houses were lost during the early part of the twentieth century. Fortunately one example remains to this day at Bicton in Devon. Its very light filigree structure is reminiscent of the delicate structure of a leaf.
6. Bicton Gardens, Devon, J C Loudon

It is interesting to contrast this to the temperate house designed by Burton ten years later, without the ingenuity of Turner, where a pre ordained architectural form takes over. The development of this very organic exploration of design solutions probably culminates in the soft rolling forms of Kibble Palace, originally constructed in 1865 and later dismantled, shipped and reassembled in Glasgow's botanical gardens. Certainly this building was enormously influential on myself during my time studying at the Mackintosh School of Architecture, and was a very strong influence as we explored solutions for the Eden Project. Paxton's ridge and furrow glazing at Chatsworth with its east west orientation minimised the loss of sunlight due to the reflective nature of glass. At the same time he had developed new mechanised prefabrication techniques that delivered an improvement in construction times and

Conservatory, Liechtenstein Castle

Leaf structure

The Palm House at Kew, built in 1848, uses the same technology and is far better known. It, also, demonstrates the value of collaboration, this time between the architect Decimus Burton with the intuitive engineering expertise of Richard Turner. They both answer the challenge from

Paddington Station, 1 K Brunei. 1854

77

construction efficiencies. The influence of the Great Exhibition (1852) on I K Brunei's Paddington Station (1854) is well-documented; both were built by Fox Henderson. Undoubtedly the techniques that had been developed for the glass house were transferred and developed for the demands of the latter nineteenth century: the large Railway sheds. This sequential influence took root in our own work when we came to design Waterloo International Terminal . There had in fact been little in the way of ail way architecture since the nineteenth century in the UK. The terminal would symbolise not only a new renaissance of high speed rail travel but also a new permanent connection and gateway to mainland Europe. The roof, although only 10% of the overall building budget, was to be the signature and emblem of this new service.

This is in fact where the line of thought turns full circle. Glass Houses to Railway Halls and back to Glass Houses. It was because of our work at Waterloo that the team were asked to prepare proposals for the Eden Project. Eden

International Terminal, Waterloo at night

EDEN
Our brief was to create a showcase for global biodiversity and human dependence on plants. The structures were to be large enough to allow the exhibition and study of a range of plants on a hitherto unachievable scale. Within the first phase two climate capsules were to be recreated from different world environments (biomes). The humid tropics (rain forest) and the warm temperate (Mediterranean) biomes were to be constructed as enclosures; a third zone, the temperate, was to be in a sheltered external area. It was the team's goal from the outset that the project should both entertain and educate at the same time. The creation of natural habitat zones that have the height and volume to allow plants to grow in a natural way, to their full mature height, has seldom been done; we believed it required fresh thinking. With the humid tropics this required an enclosure that would allow trees to mature and form a canopy at forty metres in height, setting a clear span building height of fifty metres. Botanical science has developed from the nineteenth century encyclopaedic cataloguing of specimens. Now, in the twenty first century, science is exploring our understanding of biodiversity and the importance of genetic grouping and ecosystems. Our goal was to develop an architectural response that was informed by these new demands in the same way that our predecessors had, almost two hundred years ago. Finding the right site for this new botanical garden was critical. The original sites that were considered were the clay workings at Roche. The 'Clay Alps' were mountainous heaps of clay and spoil; apart from the difficult ground conditions they were also highly exposed. The final chosen site was an old clay pit that was coming to the end of its useful life. This hollow in the ground provided the inspiration for our design solutions.

International Terminal, Waterloo Station, 1992

The technical challenge was to design a roof structure and envelope that could deal with the twisting and diminishing geometry of the track alignment. For speed and economy the glazing had to use standard rectilinear sheets of glass. In collaboration with the engineer Anthony Hunt Associates we developed the steel and glass roof that took on the sinuous shape of the tracks below.

Aerial view of the International Terminal, Waterloo

78

The Bodelva china clay pit, St Austell, Cornwall

TOPOGRAPHY AND FORM


Our starting point was to use the contours of the clay pit as an integral part of the architecture, using the quarry wall as one side of each biome. This had the advantage of creating great spatial drama and a terraced profile as staging for the planting , thus creating drama from day one, even when many of the plants would be relatively immature. A three dimensional model was created on the computer to explore the potential sites for the biomes. This was assessed by looking at both the topography and potential solar orientation. Sun path analyses were used to find the optimum location for each biome.
Early concept of 'arch' scheme Early concept sketches Ground model of existing site topography

The inevitable protracted funding process gave us time to thoroughly evaluate our proposals. There were the logistical problems of transporting large steel trusses in Cornwall. The quarry was also changing shape as the last of the clay was extracted, effectively meaning that our ground terrain was constantly changing as we tried to complete our proposals.

NATURE AND EFFICIENCY


There are many influences during any design process. During the development of Eden we often referred to the Science Fiction film from the early seventies entitled 'Silent Running' . This centres around a series of very light weight biomes replicating the earth's principal climate zones, all floating in outer space. The concept of these biomes helped to encourage us in our conviction to explore new and innovative technical solutions for the structure and envelope. At this time we were appraising light weight foil as an alternative to glass and we wanted a solution that would capitalise on the properties of this ultra light weight material, in a similar way as Louden's wrought iron system did for glass.

Model of 'arch' scheme

Roof of 'arch' scheme

Our first proposals built upon our work at Waterloo with a series of diminishing primary steel trusses connected to each other with a secondary system, supporting a ridge and furrow glazing system that would have been familiar to Paxton. However at this stage we had established the idea of a free form in plan and section that hugged the contours of the pit.

" % r f i * > i r r -

Image from the film 'Silent Running'

79

K^^M^t^S

honey combs because they are 'busy bees', trying to achieve the maximum with the minimum effort. Nature seems to continually form hexagonal structures as the most efficient way of absorbing stress.

THE EDEN SPHERE


We resolved our concerns by finding a simple and direct solution to the geometry. David Kirkland redefined the generation of the biome forms as a series of interconnecting spheres. We took this computer model and intersected it with the terrain model of the clay pit, which in turn defined the final form of each biome. This allowed us to develop a proposal that was independent of the exact quarry profile. It also allowed us to define the surfaces as geodesic shells, that could get the most out of the long span ultra light weight nature of foil pillows.

Responsive structural system, Andrew Whalley & Chris McCarthy

On a previous project I examined, with the engineer Chris McCarthy, a responsive structural system that could adapt to changing loading conditions. Forces could effectively be moved around the system in the way a body does with bones, muscles and tendons. This skeleton carried an ultra light weight skin formed from a series of pneumatic pillows using layers of transparent foil and spluttered metal coated foil. Again loads could be responded to by varying the pillow's air pressure, resulting in an extremely light weight dynamic enclosure. Ideas that start as theoretical exercises can help develop and inform later projects. Nature has many lessons to teach both architect and engineer; most obviously nature is based on the minimum use of energy and the careful use of resources i.e., efficiency in metabolism. What often appears to be fragile is actually robust as it has an ability to adapt.
Radiolaria Honeycomb

Connecting spheres

Sketch of siting

Spheres & structure

As with the Radiolaria the geodesic shell is formed from hexagons to minimise on tube length to surface area. The size of the hexagonal grid is a proportion of the diameter of the sphere, with the largest dome being subdivided into pillows with a diameter of approximately eleven metres.

^^mESv' ft" ' -*WW


0 * 0 .,; ^MWW\

Computer-generated image of the first 'geodesic' scheme

H O W TO BUILD A SPHERE?
Everybody is familiar with the problem of representing the surface of a sphere on a two dimensional plane. An orange skin can not be rolled out flat on a table, and the attempt to represent the surface of the earth on a sheet of paper leads to great distortions as far as the size of the land mass is concerned. However the surface of a sphere can be subdivided into planar triangle-based surfaces similar to a football. The earliest example for the realization of a geodesic sphere is Walter Bauersfelds Zeiss Planetarium in Jena, Germany from 1926.

Microscopic photograph of a fly's eye

An excellent example of these efficiencies can be found when examining the one-celled creatures Radiolaria. As they grow through centrifugal force the silica that they are formed from takes the geometric form of the minimum length hexagonal pattern. In just the same way bees build

80

Later, Richard Buckminster Fuller carried out substantial research into geodesic spheres and their underlying geometry. The problem has always remained the same. How can the surface of a sphere be subdivided into a number of building elements that: can be easily constructed with available construction methods are ideally selfsimilar in order to reduce the number of different components preserve the structural integrity of the overall structure

layer consists of triangles below the node points of the outer layer which circumscribe hexagons themselves. The connection between the two layers is established with diagonals which connect the node points of both layers. We call the resulting geometry a 'hex-tri-hex' grid. Although on first appearances this is a very ordered geometric solution there is a complication in resolving the interconnection of the spheres as each has a different diameter and sub divisional grid. Again a solution can be found in nature. A dragonfly's wing is constructed from a series of very light weight skins with a hexagonal cell structures. These are connected to the body by a series of primary sub dividing elements. When the cell system meets the primary system it simply connects in a perpendicular relationship. Exactly the same can be seen with soap bubbles. If they are approximately the same size they take on a hexagonal geometry. If you sub divide them then again they join the subdivision at right angles. Nature does not have formalist architectural hang ups!

THE GEOMETRY OF THE I C O S A H E D R O N A P P L I E D AT E D E N


Like earlier predecessors our geodesic domes are based on the geometry of the icosahedron, an element with 12 corners and 20 surfaces. Circles drawn through two adjacent comers of the icosahedron result in 'Great Circles', because all corners of the icosahedron are positioned on the surface of a sphere. These Great Circles intersect in such a way that 5 warped triangular and selfsimilar surfaces are generated around each comer of the icosahedron. Because each comer of the icosahedron is surrounded by 5 triangular zones the element directly on the comer is a pentagon. The subdivision of the sphere's surface into triangular zones with equal side lengths is the key to finding a construction method that applies selfsimilar sticks and varying nodes to form the structural 'net'. The triangular base zones described above can be further subdivided into triangular elements at a selected frequency. The body formed with planar triangles approaches the spherical shape more and more the smaller the subdividing triangular elements. At Eden we have omitted triangles in such a way that the zones between the pentagons are filled with hexagons very similar to the surface of a football. Unfortunately these hexagons are not planar in a geodesic sphere based on an icosahedron. This was obviously a problem keeping in mind that the hexagons are the basis of our cladding panels. Developing our ideas futher with Mero GmbH a solution arose. This was to apply a recently developed theory by the Russian scientist Pavlov who managed to work out a geodesic sphere with hexagons that are planar and the geometry still based on the icosahedron.

Dragonfly wings

Soap bubbles

As we developed the geometry the computer was invaluable not just as a number cruncher but as a way of exploring the spatial forms that the combination of differing diameter interconnecting spheres and the pit topography created. We undertook a series of computer studies that culminated with a series of 'fly through' animations.

T H E HEX TRI HEX GRID


The resulting net of hexagons alone could have formed the envelope for the Eden biomes. However the stick diameters for the individual elements would have been around 500mm. It was felt that by introducing a second, inner layer, of structure the member sizes could be reduced substantially leading to a far more economic structure with a more light weight appearance. The inner
Still from 'fly-through' animation in the Humid Tropics Biome

81

ENVELOPE
The skin of the biomes utilises Ethyltetraflouroethylene foil, ETFE. It was selected as its performance was far better than glass in both horticultural and energy terms. It allows a far greater range of daylight to pass through in particular the Ultra Violet part of the light spectrum. Its light transmittance quality is further enhanced by its long span characteristics: the largest pillows at Eden span eleven metres without any secondary structural system. Consequently there are very few light-blocking structural members.

Single-chord & bowstring geodesic

Hexagon with ETFE cladding

Efficacy (relatival

Transmission

Our original design was for a single chord hexagonal grid geodesic form with a secondary bow string cable stay support. The geometry and pillow sizes were informed through discussions with the two principal foil supplying companies. A whole series of solutions were then developed and considered including timber/ glue-lam and aluminium for the geodesic structure. The eventual winning contractor was Mero GmbH who offered a combined structure and envelope package. As the two are intrinsically linked this was a significant advantage. Their experience with this type of structure brought considerable benefits to production and assembly. Their preferred material was steel and by adopting a double chord system the tubes could be kept below 200mm despite the fact that the span was 100 metres. Mero's sub contractor for the envelope was another German company, Foiltec GmbH, and over a period of nine months we have developed and tested the pillow solutions. Initially the size for the biggest pillow was based on intuition - a feeling for what the largest achievable span was likely to be. There were incentives on all sides to design as large a hexagon as possible: the larger the hexagons were, the lighter the steel would be and thus the cheaper the overall building. bigger hexagons also meant more light and, because the frames are relatively less insulative than the pillows, better thermal performance.

L ultraviolet tuv> -L visible light (VIS) 0.00001 + ^ , , i ~ ! 1i 1 i iiii. 250 300 3B0400 500 600 Wavelength

1i

rii"-^H O TOO nm 780

i.

Light transmission of ETFE/glass compared (information from Dyneon GmbH)

The pillows are up to two metres deep and are formed from three layers of ETFE foil. The two air cavities are pressure equalised by means of a small connecting aperture but in terms of thermal transmittance they are effectively separate. The complex geometry of the pillows - hexagonal on plan and double-segment-shaped in section - makes U-value calculation by conventional methods impossible. Much of the cavity space is large enough for significant convection currents to be set up. A combination of theoretical analysis (computational fluid dynamics / finite element analysis) and empirical testing (by means of hot-box experiment) determined that the Uvalue was approximately 2.7W/m2K. Therefore, in spite of the material being 200 microns thick or less for each layer, it performs better thermally than double glazing. As with glass it would be possible to apply low emissivity coatings to one or more of the ETFE layers to achieve even better performance. Our goal was also to create a solution that embodied Eden's environmental ethos. The embodied energy is substantially better than a glass solution. In material terms it uses less than 1 % of the volume of material that would have been used in a double glazed solution. This, coupled with a proportionate reduction in supporting framework, again substantially reduces transportation impact and costs. The material is very light to erect so again there is a reduction of site equipment. Most importantly it can be recycled. This desire to optimise on the properties of foil, i.e. large span pillows, has meant a great deal of research and testing.

A hexagon side length of 5.5m, equating to a diameter of 11 m, was set as the target. This resulted in pillow sizes of over 75 square metres. The first stage was to establish the wind loads that needed to be accommodated. A scale model of the biomes and a few square kilometres of the surrounding terrain were built and tested in a wind tunnel. This was followed by full-scale tests to establish the behaviour of the ETFE under dynamic biaxial loading at varying temperatures, the strength of the welds (ETFE is manufactured in 1200mm wide strips) and the strength of the connection to the extruded aluminium frames. Parallel to the empirical testing, computational non-linear analysis was carried out based on the material's physical characteristics (established by Instrom testing). By comparing the results of the empirical and theoretical analysis a clear picture was built up of how the overall system behaved under the design loads.

82

plants' environmental requirements for each biome. Within this context we defined the design parameters with Ove Arup. We did not want to have any sun-shading devices on the envelope and we aimed for the minimum of any mechanical devices or plant within the biomes. In principle this was achieved by using the form of the biomes for warm air reservoirs; the curved biome form would assist the natural convection currents created with air jets. 'Traditional' calculation methods were used early in the design process to establish the number of air-handling units needed to supply the large volumes of heated air required. Similarly the strategy for ventilating the Biome enclosures was established, involving a combination of opening glass louvres at low level, with hinged panels at the top of each dome to exhaust hot air. As the design progressed, Ove Arup & Partners undertook a number of more detailed studies to refine the environmental systems. In particular a Dynamic Thermal Model was used to calculate the thermal conditions within the Biomes for typical days in selected months. This study then provided the data to allow a detailed study of air movement within the Biomes using Computational Fluid Dynamics (CFD). CFD analysis was carried out for the Humid Tropics and Warm Temperate Biomes during Winter and Summer conditions. These studies produced predictions of environmental conditions in terms of air temperatures and velocities; the results broadly followed the conclusions of the initial calculations. Critically, however, the CFD model allowed us to assess the effects of removing, or relocating, some of the air-handling units. This proved important as the terrain around the Biomes makes access to some areas difficult both for initial installation and subsequent maintenance of the mechanical plant. Consequently we were able to fine-tune the mechanical systems, omitting a number of air-handling units and relocating others to areas of easier terrain, whilst maintaining the required environmental conditions within the Biomes. One interesting result of the CFD study relates to the effect of air movement on the plants within the Biomes. In normal greenhouse conditions plants tend to grow with relatively weak stems due to the lack of wind. We have discovered that air speeds within the Eden Biomes will tend to strengthen the plants; ripples of air movement recreating natural conditions to produce specimens as close as possible to their counterparts in nature.
Test pillow in Bremen, Germany

Testing an ETFE foil pillow in Bremen, Germany

It emerged that the 3-layer cushions were not robust enough to construct hexagons with a 5.5m edge length! The first solution that was attempted was to install reinforcing cables over the pillow to reduce the effective span of the ETFE. While this was workable in principle there were a number of disadvantages: it was a relatively expensive solution, and was likely to cause chafing between the steel cables and the ETFE. Significantly increasing the thickness of any individual layer was not feasible because it resulted in embrittlement: to work effectively the ETFE needs to maintain elasticity. In one of many round-table discussions a solution emerged to use a double layer of foil for the externalsurface (the part subjected to the most onerous suction loads). The two layers would work together to share the load. A second series of tests was then carried out which established that the proposal would work with a margin of comfort.

THE GARDEN ENVIRONMENTAL SYSTEMS


Greenhouse design has frequently suffered from the environmental control systems, both from a successful operational aspect and from the visual impact of the system itself. Very early on, with Peter Thoday, the horticultural consultant for Eden, we established the Throughout the project we have continued the same philosophy towards sustainability, challenging the way we have previously considered designing buildings. The visitor centre is a prime example. Again it is built into the topography of the pit with part of the building sunk into the terrain with a grass roof. To keep the embodied

83

energy of the building to a minimum we have used materials from the site in the form of gabion walls using site rocks. Soil from the pit has been used to construct rammed earth walls and much of the cladding utilises light and economic cedar wood shingles.

The perimeter glazed wall achieves solar control by the use of external cedar louvre screens. These allow views directly in and out of the building, whilst shielding the internal public areas from strong direct sunlight. They also lend a layering to the facade that helps to break down the boundary between internal and external spaces.The Biome Link has a turf roof that curves to align with the adjacent geodesic structures. It slopes down at each comer of the building where the surrounding ground slopes up to meet it, so that the overall effect is of a "saddle" of ground covering the Link. The roof structure is an array of steel bowstring trusses. These vary in shape to accommodate the profile of the curved roof as it transforms from the Humid Tropics arch geometry to the Warm Temperate arch geometry. This self effacing approach hopefully puts the aspirations of the Eden project before any preconceived architectural metaphors. We hope that in the same way as the technology of wrought iron brought about a revolution in the design of the glass house at the start of the nineteenth century, the technologies adopted at Eden will be part of a major step in the development of greenhouse architecture at the beginning of the twenty-first century.

The Visitor Centre

A very similar approach can be found in the "Biome Link", the connection and entrance building for the two biomes. The Biome Link responds in plan form to the same rhythm of interlocking spheres exhibited by the geodesic structures that it connects. Dining and exhibition areas are bounded by the sweeping perimeter of the front glazed wall, and separated from back-of-house facilities such as kitchens, toilets, offices and "plant holding areas" by a continuously curving double-height earth-rendered wall. Access from the external gardens into the building is via an elevated walkway, which passes over the external terrace then penetrates the glazing where it splits, taking visitors at high level either to the Humid Tropics or the Warm Temperate biome.

Model of the Humid Tropics Biome

jtxjjSw

We have been inspired by the elegance and economy of design of airships, more than any other type of constructed form they have to explore maximum efficiencies. Perhaps this can be best summarised by the weight of the biomes. The humid tropics biome weighs approximately 450,000kg; this is actually less than the weight of the air that the envelope encloses. Fortunately it is firmly bolted down to the ground!

The Biome Link

Luftschiffbau Zeppelin airship 129 Hindenburg, 1936

84

Dome A, Humid Tropics Biome, February 2000

85

PLANTING ENVIRONMENT FOR THE EDEN PROJECT


Peter Thoday

INTRODUCTION
This paper was prepared in the first days of 2000. The planting of the Eden Project Biomes is programmed for October 2000 in time for the public opening in April 2001 from which date plant growth must continue in a satisfactory condition for many years. What follows is therefore a resume of what is known of the biological response of tropical and sub-tropical plants to greenhouse conditions in northern Europe, the client brief setting out the conditions required in Eden's structures, the anticipated physiological response of the plants to the environmental conditions within the structures and some reference to the aesthetic and presentational form of the plantings.

RELATIVE HUMIDITY AND TEMPERATURE


The 20th Century saw massive studies on the relationship between a plant species' biochemical processes, including those that control reproduction, and abiotic environmental factors such as day length, diurnal and seasonal temperature fluctuation and water stress. This understanding of plant response to environment had a significant influence on Eden's strategy for the cultivation of its Living Plant Collection and formed the basis of the client brief regarding environmental conditions within the Biomes. Eden's strategy is to present as accurate an image of the interaction between plants and man as possible. For that reason the surface area and height of each Biome is extremely important, so it was decided to keep the number of structures, and hence the separate encapsulated micro-climates, to a minimum. The target number of biomes is four, these are the Humid Tropics, Semi-Arid Tropics, Sub-Tropics and Warm Temperate to which one should of course add the outdoor climate of Cornwall. Phase one achieves two Biomes; the Humid Tropics and the Warm Temperate. In common with most botanic gardens Eden wishes to grow a very wide range of species in each structure. This eclectic approach clearly mitigates against the possibility of being species specific in defining and prescribing environmental conditions. The reasoning behind the client brief is therefore to achieve a tolerable set of conditions for a wide range of species whose natural habitat falls within the two broad climate zones "humid tropics" and "warm temperate". The great advances in environmental engineering made possible through the development of computer controlled systems should provide the managers with the ability to adjust and vary conditions within a prescribed band which should, in turn, lead to greatly improved growing conditions. The clients are not searching for constant conditions either on a seasonal or diurnal time scale. Unlike growers managing commercial glasshouses dedicated to a single crop we

HISTORICAL PERSPECTIVE
Tropical and sub-tropical plants (frost tender species) have been cultivated in northern Europe in glazed structures for more than 300 years. By the middle of the 19th Century glasshouses of a size capable of accommodating shrubs and small trees to maturity were being constructed in most National Botanic Gardens. By the close of the century the following points pertinent to this paper had emerged: 1. 2. 3. That it is possible to keep such vegetation alive. That only a small percentage of the subjects flowered and even fewer fruited. That growth was atypical when compared with that collected from either wild or field grown specimens.

As biological science progressed in the early years of the twenthy century it became clear that in general the conditions limiting plant performance were atmospheric, abiotic factors rather than edaphic. They were light, temperature, and humidity. These findings remain true today.

86

look forward to being able to exploit small differences in temperature and humidity between locations within the biomes. The greatest advantage we anticipate coming from Eden's Biome design is in light transmission. This is achieved through a combination of the translucency of the cladding material and the reduction in solid steel work superstructure.

TEMPERATURE CONTROL
Heating and ventilating are the most challenging environmental factors relating directly to the Biome design. Ove Arups, the Project's environmental engineers, have carried out exhaustive computer studies that have provided models of the expected air movements when either the blown air heating system or the ventilation system are in operation. Clearly it is too early to comment on those predictions. However it is reasonable to anticipate that air movements and therefore temperature conditions are likely to vary between those experienced in the empty structure compared with those recorded in future years when a significant percentage of the volume of the space is "filled" with biomass. A condition that will take some 10 years to achieve.

One of the more difficult aspects of the project has been to quantify the "optimum but unnatural" conditions that have proved satisfactory in the cultivation of tropical plants in traditional glasshouses. Understandably, some aspects of this vague quantification seem to baffle engineering colleagues. One is the concept of biological tolerance allowing conditions to be "satisfactory" for a considerable distance from a theoretical optimum point, a tolerance that decreases with time. Another is that a change in one environmental parameter can cause a shift in the optimum condition in other environmental factors, for example the optimum temperature is influenced by light availability, i.e. less light, lower temperature. This linkage and the fact that plants have considerable short term tolerance, negates the need to build in massive and expensive capacity to hold optimum conditions during exceptional weather. A third and most damming of all to the client's professional image is that optimum conditions for some 80% of the plant species we intend to grow in the Biomes at Eden are simply not known! The clients best endeavours to express these considerations as environmental guide lines are as follows:

TEMPERATURE REGIMES HUMID TROPICS BIOME Dormant season Optimum day Optimum night Maximum day if sustained for several hours 0 20 18 23 0 28 25 33 Growing season Optimum day Optimum night Maximum day if sustained for several hours Minimum day/night if sustained for several hours Irreversible damage to many species

Minimum day/night if sustained for several hours

15

19

Irreversible damage to many species

10

12

TEMPERATURE REGIMES WARM TEMPERATURE BIOME Dormant season Optimum day Optimum night Maximum day if sustained for several hours Minimum day/night if sustained for several hours Irreversible damage to many species 0 16 12 20 5 2 0 26 19 34 13 4 Growing season Optimum day Optimum night Maximum day if sustained for several hours Minimum day/night if sustained for several hours Irreversible damage to many species

87

H U M I D I T Y A N D AIR M O V E M E N T
The role of atmospheric humidity in plant performance is complex. Its direct influence is on the rate of transpiration of water from the leaf stomata. Any resultant internal water stress effects many physiological processes. Such humidity also plays an important role in the spread of some plant diseases. Typically the horticulturist tries to match the humidity of the greenhouse air with that found in the plant's natural habitat and to which it is assumed to be adapted. In common with other people growing plants from the humid tropics we aim to run the humid tropics Biome at a minimum of 80% Rh. In contrast the warm temperate Biome will operate more successfully at between 40 and 65% dependent on season. If, as anticipated, these figures are achieved for a high percentage of the time, conditions will greatly assist the cultivation of the collection. Air movement not only promotes equable conditions but also agitates foliage. Even minor movements may trigger desirable effects. It is planned to utilise the anticipated air movements associated with both the heating and ventilation systems to influence plant morphology and growth form.

The translucency of the cladding when used as designed as three sheets totalling 0.5mm and has a similar value to 4 mm of greenhouse quality glass. A great benefit will result from the massive reduction in the structure of the building when compared with other large show greenhouses [conservatories] in European Botanic Gardens. The quality of the light admitted through the cladding fits well with the spectrum used in photosynthesis. Currently there is a great deal of interest among plant physiologists in the role of specific spectra in influencing other growth processes. However these remain to be fully understood and the relative benefits / disbenefits of glass or any other cladding compared with open air sites cannot be evaluated at present.

SPACE
It is surely a truism that widespan structures provide optimum room and flexibility of design even though the client's choice of the sloping margins of an old clay pit mitigate against this! A large surface area is of course a common feature of commercial glasshouses many of whose repeated modules cover more than One hectare. In the case of the Eden biomes height is the most striking, and from a horticultural point of view, novel feature. With up to 50m of available "head room" we will have the opportunity to grow a few specimens of selected species of tree to maturity. These dominant, emergent, light demanding species should flower and fruit, in many cases for the first time in greenhouse cultivation. The amount of available space will allow them to perform their natural role in the forest ecosystem by providing the over-shading canopy below which other species have their ecological niches. These trees also offer the great long term benefit of being able to "host", as in nature, a population of epiphytes and climbing lianes. Ironically the absence of internal supporting columns presents a short term problem. Until the trees are well developed the support for both humidification pipe-work and display lighting must be achieved from masts. A basic tenant of the design is to deliver services "from the ground up", as pendent fittings will not be used. From the design point of view the advantages are huge and obvious. An unfettered, unobstructed space allows the positioning of plants, paths and water features wherever appropriate. Although the curvilinear nature of the structure appears to restrict options near to its sides the scale of the buildings are such that within 3m of their southern edge it is possible to accommodate a rich palate of plants. The wish to optimise light entry mitigates against any desire to plant trees along this zone.

LIGHT QUANTITY AND SPECTRAL QUALITY


The prime concern of those growing plants under protection in Northern Europe is the availability of light during the late autumn, winter and early spring. At these times light levels are frequently so low as to prevent "normal" growth forms and to inhibit some physiological processes. In glasshouses the light arriving at the crop canopy is typically some 65% of the value recorded outside the structure. The quantity of natural daylight entering the Biomes, relative to local solar radiation under prevailing climatic conditions, depends on:i. ii. iii. The position of the structures within the crater of the old china clay pit The translucency of the cladding The obstruction caused by the steel superstructure.

Light intensity will be considerably reduced in both the early and late daylight hours when the sun is below the East and West rims of the pit. This will be, of course, for longer periods during winter.

88

SUMMARY
The Eden Project recognises the challenge it has set its design team in providing satisfactory conditions for the cultivation of tender plants from a wide genetic base. Typically botanic garden "conservatories" provide far less satisfactory conditions than commercial glazed structures particularly if these have been designed for a named single crop. The staff at Eden look forward with some confidence to curating our tender plant collection in the Biomes where we anticipate that growing conditions will be considerably better than in most botanic gardens and we are confident that the structures ensure that the visitor experience will be superb. Eden staff wish to record their thanks to the design team for the interest they have shown in becoming familiar with the environmental requirements of their clients - the plants.

89

CIVIL AND STRUCTURAL DESIGN OF THE EDEN PROJECT


Alan C. Jones BSc(Hons) FIStructE Project Director Anthony Hunt Associates

ABSTRACT
The paper describes the civil and structural engineering design and construction of the project from early inception in April 1995 through to the present time. Construction has now commenced with the substructure substantially complete and the envelope well underway. The buildings are programmed for completion in October 2000 when planting will commence for a full opening of the project in April 2001. and Partners and Anthony Hunt Associates to act as architect and civil/structural engineer respectively. The design team also included Land Use Consultants as Landscape Architects and Ove Arup and Partners providing Mechanical and Electrical design. Initially, the brief stated the project must be located in Cornwall, for a number of reasons: The mild climate is unmatched in Europe; it is an area of high unemployment which will benefit from the presence of the project; a number of sites could be identified which were relatively easy to access by road and rail and were close to centres of population. After reviewing the available sites the Bodelva Pit was chosen, located about 5km east of St Austell. This China Clay working was approaching the end of its commercial life and was due for decommissioning. Hence, the benefit of land reclamation was added to this list. Eden is seen as a considerable commercial investment in Cornwall and is supported by The Millennium Commission; English Partnerships; The European Development Fund and many local councils and other organisations.

INTRODUCTION
The Eden Project is one of the UK's top landmark Millennium Projects created to tell the fascinating story of man's relationship with plants. It is a non-profit making, charitable, scientific, organisation with a commitment to communicate with the public through entertainment, education and involvement. At the outset it was acknowledged that such a creation would require an architectural input of the highest quality to provide the dramatic and inspiring setting. The project principals, Tim Smit and Jonathan Ball approached Nicholas Grimshaw

Humid Tropics Biome


Plants from Amazonia, West Africa, Malaysia and Oceania.

Warm Temperature Biome


Plants from California, Southern Africa and the Mediterranean.

Landscaped Grounds
The story of plants from our own temperature climate.

Visitor Centre
Gallery, restaurant and shops.

Lake Main car park


Amphitheatre (2300 seat)

Restaurant at the Centre of the World.

Plan layout and main features of the project

90

THE SITE
The Bodelva pit lies within the eastern end of the St Austell granite in a zone of biotite granite. The granite was emplaced with the surrounding shale sedimentary rocks during the late carboniferous/early Permian geological periods some 280 million years ago. The granite has become variably kaolinised. Kaolinisation is the process of chemical alteration of the feldspar component of the granite rock mass. The other components, quartz, tourmaline and mica remain unaltered. The feldspars are leached of potassium and are broken down to a plastic clay known as kaolin or china clay. The rock mass encountered on site has been classified into four different grades. Grade 1 which is fresh, unaltered rock through to Grade IV, a highly kaolonised plastic material which can be moulded by hand. An option to purchase the pit was negotiated with the owners at an early stage. However, whilst the scheme design progressed and funding was secured, the owners continued to work the pit "winning" china clay from any areas of their choosing. The pit covers an area of about 22 hectares and varies in depth from 30 to 70 metres. The dramatic landscape, which made the pit so attractive to all who set eyes on it, presented the greatest challenge to the engineering team. The brief was to preserve as much of the wild, rustic appearance as possible while creating a safe stable environment for visitors. A digital ground model, based on arial survey data was used to sculpt the pit. The model, created with Microstation and In-roads software, allowed a detailed evaluation of cut and fill operations to take place. As part of the environmental statement of the project, an undertaking was given that the exporting or importing of fill material would be kept to a minimum. This obviously also had important commercial advantages. The landscape architect developed his scheme and provided outlines to be fed into the model. After several iterations a balance was obtained. Approximately 800,000 cubic metres of material was moved to create the final

profiles. No material was exported. The only material imported was organic matter to construct the topsoil and planting medium.

"Winning" the china clay using high pressure water monitors.

Water management was also important to the long term stability of the site. The base of the pit is around 30 metres below the natural water table in the area. During quarrying operations the water was used as an essential part of the clay extraction process. High pressure hoses are directed onto the quarry faces releasing the kaolin and creating a suspension, which is pumped onto settling tanks remote from the site. As the Eden project will not continue to extract water in this way, a new storage and pumped drainage system was constructed. Run off from the car parking and landscaping is controlled by incorporating swales alongside all the flat areas. These store water in the short term so that the pumped outlet into the Bodelva brook to the south of the site, can be strictly regulated to prevent flooding downstream and comply with The Environment Agency requirements.

POSITIONING THE BUILDINGS


The principle buildings are the greenhouses, or biomes. They are designed to provide two climate zones, one modelled on the Humid Tropics (HTB) and the other a Warm Temperate (WTB) or Mediterranean climate. Plans for a third zone providing a desert climate were postponed until the next phase of the project due to funding constraints. All of the zones require high levels of natural daylight but the Humid Tropics is the most critical. Following a sun path study, undertaken by the architects, the HTB was positioned against the north face of the quarry with the WTB curving away to the south east. In this way the environmental control system benefited from maximum sunlight and the thermal mass of the rock face. The position of the buildings was therefore dictated by criteria outside the civil engineers control and led to considerable complexity in the foundation design. The HTB is up to 110 metres wide, 55 metres high and 240 metres long. Within its length it rests on all four designated grades of granite, passing from unaltered rock, across the

The Bodelva pit, November 1998. Before construction commenced

91

kaolanised clay materials and back to unaltered rock. At the base of the pit it sits on up to 12 metres of fill material. Whilst, slightly smaller in form, the WTB (150 x 35 x 65) has a similar formation. The Visitors Centre, positioned on the South West face, to give views over to the biomes and external landscaping, sits on a new plateau cut into the side of the pit and extended with fill material reclaimed from the car parking areas.

spanning from the base of the pit onto the cliff face. Each truss therefore had a unique profile and associated span. It was critical from both commercial and aesthetic criteria that

PROCUREMENT PROCESS
The whole project was driven by the many funding initiatives required to raise the 75 million total cost. The initial programmes were amended time and time again by revised applications for grants and sponsorship. The tender was issued using a preliminary design in late 1996 and the constructors were invited to assist with funding. The successful tenderer, a joint venture between Sir Robert McAlpine and Alfred McAlpine was given preferred contractor status in early 1997. The form of contract used was the Engineering and Construction Contract Option C. target contract with activity schedule, amended to make it a Guaranteed Maximum Price (GMP) contract, with profit sharing. The latter was split 50/50 between client and constructor and is considered to be extremely beneficial to the client. A protracted negotiation period followed during which the design was modified on numerous occasions to bring the cost plan/GMP into line with the target figures. The McAlpine JV lead the tender process for the envelope and mechanical and electrical services and negotiated the final contracts. Both of these major packages were let as contractor designed works with detailed design being the responsibility of the works package contractor.

Initial scheme used to secure early funding

the pit shape was maintained whenever possible. Wholesale reshaping was not an option. Very little repetition or rationalisation of trusses could be achieved. With the constantly changing topography, as iruning continued, it was inevitable that several redesigns would be necessary before the final geometry could be confirmed. At this point, the design team proposed a radical change to the basic form of the Biomes. A series of intersecting domes, of varying diameter, were developed. The concept being that once the size and relative position of the domes had been determined the shape of the pit became of a secondary concern. The structural form of the domes could be coirfirmed and the intersection line between superstructure and ground determined the position of foundations and

Rendered images of the digital ground models showing the original survey data on the left and remodelled site on the right

STRUCTURAL FORM
During the early scheme development period it became apparent that the exact profile of the pit would not be known until construction commenced. Initial schemes for the biomes used curved, arched trusses at regular intervals

extent of cladding. This enabled the team to proceed with design development of the biome envelope before the final survey of the pit was available. One of the principle criteria in the clients brief was to maintain the transparency of the envelope at a maximum. In order to achieve this, the selected cladding material had to

provide very high levels of light transmission and the structure elements had to be kept to minimum size and number.

successful contractor, Mero Gmbh, offered a combined package supplying both frame and cladding. Their proposal incorporated a space truss system, developed over many years. The two layer structure used the geometry given in the tender for the outer layer with a combination of hexagons and triangular elements forming a semi-braced inner layer. The combined system is referred to as a hex-tri-hex arrangement. The outer members are 193 mm diameter circular hollow section with semi-fixity developed at the nodes whilst the inner members are around 114 mm diameter circular hollow sections with pin ended connections from the Mero system. The proposed alternative offered considerable reductions in the weight of steel although fabrication complexity and the number of nodes increased considerably. The number of nodes in the system has a significant effect on cost, as does the size of the cladding panels. The larger the panels, the fewer number of nodes and generally, the cheaper the cladding. Hence the objective was to develop as large a cladding panels as possible.

Dopuble layer space frame referred to as the "Hex-tri-hex" arrangement

After a prolonged study of various geometrical arrangements for a spherical surface, a geodesic arrangement was selected. By adopting the hexagon form derived by Buckminster Fuller an even distribution of structural members could be

THE CLADDING SYSTEM


The performance criteria that developed for the cladding system were extensive. The size of panel was very important, glass would be: limited to panels no greater than 4 metre x 2 metres and double glazed, would weigh up to 75 kg/m2. The installation, cleaning and replacement of large glass units requires careful planning and large crane capacities to reach over the buildings. The system chosen to clad the biomes is a pneumatic structure of "cushions". Each cushion is contained within

Scale model undergoing wind tunnel tests.

achieved. By varying the frequency of sub-divisions in each dome, cladding panel sizes can be adjusted to give optimum form and light levels. The original scheme utilised a single layer, unbraced three dimensional space frame structure with 500mm diameter circular hollow sections. The envelope was tendered as two packages, steel and cladding. The

93

one module of the structure in the form of either a hexagon, pentagon or triangle. On the largest domes hexagonal cushions up to 10.9 metre across points are used. The panels are formed from multiple layers of Ethyltetraflouroethylene (ETFE) foil. The foil is extremely thin, each layer being between 50 and 200 (im thick, giving very high levels of light transmission in both the visible (94-97%) and ultra violet range (83-88%). The cushions are held in extruded aluminium perimeter frames using a "luff' grove and bolt rope type detail, known as a keder, derived from sailing and fabric structure technology. The frames are in turn bolted to brackets on the tubular steel structure at regular intervals. Even with such large panels the whole cladding system only weighs around 15 kg/m2 - a considerable weight saving on the equivalent glass envelope. Thermal insulation valves are as good as double glazing and in some instances better than triple glazing when used horizontally.

The ETFE is a modified copolymer which is extruded into a thin film. This means the surface is extremely smooth and when coupled within the anti-adhesive properties of the material, gives a self cleaning surface. Dirt such as bird droppings is washed off by rain and the requirement for regular cleaning is minimal. The material is unaffected by UV light, atmospheric pollution or weathering and extensive testing has been shown to give an anticipated life expectancy in excess of 40 years. At Burgers Zoo in Arnhem, Holland, buildings with foil roofs have been in use as plant houses for over 20 years. The foil panels themselves only weight up to 50kg making replacement a much easier operation than glass. It is also possible to effect short term repair insitu using adhesive ETFE tape. The system is considered to be environmentally friendly. Although, the raw ingredients include natural resources such as gas, oil and other minerals, the quantities used are relatively small per square metre of envelope. The

94

manufacturing process does not involve significant use of additives (unlike PVC for instance) and the foil is recyclable. The inflation units consume energy to maintain the air pressure within the system but the increased light transmission compensates for this in reduced artificial lighting requirements.

transmission and to minimise cost: large cushions mean fewer connections in the steelwork and reduced length of aluminium framing. The subcontractor has undertaken the detailed design of the cladding system and the development of the systems took many months. Limited guidance on economical cushion sizes indicated that side lengths up to 5.5metres were possible but would necessitate reinforcement in the form of a cable net, above and below the cushions. In order to avoid this added expense, the geometry of the domes was scaled to suit their size - smaller domes incorporating smaller cushions. As a result only the largest cushions, on dome B in the HTB and dome G in the WTB required reinforcing. Table 1 gives the final geometrical arrangements for the biomes including indicative side lengths for cushion elements. The line of intersection between domes posed a particular problem. It was not possible to align the nodes on either side and this was exacerbated as the geometry of each dome had been scaled to give suitable cushion dimensions. Tubular lattice arches were introduced to accommodated this and pick up individual node points. The arches are fabricated in segments from curved tubes and site welded together. When the pit was handed over for construction a full topographical survey was performed to confirm the shape of the areas most recently worked. The digital ground model was integrated with the superstructure model to give an intersection line which formed the setting out for the foundations. The environmental control systems are based on a combination of natural ventilation and blowers. Consequently, the hexagons at the apex of each dome are subdivided into triangles to form opening vents. Air is introduced at the base of the domes via louvered panels in full and part hexagons. These panels also allow the passage of air ducting from the air handling units through the envelope.

ENVIRONMENTAL LOADS
The combination of a lightweight steel frame and cladding system (with a combined weight of around 40 kg/m2 of surface area) makes the effect of environmental loads on the structure all the more critical. To achieve the most efficient solution possible, snow and wind loads were assessed in detail in accordance with the current British Standards. The consequences of drifting snow accumulating between cushions, or in the valleys between domes were evaluated. Wind loads were impossible to accurately assess from the standards because of the unique topography of the site and the complex geometrical shape. A detailed study was conducted using scale models of the development in the wind tunnel at British Maritime Technology Ltd. This demonstrated that the profile of the pit shelters the buildings from the extremes of wind. As the pit is over 60 metres deep and the highest biome is only 50 metres to the apex, the whole development could be considered to be below ground level. The results of the tests supported this, giving design wind pressure well below those initially predicted.

GEOMETRY
Once the type of cladding and intensity of environmental loads had been established, the design team concentrated on deriving the optimum geometrical arrangement for the spherical structures. As previously stated, the object was to utilise the largest cushion possible in order to maximise light

Table 1 Biome Geometry

Bubble ref:

Radius (m)

Frequency side length approx. (m)

Hexagon depth (m)

Structural Cusshions

Number of (or part)

Hexagons (or part)

Pentagons

Triangular opening vents)

Principle tube diameters Top Boom Bottom Boom

Human Tropics Biome


49.972 62.465 49.972 24.986 15 15 15 9 4.391 5.488 5.4391 3.634 1.6 2 1.6 1.3 129 163 155 58 103 137 128 42 25 25 25 15 193.7 193.7 193.7 193.7 88.9/114.3 114.3 114.3 114.3 76.1 88.9 76.1 76.1

D
Total

Warm Tropics Biome


E F G H Total 18.74 28.109 37.479 28.109 9 9 9 9 2.725 4.088 5.451 4.088 1 1.5 2 1.5 67 88 96 75 326 40 60 68 47 215 2 3 3 3 11 25 25 25 25 100 193.7 193.7 193.7 193.7 60.3/76.1 76.1/88.9 88.9/114.3 60.3 60.3 60.3 60.3

96

ANALYSIS
The biomes are supported on a concrete "necklace" which follows the perimeter of the buildings, hugging the contours of the site. The insitu reinforced concrete strip is generally 2.0 metres wide by 1.4 and 1.0 metres deep in the HTB and WTB respectively. A three dimensional analysis model was built to investigate the interaction between the foundation and the rock mass. Equivalent spring stiffness was used to model the varying soil

cushions in the event of a long term power failure. Whilst back-up pumps and power supplies are provided for the inflation system, air can be diverted From one pump unit to the next and it would take many hours for the cushions to deflate, it was still considered necessary to assess this condition. Calculations indicate that the steel frame is capable of supporting up to six flooded cushions without any failure predicted. Further investigations using non-linear techniques were undertaken to demonstrate that the risk of progressive

Humid tropics biome. Structural anlysis deflection diagram for snow loading

properties of the four grades of granite. In the case of the HTB the building passed from Grade I & II granite at the' West End, through a band of Grade III and IV and back onto Grade I & II granite at the East End. Predicted settlements were limited to 25 mm overall with a maximum differential between adjacent node points of 15mm. These values were incorporated into the detailed design of the superstucture. At the locations where the foundations passed from the hard granite onto the softer clays, articulation joints were introduced to prevent abrupt changes in level During the early foundations design stage the final loads from the steel superstructure were not available from the frame contractor. Anthony Hunt Associates constructed a full three dimensional model of the biomes and performed a preliminary analysis to estimate the foundation loads. These loads were used to design the foundations. Once the final loads were available a further check was undertaken to ensure the foundation design was acceptable. The model created for the superstructure was then used to perform an independent category III check on the sub-contractors design. The final design of the biome structure incorporated a full linear elastic analysis of all environmental loads and combinations of loads. It was also necessary to check exceptional loads such as rainwater ponding in deflated

collapse in this condition is acceptable. The asymmetric geometry of the HTB produces a significant imbalance of loads within the structure. The space truss is subjected to out of plane bending moments which induce large deflections up 200mm, these are however well within "normal" limits when compared with the clear spans of up to 100 metres.

Cushion under test at Foiltec Gmbh, Bremen, Germany

97

increasing the patterning of the profile to give a rise of 15% of the span under full load conditions. Cable net reinforcement on the inside of the cushions is still required adjacent to the valleys between domes to support the high loads generated by drifting snow. Under normal operating conditions the cushions will be inflated to a nominal pressure of 250 pascals above atmospheric. In the event of heavy snow the pressure will be increased to 400 pascals to prevent deflation under sustained load.

SITE W O R K
Work commenced on site in November 1998 and the first operation was the construction of structural embankments to support the biomes. To create the new ground profile, up to 15 metres of fill was placed in the base of the pit, in an operation described by the client as "reversing the mining process". The principle reason for this was to achieve acceptable gradients on paths and access roads down into the pit and to provide level areas for planting within the landscape. The lower foundations of both biomes rest on top of the fill. Without special measures the settlement under load would have been far in excess of the limits set for the envelope design. There

Cushion mock-up at Foiltec Gmbh

CLADDING DESIGN
The design of the ETFE cushions was initially based on theoretical analysis using an iterative solution. It was accepted that the size of the panels was greater than any system previously built by the supplier so a series of physical tests and mock-ups were included in the

Site progress in January 2000

contract. As a result of these tests, the degree of patterning used to form the cushion profile was increased and the cable net reinforcement was omitted in favour of a second top layer of foil. Under negative wind pressures the principal loads are applied to the top layer of foil and the load capacity can be increased by either providing thicker material or a greater rise in the profile. The former was not recommended as the foil was already 200 um thick and above this limit it can become brittle and difficult to weld. As a general rule the cushion profile is based on a maximum of 10% of the span and above this, the lateral stability of the cushion under transverse loads is unpredictable. Hence the adoption of two separate layers of foil. On the inner layer, which experiences principally snow and positive wind pressures but no transverse loads, the load capacity was achieved by

were two separate effects to address: the undisturbed ground below the fill had experienced pressure relief from the removal of the overburden during mining, allowing it to absorb ground water; although thoroughly compacted in layers the fill held water and would take many months to consolidate if left untreated. All fill material was selected from "as dug" on site. A site laboratory was established to constantly monitor the quality of the fill and grade it generally into two forms: acceptable fill for the embankments and roads and general fill for other areas of soft landscaping. Vertical band drains were installed in the virgin ground, to a depth of 15 metres before any fill was placed. The horizontal spacing was calculated to achieve an acceptable level of

98

Dome A nearing completion in January 2000

consolidation within 60 days. As the fill was placed and the ground surcharged, the water was forced up the band drains, into drainage layers incorporated into the base of the fill, before being pumped off site. Horizontal drainage layers were also laid in the fill material making up the structural embankments. A system of ground anchors was installed, through the foundation strip into the fill, so that the foundation could be preloaded before erection of the envelope commenced. This removed any short term settlement within the embankments. Consolidation of the fill was estimated to take up to 30 days. Inclinometers were installed in the embankment to monitor the settlements and determine when it was safe to proceed with foundation construction. In actual fact the consolidation of both the naturally occurring material and the fill took far less time than expected, allowing construction to proceed unhindered. Once the embankments had been formed and the consolidation process had commenced, work began on the slope stabilisation and reprofiling of the pit walls. All slopes in the Grade III & IV granite up to 50 were achieved by battering the profile and spray applying a carefully selected mix of grass, shrub seed and fertiliser in an hydrated adhesive gel, to bind the surface. This planting established itself very quickly, taking only days to green up the slopes and becoming a firm carpet of grass in a few weeks. Using a similar technique slopes up to 35 could be achieved in the fill material. In many areas inclinations much greater than this were required, approaching the vertical in some cases. In the Grade I & II granite some blasting was required and due to discontinuities in the rock mass (including tourmaline veins) rock anchors were required to prevent topping of large segments and create a safe rock face. In the Grade III and IV granite, the faces were also reinforced with rock anchors, which were then used to restrain a sprayed concrete facing. As this produces an unnatural looking finish which is difficult to plant, the landscape architects attempted to minimise its use wherever possible.

On completion of the slope stabilisation and when settlement monitoring indicated that the consolidation of the fill had effectively ceased, the construction of the foundations necklace began. The complex geometry made conventional drawing methods insufficient as a form of communication with site. A series of three dimensional co-ordinates were provided, giving alignments similar to those used in road construction. These were fed directly into electronic distance measuring equipment and set out point by point. Despite the horrendous access problems the foundations were completed on programme, in 12 weeks. This was followed by the assembly of a vast bird cage scaffold used as temporary support for the erection of the biome frames. Beginning with dome A of the HTB in the west and dome H of the WTB in the east, the space frame was assembled from a kit of individually labelled parts. The primary elements were shipped to site prefinished by hot dip galvanising. Nodes were pre-finished with a zinc rich paint system. Elements were lifted into position by mobile crane initially and on the larger dome in the HTB using a tower crane on piled foundations.

Aluminium cladding frames and corner units ready to receive cushions.

The aluminium cladding frames are bolted to the top boom elements on the ground before erection, leaving only the comer units to be installed at high level. As each dome and sufficient areas of the adjacent dome to ensure stability, are completed, the bird cage will be removed to allow installation of the pneumatic cushions. Cushions are installed using rope access techniques, working from the perimeter up to the apex. Each cushion is flaked (loosely folded) then stowed into a PVC bag before being winched to its location on the frame. Teams of rope access technicians fit the cushions within the aluminium perimeter frames by slipping lengths of small aluminium extrusions on to the keders and clipping them into the cladding frames. Final weather seals are effected with comer node and extrusion capping pieces, secured by self-drill/self-tapping screws. Once in place the cushions must be inflated immediately to ensure their structural integrity in maintained. Any flapping in windy conditions will inevitably lead to damage.

99

FUTURE USE A N D M A I N T E N A N C E
The buildings have been designed for a 50 year life span. The meaning of this has been debated long and hard, as in all projects! The galvanised finish to the steelwork should last at least this long without major treatment or repainting other than areas of damage. It was selected for this reason and the relatively low capital cost. The cladding system will require regular maintenance of seals etc but the basic components are expected to last longer than the stated life span although no guarantees can be given on this. Replacement of cushions will follow a similar approach to the installation and permanent man-safe systems are being provided to achieve this. Internal gantries will be hung from the steel frame below the apex of each dome to allow access to the opening roof vents at high level. The building management team will be briefed to undertake regular inspections of the cushions to identify any damage of air leakages. Any repairs that are necessary will need to be executed promptly to ensure the integrity of the envelope.

CREDITS
Client Civil & Structural Engineer Architect Landscape Architects Mechanical & Electrical Engineers Cost Consultant Project Manager Constructor The Eden Project Anthony Hunt Associates, Cirencester Nicholas Grimshaw & Partners Land Use Consultants Ove Arup & Partners Davis Langdon & Everest Davis Langdon Management McAlpine Joint Venture

100

FOIL CLIMATIC ENVELOPES

Ben Morris, Vector Special Products, UK

Fig 1 Eden Project

FOIL TECHNOLOGY
Over the last twenty years Vector Special Projects in the UK and Foiltec in Germany have invented and developed the use of a modified copolymer Ethylene Tetra Flouro Ethylene ( E T F E ) to provide a unique high performance cladding technology. The material is a Fluoro-polymer and shares many of the properties of PTFE, the latter being used for weathering stressed membrane glass fabrics. Properties shared include resistance to UV degradation, self cleansing properties, and resistance to chemical or environmental attack. Unlike PTFE however ETFE is a thermo-plastic and can thus be extruded into thin foils, of precise thickness. This means that the material properties are both known and consistent and can therefore be engineered. In addition ETFE is used monolithically, obviating many of the problems of composites, such as hydroscopic wicking and the like. We generally use the material by fabricating pneumatic cushions comprising of between 2 and 5 layers of foil. These are restrained in an aluminium extrusion, which in turn is supported by the primary structural frame. The cushions are inflated by air to give the foils structural

Fig 2 Facade of the Duale Building at the Hanover EXPO

stability and give the cladding high insulation properties. A standard three- layer cushion has a U value similar to triple glazing and this can be further enhanced by the addition of further layers of foil and/or selective surface treatments. ETFE Foil is not only more insulative than glass but is considerably more transparent having a light transparency of approx. 94-97% of total light. Transmission across the ultraviolet range is also very good (83-88%). This is one of the reasons why the material has been used extensively in large span botanical houses across Europe to grow plants ranging from communal garden grass (important for sports facilities) to tropical rainforests, such as the Eden project (Fig 1). Whilst the base material is very transparent, ETFE Foil can be treated in a number of different ways to manipulate its transparency and radiation transmission characteristics. The Foil can be over printed with a variety of surfaces to affect transmission, printed with graphic patterns to reduce solar gain whilst retaining transparency, or can incorporate a body tint to render the foil translucent. The degree of transparency can thus be

manipulated to give the designer whatever level of solar shading is required, obviating the use of internal blinds or shading devices. Furthermore cushions can be constructed with variable shading and reflectance by differentially pressurising air chambers in cushions to cause opaque graphics on intermediate layers to alternately cover or uncover each other. A good example of this variable skin technology is the facade of the Duale Building (a German Recycling organisation) at the Hanover EXPO (Fig 2). The use of ETFE foil is environmentally friendly, both in terns of its manufacturing, and its enhancement of buildings function due to its superior insulating qualities. As a fabricator we recycle all our waste and actively import waste from other industries. Being so light the embodied energy of Foil cladding is approximately 1/100 that of a glazed structure.

102

building movements.Consequently a different engineering and detailing approach must be taken than with conventional cladding technologies, enabling the development of light flexible large span structures. Typically, conventional cladding needs to deal with the problems of isolating the cladding from structural deflections and accommodating differential movement at cladding interfaces through complex detailing. This is usually achieved by the development of a secondary structure of aluminium extrusions with complex systems of gaskets and movement joints, and frequently the liberal application of Silicon. It is usually the resolution of these interfaces that give rise to problems during the buildings life and accounts for the limited life to first major maintenance of the envelope. ETFE foil cladding can span much larger distances than glazing or other transparent cladding materials. The material can naturally deal with very large deflections in the support structure, as the material is inherently flexible. This innate material property can be further exploited by detailing the interfaces to ensure that building movement is accommodated across the whole surface of the cladding rather than concentrated at interfaces. This enables us to integrate the design of the primary structure with its cladding, obviating the need for a secondary structure.

Fig 3 The Hampshire Tennis and Health Club

CONTRAST WITH CONVENTIONAL CLADDING


In many ways ETFE foil's transparency and insulting properties make it similar to high performance glazing. There are however a number of important differences. Firstly, ETFE due to its unique chemical structure, self cleanses under the action of rain. Secondly ETFE Foil cladding is acoustically transparent, acting as an acoustic absorber in respect of room acoustics. Thirdly, and perhaps most importantly, ETFE cushions are light, highly flexible and consequently very accommodating of

Fig 4

One of eight inter-linked hex-tri-hex biomes, Eden Project

103

Fig 5 The Duale building Hanover clad in a foil 'variable skin' facade

Furthermore the inherent flexibility of ETFE enables the development of new aesthetics and engineering approaches. The development of uni-axial cable nets such as Wildscreen at Bristol and the Hampshire Tennis and Health club (Fig 3) have pioneered the integration of cladding and tensile structures whereby the structure not only supports the cladding but is itself stressed and damped by the pressurised cushions.

short term loading and our bespoke form finding non linear analysis programmes, enabled us to develop a technique for patterning load sharing foils to combat the environmental loading. The Duale building Hanover (Fig 2) is roofed by a single 30m-diameter cushion punctured by a large fan. The building is clad in a foil 'variable skin' facade (Fig 5) and houses a large exhibition on the environment. The exhibition cycle includes the generation of an internal twister, which is created within the building and exits through the roof. In order to accommodate the large loads inherent in a 30m diameter free span cushion, we developed an approach whereby the roof cushion both supports and is supported by two radial cable nets. The cushion sandwiches a shading device consisting of pneumatic tubes enabling the clear cushion roof to be blacked out in conjunction with the variable skin ETFE facade as the building runs through its exhibition cycle. The Magna project at Rotherham (Fig 6) takes a different engineering approach. Here the original brief was to construct a Zeppelin shaped enclosure supported on a compressive structural system of ribs and stringers. The use of foil enabled us to omit the support steelwork and replace it with an external tensile system of cable rings

RECENT PROJECTS
These engineering approaches have informed many of our more recent projects in particular the Eden project, Cornwall, the Duale building, Hanover, and the Magna Project Rotherham. The Eden project required the use of over 750 cushion some up to 11m diameter to clad eight inter-linked hextri-hex biomes (Fig 4). We conducted an intensive research programme to develop ways of designing the cushions in order to combat the wind loads, whilst dispensing with the need for external cable nets. This coupled with an analysis of wind spectrum patterns, further research into the Foils properties and reaction to

Fig 4 The Magna project at Rotherham

and hangers. The 50m long cushions are specially patterned and engineered to free span between longitudinal aluminium extrusions, which in turn are hung off the cable rings. The inflation of the system tensions and stabilises the net forming a structure of minimal weight. This creates a symbiotic relationship where neither structure nor cladding can exist without each other. It is this approach to structures which utilise the material properties and loads inherent in cushion structures where solutions of great lightness and elegance can be achieved. The benefits of ETFE Foil cladding has now been proved by the test of time. As confidence has grown the use of the technology has widened to encompass hospitals, government buildings, research institutions, housing and community facilities, sports facilities and leisure buildings as well as city office atria, commercial and* retail developments. This makes Foil one of the few 'innovative' technologies which has the technical superiority to service a mass market.

105

THE ENVIRONMENTAL CONSEQUENCES OF A BUILDING WITH A WIDE SPAN


Max Fordham Max Fordham & Partners

Fig 1

Model for a competition in Pottsdam designed by Straub & Vogler

This paper examines the environmental consequences of a building with a wide span. A bridge is wide span but it does not make the kind of environmental impact which concerns me. I take it that much of this symposium is concerned with structures like the Dome :Sports arenas (Sydney 2000 Stadium), cricket schools, greenhouses (the Great Glasshouse at the National Botanic Garden of Wales), cities in Alaska, garden centres (Manheim), the Albert Hall, the Crystal Palace, the Pantheon, Gothic Cathedral. Enclosure of Manhattan (Buckminster Fuller)

A wide span building is a building where the enclosing envelope is the top surface and has a wide span. This means that the top surface is light in weight and is designed to carry only the minimum load. The wide spanning surface can hardly carry the load of another storey of accommodation and so we are considering single storey buildings. I am not a structural engineer but I suspect there are reasons for wide span structures to be tall at least in places. Even though the enclosure of a wide span structure may be light, gravitational forces will be developed. These forces are vertical and a vertical component of forces generated in the structure will have

106

at least to equal the weight. Members which are inclined to the horizontal can generate vertical forces even without bending moments, so most wide span structures seem to be carried with arches and catenaries. I dare say the structural papers will expand on this theme and give some credit to Frei Otto who developed ideas for forming and finding shapes which could support wide span enclosures with bending moments only needing to be developed for perturbing forces. The wind is a big perturbation. So we get single storey, wide, high spaces with lightweight cladding.

At a smaller scale there are cyclones which are turbulent and chaotic and provide the variability in our weather. The scale here is about 1000km. Cumulo nimbus clouds are limited by the height of the atmosphere where the plumes of buoyant gas fan out when they reach the tropopause. The point about this introduction is that the air in very large spaces is mixed by turbulent convection currents. The other point to notice is that at the tropopause there is a temperature inversion (the temperature increases with height), and the atmosphere is stratified. In a large span structure, convection cells are likely to develop, with a scale characterised by the size of the space itself. When a strong temperature inversion develops at the top of the space, a stratified layer is likely to develop.

VENTILATION
The world is ventilated by natural movement of air. Inequalities of heat distribution drive the climatic air movements. The convection currents in the atmosphere are generally powerful enough to dilute the pollutants which we generate. However, pollution due to industrialisation is not always removed by the wind. The great fog in London in 1952, which killed in excess of 4000 people who had vulnerable lungs, and the photochemical smog in Los Angeles are examples where the natural air movements in an anticyclone are not strong enough to ventilate a city. When considering the ventilation of individual parts of a town, there is the concept of canyon streets where pollution is dispensed slowly and the concentration of Carbon Monoxide is a problem.

Hot gas<ts f l o w i n g through vent


o u t

1 >
V"

twfyl"
%y

ntrc.nad by r i s i n g s t r c c m ot goscs

^LAl
Formation of a layer of hot gases

Fig 2

The thermal equilibrium of any part of the world is affected by ventilation. We do not particularly think of the hottest parts of the world as being places with inadequate ventilation, but the hottest parts of the world are generally places subject to anticyclonic weather with low air movement, where the temperature builds up. A wide span structure clearly modifies the ventilation of the space it encloses. Ventilation is needed to disperse pollution and to control the thermal conditions in a space. The ventilation of a fire is a particular case where these two requirements are combined.
Fig 3

Design of roof-venting systems for single storey buildings, Fire Research Technical Paper No 10; 1964, Fig 2 page 4 Picture on Fire Research.

W a r m air discharge

Heat source forming a plume in a dome

I believe we have to understand some basic principals about fluid flow in enclosures so that we accept the possibility that ventilation can look after itself. Imagine looking at a plan of Paddington station and wondering how to provide adequate ventilation in this deep structure with combustion engines inside. In fact, the enclosure is adequately ventilated by the openings in the top and at one end. We live in a very large enclosure. It is about 12km high and 7000km radius. Convection currents drive strong air movements. The major convection cells (the trade winds) have a scale of about 10,000km, even though the atmosphere is only 12km high.

Fig 4

Multiple heat sources and plumes in a dome

107

If we need to ventilate a space so as to remove heat from 'h' the lower occupied section of a tall space ' H \ then the amount of air rising in the plumes to height 'h' must be extracted from the upper reservoir and replacement air allowed to enter at the base of the space. The entering air must not cause uncomfortable air movements in the occupied zone and it must not disrupt the stability of the hot air reservoir. Most of the envelope of a wide span structure is likely to be subject to a low pressure zone while the elevated pressure stagnation zone will be developed at low level on the windward face. It may be possible to rely on input air flowing only from the windward with discharge at the top. Often with strong winds, too much air will come from the windward, and it will flow out to leeward. If the resulting air currents are tolerable then the wind driven ventilation is a good solution. If the air currents are too strong then the air inputs on the windward side have to be throttled off so that the air enters on the leeward faces driven by the thermo syphon effect of the reservoir of hot gas. This pattern of air movement requires input around the perimeter and air outputs at the top. I was tempted to illustrate how this thinking might be applied to the Millennium Dome.

During sunny weather the temperature of the skin will build up. The temperature build up depends on the wind speed, the reflection coefficient to short wave radiation - light, and the emission coefficient to long wave radiation. It is likely to be 10C to 20C above ambient. It is important that the warm air is stratified and stable above the occupied zone of the building. Where a hot zone of air is lying above a cooler zone, there is a region with a strong vertical temperature gradient. If this region has strong air movements with the air speed changing with height, then the stratification and the turbulence will act in opposition. The Richardson number (R S Scorer 1) is defined as :dd_ dU

lz

where,

6= z = U=

temperature height horizontal velocity acceleration of gravity

If Ri > 1 then the turbulence will die down and the horizontal air flow will be restrained below the stratified layer. It is tempting to apply the ideas of stratification and displacement ventilation to the heat loads in the Dome, but the following calculation shows how quickly a plume cools down and how much input air is needed. Say there are 50,000 people in the centre of the Dome (25,000 people in the exhibition buildings, and 25,000 people clustered into 13m diameter arrays of 400 people each, producing 53kW per array, then with the addition to solar gain reaching the floor at, say, 100W/m , the temperature rise in the plume is much less than 1C.
2

Qf heatsource

r = 5a
0

Fig 5

Hypothetical flow for the Millennium Dome

r du o has uc a i s f e to re
Fire Research Paper No 7 HMSO 1983 1968 reprlr

The internal area is, say, 80m high. Strong air movement can be tolerated round the perimeter, say 0.5m/s at the centre. The heat from the exhibition buildings will be given out as driven ventilation plumes at a temperature of, say, 25C. I suggest these discharges should be ducted above the reservoir.
Fig 6

acceleration temperature temperature heatsource

due to gravity above ambient C "K kW


:

9 8" m-'sec'

density of air at source temp 1 2 kg/rrf' specific heat of a>s kJ/kg height from point source to plane defined by "y

Fire Research Paper 7

108

The air flow into the reservoir would be over 4000m /s. In fact the ventilation requirement for 50,000 people is nearer 1000m /s and the plume is likely to recirculate inside the building.
3

National Theatre warm stage - cool auditorium smoke flows towards fly-tower

lOOOmYs of ventilation picking up a heat gain of, say, 100W/m plus 50,000 people is 12,000kW and raises the temperature of the air by 10C.
2

If this stratifies in the top 30m of the Dome, the stack effect is about lOPa, giving a velocity of about 3m/s through the vents. So 300m of roof vents, which need to be controlled to prevent too much ventilation in cold and windy weather, should be supplemented with, say, 400m of vent round the perimeter.
2 2

cool stage - warm auditorium smoke flows towards auditorium

Of course, keeping the rain out has to be addressed. This crude analysis should be a precursor to more modern techniques such as salt water modelling and CFD (computational fluid dynamics). Salt water modelling gives an accurate representation of convective turbulence, but it is not able to model the momentum of incoming air, nor the thermal capacity of the bounding surfaces. The modelling of a thermal plume has to be carefully considered. A person is modelled as a 100W heat input. The model should equate to 100W emitted from, say, a 450mm diameter source with an initial plume of, say, 20 1/s at a 5C temperature rise. It should not be a 100W gls lamp, say, 100mm diameter with a plume of, say, 5 1/s at * 20C temperature rise. The difference here is represented by different flow rates of salt solution at different densities. Computational fluid dynamics cannot deal properly with turbulence. Assumptions about the amount of turbulence have to be built into the finite difference equations as dummy transfer constants. The constants have been developed for heat transfer in jet turbines, nuclear power station boilers, meteorological forecasting, and other rich applications. The constants for buildings certainly need examining as far as I am concerned before I am happy to have them used in very large and very small spaces with very different Reynolds numbers. These ideas about ventilation and fluid flow are based on experience. When commissioning the National Theatre, the convection currents seemed to be preventing the ventilation operating as required by the fire officer. We turned up the thermostat controlling the stage fan convectors and the convection currents reversed.

Fig 7

Air flows in the National Theatre

Then at the Royal Exchange Theatre Manchester which is a very large enclosure with a theatre inside it We worked with Professor Manfredi Nicoletti on an entry for the Cardiff Bay Opera House Competition which was to be a large glass enclosure.

Fig 8

Competition entry for the Cardiff Bay Opera House - Interior. Designed by Manfredi Nicoletti

Fig 9

Competition entry for the Cardiff Bay Opera House Thermal Balance

109

LIGHTING
One of the immediate consequences of the single storey property of wide span structures is that the spaces can be roof
lit

The requirements for light to enable people to see are not to be defined too simplisticaUy. The eye is like a camera with a variable aperture pupil to control the amount of light entering and with a light sensitive membrane retina to generate signals which are focused and transmitted to the brain. The signals are processed by the brain and the nerves in the retina to generate sensations which we interpret as "seeing". The retina can integrate the photons it receives in a variety of ways. I am sure I will raise as many questions as I answer in this brief summary of seeing but I am presenting you with a working designer's basic knowledge.
I

Fig 10 CFD for the Cardiff Bay Opera House - Velocity Vectors

In the absence of light the pupil is wide open and too few photons are received to see anything. The sensitivity of the eye/brain system is increased and it takes about half an hour for full sensitivity to be developed. The part of the retina most sensitive to low light levels is not in the same location as the area most sensitive to detail, so at night air force pilots are trained to look away slightly from a dim object. At low light levels, the retina integrates different coloured photons and integrates them over a period of time before transmitting a signal to the brain. Thus the image seen is monochrome and in these conditions the retina cannot catch fast moving objects. The sensitivity varies with age, but a fully adapted night eye can "see" a light coloured object under an overcast night sky at 0.001 lux.
SECURITY (;N'IM;t.;K]Ni; Uppf limit of V io tolerance Fi n Ronge of 'Vidicon' EOffiWa ojtenuW by doling leni iris

CflRDff 8 6 V OPERA MOUSF 'eMPERATI

Fig 11 CFD for the Cardiff Bay Opera House - Temperature Profile

The air flow and temperature were modelled using CFD as shown on the figure above. The air flow model immediately suggests that the air entry slot should be above head level and slanted upwards. There was another issue about the temperature plot. We initially modelled the solar gain onto every node of the lowest part of the floor level. The temperature plots showed very high temperatures on the floor which we couldn't understand. However, I vaguely remembered a lecture during which it was explained that mirages in the desert sometimes occurred when the morning sun heated the desert surface and the invert layer of air to a high temperature causing a very strong temperature gradient close to the ground and making the mirage. The hot air did not rise as a convection current because there was no trigger to start the convection at any point. The relaxations calculation process of CFD would be similar if all nodes were at the same temperature with the same heat gain. The heat gain would be equated to the temperature rise and the relaxation process might be stopped before getting a proper answer. We changed the input, putting in a double batch of solar input to half the nodes and immediately the answers looked sensible like the figure above.

Operating theatre Wcli iH chart

Drawing ofiice Offices, shops Stairs, corridor

Approximate rong of 'Newvici camera, using auto ilia .n*

Weil E s H

(0-4t*| f * 10
?

Range of 'V.dicorV and extended b , infra-red loop*

10"*'

i i

Clear nsght

|2r,lx| | 102ml| 1 iO Oorolsjj

10"*-

OttKCOtt slight Ay

10 '

Fig 12 Figure 4.8 from the CIBSE Applications Manual AM4: 1991 Relative light level chart showing operational ranges of several types of camera.

110

"See" must mean differentiate between two levels of brightness of objects of a certain size (not too small) so the reflective properties of the field of view are important. When I say "see" in quotes I mean the whole eye/brain process. As the general light level increases, the eye/brain adapts to the changed signals. With more photons arriving the cells can differentiate between different colours and can send/fire off signals to the brain at shorter intervals and the brain can "see" colour and fast moving images. There is a limit to the rate at which a cell can send signals to the brain. After receiving light and sending a signal it has to reset itself. So, as the light received by a cell increases, the frequency of sending signals increases until it is saturated and can simply report maximum brightness. Indeed, if too much light is received the cell can overheat and die. I was reminded of the ancient Egyptians who worshipped the sun god Ra. A person accused of offending the god was tied down on their back in the open and their eyelids cut off. At the end of the day, if they could still see they were innocent but usually the retina was burnt out by the bright sun and this was taken as a proof of guilt. Bright sun at 100,000 lux (1000W/m ) is definitely too much.
2

Imagine looking at a hemisphere on a table. In diffuse light where every surface is uniformly illuminated the hemisphere looks like a disc. In purely vector light, the shaded part looks black (like a half moon). A mixture of diffuse and directed light is needed in order to perceive the shape of the hemisphere.

Horizontal light from right 30% diffuse

Generally diffuse light

Fig 13 From Table 10 - Relationship of vector/scalar ratio to assessment of directional qualities of the lighting. IES Code for Interior Lighting 1977

The overcast sky is very diffuse and the design minimum figure is taken at about 5000 lux. Under an overcast sky the field of view reflects back a diverse field of light because of different reflection coefficients and different colours. The lighting/seeing is pretty adequate despite being viewed in very diffuse light. Inside a building the outside light is usually introduced from a transparent window. The light then has a vector component, and shapes and shadows can be seen even on an object with a uniform surface. Compared to outside the light level is reduced but the "seeing" is improved because the light has a stronger vector component. As an illustration of the need for diffuse light, think back to the difficulty of seeing anything in the region of a matt black motor car engine by the light of a single torch, even when it doesn't wobble. The reason for this discussion is to try to get you away from the view that the amount of light needed in a space can be defined by a single, simple figure of the amount of light. The contrast between inside and outside is important. The shading over a motorway underpass assumes that the speed limit is being kept and the light level can be halved every 3 seconds. In the Mediterranean one walks from the bright sunlight outside at 100,000 lux into a room with the shutters closed. It takes some time before you can see anything.

If there is an unevenly lit field of view but all the cells report saturation, the field of view is not being "seen" very well. The eye needs to adapt so that the bright part of the field of view just fails to saturate any cells. Of course the dimmest part of the field must provide enough light for good, fast colour vision. Thus, there is a maximum ratio of brightest to dimmest light for good vision. The eye adapts to the average illumination. If a field of view is uniformly lit then the eye simply receives a uniform vision and there is very Utile differentiation between planes and objects. I experienced this once when I went very early to an exhibition at the Satehi gallery. The gallery was painted white all over and lit with fluorescent uplighters. Light was diffusely reflected off the ceiling in all directions. There were very few exhibits and very few people so everything was white and I felt very disorientated. In order to "see" the field of view must not be uniformly bright. Hold a pencil vertically up on a sheet of paper and look at the shadows. I hope there are some - usually several. A shadow represents a place where one of the brightest light sources cannot shine on the paper behind the shadow. The light in the shadow comes from all other sources. The contrast between the brightness of the paper generally and the brightness in the shadow represents the contrast between the general diffuse light and the directional light from the source. Set this up in a more disciplined way and we can define the vector/scalar ratio of an illumination field.

I guess the shuttered room allows about 0 . 1 % of the diffuse sky light (10,000 lux) into the room so that the light level is about 10 lux. So on entering the room the light level drops from bright sunlight at 100,000 lux to 10 lux or 1:10,000, ie 213.3 = 10,000. It takes 3 seconds for the eye to adapt to a halving of the light level and therefore takes 40 seconds (3 x 13.3 + 40) to adapt to a light level of 10 lux. In the UK, well designed rooms with fixed windows keep most direct sun out and then provide a daylight factor of about 1% or 2%. I believe this should be increased for new buildings so as to reduce the amount of fossil fuel and electricity used for lighting. However, this aim of mine carries an increased risk of buildings getting too hot in summer. The indoor cricket schools at Lord's and Edgbaston try to exclude most direct sun and to provide a daylight factor of 5% to 6% and so give 1000 to 1200 lux on an overcast day. These buildings do not fit my idea of wide span enclosures but the cricket school at Lord's was won in a competition where we expected the opposition would offer air supported or other lightweight solutions. We put forward a case for natural lighting.

A diffuse skin with a transparency of 20 to 25% provides a light level of 1000 to 1250 lux as required. However, in strong sun the internal light level rises to 20,000 to 26,000 lux and the direct solar gain rises to 200 to 260W/m as well as long wave radiation from the translucent skin.
2

Another disadvantage of the overall diffuse skin is that it gives a very diffuse light inside. The lighting inside a tent or marquee is very diffuse and gives poor figuring to three dimensional shapes. In saying this I am offering opinions which could inform the development of the design of lightweight, wide span enclosures. I realise that diffuse skins are provided for indoor tennis centres, millennium domes and so on. As structural engineers gain confidence in making lightweight wide span structures using glass as the membrane, then I think the design, for example, of ventilating roof lights will be able to be developed. The thinking about the type of lighting needs also to be developed. A solution with 75% to 80% opaque area with 20% to 25% horizontal, transparent area supplies the same level of light as a diffuse skin. It is then possible to insulate the opaque area.

1 0 0 , 0 0 0 Lux in

Light Cloud 2 0 , 0 0 0 Lux

Blue Sky 10,000 Lux

Overcast Sky , 0 0 0 Lux

direction of sun

No to option one - dark building fitted with fluorescent lighting. Poor ambience, and lighting consumes 4 5 0 K W h / m a year.
2

No to option two - fabric/translucent roof In sunlight 24,000 Lux internally can quickly be reduced to 2400 by a small cloud.

Fig 15 Interior of the Indoor Cricket School at Lord's Ground. Photograph by Dennis Gilbert.

l> 7 ;>

JJJJlJJJJJhiflHB
I

\
1 1200 - 4 8 0 0 Lux

The transparent areas possibly need blinds or shades to protect the space from direct sunlight. Or the transparent areas can be diffusing with high transparency.

Yes to option three -north roof light. Only diffuse light admitted. Light levels in excess of 1200 Lux except after sunset in winter.

Fig 14 Sawtooth roof arrangement at the Indoor Cricket School at Lord's Ground designed by David Morley Architects

112

HEATING
J H K

The air movement in a large, tall space is likely to be violent. The subject of ventilation and air movement has been addressed. The roof of the structure is likely to subtend an angle of 2p steradians from a person so that its radiant temperature will be an important factor.
Fig 16 Interior of Bespak Stage 1 showing diffusing roof lights. Designed by the Cambridge Design Group.

One of the main issues with wide span buildings is that if the skin is to be light and transparent then a single or possibly double skin of fabric is unlikely to meet the Building Regulations for energy conservation. A justification might run :1. A conventional building deemed to satisfy the Building Regulations We need a light level of 1000 lux and a building with an opaque roof will need electric light at 30W/m , using fossil fuel at a rate of 90-120W/m .
2 2

An opaque roof is allowed to lose 3W/m of heat generated by fossil fuel, ie a U-value of 0.3 x 10C mean. Total energy requirement 93-123W/m . 2. Lightweight wide span skin During a 24 hour mean day with a mean inside to outside temperature difference of 10C, we have :Opaque Conventional Roof single skin
U-value Temperature Difference Heat Loss Fig 17 The Menil Collection, Houston. Designed by Renzo Piano. Photograph by Hickey Robertson. Heat Loss in 24 hours Electric Light (12 hours) (at 30W/m
2>

Transparent Roof double skin


6 10C 60W 1440Whr 3 10C 30W 720Whr

0.3 10C 3W 72Whr lOoOWhr

The lighting solution for the Menil Gallery is a special case where the external condition was generally strong direct sun (100,000 lux) and the light level inside had to be kept low for conservation purposes (50 to 100 lux), so 0.1% of the light was required and multiple reflections provided a really clever solution. So a wide span, single storey building can easily provide adequate light. A light transparency of 2 to 25% is feasible and provides adequate light. A lighting strategy can be developed to provide shading for direct sunlight and improve the overall thermal efficiency of the skin at the expense of a less homogenous solution.

Watt hours per day

1152

1440

720

This argument can be developed for different conditions. The light saved in the summer will improve the argument but for a lower light level, say 500 lux, the electrical energy saved is less impressive. Of course if the space is unheated the insulation value is unimportant. For a competition entry for the Cardiff Bay Opera House (shown previously in Figures 8, 9, 10, and 11), we had postulated a roof of 50% double glazing and 50% insulated panels. The enclosed space was a foyer so the electrical energy saved by lighting to 100 lux or so was not significant.

113

However spaces which formed the brief were huddled together rather like a Greek village or the National Theatre.

Heating large high spaces depends on several levels of consideration :Heat Loss Firstly we have to decide on the heat loss. The U-value of the cladding is important. Next, the amount of winter ventilation; how airtight will the enclosure be and what stack effect is likely. How much temperature gradient will there be to increase the heat loss at the top. Ventilation air will tend to come in at the bottom and on the windward side. The incoming cold, fresh air needs to be heated before it can lead to discomfort. A 4 or 5m/s wind speed coming through the windward cracks or open doors must be heated. Most 50m high buildings have lobbied entrances A 50m high stack with a 20C temperature difference will produce air movement through openings of about 8m/s. The temperature gradient in the space depends on the types of heat source. It is difficult for any part of the space to get hotter than any individual object inside. Direct fuel fired warm air heating is designed to be cheap by recirculating air into a space at around 70C and using heaters of 300 to 600kW capacity. The air flow is of the order 6 to 12kg/s and the air has to be supplied at a very high velocity to ensure that it mixes into the room before losing momentum and drifting up to the ceiling. At the necessary velocity noise generation is the problem. The parameters of air flow, heat load, noise generation, and temperature gradient have to be considered. In working out warm air heating we have relied on a hypothesis advanced by Holmes and Caygill [2] and repeated by P J Jackman [3], that :-

Fig 18 Sketch by Max Fordham for Cardiff Bay Opera House competition with Manfredi Nicoletti.

The surface area of this convoluted shape implied a heat loss through walls and windows with ventilation which we evaluated and compared to the heat loss of the simply shaped envelope. The envelope had a lower heat loss than that deemed to satisfy the building so the Building Regulations were satisfied. The internal buildings could have simple, un-insulated walls which notionally helped to pay for the whole scheme.

if thermal forces are not to dominate the pattern of air circulation. This relationship was originally postulated for a specific set of conditions but we have used it successfully in much more extreme situations. Where a heating system provides W kg/s of air at qC specific heat c kJ/kg = 1 at velocity V then the momentum M = WV and the heat load q = Wcq The relationship, where H = height, becomes :-

Fig 19 Buckminster Fuller Dome over mid town Manhattan

This argument follows Buckminster Fuller's for the dome over mid town Manhattan where the extended heat transfer of the buildings is replaced by the smooth, reduced envelope of the dome. In both cases there is a problem enabling heat and pollution to escape from the inner layer of buildings and this is basically the ventilation problem addressed in the next section.

114

wv f WOH

0.07

The plume is a particular case of jet flow. It is a hot jet. Jets are also described by Scorer and are very important to HVAC engineers in considering how air flows in space and how grilles need to be sized. The best visualisation of a jet which I know and which I expect most of you can visualise is a stream running under a humped back bridge. I idealise the flow in the following figure.

or V > qH 0.07 We have used the relation at Churchill College, St Mary's Church Barnes, and the CZWG office in Bowling Green Lane. At St Mary's Church Barnes we deliver 4mVs of air at lOm/s from a nozzle at 70C into a 10m high space. This does not generate a noise.

i, i i i i * i

^ ^

Streamline flow Flow starts to converge Flow remains radial Velocity on circle/sphere Stagnant: weeds thrive Rapids form at centre of stream Level drops Eddies feed flow into side of jet Jet expands Velocity drops Momentum is conserved Quantity of flow in jet increases then reduces as flow is bled off to serve the eddies

V i

^ ^

r 1 I
1

Status quo reinstated

Fig 20

St Mary's Church, Barnes Fig 21 Idealised flow at a hump backed bridge at a quiet stream.

At Churchill College, the space is 10.5m x 18.5m x 22m and the air supply to a dining room is at 12m/s.

CONDENSATION
I have started with crude warm air heating because I believe it is suitable for large open space of indeterminate use. Of course, radiant heat has its advocates for tall spaces. I don't want to give a detailed case as to why I am not in favour. Where competing design solutions coexist in a market then the reasons favouring one rather than the other are probably marginal. Of course, if a group of open air dining spaces were under a wide span canopy, radiant heaters to each space would be a good solution. The behaviour of the air in a space with heat sources inside it is largely defined by the behaviour of the plumes. A plume is a rising current of air which is warmer than the surroundings. The behaviour of plumes is described in the book "Environmental Aerodynamics" by Scorer and it has become a very important topic for fire engineers. Moisture movement in buildings is not perfectly understood. We should remember that the moisture content of air and water vapour has an upper limit. The upper limit is a function of temperature.

Air and water separate out into cloud/mist ,

. Water content

Upper limit of water content Temperature

Fig 22

115

The air in a building may be represented by point 'A' at a certain moisture content and temperature. If the air is cooled down until it is saturated - at the limit line - moisture will deposit as condensation or dew. The temperature is then a measure of the moisture content. It is called the dew point. The dew point of the air inside a building will be the same as the dew point outside unless moisture is added. Generally in well ventilated slightly wanned spaces condensation rarely occurs even on single membranes. For condensation to occur, the dew point of the air has to be above the temperature of any surface. On a clear night, heat is radiated to the sky and a lightweight surface quickly cools down to below the air temperature and often to below the dew point so that dew forms. Of course, dew will also form on the underside of a single membrane under this condition. As moisture condenses on the surface, latent heat is released so that the surface temperature tends to stabilise a little below the dew point. The boundary layer on the outside has a higher conductivity than the boundary layer on the inside so more dew will form outside than inside. The conductivity ratio is about 3 so there will be about 3 times more condensation outside than inside. As condensation takes place, moisture is taken out of the air. The slope of the saturation curve varies with temperature but at 10C 10C the moisture content changes by about 5g/kg for a 10C change in dew point (E B H Stevens and M Fordham ).
4

Condensation on the underside of lightweight sheeted roofs - warehouses, dutch barns, and so on - is not a problem which has been recorded in the literature. In buildings which are enclosed, moisture is likely to be produced. A person produces 0.022g/s of water and if ventilated at 10 1/s raises the dew point by 3-4C. The surface temperature of a membrane has to be raised by this temperature above outside to prevent condensation. A person raises the temperature of 10 1/s of air by about 8C and the temperature of a single skin is raised by about a quarter of the temperature difference between inside and outside. It is rare for condensation to form on the inside of single membrane surfaces unless there is a high degree of moisture production inside, say, a laundry, kitchen or bathroom. If the outside heat is transfer suddenly increased or the outside temperature suddenly reduced by a hail storm, condensation is sometimes seen on otherwise clear building surfaces. Condensation is more complicated in more complicated constructions. In a double skin membrane the heat transfer to the upper skin is reduced while the moisture transfer through a porous membrane is hardly reduced. Moist air is circulated through some constructions and this has been analysed in some detail by Stevens and Fordham. Membrane structures need to be developed with improved thermal resistance. The main complication arises because of handling interstitial condensation problems. An insulating layer is typically going to be 100mm thick and of mass 3kg/m .
;

The volume of air associated with each 1 sq m of roof is roughly the height of the space. To reduce the dew point of 50m of air by 1C and so prevent further condensation inside, we need to condense 25g of moisture per sq metre. lOg/m* looks like this.
1

OPTION 1
In cold climates the insulating layer can be resistant to rain penetration and placed on top of the waterproof membrane. The membrane is then kept warm and condensation is prevented.
Gravel, paving slabs or other means of holding insulation in place

Roof membi
Fig 23 Fig 24

116

In warm moist climates with air conditioned enclosures condensation is likely to occur on top of the membrane but that will be drained away as though it were rain.

OPTION 2
The insulation layer below the membrane. With this option, the membrane is at or near the outside temperature. Any moisture which flows through the insulating layer has to be removed by a ventilation path to outside. Standard UK practice (Building Regulations ) gives as guidance that vnetilation openings should be about 0.6% of the roof area. I think I came to the understanding while at the conference that wide span structures really do have to be designed to minimise the weight. On the other hand, some resistance to wind uplift is a benefit. For a target insulation resistance of R = (say) 3C/m W the mass per unit area = p R/k where p is the density and k is thermal conductivity. So p/k is the significant property.
2 2

The special feature of wide span structures are that the spaces tend to be tall, rooflit, and poorly insulated. The size is not a reason for abandoning natural ventilation because air movement is more likely to be turbulent in a large space than a small one. I would like to see light admitted as a series of discrete, transparent areas rather than an overall diffusing surfaces. Then the opaque areas can be properly insulated.

REFERENCES
1. Scorer, R S. Environmental Aerodynamics. Ellis Horwood, Chichester, England. 1978 Holmes and Caygill. B SRI A Lab Report 83. 1973 Jackman, P J . B SRI A T N 3/90 E B H Stevens and M Fordham: "Interstitial Condensation in Building Structures" in Building Services Engineering Research & Technology, 1996. Building Regulations Approved Document F, 1991

2. 3. 4.

5.

117

CONTROLLING THE INDOOR CLIMATE IN WIDE SPAN ENCLOSURES 4 CASE STUDIES


Nick Cullen Hoare Lea & Partners - Consulting Engineers

SYNOPSIS
This paper presents four case studies of different large span structures, describing the characteristics of, and the systems used to control, the indoor climate.

The temperature within a large space can be controlled using air systems or radiant systems. Indoor Air Quality (IAQ) can only be controlled using 'fresh air' (usually outdoor air). Many systems tend to combine the temperature regulation function with the IAQ function. The problem faced by Engineers is that hot air rises, or more accurately, cold air falls and forces warmer air to high level leading to temperature stratification within the space. This fundamental law of physics can work to the Engineers advantage. A case in point being Displacement Ventilation Systems (natural or mechanical), which rely upon buoyancy and gravity forces to drive them. However displacement air systems require the supply air to be introduced at low level and at regular -albeit fairly large -intervals. This is rarely compatible with the needs of large span structures and indeed is often in conflict to the use of such structures. The consequences of stratification are twofold. Firstly, the increased temperature differential at roof level results in a greater heat loss increasing energy consumption and thereby C 0 emissions. Secondly, thermal conditions within the occupied zones may at times be unsatisfactory, depending of course, on the location of the occupants.
2

The first two studies consider the difficulties inherent in designing systems that 'fight' against the basic laws of physics. The first of the two, the British Aerospace Aircraft Assembly Hall is based on work undertaken in the 1980's and highlights the significance of buoyancy forces and the difficulty in mixing airstreams of different temperatures. The second case study, the ExCel exhibition centre in London's Docklands, highlights the need for compromise in the design of Engineering systems. The second two studies review projects in which the designs made use of the natural forces of gravity and buoyancy in order to maintain thermal and Indoor Air Quality (IAQ) conditions. The first, the Millennium Stadium Cardiff, features a fully retractable roof and relies upon Natural Cooling and Ventilation enhanced with the operation of the smoke extract fans as necessary. The final Study details the work undertaken at the House of Representatives, Brasilia the Capital of Brasil. It discusses the significance of control and alternative strategies.

INTRODUCTION
To the Building Environmental Engineer it is generally not the overall size of a building that creates the challenge it is the internal height and the lack of suitable locations for indoor climate control systems. Large span structures are synonymous with high open spaces. The Engineer seeks to control not only thermal conditions but also Indoor Air Quality (IAQ) both to achieve comfortable conditions within the occupied space and to maintain a healthy environment free from pollutants (of which there are many). Ideally the Engineer would seek to condition the occupied space rather than the whole volume and hence benefit from both reduced plant capacity and reduced energy consumption and C 0 emissions. This is not always possible.
2

High spaces are generally conditioned using mixing systems with the supply air introduced at high level, the objective being, to minimise stratification by producing a fully mixed environment. The designer has to ensure that when heating the supply air can deliver heat to low level and when cooling the air arrives at low level without causing discomfort due to cold drafts. In the process of creating a mixed condition, pollutants, produced within the space, are diluted by 'fresh' air. The alternative, that of displacement ventilation, seeks to condition and removes pollutants only from occupied zone.

118

CASE STUDY NO.l BRITISH AEROSPACE AIRCRAFT ASSEMBLY HALL, BRISTOL "THE BRABAZON HANGER"

BACKGROUND The aircraft assembly hall was constructed in the 1940's for the specific purpose of constructing the Brabazon aircraft, the largest aircraft in the world at the time. The building's clear height (23m) was determined by the height of the Brabazon tailfin and its clear internal span, by its wingspan. Its overall internal height reaches 35m. At the time the building was completed, it was one of the largest clearspan structures of its type in the world. Its floor area was approximately 30,000m and enclosed a volume of 1,000,000m3 (Figures 1&2).
2

Height to Eaves 26 m Height to Apex 35 m Floor Area 30,000 m


2

Total Volume 1 mi 1 ton . 1

EXISTING HEATING SYSTEM The original (1940's) heating system comprised steam unit heaters at catwalk level blowing air vertically down into the space. At the perimeter of each bay were located a row of "swan neck" steam heaters which drew cool air from low level, heated it, and discharged the warm air down towards the hangar floor from a height of about 10m (Figure 3). By 1980 the steam pipework was beyond its useful life and had significant leakage problems. The pipework was poorly insulated, mainly with asbestos and as a consequence, apart from the health issues of asbestos the operating efficiency of the system was extremely poor. Furthermore, under test it was found that the unit heaters at catwalk level gave insufficient velocity to the hot air to overcome its inherent buoyancy. The heated air lost any momentum after the first few metres and rose back up to high level. Thus, only the perimeter "swan neck" heaters provided any useful heat to the hangar floor, the remaining capacity being used to heat the roof space. Temperatures at roof level rose regularly towards 40C in the vain attempt to hold a comfortable temperature within the occupied zone (Figure 4).

He,! Lot* through Roof

Down draught heaters dine! Fig I The Brabazon Hanger - Exterior View ar o t > r 3 id ww n y m Swan Neck Darcharge

C R O S S S E C T I O N - C E N T R E SPAN)

Fig 3

Existing Heating Sytem

Improved thermal performance Reduced heat Ion leading to increase tn temperature

Fig 2

T h e Brabazon Hanger - Interior View CROSS SECTION -CENTRE SPAN) ORIGINAL ROOT 4 HEATING SYSTEM CROSS SECTION -CENTRE SPAN) NEW ROOF 4 HEATING SYSTEM

The building has always been difficult to heat effectively. In the early 1980's a complete re-cladding of the building was undertaken to upgrade the performance of the building envelope to comply with the Building Regulations standards of the day. Sadly, the cost of upgrading the doors was prohibitive, a feature which we will return to later.

Fig 4

Temperature Profiles

NEW HEATING SYSTEM Immediately following the recladding contract, Hoare Lea & Partners were commissioned to design a new direct gas fired heating system to replace the original steam fired system. The concept was to replace the

119

existing steam heaters at catwalk level with direct gas fired unit heaters, blowing vertically downwards from a height of 23m (Figure 5). The existing perimeter heaters were to be modified, and instead of blowing warm air down to low level, they were to draw cool air from low level and to discharge the air vertically upwards, mixing the cool air with warm air at high level, inducing destratifying circulation currents within the space.

PERFORMANCE After completion of the installation, the performance of the heating system was monitored to assess whether the predicted performance was achieved in practice. The results were dramatic. The delivery of air at 45C to the hangar floor from a height of 23m required a substantial discharge air velocity. At part load conditions, when the discharge temperature was lower, the high discharge velocity was not dissipated, so that a very high air movement occurred at low level. It was decided to accept a restricted turndown ratio on the units, typically to a minimum of 80% of full heat output, the fans being controlled "on/off below this level. The building fabric, and particularly the old hangar doors, were found to allow a considerable amount of cold air to infiltrate into the building. As a consequence of the density of this cold infiltration, it tended to collect at low level creating a cold "lake" of air at about 10C in the first 2m above the hangar floor, the very zone that was required to be heated. Under full load output from the heaters, operating in response to temperature sensors located in the cool occupied zones, the buoyant warm air was found to have lost most of its momentum by the time it arrived at the bottom 2m zone. The discharge air suddenly moving in 10C set, not 20C ambient air, effectively "bounced" at this 2m level, providing very little heating effect in the occupied zone. As a consequence, the whole volume of the hangar was being heated to a temperature of 20-25C, in order to maintain I0C in the occupied zone (Figure 6)!

CROSS SECTION -CENTRE SPAN)

MODIFIED HEATERS

Fig 5

Proposed New Heating System

The concept had been developed in conjunction with Bristol University who carried out performance monitoring on the existing system and then on a trial mock up, modifying one of the perimeter "swan neck" heaters. Initial results were promising, showing a much reduced temperature gradient in the space. The team identified the proposals as carrying significant, technical risk, there being no precedent for use of reverse destratification system, least of all, on a building of this size. In order to offset this risk, the team applied to the EEC for a Thermie Grant which was subsequently awarded, in recognition of the innovative nature of the project. The client embarked on a significant construction contract, comprising the removal of the existing heating system, including the steam pipework installation, asbestos insulation and heaters. In its place was installed a new gas pipework, new power distribution system, fan powered unit heaters complete with discharge jet nozzles. The complete installation was undertaken, at a height of 23m, whilst maintaining production on the factory floor. This required significant protection measures to be provided to allow the building occupants to continue working safely. Key design considerations involved reducing C 0 , and moisture levels in the space to acceptable levels by introducing fresh air through perimeter units. The design of the heaters, and "swan neck" discharge nozzles was also critical to give good air mixing and air distribution.
2

^t^+HMt Low truh Roof hog

""1

hg lvmixes ih e e l 20*C t Wt rse dctg Air ih atr Bhre


a Entrained Air item

'

CROSS S E C T I O N - C E N T R E SPAN)

MAIN ACCESS DOOR

Fig 6 Actual Performance

The designers struggled to balance the design parameters of heat input, air velocity, noise and power consumption and cost and eventually arrived at a "best fit" solution.

Paradoxically, the solution to this problem was to reduce the maximum heat output of the gas heaters, lessening the buoyancy of the supply air, which enabled proper penetration by the supply air into the occupied zone, and good mixing in that space.

120

The modified "swan neck" destratification units were found to have minimal effect in destratifying the space, the temperature profiles and airflow patterns being determined primarily by the velocity and discharge temperature of air from the direct fired gas heaters. Of course with hindsight the solution should have included: (i) an increase the thermal performance of the doors a reduction in the infiltration leakages of the building.

CASE STUDY NO.2 EXCEL LONDON, ROYAL VICTORIA DOCK

(ii)

Had it been practical within the constraints of an operational production facility, the provision of a warm floor by embedded piping or by overlaid radiant heaters, may have overcome many of the problems.

INTRODUCTION Across the river from the Millennium Dome on the North side of the Thames a New "State of the Art" exhibition centre is about to open. Phase 1 of the project will provide 93,500m of accommodation including 64,000 m of exhibition space split between two halls. Each hall is designed with a minimum clear height of 10m. The entire exhibition space is located above a car park. A boulevard running the length of the building separates the two column free halls. The whole building can operate as a single exhibition space or be sub-divided down into individual halls each of 4000m (Figure 7).
2 2 2

Fig 7 Excel Exhibition Centre - London Docklands

VALUE MANAGEMENT The indoor climate control system was divided according to the minimum module size. A single air-handling unit serves each module and is located at high level within the structural depth of the roof. Supply air ductwork from the air-handling unit is distributed at high level (Figure 8). Out door air is drawn in via a 'beehive' air intake the amount being determined either, by Indoor Air Quality (IAQ) as measured by C O sensors, or according to the free cooling opportunities. As extract air is drawn it passes directly from the space and discharges to out doors.
z

Itk Ar nae i
^s- .,.,-

Ehut Ar xas i
"7\

)IINI?jahu|

Spl Ar va Ln trw Dfies upy i i og ho ifsr EUilna] zbtoHl


r

I Clm

Buead olvr

90m

Fig 8 Excel Exhibition Centre - London Docklands - Diagramatic

121

The exhibition space required both cooling and heating. The supply air system therefore had to operate to deliver warm buoyant air to low level during heating, and cool non-buoyant (heavier) air during cooling. The obvious answer was to vary the trajectory of the supply air according to the supply air temperature by using adjustable geometry diffusers. This however proved to be too costly and would probably prove to be unreliable and an alternative approach was required. The alternative proposal envisaged a fixed airflow trajectory with long throw nozzles fixed directly into ductwork and arranged in groups. With volume flow rate and design supply air temperatures, fixed, two variables remained under the designers control, discharge velocity and trajectory (Figure 9). Using Computation Fluid Dynamics combinations of the different parameters were tested in both heating and cooling modes. Computational Fluid Dynamics, not available at the time the design of Brabazon Hanger Design was employed to assess options and performance of the design (Figure 9).

Figure L5 (a) Temperature distribution at height of 1 5m

Fig 9

Results - Computational Fluid Dynamics

23B

RESULTS The results from the analysis showed that the cold slab (due to the unheated car park below) would create a 'lake' of cold air at low level which could be reduced in depth by increasing the momentum of the supply air, but could not be completely overcome. Once again the conclusion pointed to the need for a warmed floor which was beyond the budget. (Figure 10). The CFD modelling images brought instance 'Deja vu' to the (by now Partner) engineer who years earlier had experienced the Brabazon hanger or refurbishment and its outcome. It was recognized that the primary circumstance likely to occur was that of cooling and so parameters were selected to satisfy the associated thermal comfort conditions. Engineering designers learn very early that compromise will be called for, that compromise often involves designing to satisfy the primary circumstances. When warmth from exhibits and people will require a cool air supply from the building systems. That lessens the outstanding probability that when a few people rent a small amount of the space in colder weather they may find a bracing experience requiring a pullover. Satisfying the majority that is now called value judgement and is an essential part of an engineer's experience.

Winter model, no occupancy

Winter model- Low level occupancy

Fig 1 0 Results - Computational Fluid Dynamics

122

Fig 11 Millennium Stadium Cardiff - Exterior View

CASE STUDY NO.3 THE MILLENNIUM STADIUM CARDIFF

Fig 12 Millennium Stadium Cardiff - Interior View

INTRODUCTION The 120million Millennium Stadium Cardiff has a capacity of 72,500 people and is the first UK arena to have a fully retractable roof. It provides a multi-use all weather venue with completely un-restricted views. The grass pitch is completely removable allowing the arena to be put to use as a concert venue. The stadium takes the form of a bowl complete with retractable roof. This form clearly limits the Natural ventilation and cooling mechanisms that act around stadia with open corners. The retractable roof (Figure 11 & 12), when closed, created a number of problems that the designers needed to resolve. Firstly the space needed to be ventilated to remove unwanted heated and metabolic pollutants. Ventilation was also an important factor in maintaining a healthy grass pitch. Secondly it had to be safe, allowing spectators to escape in the event of a fire.

123

The arena was conceived as being Naturally Cooled and Ventilated using the vomitory passage ways and a highlevel louvre system as air paths. Numerous different scenarios were considered using Computational Fluid Dynamics (CFD). The Criteria set for the Stadium was for all occupied areas to remain below 28C at design summer conditions (26C). The effect of different sized openings, their number and location were investigated. The initial analysis assumed a worst-case scenario of stack driven ventilation only without wind assistance. The analysis showed the need for two sets of parallel louvres running at high level , one at the junction between the retractable roof and the fixed roof and the around the back of the upper tier seating. Temperatures at high level varied only slightly between the various options (Figure 13). The arrangement operated primarily using Natural buoyancy effects and, when available, wind pressure to drive air through the arena. The smoke extract fans are made available to guarantee a minimum volume of fresh air movement through the arena. CFD modelling showed that the combination of vomitary and high level openings produced acceptable conditions with the roof closed even without the beneficial effects of wind or with the fans running.

FIRE The fire engineering for public arenas is vitally important. The objective was to determine whether, in the event of a fire, there would be sufficient time for the audience to escape. This time for full evacuation from the arena was calculated as 12 minutes taking into account detection, investigation, action and evacuation times. In addition a smoke temperature limit of 200C and a visibility distance of 25m to a reflective sign were adopted, as design criteria. Being primarily a sports stadium the potential fire load was minimal. It was considered that a pop concert with a stage located at one end of the pitch was the worst case scenario. The effect of the operation of the mechanical extract system was investigated using Warrington's Fire Research CFX CFD software. The results highlighted two important factors. Firstly that the depth of the smoke was worst at the end of the stadium closest to the fire (Figure 14). The time available for escape in these areas did not meet the design criteria and people could not be located in these areas. Secondly the operation of the fans provided an additional 2 minutes escape time extending the period to 14 minutes for the topmost seats. The extract temperature of the smoke was estimated as being between 39C and 43C, well within the operational capability of the fans (Figure 15).

& PARTNERS

Time: + 1 2

minutes

0.0010

26 0

Cardiff Millennium Stadium

0.0000

Fig 13 Results - Ventilation and Cooling CFD Results

Fig 15 Computational Fluid Dynamics - fire/smoke - fans operational

Fig 14 Computational Fluid Dynamics - fire/smoke - no fans

124

CASE STUDY NO.4 HOUSE OF REPRESENTATIVES, BRASILIAN CONGRESS BUILDINGS, BRASILIA, BRASIL

In late 1997 Hoare Lea & Partners Research and Development group were asked to offer advice on the problem of acute 'Sick Building Syndrome' in the House of Representatives at the Brasilian Congress. The particular Building, is that pictured and constructed in the 1960's to designs by the renowned Architect Oscar Niemeyer (Figure 16).
Fig 18 House of Representatives, Congress Building, Brasilia Interior View towards Podium

The building had been reported as 'sick', indeed a Government Minister had passed away it was said, "because of the amount of his time he had spent in the building". An initial visit and inspection of the air supply system indicated that the system was clearly at the end of it's serviceable life. It also had some inherent design problems most notably the absence of any system of air extraction other than by tortuous route out of the chamber via the main entrance doors which had to be left open. (Figure 19).

Fig 16 House of Representatives, Congress Building, Brasilia Exterior View

The House of Representatives is one of two chambers (plenaria) in the Congress building complex and it measures some 30 m in diameter and 15m high. The plenaria has capacity for up to 550 people made up both of Representatives and a smaller number of journalists. A raked gallery for 'spectators' overlooks the chamber, encompassing / of the high level perimeter, but this is isolated from the chamber by a glass screen (Figure 17 & 18).
3 4

Fig 19 House of Representatives, Congress Building, Brasilia Schematic representation of existing ventilation and cooling system

Hoare Lea and Partners were asked to put forward a scheme which after much consideration was based on Displacement Ventilation principles. Unlike the first two case studies displacement ventilation is a system that relies upon natural forces to function. Cool fresh air is introduced at low level and is drawn towards any heat source where is warms and is 'displaced' to high level taking with it unwanted heat and pollutants. The polluted air can be extracted and thrown away having first passed through heat exchangers.

Fig 17 Plan and Section through House of Representatives

Two alternative schemes were studied and each was modelled using Computational Fluid Dynamics. The favoured scheme envisaged the installation of a compartmented raised floor through which air would be delivered to air terminals integrated into the seat. The floor would double as a conduit for power and data cabling (Figure 20).

Ka.-il.aii loi'iKc&t - under Lest tolut> . . i . ->' - of Temper&l'j'e tl)

Fig 21

CFD Results - Velocity Vectors - Temperature Supply

Schematic of Proposed New Displacement Ventilation for

dulled AHeattng Water from Existing Central Plant -

Fig 20

Schematic of proposed new displacement ventilation

The alternative method was to introduce the air around the perimeter of the space a scheme that would have required only a small raised platform. The size of the space highlights another inherent problem of large spaces not so far mentioned, that of locating control sensors. This problem exists irrespective of the parameter being measured. Ideally the sensor should be located at regular intervals within the occupied zone, but without a surface on which to mount the sensor an alternative strategy is required. The walls around the chamber offered possible locations but were rejected due to their variable surface temperature and unrepresentative location. The main concern was the IAQ within the space and the main pollution sources both of heat, chemical and biological contamination were the occupants themselves. The quantity of air could therefore be varied according to the number of occupants within the space. Whilst C 0 sensors are regarded as a good measure of IAQ when people are the main pollutant source, they were considered to be too much of an on-going maintenance item requiring regular re-calibration.
2

Riri/'iian Conor Case 2S Vekioty Victors Gotwrt C> Vckc-oty Magn-tuOo (nvs)

RMHAJNS 4.2 pa to i :

Thu Apr te-W8 ;

Fig 22

CFD Results - Velocity Vectors - Perimeter Supply

Rrf.7tu*nCuiiti(e** C&ce 2 Vei:*v Victors Ccwaoa Dv V**fflv Magirtudo (m/s) Cioss section at tortus 20m

eiuonfUNS 4.2 0(1. he) Thu Ap 16 -aSK) Fluefil inc.

Fig 23

CFD Results - Velocity Vectors - Perimeter Supply

Two alternative strategies were conceived. The first was the inclusion of a variable volume damper within the construction of the seat itself. This would enable the associated diffuser to deliver fresh air only when the seat was occupied. A background supply would be guaranteed through other diffusers. The alternative was simply to count, electronically the number of people within the space and then deliver an appropriate volume of fresh air. This would rely upon the characteristic of displacement ventilation for the air to be drawn to the heat sources within the room. Both these options would have resulted in energy and C 0 consumption reductions.
2

B'tU'haii Cu tyieu Cass 1 '->:< V6CKHSCokxM r. . e . Cressfreenon at racJus - 25nt

"

H1H. NS4.2(3J M2.W. unstMQY Fn Apt 17 1998 Hurt IK,

Fig 24

CFD Results - Velocity Vectors - Under Seat Supply

126

RESULTS The results confirmed the design supply air volume was sufficient to maintain thermal conditions within acceptable limits in both cases (Figure 21). It did however identify that the alternative perimeter supply solution generated a 'dough-nut' vortex which had the effect of driving high level polluted air to low level back down into the occupied zone. This was due to three factors. Firstly the massing of heat sources created a coalescence of individual plumes which rose to high level. Secondly the thermally cool surfaces of the glass divide between the gallery and plenaria generated a down flow of air. Thirdly the rising plumes of air drew air from the perimeter supply points. The combination of these three characteristics generated the vortex (Figure 22 & 23). In contrast the favoured option with the supply air introduced on a seat by seat basis showed a less vigorous air movement with a general, albeit un-steady drift of air flow to high level (Figure 24). The project proposals await approval and finance from the government which, unlike our own, of whatever party, is very concerned not to spend money on it's own accommodation whilst there are calls for money from its populace.

CONCLUSION Wide span structures enclosing large volume high spaces present the Building Engineer with significant challenges. The Building Environmental Engineer seeks to control the conditions within the occupied space with the minimum of 'environmental impact'. Numerous different scenarios often need to be considered The function of the space along with cost restrictions often force the Professional Engineer to design systems that fight the basic laws of physics and to seek compromises in performance. The advent of CFD has given the Engineer an invaluable tool enabling the prediction of the performance and comparison of different engineering systems. Despite the rapid growth in computer power we are still limited to making only global assessments of large spaces.

SECTION III Mezotecture and the Millennium Dome

The Millennium Dome: Introduction to Client Concept Mezotecture

Servicing the Dome Environment Fire Engineering The Roof Structure of the Millennium Dome Principles of Construction Construction Management of the Dome

129

THE MILLENNIUM DOME 'INTRODUCTION TO CLIENT CONCEPT'


Jennie Page

INTRODUCTION
The Millennium Experience incorporates the Dome at Greenwich and a linked National Programme of events and activities which began in late 1998 and will run through the year 2000 and, in some cases, beyond. These events include programmes such as Children's Promise, Tesco SchoolNet 2000 and McDonald's Our Town Story. It also includes partnership with the eleven National Lottery Distributors, in the Millennium Festival. The Millennium Experience was created to attract, entertain, educate and involve both the visitors to the Greenwich site itself and participants throughout the UK, with the goal of influencing positively each individual's view of themselves and the world's view of them. It is estimated that, if all goes well, by the end of the year 2000 forty million people will have been touched by the Millennium Experience. The iconic centrepiece of the Millennium Experience is the Dome; not just a potent symbol for the New Millennium, but the best known of all wide-span enclosures.

Chairman of the Commission, Peter Brooke. At that stage no one imagined that setting out to create such an exhibition would result in the construction of a globally recognised landmark building, seen by many as one of the new wonders of the modern world. It was the Commission's role to define the objectives of this exhibition, and in spring 1995 the key principles were opened to consultation with official bodies. These principles were that it should be inclusive and accessible to all, be based on the use of public transport, encourage urban regeneration and should leave a lasting legacy. The Commission laid down one final objective, with what would later turn out to t>e an amazing act of foresight; that the exhibition should include an all weather venue capable of accommodating a large audience for one off events and entertainment. The consultation received only desultory responses, but nonetheless the Commission persevered with its plans. Two competitions were set in motion, the first to find suitable locations and the second to identify organisations with the potential to run a national exhibition. Fifty-seven sites were identified by owners, local authorities and enthusiasts, and these were reduced to four which best met the criteria. Detailed information on these four was provided in a two stage bid process to potential operating consortia reducing from fifteen to four and subsequently by withdrawal of one and amalgamation of two others to two submitted bids at the end of 1995. The bid from this design and exhibition company, Imagination was quickly recognised as the most likely to meet Commission requirements, and the selection of a site became a straight choice between Birmingham and the Greenwich Peninsula. Both had their problems, and both opened up different opportunities. Greenwich was contaminated and derelict, but paradoxically this offered the opportunity to kick-start the regeneration of the largest remaining development site in central London. This would provide a very tangible legacy from the exhibition. In addition to this Greenwich had the winning attraction of being the "Home of Time", generator of the prime Meridian. This meant an exhibition in Greenwich would welcome the new millennium on behalf of the world. The announcement that Greenwich had been selected as the site for the exhibition was made on 21 February 1996.

A NATIONAL MILLENNIUM EXHIBITION


When John Major's Government decided to establish a National Lottery in the UK, it decided that the proceeds after tax and operator profit would be equally distributed between five good causes, one of which would be to mark the passing of the second millennium and the arrival of the third. The body created by Parliament in 1994 to do this was the Millennium Commission. I became its second Chief Executive in March 1995. The Commission's first task had been to determine what it was actually going to do. There was no precedent for celebrating a millennium and a variety of possibilities were explored within the legislative constraint that the majority of the money would be distributed by grant to support capital projects proposed by not-for-profit organisations. However the idea of organising a national exhibition had been mooted as early as June 1994 in a public speech by the then

130

T H E M I L L E N N I U M E X P E R I E N C E AT GREENWICH
British Gas as owners of the Greenwich Peninsula site had achieved the diversion of the Jubilee Line Extension to provide a station on the peninsula as a precursor to developing value for the site. As 1996 progressed they negotiated with English Partnerships for the sale of their entire holding on the peninsula so that English Partnerships could provide essential infrastructure. British Gas, with the agreement of English Partnerships, appointed the Richard Rogers Partnership to develop a masterplan. Imagination used this masterplan as a template for their exhibition proposals which they developed to comprise 12 pavilions arranged in a circle in place of the linear arrangement of 10 pavilions which characterised their Birmingham design. Imagination liaised closely with the Richard Rogers Partnership as they developed their ideas and it quickly became apparent to both teams that the constraints of the Greenwich site made the creation of shelter vitally important. Surrounded by low-lying land and with the Thames on three sides, conditions at the Greenwich Peninsula can be cold, windswept and inhospitable. These climatic considerations were compounded by the fact that, unlike the Great Exhibition and the Festival of Britain which were open only during the summer months, the Greenwich exhibition was to operate through the year, opening and closing in the depths of winter. Imagination and the Richard Rogers Partnership set about this challenge with tremendous energy and creativity and, within a month, the breakthrough was* made, and Buro Happold, already working with Imagination on its designs, became fundamental to further development of the project. The circle of 12 pavilions become 12 zones covered by a single span structure that was weathertight, efficient, innovative and highly attractive. History has recorded that the Dome was born on 22 May 1996.

captured the imagination. It has become an iconic symbol which, like all icons, can be immediately recognised whether drawn by an artist or a child. It can be lampooned as easily as it can be lauded. It is also a building that is difficult to grasp. We have no reference against which to compare it. At twice the size of the Georgia Dome in America, the next largest building of its kind in the world, the Dome cannot really be said to have a kind at all. As an enclosed space it is a quantum leap, and a million comparisons have been created to fill our experiential gap. It covers an area more than 3 times that of the Coliseum in Rome and 50 times that of the Dome of Florence Cathedral. It can hold the water contained in 1,100 Olympic size swimming pools. Finding new ways of describing the Dome has become a national pastime, but there are no words that can truly give a sense of its vastness. The Dome is a paradox; there is more to its impact than simply its scale. Minimalist but monumental, dynamic but eternal, translucent but concealing, expansive but restrained. It may be a long time before we are able properly to comprehend the Dome, a space so unlike any other that it has itself become a major part of the visitor experience at Greenwich. From its birth on 22 May to the submission of the planning application to the London Borough of Greenwich on 31 October 1996, the Dome developed rapidly. The planning process was equally fast. Temporary planning permission was granted in January 1997, by which time, following the failure of earlier plans for the management and financing of the project, I and three colleagues were persuaded to leave the Commission to create the management and corporate structures to take the project forward. By this time too, the project had had the first of its many bruisings in the press because of political and financial difficulties, and it was still not clear that an incoming Labour Government would finally give it the go ahead. All decisions in the first half of 1997, including the choice of roofing material, were on the basis of possible cancellation and, if not, then a short building life on that site. In this environment the Dome was a crucial part of the project strategy. As a low cost enclosure it minimised the upfront costs at risk. It also gave phenomenal programme advantages. By focussing on the Dome as an enclosure first we were able to begin construction to a programme and budget without restricting the way that the exhibition itself continued to develop. It bought time to develop content to suit the public and the all-important financiers - the sponsors without prejudicing the end date. With 20 acres under cover, we could look forward to building the content free of weather constraints; without this flexibility the programme would have been unobtainable.

THE DOME
It was immediately apparent that the Dome would be supremely fit for its purpose. A vast shelter of infinite possibility, it would provide both a construction envelope and a protective umbrella within which the exhibition could change and grow. The design development of the Dome took it rapidly from the initial concept to the present synthesis of architecture and engineering, not simply a structural colossus but also a festive, delicate spider's web that has

131

THE CONTENT
During early 1997 we undertook market research into the nation's expectations and aspirations for the millennium. The findings were revealing and showed that the general population had a very strong sense that the new millennium should be marked by a change for the better, a collective notion that this special time offered us all a "Time to Make a Difference". This proposition was later broken down into three components that together reflected our every day lives; who we are, what we do and where we live. These three areas where used as the brief for the design competition for nine of the original exhibition zones; Body, Mind and Spirit (who we are), Work, Rest and Play (what we do) and Local, National and Global (where we live). Design responses to our brief were as creative as they were varied, but immediately apparent in each of them was the freedom that the Dome's structure gave to the exhibits. Architects and engineers, unhindered by the normal constraints of wind, rain, and extremes of temperature, and just as importantly, concealed from the eyes of the local planners, were able to develop extraordinary structures that not only housed exhibition content, but became a part of that content themselves. Each zone has become unique and immediately recognisable in a way that would not have been possible had it needed to keep out the elements. Within the fixed budget available to us, the constraints of keeping off the rain, keeping out the wind, keeping in warmth and so on could otherwise only have created a series of more conventional pavilions that were a variation on a theme. The Dome allowed us to create zones with character and identity. Given this freedom, the Dome has begun to answer one of the long-standing questions of wide-span enclosures: "If a building did not have to be weatherproof, what would it be?" It does not need a roof, it does not need windows, doors, insulation, it does not have to withstand the same loading of wind and snow - it can really become your imagination.

Experience, the client was a start-up company established long after its consultants had begun working on the project, and after its original begetter, Imagination had indicated that it wished to limit its involvement in management and delivery. Millennium Central Limited, later renamed the New Millennium Experience Company Limited, did not come into being until 13 February 1997, more than 3 months after the planning permission for the project had been submitted. In the company's earliest days the team was literally operating out of cardboard boxes, a private sector company, but with ownership of its single share remaining in the public sector. We were faced with the challenge of meeting an unprecedented deadline and creating something that was unique. Our engineers, Buro Happold, will probably tell you that the Dome is the culmination of decades of work; the next logical step in a progression of ideas, technology and projects: not rocket science. The Dome sits comfortably too as a natural step in a succession of projects realised by the Richard Rogers Partnership, but from the perspective of a new and still emerging client, the Dome was not just a first project, it was an entirely new world. The process of learning this world took place on a very public stage. As a new client we had not only to reassure ourselves, but to do so with sufficient confidence to carry along the politicians, the sponsors, the public and the press. It is difficult to see how this would have been achieved without the Dome itself. The creation of the structure before the contents were finalised was crucial to timescale and funding; the immediate identification of the project with a major building with no precursor both gave it a heroic status and stimulated significant attention on the construction achievements. The opportunities, by virtue of scale, to make major policy initiatives, for example on the environmental front, also gave the project additional relevance to Government policies. In essence, the Dome became the Millennium Experience, at least until 1 January 2000. To undertake such a project in such a short time frame, and with such an immovable deadline was a considerable leap of faith that was only managed at each of its difficult stages through teamwork. In every area into which we ventured, we had to find the experts, enthuse them with our love of the project and integrate them into a specialist team capable of withstanding the trials ahead. This has become very much the success story of the project. From abseilers to engineers, project managers, quantity surveyors, electricians and many more than I can possibly mention, the remarkable achievement that is the Dome has in the end only been made possible by the coming together of the widest possible range of exceptional people.

THE CLIENT PERSPECTIVE


The client's particular perspective on the realisation of such a monumental project is difficult to quantify, not least because the Dome has served so many masters during the course of its development, but also because of the special nature of the client organisation. Generally, the client of a major building project of a major building project has the luxury of an established company history with a tried and tested operating structure and past experience of procuring a particular type of development. In the case of the Millennium

132

FUTURE THINKING ON WIDE-SPAN ENCLOSURES


The Dome is by no means the only wide-span shelter on the Greenwich site. Like many of the great Expo sites of the past, the Millennium Experience at Greenwich has relied heavily on wide-span enclosures. Not just for the flexibility that they allow, but for the very character of the visitor experience. Operational support buildings, reception buildings, entrance canopies and theatres have all been provided using lightweight fabric structures. It is a building type that has proved invaluable in providing fast, economical cover for large areas of an exposed site. The Dome also has its precedents in the Crystal Palace and the Dome of Discovery, each a very particular reflection of the architectural and engineering capabilities of the day. Perhaps the most remarkable thing about the Dome however, is that it was in the end actually built. There have been a great number of monumental projects that came before the Dome, bigger, more challenging and more revolutionary, and yet this is how they have remained - as projects, fragments of ideas with aspirations that have never be achieved. From the megalomaniac visions of Albert Speer's vast cloud-filled Dome to Buckminster Fuller's mile-high bubble for Manhattan, each has foundered. The Dome is not the biggest idea, nor the most groundbreaking concept, but unlike the theoretical projects that preceded it and despite its many near death experiences, this Dome has happened; and through its creation the Dome offers us many lessons. It has taught us most importantly that the widest of widespan enclosures, the meso environments, can be made to work. Designed properly and with attention to detail as well as size, clouds don't form, snow doesn't fall, our physiology does not suffer, internal conditions do stabilise - vast enclosures do work, and in so doing they open a new world of design, freed from the constraints of the outside environment. As for the future: teams are already hard at work devising ongoing uses for the Dome, under similar constraints of time and budget, and with similar pressures to deliver a project fit for the nation. I feel sure our successors will go through many of the same experiences that we have, but with one comfort that was denied our team - they know that it works. Doubtless, with the Dome in existence, new projects will be devised for more astonishing, more revolutionary wide-span enclosures. It remains to be seen whether any such projects are actually realised, whether the Dome will be surpassed to become simply a step in a progression, or whether it remains a momentous one off.

133

MEZOTECTURE
Mike Davies Richard Rogers Partnership

Walking in the open across the valley to the forest one is buffeted by the chill wind and the touch of rain. On entering the forest the rain stops, absorbed by the foliage canopy, the wind dies away becoming just a rustle in the tree tops. One is 'outside' but somehow more 'inside'.

The forest is a sort of outdoor enclosure - a world of its own - a mezo-environment - mid way between truly outside and truly inside. In cities one finds the same effect. Big covers like the great Victorian railway station roofs shelter passengers from the rain and reduce wind effect. They are not heated nor really enclosed but with reduction of wind and no rain one is definitely more comfortable than standing outside. In Covent Garden Piazza the big pavilion provides cover for the shoppers and strollers. The big stations and the market halls of yester-year were unheated enclosures built to improve comfort and to protect the activity of the exchange of goods. These big sheds took the edge off the weather. No energy apart from lighting in various forms was provided to keep them operational. At the other end of the scale from the simple sheds, we have large buildings with fully heated, cooled and air conditioned interiors which are climatically controlled to fine limits and to normal human shirt sleeve comfort levels, with sophisticated control systems, large services and plant areas and considerable technical complexity. On the one hand we have simple covers and on the other hermetically sealed, fixed use, energy hungry buildings. Somewhere between the two lies Mezotecture. As time and technology march on there is an inevitable tendency for the simple covers to become more sophisticated and more daring. Over the last decades structures even larger than the great railway station roofs have been constructed for more concentrated human uses. Open stadia provide shelter for huge numbers of spectators under covered terraces. In more sophisticated stadia, deployable roofs have appeared, opening and closing as required and sheltering spectators and pitch alike in inclement conditions. In a further step, very large fully enclosed stadia have been built which offer complete protection from rain and wind for all internal activities. Open and enclosable stadia make few energy demands in order to operate, but the moment a large public space becomes enclosed, it brings shelter benefit

Forest

The interior of the forest creates its own local climate. A copse with fewer trees does not provide the same sense of shelter. The forest is a very large entity and its sheer size creates and sustains the shelter effect. The weather outside the forest zone is buffered by the forest, and changes outside take time to have an effect on the forest interior.

134

but also brings with it another consequence. The enclosed stadium inevitably requires some form of environmental control in order to ameliorate local conditions which it has itself created and which is dependent on its local geography and climate. Whereas the shelters have no energy requirement - enclosed stadia such as the Houston Astrodome incorporate the expectation and the manifestation of far greater environmental control. The Astrodome being situated in Southern Texas is geographically placed in an uncomfortable, high temperature, high humidity environment. It must be cooled to be viable and to increase public comfort. The moment the big cover becomes enclosed, the act of enclosure brings with it service plant, distribution systems, extra space demands and big energy bills. The loose fit cover evolves into something else. Somewhere between the simple cover and fully environmentally controlled building lies the generic archetype that I call mezotecture - very large scale enclosures - but not big sophisticated buildings. Mezotecture is not simply weather protected space nor fully air conditioned interior space, but mezotecture is big and like the forest, its size is the buffer that assists in tempering its environment and which also gives it a different performance expectation than that of a traditional building. St Pancras and the other great stations were the large span covers of the 19th century but the Crystal palace and the Palm House were the mezotecture of their time. Buckminster Fullers' cover for New York captured the spirit of the mezo-environment whilst the Houston Astrodome and the Crystal Cathedral in Los Angeles put the concept into working practice in the last decades. Other examples proliferate, including the giant covered aqua park in Japan and at mega-malls such as the Edmonton Mall in Alberta where people live in the mezo space for a week at a time.

A curious side shoot of mezotecture as a concept, although raising other puzzles, are the networks of the Dallas Skyway and the Minneapolis/St Paul shelter system. In Dallas and Minneapolis, medium scale internal atria within individual city blocks are linked together by a series of subways and skyways between blocks. By this means very large scale multi-block, enclosed and partially controlled public realms are created by the coalescing of individual parts, in one case as an air cooled network to avoid going outside to the extreme heat of the southern Texas summer and in the other a heated city armature to avoid freezing to death at a Minnesota bus stop at temperatures of 30 degrees below zero. Other recent examples of mezotectural concepts are RRP's own proposal for the South Bank Arts complex in London, the Biosphere II project in Arizona - USA, Eden, the new eco-centre in Cornwall and Gilles Perraudin's new college mega-cover in Westphalia. All of these structures have different objectives and performance criteria than simple covers or orthodox complex buildings. The largest single space of all, the Millennium Dome in Greenwich is also Mezotecture. It is interesting to note that these examples are very site specific, being tailored individually to their local conditions. Thus although the general principles that apply to these huge mezo-environments are well understood, it is difficult to extrapolate firm environmental conclusions or apply hard data from these examples to new applications in other locations. Each case has to be derived very much from first principles with the specific ingredients of site, external and environmental parameters and geography, internal requirements and performance and specific operational issues all being studied holistically together. Only by an iterative process with complex environmental analysis and CFD studies is it possible to predict the internal performance of these mezo-environments with any degree of precision. The science of the mezoenvironment is still young. The RRP proposals for the South Bank Arts complex in central London were based upon the concept of a universal cover that linked most of the buildings on the site, enclosed them all and their surrounding terraces providing an environmental buffer zone between the hostile external South Bank winter world and the real interiors of the arts and music buildings. By providing a new mezo-environment, all the interstitial spaces between the buildings would become more viable civic areas offering some protection and environmental comfort, rather than being hostile and unusable for much of the year.

The Palm House, Kew

It was calculated that the addition of the transparent and translucent envelope around and over the South Bank buildings and their large areas of underused terraces

135

The design team of Richard Rogers Partnership as architects and Buro Happold as structural and services engineers designed the structure and developed the environmental strategy for this vast mezo-environment. Tony McLaughlin the Buro Happold team leader explored and analysed means of tempering this huge environment rather than providing total internal environmental control. The team developed a concept whereby the Dome is heated in winter but would not be able to fully compensate for the worst outside temperatures and in summer would be cooled and ventilated by active plant and by means of natural ventilation but again only providing partial cooling. The internal environment of the Dome would act as a buffer between ambient external conditions and more traditional internal comfort conditions. The Dome interior is a tempered microclimate rather than a highly regulated building, a mezo-environment which consciously counts on a concept that Buro Happold and RRP dubbed "clothing co-operation". In winter, visitors would arrive well wrapped up and are asgumed to keep all or some of their warm outdoor clothing on, regulating their own individual comfort levels. Similarly in the summer, visitors would come more lightly clad and take off jackets and light overclothes. To date, this "clothing co-operation" has worked well, allowing Dome visitors a reasonable degree of comfort except in the most extreme conditions. The scale of the Dome mezo-environment is such that the temperature change is slow and needs to be planned for. External temperature and humidity, internal conditions, heating and cooling and thermal load of internal lighting are all looked at holistically, in addition to the dramatic thermal load of more than 3 megawatts of up to 30,000 visitors. There is no exact computational model for all these interactions but practical experience within the Dome demonstrates that expected conditions follow earlier CFD analysis closely. The Dome mezo climate provides a tempered environment for a round million square feet of enclosed space based upon a holistic balance of all environmental inputs. The micro climate was provided extremely economically, within budget, on time, and has so far achieved virtually all that earlier environmental control calculations indicated were possible. The team awaits the British Summer with bated breath. Further ideas need to be explored, further opportunities may be lurking. Could the Dome mezotecture be pushed further?

South Bank Centre

would have the same environmental impact as transporting the whole complex to Bordeaux. By creating a mezo-environment, a seven degree increase in local ambient temperature could be achieved without compromising the buildings or their surrounding spaces whilst bringing substantial advantage and new life to the arts complex. Sadly, arts politics prevented the scheme proceeding but at the same time a much larger mezo-environment emerged which would later be built at Greenwich, the Millennium Dome. The Millennium Dome was proposed in response to the natural criteria of the Greenwich site which was characterised by strong but warm south-westerly winds in summer coming up the long river reach from Greenwich and by bitterly cold north-easterly winds in the winter flowing directly from the Baltic and Siberia. With these winds and no land over 60 metres high between the Greenwich site and Russia the peninsula could be completely unviable as a site for a year round national celebration. The principal design driver was therefore shelter and enclosure; these to be created at an enormous scale, to be built very quickly, at very low cost, with much spectacle and designed to have an unmistakable national identity. The Dome is not simply a cover, it is the largest single enclosed public space in the world - and certainly the lightest. Once designed, erected and enclosed, it became a huge mezo-environment, not just a cover but not a true building either.

V lii-J!

^ / ///

Exterior shot of Dome

136

This will affect all the other parameters within the Dome. How strong will the summer stack effect really be? Will local rain occur in the Dome? The dome contains a town under cover and develops its own convection cycles and internal weather patterns. There is no fully measured precedent for the Dome. More information needs to be collected from weather sensors in around the Dome in strategic locations. We must build up our knowledge of giant mezo-environments. The current skin of the Dome is only one of many possible enclosures ranging from porous sun shading at one end of the scale via glass or polycarbonate membrane scales, to energy collecting adaptive intelligent skins at the other, which themselves carry out the active environmental control functions. The intelligent mezo-environment of the future may be largely self-regulating. The concept of permanence versus the temporary may have to be explored further. We may stretch planning legislation with new approaches. Is a large lightweight mezoenvironment a building, a tent, a shed. Is it temporary, is it permanent? Furthermore, enclosures within large mezoenvironments do not have to be weather proof or very highly insulated and are thus highly flexible and adaptable in nature and form. Are they buildings or are they sets. Do they require planning permission or are they interiors. There is no reason to assume that mezo-environments have to be as formal as the architecture of the Dome. Lightweight covers and enclosures of all types will become an increasingly significant part of our previously outdoor urban fabric. Environmental conditioning of quasi-external spaces will become more common. The construction of the Dome has demonstrated that very large scale mezoenvironments can be built economically, rapidly and with very low embodied energy. Passive environmental conditioning which has after all been in use for thousands of years will be increasingly employed in modem applications as the pressures of the energy issue increase. Active natural environmental conditioning in mezo-environments will proliferate. The line where the outside world stops and internal space begins will become increasingly and maybe beneficially blurred. Macro covers and mezo-environments will proliferate for both specific uses and for city block scaled and urban scale environmental interventions - informal, sinuous, energy effective and comfortable, wrapped amongst the urban fabric. Where architecture and urbanism meet there is mezotecture. We will be wise to explore its potential further.

Interiors of Dome

If the team had had the money and the appropriate permissions a major heat pump energy exchange with the river Thames beside the Dome might have been possible with a hundred and fifty million tons of water flowing past the Dome each day. What effect would evaporative cooling by roof sprinkler have in hot summer conditions. What effect could be gained by solar shutters in the external skin. Thermal storage in the floor slab and under slab could be investigated, possibly in combination with heat pump storage. Questions still remain to be answered in the real situation. What part will the enormous thermal inertia of the slab and ground within the dome play over the course of the year? It is certain that the Dome footprint will be storing more thermal energy at the end of the year than on opening night.

137

SERVICING THE DOME ENVIRONMENT


Tony McLaughlin BSc. C Eng. MCIBSE. MASHRAE. M Inst E. Partner, Buro Happold Consulting Engineers

SYNOPSIS
The Millennium Dome is a fabric clad structure covering some 80,000m2 which is to house a spectacular exhibition for the duration of the Millennium Year. This paper describes some of the constraints on the environmental engineering design, the servicing strategy adopted, the cooling and electrical loads determined, and how the environmental design evolved to meet the changing development of the exhibit designs.

In May 1996, faced with time running out, Gary Withers of Imagination and Mike Davies of the architects Richard Rogers Partnership suggested covering the whole site with a giant umbrella. This would create a protected environment in which exhibition structures could be designed specifically for the exhibitions and be rapidly erected without the necessity for weather tight cladding. We in Buro Happold picked up that idea and suggested a fabric clad stressed cable-net structure supported by 12 masts. This concept was welcomed by the client and engineering work got underway.

INTRODUCTION - THE MILLENNIUM EXPERIENCE


The idea of holding a celebration for the millennium had been talked about since 1993 and even before. The Greenwich site had always been a possibility but other sites were also under consideration. In the latter part of 1995, the Millennium Commission invited bids with design proposals for several sites. Imagination Ltd joined with the NEC and Birmingham City Council to put forward a proposal for Birmingham. Buro Happold assisted Imagination in that bid. Imagination's proposal for content and design ideas was judged the best and they were subsequently asked to consider how they would transfer it to Greenwich. In the first months of 1996, Imagination, with assistance from Buro Happold, put forward a number of proposals for housing an exhibition in pavilions with a large arena for shows and displays. Richard Rogers Partnership was at that time working with British Gas and English Partnerships on the master plan for the whole of the gas works site. Their master plan had a circular road pattern at the northern end, which Imagination had incorporated into their exhibition plan. The separate pavilions were four generous storeys high and involved a considerable amount of construction work leading to a difference between the costs of the designs produced by Imagination to meet the brief, and the Millennium Commission's budget. The site was very exposed to wind and rain coming off the river and there was a worry about the impact of this on visitor experience in the winter months. Imagination was trying to deal with this by covering the spaces between the pavilions, which were arranged around a central show arena.

THE SITE CONDITIONS


Greenwich peninsula is an exposed site with the river Thames on "three sides "of the Dome site leaving it vulnerable to the winter winds from the east. In the depth of winter it can be an inhospitable place when the wind is in the wrong direction. Mike Davies reminded us many times that to the east there is no ground over 100m between Greenwich and the steppes of Moscow. But like most southerly UK sites the met office offers the following synopsis, the prevailing wind is south westerly, the coldest month is January with a mean monthly temperature of 4oC and July is the warmest month with a mean monthly temperature of 17.5oC. As a matter of interest a temperature of 37.8oC was recorded at Greenwich in 1911. Greenwich is 7m above sea level.

ENVIRONMENTAL ENGINEERING
THE 'UMBRELLA' CONCEPT The concept of developing an "umbrella" environment is nothing new, as many of the mainline rail stations demonstrate. What makes the Millennium Dome different is its physical scale and its intended purpose. The dimensions of the Dome are huge: 320m in diameter, 50m high in the centre, with an enclosing wall structure 10m high and 1km long. The contained volume is approximately 2.1 million cubic meters, which leads to a many well published and interesting statistic eg the weight of air inside the Dome is actually greater than that of the structure that encloses it, not to mention the fact that it could contain 3.8billion pints of beer... It is twice the size of Atlanta's Georgia Dome, previously the largest tensile-roofed structure in the world.

138

mm
Fig 1A The Environmental Concept

When we first started work on the building services and environmental systems, it is fair to say that the Clients brief was somewhat lacking, both in terms of what was likely to happen inside the Dome, and even more so, what its form and operation was likely to be. The driving issue was time. The only design guidance we had, was to draw on our own previous experience and look at precedents. The idea was to provide a services back bone which would give the desired flexibility for an exhibition theme which was still very much in the melting pot. Naturally this flexibility would also have to have the capacity for the likely energy demands of the future exhibitions, all of which would be unique. At the time, Imagination were leading the team, and it was with their extensive knowledge of past and present exhibitions, as well as a lot of research into utility loads for existing exhibitions that we established the following energy demands: w> Power supply Cooling demand Heating demand 35MW 18MW 2.5MW for the Dome air intake systems.

320m

How these were delivered is addressed later in the paper.

ENERGY BALANCE
Fig IB The Problem of Scale

Solar Reduction

Initially, one of the design team's primary concerns was the environmental implication of putting such high heat loads together with 35,000 visitors under a transparent roof. What were the environmental conditions likely to be experienced by the visitors? What would visitors expect? Would conditions be acceptable? The roof is a double skin fabric, which allows 12% light transmission and has a shading coefficient of 0.08. Externally, the fabric is highly reflective (white), whilst internally the fabric is a white "matt" finish. Initially, the architects and cost consultants preference was for a single skin structure but this had to be rejected environmentally, due to the need for increased solar protection of the double layer and, just as important, the need to provide some thermal insulative properties for the winter conditions. The convincing argument was the much-reduced risk of condensation. The inner fabric liner - which is not structurally taut as the outer layer - also assists in "softening" the enclosure's acoustic characteristics. The following energy balance diagram was first used to illustrate the problem and was the first simple step in our environmental analysis of the Domes environment. One of the underlying design objectives for the "umbrella" environment was that it should use the niinimum amount of energy to provide the transient environment within which the exhibition would operate. Another advantage of the "umbrella" is that it allowed the construction of the exhibition and core buildings under cover, free from the extremes of the British climate. The downside of this latter point was that construction dust etc was trapped which eventually stained the inner liner.

The Creation of a M e s o E n v i r o n m e n t under the Dome roof from w h i c h other structures (core & exhibition buildings) c a n s p a w n . FiglC

design and passive control systems: protection from solar radiation in hot weather protection from precipitation in wet weather natural ventilation in hot weather wind protection in cold weather a smoothing of temperature or humidity

mechanbal control systems: fresh air ventilation air movement heating in winter comfort cooling in the core accommodation and exhibits

Fig ID

Under the design conditions assumed for ventilation, occupancy and internal gains, there is a low risk of surface condensation on the inner surface of the roof. Whilst some condensation is likely it should not be sufficient to cause drips and will quickly evaporate as conditions improve. Condensation risk in the Dome is however very sensitive to the ventilation of the space. If ventilation rates do not reach those assumed, particularly in winter, there is a very high risk of severe condensation on the inner skin. As occupancy moisture builds up. The simulations described in this report indicate that ventilation rates should be greater than 0. 3 air changes per hour. Thus natural ventilation alone may not be sufficient. Condensation risk may be significantly reduced by continuous overnight ventilation, even in winter.

Fig 2 Energy balance diagram

CONDENSATION
The thermally lightweight, low insulation, high occupancy characteristics of the Dome did give the design team some concerns regarding the risk of condensation. The arguments for the inner liner were almost entirely based upon environmental issues of which condensation was one. The others being solar protection, heat insulation in winter and acoustic absorption. Buro Happold's TAS analysis of the volume indicted that with the installed mechanical ventilation systems operating it was possible to limit the build up of condensation. We looked at a number of operating scenarios which indicated to us that the greatest risk of condensation on the internal skin was late in the evening on a cold winters day with a reasonable attendance. We calculated that the build up of moisture to be in the order of 30g/m2, this equates to a very thin wetted surface less than 1mm thick. Condensation dropping is therefor unlikely. To validate our work we commissioned "The Centre for Research in the Built Environment" at Cardiff University to carry out an independent check. This study made the following observations;

Condensation risk may be reduced by reduced by increasing the inner surface temperature. Condensation risk will be increased by introducing internal water features, planting and other water vapour producing processes. There is also the risk of condensation on the inner skin of the outer surface this could lead to staining or mould growth.

VENTILATION STRATEGIES
Initial thoughts were for a totally naturally ventilated building, similar to the railway station environment referred to above, but this alone would never be sufficient because of the scale and usage of the enclosure.

M M

This picture was taken from Building Services Journal April 1899

Fig 3 Ventilalion systems cross section.

140

Natural ventilation could not penetrate the depth of the dome (particularly so in summer) and entering fresh air would tend to rise a short distance in from the perimeter as it picked up the internal heat gains. Secondly the flow restrictions imposed by the large scale perimeter exhibition buildings would prevent air reaching the centre. Next, the team considered the use of underground air ducts supplying a large displacement system. However, the desire to reduce the amount of ground excavation to the absolute minimum due to the costs of excavation in contaminated ground, made this uneconomic. Further, such a major displacement system imposed on the plan at such an early stage of the design process could impede the future placement of the exhibition structures, so this solution was also rejected. The adopted ventilation strategy relies upon the perimeter zone being naturally ventilated via open doors, a permanently open strip at the top of the perimeter wall and the natural leakage of the structure itself. To move air into the centre of the Dome, two 25m /s air handling units are located in each of the six core buildings providing 300m /s in total. These air systems have a modicum of heating, equivalent to the Building Regulations uninsulated structure, which allows 25W/m of heating input. This input is just sufficient to take the chill off the incoming air. The same applies in summer when again, a modicum of cooling is added primarily to assist the air in dropping into the occupied central zones. There is no attempt made to control the Dome environment as a whole. It will be a few degrees warmer than the external environment in both winter and summer. The same large air handling units have variable speed drives and can operate in full re-circulation mode. The full re-circulation option is used during shut down" and rehearsal hours during the winter months.
3 3 2

DESIGN VERIFICATION
TWO DIMENSIONAL MODEL To verify the team's proposal, a three dimensional 360 CFD model was developed with AEA Technology in Didcot, Oxfordshire, acting as a sub-consultant to Buro Happold. The model went through a number of refinements as the information on the exhibition structures began to filter through from the exhibition design teams. As it stands today, the model has been generated by 700,000 cells, takes 450 MB of memory, and to run one scenario on AEA's most powerful machine takes approximately four days. Outputs are air speed, air temperature, resultant temperature and a "Comfort Index". As stated earlier, the question posed by our client was what would the internal environment be like and to what is it comparable. Buro Happold set about trying to establish comfort criteria for the space. Fangers or Bedfords comfort criteria were not considered appropriate as they related primarily to "static" office environments. Instead criteria established by United States Department of Transportation called the "Relative Warmth Index" (RWI) was adopted. This was developed for subways and train stations, so it was felt to be the most relevant and appropriate for the Dome environment.

RWI = M(Icw +Ia) + 1.13(t - 95) +RIa 74.2 where: M lew Ia t t-95 = metabolic rate = insulation effect of clothing, clo. = insulation effect of air boundary, clo. = dry bulb temperature = difference between dry bulb and average skin temperature. = mean incident radiant temperature from surrounding surfaces.

The following diagrams illustrate the applied layers of ventilation.

The above are all imperial units.

Explaining our results and "comfort" to our lay client was not an easy task, so the following diagrams were used to illustrate our results.

141

Fig 4 CFD plots with comfort criteria.

Its worth noting, that as the model became more accurate, and our Client became more knowledgeable, we reverted back to simply stating Resultant temperatures as our measure of the Domes environment. The following diagrams illustrate some of the CFD model outputs.

Hot S u m m e r Day

Fig 5a,b & c CFD plots

Ventilation tests to date on the installed systems (Andrew Cripps paper)

142

3010

302.5

304.0 305.5 Temperature (K)

307.0

The following figure gives the current break down of areas:

SPATIAL BREAKDOWN OF AREAS WITHIN THE DOME m? Exhibition Area 35,000 Central Show Arena 14,750 Catering 3,900 Circulation 11,350 Retail 1,350 Toilets, plant, support 10,300

ENERGY DEMANDS
PREDICTING THE LOAD Formulating the Dome's energy demand twelve months in advance of knowing what was going to happen within the building, came to down to guesswork, albeit educated guesswork. We talked to a number of organisations who had done "something" like it before, even if not on the same scale, we did a lot of reading and research into major exhibitions throughout the world, including asking the Disney Corporation, whose advise was particularly comforting - "Get your exhibition designs first before establishing your energy loads". The result is we have an all-electric building, this decision taken against a brief for a temporary exhibition (and at the time of the decision, also a temporary ^ building). Other energy supply methods were reviewed, including Combined Heat and Power (CHP). The CHP scheme was to be part of the total Greenwich Peninsula development, initially be used to serve the Dome (this being first load on-line) before commercial and domestic loads came on line in the future. Time, cost and lack of funding saw this proposal stranded. Gas heating was excluded because of programme and the cost of reinforcing the mains for the given demand. At the Dome, gas is only used for catering. At first sight the use of electricity as the Dome's sole energy source is questionable, but when viewed against the Clients programme, costs, the temporary exhibition brief and the post exhibition legacy (an electrical infrastructure in place for the future development of the Peninsula), electricity proved the most attractive option.

COOLING LOAD The 18MW of cooling is split between the following functions as follows: Exhibition Structures Central Arena Baby Dome Core Buildings Dome air supply External Buildings Spare (April 1999) Total 5.5MW 3.0MW 1.5MW 2.5MW 1.5MW 1.5MW 2.5MW 18.0MW

ELECTRICAL LOAD The installed capacity of electrical power for the landlords' supplies, at 57.5MVA, is only slightly higher than our original estimate. In providing this demand, the team set about standardising the size of transformers, the set up being: 2x 1.25MVA transformers for each of the six internal core buildings. l x l . 2 5 M V A transformer per exhibition 4x 1.25MVA for the central area buildings 4x 2.50MVA for the central show 9x 1.25MVA for the external buildings and landscape including "Baby Dome"

PLANT SELECTION This standardisation of the primary M&E equipment was an early design objective set by the Client and the design team, based on the directive of an exhibition lifetime of two years, and the short design and build times available. This directive was fundamental in the selection of services plant. Tried and tested technology was used, albeit on a very large scale. Standard off-the-shelf plant was selected and positioned around the site.

143

The London International Financial Futures and Options Exchange (LIFFE) was consulted by Buro Happold to evaluate resale values of plant on the futures exchange, and items were selected on this basis. The type of transformer specified, for example, was changed as a result of this input as was the rated output of the packaged air-cooled chillers. Equipment is standardised throughout in order to increase the likelihood of resale and minimise downtime if repairs are required.

As the siting of exhibits and public services was devised during construction, it was necessary to ensure that services would be available throughout the site when required. Despite the uncertainties, the entire M&E services were designed in only nine months and their installation is now complete and commissioning started in March 1999

SERVICING STRATEGY
FUTURE FLEXIBILITY At the beginning, it seemed that we were faced with a seemingly impossible task. The site was a barren and formless landscape, with little, if any, infrastructure, and there was increasing public debate as to what to put in it and if it should be built at all. But with an immovable completion date looming, design and construction work on the Millennium Experience had to start. Faced with the uncertainty on the exhibition form and content, the design team had to make some fundamental decisions on how the services should be planned so that the design and construction could progress well in advance of any exhibition designers been appointed. There was much discussion and debate on the servicing strategy and its intended flexibility as it was soon realised this would have a major influence on the exhibition layout and size. A number of scenarios were tested, including services within raised floors or above ground service beams. The adopted solution takes a very pragmatic approach. The dome is split into six equal 'pie' segments, each a mirror image of the other, the core building being the 'heart' of each segment, with plant in a pair of cylinders feeding into each segment. External plant such as the air cooled chillers, HV switchgear, standby generators and water tanks are contained in the twelve prominent service pods, or cylinders, around the perimeter, these operate in pairs to service each of the six segments. All segments have equal capacities although each pair of pods holds slightly different plant. From the cores, the services are distributed into a series of six radial trenches, each six meters wide and 900mm deep, and three circumferential trenches which run under the Dome's ground slab and carry all cable and piped services, including drainage. The radial trenches are generally arranged so an exhibition lies on either side with an access route directly overhead. Two major exhibits are serviced from each, with any excess continuing around the circumferential trenches to pick up any secondary loads. The radial trenches continue into the centre of the Dome to supply the demands of the central arena.
Fig 6 Services Strategy

Fig 7 Early concept diagram for the service pods

THE EXTERNAL SERVICE PODS


Included in all views of the exhibition since they first entered the public domain, the cylindrical pods surrounding the Dome are now entrenched in the minds of everyone who has seen any photographs or models of the structure. Housed within these aluminium-finned cylinders are all the primary services for the Millennium Experience. Originally intended to be spherical, creating a futuristic space-station look to the structure, they were to hold part of the exhibition. However, space constraints meant that the plant had to be moved outside of the Dome. The team liked the idea of using the now defunct exhibition spheres. However, it became increasingly difficult to fit square plant into a round space, so the spheres are now cylindrical.

144

ACKNOWLEDGEMENTS
Client:
The New Millennium Experience Company Jennie Page, David Trench, Richard Coffey, Peter English

Architect:
Richard Rogers Partnership Richard Rogers, Mike Davies, Andrew Morris, Stuart Steve Martin, Adrian Williams, Mike Elkan, Laurie Fig 8 Architectural image Forbes, Abbott.

Construction Manager:
MacAlpine Laing Joint Venture Bernard Ainsworth, Gary Nash.

Our sub-consultants:
Central Area: Cundall Johnston and Partners Ric Carr, Peter O'Halloran, Mike Golding.

Lighting Designers:
Speirs and Major Jonathan Spiers, Mark Major

Lift Consultants:
Fig 9 Dunbar + Boardman Peter Boardman, Chris Meering

Each pod is split into three levels, plant not requiring weather protection was left open to the elements, the remainder was enclosed in packaged plantrooms assembled off site by GEL and AC Engineering and installed complete. The contents of each pod varies due to the way the required capacities have been apportioned. Where plant is not required, the space is left empty for future expansion of exhibition demands.

and most of all, to all my numerous colleagues at Buro Happold who have contributed to this project.

REFERENCES
1 Constructing the Millennium Dome. Ian Liddell, Lecture to the RA, September 1997 2 The Design and construction of the Millennium Dome. Ian Liddell and Peter Miller The Structural Engineer, 6 April999

Two 750kW air cooled chillers are located on the top level of each pod. These are connected in parallel and grouped onto a common header. A chiller system (i.e. a 'pie' segment) consists of four units, the pods working as a pair, giving six systems in total to service the Dome and site. All systems run totally autonomous from each other. On floor two of the pods, one of each pair holds a prefabricated packaged plantroom, which contain closecoupled end-suction chilled water pumps to give a flow of 132 1/s at a head of 200kPa to a common pipe system which distributes around the Dome to null headers. A pressurisation unit and a controls system is also included in this plantroom. In eight pods London Electricity packaged HV switchrooms are sited at this level. On the bottom level, a 92.5m sprinkler water tank is sited in two pods, the volume required for the Category 3 Special system being too large for a single pod. Separate pump rooms have been installed as the tanks are more than 30m apart. A 500kW standby generator has been installed in three pods, each generator servicing two sectors.
3

Servicing the Dome Various, The Building Services Journal, April 1999

Subway Environmental Design Handbook Vol 1. U.S Dept. of Transportation

145

FIRE SAFETY ENGINEERING STRATEGY FOR THE MILLENNIUM DOME


Martin J Kealy BSc(Hons), CEng, MIFireE, MCIBSE, MIFS. Director of Fire Engineering, Buro Happold Consulting Engineers.

ABSTRACT
The Dome is designed by the Richard Rogers Partnership in conjunction with designers Imagination and consulting engineers Buro Happold. FEDRA the fire safety engineering part of Buro Happold were responsible for the design of the fire safety strategy. This paper looks at the fire safety engineering methods used in practice on the Millennium Dome in Greenwich, and examines the modelling techniques used for the means of escape design, smoke management system design and structural fire protection. It describes the benefits of using a fire safety engineering approach and how such an approach can be applied to other buildings.

1.0 T H E M I L L E N N I U M D O M E APPROACH
FEDRA first met with Greenwich Building Control and London Fire Brigade in March 1997 to present the concept fire strategy. It was agreed that a fire safety engineering approach was necessary to satisfy the functional requirements of the Building Regulations [1]. The dome can be viewed as the centre a small town with core buildings, exhibition pavilions and retail areas treated similarly to how buildings in the open would be, this approach is possible due to the considerable scale and volume of the enclosure. Working with Greenwich and the London Fire Brigade FEDRA produced an integrated design guide, a "mini Building Regulations", for the dome that takes into account means of escape, fire load control and fire fighting. Controlling the fire load of the Dome content allowed the shell and core to be designed and built long before details of the dome contents were available. Fire Brigade vehicles are able to gain access in to the dome space. The dome is provided with a wet fire main with internal fire hydrants and the internal roads and clearances allow for the largest fire brigade appliance. It is important to ensure that the open dome environment is tenable during the escape period. The volume of the space, the fire loading and the escape regime are factors

in the design of the smoke management systems. The escape period is estimated using escape models and the available safe egress time ASET is estimated using smoke modelling, both are explained in more detail below.

2.0 M E A N S O F E S C A P E 2.1 Overview


The dome will contain up to 37,000 people at peak times in a vast space. The standard travel distances applied to buildings are greatly exceeded and there is also a need to avoid evacuating large numbers of people on a false alarm or for a small controlled fire remote from the main population. For these reasons a zoned and phased evacuation regime is an essential element of the design and the detection and voice alarm systems are designed accordingly.

2.2 Simple model


The normal methods of means of escape provision would suggest setting a maximum travel distance and an escape width based on a notional escape time. The Dome is 320m in diameter so the recommended travel distance limits in Approved Document B [2] were exceeded. However, the building code criteria would only require a total exit width of 185m, the Dome has over 300m of exit width on the 1km long external wall. Issues such as local population densities, effectiveness of exits, pre-movement times and management / stewarding arrangements are not included in this simple model.

146

The simple model is the first step to gain a degree of confidence in the design and further analysis will provide greater confidence.

2.3 Intermediate model


FEDRA produced a more sophisticated model that took into account the components of the total evacuation time. These components include the following: Phased evacuation delays Detection time Pre movement time Travel time along stairs, ramps and open areas Queuing time

The Exodus model is able to randomly apply attributes to each person or dot. Such attributes include; drive, age, speed, dexterity and patience and standard occupancy profiles are available for example the general public. Exodus models people/building and people/people interactions in a more realistic way than the previous model. The Exodus model has to be run a number of times for the same layout, as with a real evacuation each event will be slightly different from the last due to the random nature of escape. A sensitivity analysis is therefore performed on each run to get a distribution of escape times. As the results are graphic it is easy to see where conflicts and bottlenecks occur. For example in the design of the central show "drum". It was clear from the playback that queuing was occurring around an intermediate exit, by making a small alteration to the geometry of the escape route overall escape times were greatly reduced. The other benefit of a graphic display is that errors in the model are very easy to spot. If people appear to be escaping in the wrong direction or re-circulating in a room it is very apparent and each person can be interrogated on screen giving details such as distance traveled, pre-movement time and escape path taken to reach that point. With Exodus the level of confidence is increased. It is intended that the model is tested by evacuation drills before the dome opens to the public.

The above times were calculated on a spreadsheet using the fire safety engineering principles given in DD240 [3] [4] and CIBSE Guide E [5]. A number of assumptions were made which are variable by their nature. The most significant time element in a non fire affected area is the phasing delay time or grace period and the pre-movement time. In an area directly affected by fire these times will reduce and the travel and queuing times dominate the escape period. The result is a range of different escape times that vary with location of fire, population loading of the area, design of exhibitions and timing of shows etc. The worst case escape times were then examined and alterations were made to the design to reduce potential bottlenecks. For example siting pavilion exits close to Dome exits that were not used to full effect and ensuring clear routes are made available from the core buildings to the dome exits. Escape from exhibition and other enclosed spaces are planned as they would be in the open air. Exits widths and maximum travel distances are taken from the published guidance and the place of safety in these cases is the Dome enclosure not external air. The intermediate model provided a much greater degree of confidence and also enabled practical design measures to be implemented to reduce evacuation times.

12 No. - Mast Fans 500mVs Natural Ventilators 400m


2

2.4 Building EXODUS Evacuation Model


The next step in gaining greater confidence was to use the Building EXODUS evacuation model produced by the University of Greenwich. Following a detailed assessment, it was agreed that EXODUS was a suitable form of escape analysis to use for the dome and an essential element given the complexity and size of the space. The benefit of the EXODUS model is that the geometry of the space can be simply imported using a standard CAD file. Spaces can then be filed with the required number of "people". The model also has an effective visualisation method that allows the escape event to unfold on screen each person is uniquely represented as a discrete coloured dot.

Inlet air supply - Doors and Fixed Openings 500m

3.0 S M O K E M O D E L L I N G
Smoke modelling is used to determine the available safe egress time (ASET). One of the main inputs into the smoke model is the fire size. In order to limit the potential fire size a combination of the following measures were applied: 1. The choice of construction material restricted to class O as far as possible.

147

2. 3. 4.

Sprinklers are an option. Fire rated enclosures are an option. Fire breaks (spacing fire loads in discrete locations) are an option

5 minutes 10 minutes 15 minutes 2 0 minutes

Smoke layer positions

The exhibition designers are used to working with treated and Class O materials and the above measures are not unduly restrictive. 3.1 Simple Model It is accepted that for large spaces with curved roofs the need for smoke extract systems is generally not required. The volume of the roof space forms a smoke reservoir that will take some time to fill and it is assumed that escape will be complete before conditions are untenable. In addition fires in the open grow slowly, compared to the same fire load in a small enclosure with a ceiling.

Fast fire growing to Plume Theory Zone Model

2W 0 M

3.2 Zone Model


FEDRA produced a zone model based on the guidance given in CIBSE Guide E. The model was produced on a spreadsheet that allowed for the following input data: Fire growth rate Smoke extract capacity Geometry of smoke reservoir Smoke yield

the model was very large Buro Happold ran it on the AEA Harwell super computer. The first step in modelling the fire case is to start at a converged environmental solution, this is important as there will be powerful air flows present in the space produced by air handling units, stack effect and lighting towers. A fire is inserted into the model using the same inputs as for the zone model. The fire case took approximately six weeks to run compared to one week for the environmental case start point. The CFD model inputs are as follows:

The model provided the following outputs against time: Layer depth Visibility in the layer Smoke temperature Fire growth rate Non-fire heat gains such as electrical loads, solar gain and people 12 Smoke extract fans extracting 500m /s 400m of roof louvres in the dome cap. 500m of make up air supplied by door openings at the Dome perimeter. Geometry of smoke reservoir and geometry of dome content External wind pressures (from wind tunnel data) Solar gain Smoke yield
3 2 2

The model was quick to run once it was set up and allowed sensitivity analysis to be carried out quickly and easily. The draw back of the zone model was that it took no account of the environmental flows in the building. The equations used in the zone model were stretched beyond the limits of validity. With this in mind a series of comparisons were made assuming a notional smoke reservoir that ranged from a 30 degree segment of the dome to a full 360 degree volume. Comparisons were made and it could be seen that the assumption of simple model held true with large factors of safety. The fact remained that the model was outside the limits of validity and so a higher level of confidence was gained by the use of Computational Fluid Dynamic (CFD) analysis

The results of the CFD are graphical the most useful output for visualising the extent of the smoke layer is to plot an iso-surface of smoke visibility, a 50m visibility iso-surface was chosen for this analysis. The major findings were: Smoke remained at high level for the duration of the model (30 minutes) A steady state condition was achieved after 20 minutes. The smoke at high level was dilute. The smoke was cool away from the fire plume. The model showed stratification and plug holing effects at high level but no downward movement of smoke into public areas. The amount of smoky gases extracted by the natural vents was four times that extracted by the smoke fans. The effects of AHU's and dome environmental

3.3 Computational Fluid Dynamic (CFD) analysis


CFD is a method by which the fluid (in this case air) is broken down in to small cells and the fundamental laws of thermodynamics are solved from first principles. The method can only be calculated on a powerful computer due to the volume of iterative calculations performed. As

148

flow dissipated in the first few minutes. The 6 lighting towers (high heat load) had a surprisingly large effect on smoke flow patterns at high level.

REFERENCES
1. 2. 3. Building Regulations 1991'(London: HMSO/Department of the Environment/Welsh office) (1991) Approved Document B: Fire Safety (London: HMSO/Department of the Environment/Welsh office) (1992) Draft for Development. Fire safety engineering in buildings. DD240:Partl:1997. Guide to the application of fire safety engineering principles. British Standards Institute (1997) Draft for Development. Fire safety engineering in buildings. DD240:Part2:1997. Commentary on the equations given in Part 1. British Standards Institute (1997) CIBSE Guide E Fire Engineering (London: Chartered Institute of Building Services Engineers) (1997).

The CFD model confirmed the results of the zone model and showed that the zone model was on the conservative side. The smoke was more dilute and cooler than the zone model predicted. What the CFD model did show was the position of the smoke and flow directions and that smoke did not drop to low level.

4.

5.

Environmental CFD Modelling

4.0 S U M M A R Y
The benefits of using a Fire Safety alternative approach in the dome were: Engineering*

Improve the quality of design Allows greater freedom of architectural design Allows flexibility for the client Minimise costs Escape tunnels and channeling screens reservoir curtains are avoided Environmental systems can be used in a fire mode Combined sound system and voice alarms Sprinkler protect areas of risk only Detection applied to areas of risk only Modelling techniques Complex analysis such as Exodus and CFD provide an increase in design confidence. The drawback of these types of analysis is that they are expensive, as they are time consuming and require expert users. Zone models and hand escape calculations are inexpensive and easy to use and areas of uncertainty can be identified early in the design. The complex modelling resources Can then be allocated more effectively speeding up the design process and therefore keeping design costs down.

149

LARGE ENVIRONMENTAL ENCLOSURES, THE ROOF OF THE MILLENNIUM DOME


Ian Liddell CBE, FREng, MA, DIC, FIStructE, MICE. Partner of Buro Happold and Visiting Professor of Engineering Design, University of Cambridge

ABSTRACT
This paper describes how ideas for enclosing very large areas have been around for some time though without being brought to completion. The Millennium dome is the first structure to be in this category. The paper goes on to describe the engineering and construction of the Dome.

INTRODUCTION
There is a long history of dreams of creating large enclosures to ameliorate the climate in inhospitable parts of the world. This trend started in the 18 century when the owners of fashionable country houses built heated glazed enclosures to grow Pineapples and grapes. These stoves or conservatories were generally built of stone with the glass in conventional wooden frames. As the glass making technology improved and iron working techniques developed the greenhouses became larger. All iron glass houses appeared at the start of the 19 century. There is a classic example remaining at Bicton in Devon. This structure is more like a shell with small panes of glass set in fine wrought iron ribs. The classic examples of this type of construction were the Palm house at Kew built in 1848 and of course the Crystal Palace for the 1851 exhibition. The Crystal Palace was a relatively simple modular construction based on an 8 foot module for the floors, walls and roof and a 24 foot grid for the columns and introduced the idea of factory construction.
th th

Iron and glass structures were used for the great 19 century railway stations and increasingly for winter gardens or pleasure palaces. The largest span was 60m at St Pancras station. In the 1950s Buckminster Fuller was working on developing larger and larger geodesic domes. Climatron at St Louis was based on his ideas. One of his futuristic ideas was to build a big geodesic dome over Manhattan. To quote his reasoning at that time. "The way the consumption curves are going in many of our big cities it is clear that we are running out of energy. Therefor it is important for our government to know if there are better ways of enclosing space in terms of material, time, and energy. If there are better ways society needs to know them. Domed cities can be illuminated by daylight without direct sunlight. That part of the dome through which the sun does not shine directly would be transparent. In summer the dome would be protected by polarised glass; during the sunny hours it would not hold heat but in winter the sun would penetrate all the dome. The atmosphere will be dust free. Controlling the environment through domes offers the enormous advantages of the extroversion of privacy and the introversion of the community" (Reference 1).

th

Fig 1 A Victorian Winter Garden

Iron and glass remained the preferred materials for large environmental enclosures and are still frequently used today. Today the glass technology has developed with large panels of toughened and laminated glass supported on ever more daring steel structures but the spans and the scale of the enclosures has not increased significantly. In thel950s new translucent polymers and plastics became available. One development in particular was coated fabrics which offered new freedoms in form and span for large enclosures. Starting in the mid 1940s Walter Bird developed air inflated structures initially for radomes but later for tennis halls and large sports halls. The main translucent fabric materials developed at this time were PVC coated polyester and later in the 1970s PTFE coated glass fibre cloth. The big breakthrough for very large covered areas was Walter Bird's low profile* cable dome (Reference 2).

Fig 4

WUS Pavilion at Osaka

In 1970 Frei Otto and his colleagues at the IL put forward a scheme for a covered city in the arctic. The covering was to be an air-supported fabric structure 2km in diameter with an area of 3,000,000 m . (Ref 3) The primary structure was to be a net of "Trevira" Polyester ropes.
2

Fig 5 Walter Bird's Cable Dome for spans greater than 300m

I helped Peter Rice with some calculations of the forces under wind load. The effects of snow were happily glossed over with the assumption that because of the smooth shape of the roof the wind flow conditions would be close to potential flow where the wind would sweep the roof clear of snow. The provision of building services and the management of an enclosure of this scale were not considered in great detail at that time. Ten years later in 1980 we in Buro Happold had the opportunity to undertake a feasibility study for covering a town in Northern Alberta. The leader of this design team was a Canadian architect called Arne Fullerton and again we worked with Frei Otto. The study was exceptionally interesting in that it included the human response of living in such a space as well as the servicing requirements and other considerations such as the impact of fire. One of the designs for this enclosure was a 150,000m air-supported roof that was to have steel strand cables at 0.5m spacing and would use ETFE foil cladding. This design was taken to concept stage and was supported by calculations and reports. This time the snow loading was not glossed over but we did not know how the building would respond to the extreme snow falls for the area.
2

Fig 3 Walter Bird's Cable Dome for spans greater than 300m

This was adapted by David Geiger for the US pavilion for the Osaka expo in 1970. The principle was then used in a reduced cost form for several large football stadia. Unfortunately these structures had large valleys along the cable lines which collected snow and caused local ponding occasionally leading to loss of pressure and deflation to a stable down-hanging position. These deflations caused unacceptable co-lateral damage to the fabric leading to the abandonment of the form. At the time the air supported structure seemed to point the way towards the city scale environmental enclosure envisioned by Buckminster Fuller and concepts were put forward by Walter Bird.

Fig 6 58N 18 Ha Air Supported Structures

Fig 8 Eastleigh Tennis Centre white ETFE foil cushions on cables

Experience with large air-supported roofs in North America had demonstrated how snow on such roofs could initiate ponding and cause severe maintenance problems. (Ref 4). Our roof would have had a much lower rise to the foil cladding and hence not such deep valleys to initiate snow drifting. The cushions would also have had higher pressures to support the snow. Even so there would have been a considerable risk of problems with snow.

PRINCIPLES OF THE D O M E ROOF STRUCTURE


The structural concept for the roof is based on the innovatory principle of using straight tensioned cables and flat fabric for the structure rather than adopt doubly curved surfaces which had become the accepted form for such structures. In 1994 Buro Happold pointed out that there was considerable advantage in using straight tensioned cables which could carry both the uplift and down loads with resultant forces in the same direction. (Ref 5) The pretension stiffens the cables against deflection allowing high strength tensile materials to be used to create very large spans. Flexible fabric or foil can be used as cladding between the cables provided it is pre-stressed in the same way as the cables. The advantages of the arrangement are: Compared with a two-way cable net one set of cables is eliminated along with the cross clamps and terminations. Whether the load is upward or downward the cable tensions are in the same direction which can be a great advantage if the tensions are resisted by a funicular arch or ring beam. Connections to the foil or fabric cladding can be greatly simplified. It was claimed that taken together these benefits would result in very economical large span roof structures. However the stressed fabric resists local loads by relatively large deflections rather than by simply increases in stresses. Concentrated snow drifts could possibly create a deflection which would be so great that water did not drain out from it. This would be potentially disastrous. The key to adopting this concept is to develop a form where there is adequate drainage to avoid ponding problems and details which will allow for the deflections. Wind stimulated dynamic oscillations are not a problem provided the fabric is fully tensioned as there is a high degree of damping from the fabric and the attached air.

Fig 7

Chelsea and Westminster Hospital Atrium, part clear and part tinted ETFE for cushions

There was no experience with the use of ETFE foil as a cladding material at that time. 10 years later in 1990 we engineered a roof for the atrium of the Chelsea and Westminster hospital with this material. Subsequently we designed a roof for a tennis hall which consisted of foil cushions on a tensioned cable structure. Now 20 years on we have built the Dome on the Greenwich peninsular for the Millennium Experience which is to be held in the year 2000. This roof is 80,000 m and is the nearest structure yet to these dreams of covered urban environments. It will enable us to evaluate the performance of such a space.
2

152

Fig 9

RSSB Tent, the fabric and cables can be seen responding to wind from the right

Between the cables, tensioned coated fabric is used as cladding. Both the tensioned cables and cladding carry the loads by deflection accompanied by increase in tension. This concept is simple but there are dangers associated with the deflections particularly ponding caused by snow or heavy rain. When loaded by wind or snow, the upper hanger, the lower tie-down and the stringer cables carry the loads from the fabric down to the ground. The stringer cables are restrained at the perimeter by the perimeter masts and large boundary cables attached to 24 anchor points. The vertical components on the forces at these points are resisted by ground anchors grouted into the London clay and the horizontal forces are resisted by a compression ring beam under the external wall. Tension structures rely on the shape of the stressed surface for their performance under load. Forces are resisted by the tension and the curvature, the greater the curvature the less the tension required to resist a given load. Both the radial stringer cables and the fabric are prestressed with sufficient tension to stiffen them against imposed load deflections. The prestress levels and cable geometry were selected to provide adequate deflection control, with the materials and structural sizes selected to provide high stiffness.

In 1994 opportunities arose to utilise this concept on two structures, the Eastleigh tennis centre and a very large demountable tent of 20,000 m for RSSB. The latter proved to be extremely economical and met the owner's requirement of ease of installation.
2

THE DOME ROOF DESIGN


The structural concept for the Roof of the Millennium Dome is apparently very simple. 72 tensioned steel stringer cables in pairs of 032mm steel spiral strand are arranged radially on the surface. The stringers are supported at a radial spacing of between 25 and 30m by an arrangement of upper hanger and lower tie-down cables that are arranged around the 12 100m tall primary steelwork masts. Circumferential cables keep the stingers on their radial lines.

Fig 11

Ring Beam

Fig 10

Dome Primary Structure

The forces in the radial stringer cables are taken by a central 30mm diameter cable ring supported by forestay cables, which run through to the centre point, where they support a flying mast, which in turn supports additional radial cables which carry the cladding. The central area is formed by a cable truss connected to a 30m diameter cable ring. The level of stress within the cable ring leads to a stiff structural form.

In the event of a loss of a panel of fabric, it is important that some of the in plane prestress is maintained in the panels adjacent to prevent the deflections rising to unacceptable levels. During the normal operation of the structure, all fabric forces pass through the plane of the surface and are resisted by equal and opposite forces that arise in adjacent panels. In the event of a panel failure (or removal) this balance is upset and the radial cable connection nodes would be forced out of line. Circumferential cables through the nodes were required to maintain their spacing and resist these forces. The dome roof shape with tapering segments has advantage in resisting ponding in that the span of fabric panels increases as their slope increases so fabric surface gets progressively softer. However, If an the the the

Fig 13 30m dia Central Cable runs

STRUCTURAL DETAILING
Fig 12 Wishbones to raise circumference cables

circumferential cables were in the surface of the fabric they would cause a dam at each circumferential line so an arrangement was required which would take these cables out of the surface. This was achieved by raising the circumferential cables above the surface with rigid members (wishbones) and connecting them to the nodes with criss-cross cables. Lower circumferential cables were also required to control the tiedown cables; these were also spaced off the surface but with out the criss-cross cables. It was also necessary to control the deflection of the radial cables. Their length is very long, 150m from the perimeter to the centre. Because of this if one 25m span were loaded the remainder of the cable in the line would act as springs so the loaded span would not be as stiff as if it was fixed at each end. The only way to gain the necessary stiffness is to use a high pretension. In fact the planned pretension in each radial line is 400kN, about 2/3 of the peak tension. The last element in preventing ponding is the patterning and prestress in the fabric panels. During the tender period some development of the design continued. We decided to change the central node for a 30m diameter cable ring. This was constructed with 1248mm diameter cables. Because of the redundancy implicit in the 12 cables, failure of one of these cables would not compromise the overall safety of the roof. These changes were brought in to the contract package before the contract was finally placed.

With cable structures it is essential that the details respect the system lines and system points of the cables and their intersections, as well as the likely movements of the cables at the connections. If the radial cables were continuous through the node points the flexing at those 'points would cause the cables to fail prematurely in fatigue. At every hanger location, the radial stringer cables are connected together at a node detail. This detail allows the high radial forces to pass directly thorough into the adjacent cables and allows the hangers (both upper an lower) to be connected into position. The vertically oriented connection plates allow the radial cables to rotate on their end fittings as they deflect under load. The flat top plate stiffens the node against shear forces and provides a surface that the fabric can be clamped onto to form a weather seal.

DESIGN VERIFICATION
As is usual for major building structures the safety of the design was verified by calculations. These relied on using our "Tensyl" program for calculating the forces in the fabric and cable structure. A Vj2 model of the cable system was modelled using Tensyl, to investigate the effects of cable prestress and the environmental imposed loadings upon the structural system. The tensyl analysis was then expanded to a V model to allow us to predict the maximum cable and fabric forces, the compression loads in the supporting structures and the deflected shape of the 'total' system.
2

Analysis of the cable system has shown that the behaviour of the structure is very sensitive to cable stiffness. The radial stringer cables rely upon a high level

154

of prestress (about 400 kN) to ensure that deflections are controlled during each load case. Under load it is critical that the stretch in the mast hanger cables is minimised. Each cable has been sized not only for strength and ultimate load capacity, but also for axial stiffness in order to ensure that the cable system does not 'go soft' when under load. This is the reason why we have used spiral strand cable, a much stiffer product than standard IWRC wire rope. Wind loads were derived initially from published data. They were then confirmed by wind tunnel testing at the BMT wind tunnel at Teddington. The safety of the components was investigated following normal design rules. Resistance of the whole structure to accidental damage is provided by redundancy, i.e. the structure can tolerate the loss of an individual component without collapse. This principle also applies to the support pyramids which are designed to withstand the removal of a leg. The 90m long masts were constructed with 8 323mm diameter tubes braced with rings at 2.5m spacing. Their overall diameter was limited by transportation requirements and a great deal of computer calculation went into verifying their load capacity. The limiting load was calculated using LUSAS in a non linear mode. Since the masts are leaning deflections under self weight and icing have to be taken into account as well as initial out of straightness. Wind loading is also significant but this does not occur with the peak down loads from snow and icing.

The cable work was subcontracted by Watson to Bridon Ropes of Doncaster. The cables have to be wound from wires that have been previously drawn and galvanised. For the dome project, class A galvanising, the lightest, was specified for cables which were beneath the roof and Galfan, a mixture of aluminium and zinc galvanising which is much more durable, for the external cables. The cable has to be pre-stretched to eliminate the construction stretch and then marked to the correct lengths under the specified pre-stress load. Most of the cables are dead length without any provision for adjustment, consequently great care has to be taken to ensure that the cables are made up to exactly the right lengths The lifting of the masts was planned by Watson with great care. This involved selecting a suitable crane and devising lifting positions which would not overstress the masts. Each mast was lifted and guyed with the two permanent backstays and two temporary forestays. There is also an intermediate position while the crane was released when only one forestay could be used and a short term guy was added from the centre of the mast to the adjacent base. While the mast is guyed with the temporary forestays the central ring is lifted by the permanent forestays. During derigging of the crane and the operations of changing the guy positions the tensions in the guys had to be carefully controlled to maintain the stability of the mast. Following the lifting of the ring, the guy system was moved so that the rest of the cable net could be assembled and lifted to its place. This was done using hydraulic cable jacks with the hoisting cables running over sheaves on the top of the masts. When the net was completely assembled and all the cable lengths checked, each of the 72 pairs of radial cables had to be tensioned. This was achieved in several steps using a 550 kN capacity 'Enerpac' pull jack in the pre-designed jacking points at the front of the perimeter masts. Because of the flexibility of the central ring and the boundary cables the tensioning of the radial cables had to be done to specified dimensions rather than to specified loads with final adjustments made at the end.

Fig 14

Masts

SELECTION OF CLADDING AND THE INTERNAL ENVIRONMENT


The roof is to provide a controlled environment for the exhibition and for what other uses it may be put. The human response preference is for a bright translucent roof with a light spectrum as close as possible to daylight. This requirement conflicts with the needs for the central show and some of the exhibits for which lower light levels would allow greater impact from exhibition lighting. Coated fabrics tend to change the spectrum to a brownish hue rather like tungsten lighting. This of course affects the perception of colours within

STEELWORK CONSTRUCTION STAGE


The selected steel contractor, Watson Steel, were obliged to develop the engineers design drawings into shop drawings for the production of the components. This process involves an element of detail design of the components and connections. The shop drawings show the cutting and holing dimensions of all the plates as well as the welding and connection details. These drawings are reviewed by the engineer and architect for approval prior to the start of fabrication of each particular part.

155

Fig 16 Fig 15 Perimeter masts with jacking points between rigging screws

PTFE/Glass fabric

temperature of around 350C. A benefit of this is that the material can be repaired on site with a permanent seam that is the same as those done in the factory. With PVC/polyester the fibres are damaged by UV light and they burn so the function of the coating is to protect the fibres from UV light as well as providing the flame proofing. The PVC itself is light stable and does not burn well but it requires a number of other compounds such as pigments, UV stabilisers, plasticisers, fungicides and flame retardants to meet the functional requirements. Since 1987 several of these compounds especially fungicides and heavy metal stabilisers have been banned and this has led to an increase in problems of fungal growth in the yarns which severely discolours the cloth. This situation has recently been improved by the use of anti-wicking treatments to the yarns. The other big problem with PVC coatings has been dirt retention. The PVC coating is porous and the plasticisers absorb dirt. This has recently been improved by the use of fluoropolymer surface lacquers which give it a durable sealed surface. After investigating the products of the three best coaters in Europe an outer fabric was selected which gave 15% translucency and an inner lining fabric which gave 75%. The combination gave the highest translucency, about 12%, and a good colour rendering. The fabric selection was changed to PTFE/glass after a political decision to build the dome with a long life.

the dome and according to our researches for the 58 N project can affect the physical performance of people within the dome. It is difficult to have a translucent fabric roof with insulation but with out any insulation condensation will occur on the underside which, in certain conditions, will fall as rain. This situation would be totally unacceptable in a building that will have a lot of electrical displays. To reduce this risk a lining can he installed under the main fabric. There has been a considerable amount of experience with fabric roofs with linings where condensation has not been a problem. Checks were run on the risk of condensation as part of the environmental modelling and they demonstrated that with two membranes the risk of condensation on the underside was very low. The available materials for cladding the dome were PTFE coated glass fibre cloth, PVC coated polyester cloth or ETFE foil cushions. Our preferred material for the roof as an environmental enclosure would have been ETFE foil. This would have provided a high translucency roof with three layers of foil which would have a considerable amount of insulation effectively eliminating the risk of condensation. Unfortunately we did not consider that there was sufficient experience with detailing this material in this situation and we considered that it would be too risky to try to develop a suitable system within the very tight time scale. The necessary properties of durability and flame resistance are provided by PTFE/glass without the need for any additives. The glass fibres are not affected by UV light but they are damaged by water. The function of the PTFE coating is to protect the fibres from water and abrasion, the PTFE itself is nearly inert and is not affected by the weather. The fabric is seamed by heat sealing using a FEP interlayer which melts at a

FABRIC W O R K
The contractor who had made the best offer for the PTFE/glass material was Birdair from Buffalo, New York State. They have been producing structures in PTFE for over 20 years including some 12 covered stadiums of approximately half the area of the dome. The fabric patterning and attachment details had to be modified to accommodate this alternative material and since time had been lost in the programme, this had to be done in a very tight time scale. Because of the

156

arrangement of the panels within the cable net, and the fact that the cloths were to be fitted in to dead lengths, the patterns had to be extremely accurate. Since the warp direction of the panels of the outer fabric ran radially on the roof with 25m long cloths, it was necessary to model the fabric as an equal mesh net to represent the warp and fill lines of the cloth. This was a much more time consuming method than the standard method of representation using triangular elements. These basic geometry patterns were converted by Birdair into cutting patterns. They also built in the stretch compensations, which were agreed after biaxial tests on the actual production cloth and added in all the edge details.which were agreed after biaxial tests on the actual productioncloth and added in all the edge details.

TUNNEL VENT AREA


A 50m diameter hole was required in the roof around the Blackwall tunnel vents, to accommodate the 'air rights' of the ventilation structure. After considering a number of ways of leaving a hole in the fabric, Buro Happold adopted a net of 8mm cables at l m spacing which would replicate the stress-carrying capacity of the fabric but would allow the vent air to pass through. The cable net was attached to the fabric with clamp bars at the edges. The net patterns were developed directly from the typical fabric patterns with the boundary line being defined to align with the top of the enclosure. The net arrived on site in rolls and was erected in the same way as the fabric using the same extruded hooks modified with a steel plate to which the cable terminations were attached.

LEARNING FROM THE DOME


There are two questions; is a large environmental enclosure of 10 or 15 hectares feasible. If so would it be energy efficient and provide a improvement it living conditions in extreme climate zones. The zones would be either semi-arctic or desert requiring either raising or lowering the internal temperature.

Form
Fig 17

The fabric attachment detail proposed by Buro Happold , was a double luff groove extrusion fitted onto the radial cable pairs to accept a roped edge on the fabric. Birdair proposed a 12mm edge cable in the fabric which would hook into special clamps fixed to the cables. The clamps were developed into a two part extrusion cut into 50mm lengths and retained by two 12mm bolts. Fabric sealing flaps were closed over the top of the site joints and sealed together using a hot iron at 380C and an fep inter-layer

The selection of the form of the Dome as a shallow spherical cap is beneficial for wind and snow loading. The smooth profile generates a smooth airflow over the surface with of local turbulence. It also reduces the uplift pressures so reducing the tensions in the cables Snow will always be a problem on transparent or translucent roofs If not from the load effects then because it excludes the light. In windy conditions the snow will mostly blow off. Dividing the roof surface up into individual panels with valleys between has been shown to give trouble because of the concentration of snow which can build up there. The smooth Dome form avoids this, it also has the useful property of having increasing slope with increasing fabric span that reduces the risk of ponding. This form could be increased to say double the area without compromising the structural behaviour. Unfortunately we may have to wait several years to get a few snow storms to prove or disprove the behaviour of the Dome under snow loading.

Internal Air
The biggest problem with the internal environment has proved to be the dust. Firstly that generated by the construction operations which began immediately the roof was completed and continued for a year. We think that the ongoing running operation of the dome will create a dusty environment generated by the visitors

Fig 18

Fig 19

moving around and by the cleaning operations with dry brushing. There is also the dust and dirt coming in with the external air in the polluted environment. The solution would be to hose the surfaces down. This is the method adopted in large covered stadia but of course it causes condensation on a single skin roof. The special qualities of fresh air are not well understood but dust and pollutants are known to make it unpleasant. In a covered environment a plentiful supply of outside air is normally required to keep it smelling sweet. If the temperature outside is sub zero any cold air introduced will flow across the ground floor tending to defeat the object of the covering. It would be interesting to investigate to what extent fountains and "rain" would clean the air so that air changes can be reduced.

Because the covered ground area is so large little heat will be lost to the outside although it might be absorbed into the ground at a low temperature. Heat exchange to the ground will change its temperature very slowly and the deep ground temperature of 8 or 10 deg C will provide limits.

Cooling
In desert conditions the enclosure would need to provide shading. This can be provided with a single skin of Teflon/glass fabric. The problems of condensation would be very much less than in cold climates. Cooling at ground level is easier to achieve since the cold air tends to stay on the ground. To gain a benefit from the enclosure it would be necessary for the ground to act as a coolth sink. Would this happen by natural means or will mechanical cooling systems be required?

Heating
In winter the warm air immediately migrates to the top of the space until it looses heat so heating the air is not very energy effective. A better way is to heat the ground to improve the local comfort and let the air look after its self. This is more or less the approach adopted in Victorian stoves where the heat was introduced via pipes in ground trenches. Ground heating can be done with low grade heat supplimented on sunny days by radiant heat from the sun and heat from internal buildings.

Cladding
In the case of the Dome the dust has made the lining fabric unacceptably dirty. This is largely due to the particular material supplied by Chemfab. Called Fabrasorb It is marketed as a sound absorbent material because of its porosity. In reality it is glass fibre cloth barely coated with teflon and the coating is easily damaged by handling. The porosity improves the

Fig 20

translucency at first but the result is that the dirt gets into the fibres and cannot be removed. There are now more translucent PTFE/glass fabrics around, the problem is that the inner layer may not bleach out because of the lack of UV penetrating the outer skin. The porous lining will also allow water vapour to pass through and condense on the outer skin. The solution for a habitable enclosure has to be to treat the roof as for a swimming pool. This means that there should be sufficient insulation to prevent condensation. The best way to acheive this is to use three layer ETFE foil as was originally proposed for 58 deg North. The benefit of this approach is that a vapour barrier would not be required on the ground so it could be treated is it is outside. Plants could be grown anywhere, water features could be used and the interior could be cleaned with a hose with water soaking into the ground and evaporating from it. The Dome and other large enclosures such as the Eden project will enable us to obtain data to evaluate the benefits of such structures and define the limits to the size. This may be determined by fresh air, roof drainage and access requirements as much as by the structure.

REFERENCES
1. R Buckminster Fuller and Robert Maries, The Dymaxion World of Buckminster Fuller, 1960, Anchor Books Edition, 1973. 2. Walter W Bird, The History of the Air Structures in the USA. IL16 Zette, publication of the Institute for Lightweight Structures, University of Stuttgart. 3. IL2, Publication of the Institut for Lightweight Structures. University of Stutgart, 1971 4. Liddell W I, 'Minnesota Metrodome' A study on the behaviour of air supported roofs under environmental loads. Structural Engineering Review 1994, Vol 6, No 3-4, pp.215-235, Pergamon. 5. C Gill, I Liddell, C Schwitter, Straight cables for tension structures. Procedings IABSE symposium, Birmingham 1994 6. Liddell W I, 'Creating the Dome' The 1997 Royal Academy of Engineering, Hinton Lecture published by the RAEng. 7. Liddell WI, Miller PW, 'The design and construction of the Millennium Dome', The Structural Engineer, Vol77, No7, 6 April 1999 8. Elizabeth Wilhide, The Millennium Dome Ted Smart 1999.

159

PRINCIPLES OF CONSTRUCTION FOR WIDE-SPAN STRUCTURES WITH EXAMPLES FROM THE MILLENNIUM DOME
Peter W Miller Watson Steel Limited

SYNOPOSIS
There are a number of key issues that need to be considered in planning the construction of any complicated structure and the Millennium Dome was no different. This paper gives a brief description of the construction method adapted for the Dome and then describes the major issues that had to be dealt with in the planning and construction of the Structural steelwork and cable net that forms the structural framework for the Dome. An explanation of the thought processes leading to the eventual construction scheme is given and it will be shown that many of the principles described here also apply to any wide-span structure

INTRODUCTION
In late 1996 Watson Steel were invited to submit a tender for the supply and construction of the steel and cable frame for the Millennium Dome. In order to begin to estimate the construction costs of such an unusual and large-scale structure a workable and economic construction scheme had to be developed. The construction scheme produced at that time, whilst just a series outline sketches, was fundamentally the same as that which was eventually used. A great amount of detailed development post contract award took place however and some of the attention to detail that was required will be demonstrated in the following pages.

it

Hi

Fig 1 A 90 metere mast is reared up prior to being placed on the pyramid

OUTLINE OF THE CONSTRUCTION SCHEME


Each of the 12 masts and the pyramids that support them were assembled and fully site welded in an area adjacent to the permanent site. The pyramids were carried to the site and were placed on the concrete piled foundations with a crawler crane. The masts were also carried with the same crane and laid out adjacent to their final position. Whilst there, they were fitted out with the temporary erection gear that was subsequently used for the cable pulling and also had the restraint cables attached, before being lifted into position with a large 1000 tonne capacity strut jib crane.

The masts were guyed off with four restraining cables before the crane was released. The front two of the cables were temporary but for the rear two, the permanent back stay cables, which ran directly to ground anchors could be utilised. In order to restrict the movement of the masthead under wind loads the four restraining cables were post tensioned to a predetermined force in the order of 200KN before the crane was released. This operation was further complicated by the fact that the mast head position in space was monitored and was required to be within 150mm of theoretical when the stressing was completed. The mast was mounted on a rubber bearing located on top of the pyramid which allowed this small amount of rotational movement.

160

When all cables had been installed they were tensioned to their final design stress by progressively jacking at the anchor points of the 72 pairs of radial cables. The stressing operation was carried out in a balanced manner around the dome in three stages. The erection of the masts commenced on 15 October 1997 and the stressing was completed by the end of March 1998. A period of only 16 working weeks.
th

PLANNING
The key to all successful site construction projects is detailed planning and the more complicated the project the more important detailed planning becomes. Inevitably wide-span projects tend to be unique and challenging where the need for lateral thinking combined with attention to detail becomes even more important. On the Dome an internal system for planning and developing the construction method was established within Watson's that proved to work very well. This was based on regular brainstorming sessions where the most experienced and practical engineers within our business with both fabrication and erection experience debated specific topics and put all the ideas on the table. The project team would go away and examine all the options and develop the ideas. The project team would then present their conclusions back to the gathered engineers for critique. Meetings were held weekly at the outset of the project when the construction scheme was formed and thereafter as and when required. When the options had been narrowed down to a few options they would all be costed out to determine which was the most economic.

Fig 2

These three sketches show the rearing procedure for the masts

The cable net, which consisted of over 2600 cables was assembled and lifted in four main sections. Each section formed a large concentric circle which when pulled simultaneously at 36 positions was elevated to its final height. The missing infill cables between the circular sections were installed individually using a combination of abseiling techniques for the higher locations and powered access equipment where practical.

Fig 4 A full scale trial was carried out on the first mast section to 'prove' the planned net pulling method

Fig 3 The second cable net being reared up with 36 pulling jacks.

161

Quite often practical trials were required to determine the best options. When the stage had been reached where the scheme was finalised on paper it was decided to carry out a full-scale trial. This was carried out in the Bolton factory using the first of the mast head sections that was fabricated in July 1997. This trial proved to be a good investment. The outcome of this planning / development stage was a detailed method statement which was developed gradually over several months.

PRE-ASSEMBLY
For any construction scheme to be successful it must be: a) b) c) d) Safe Economic Fit within programme constraints Comply with specification

One method commonly employed to ensure that these criteria are met is to pre-assemble as much of the structure at ground level as possible. In the case of the Dome this principle was applied to the pyramids, the main masts and to the erection of the cable net that forms the structural framework of the Dome and supports the fabric covering. The final decision on which elements and to what extent they should be pre-assembled depends on many factors including cost, availability of large cranes, programme, height above ground, and the alternative safe means of man access. In the following paragraphs the decisions that were taken on the pre-assembly of the principle elements of the Dome are described.

Fig 5 One of the completed pyramids being placed on to the foundations

PRE-ASSEMBLY OF MASTS
It was a fairly obvious solution to build the main 90 metre masts on the ground and to lift them with a large crane. The original cross sectional diameter of the masts however was greater than we could transport on the public roads by conventional trailers. This would mean that there were either very expensive transport costs or else only half the mast cross section was fabricated in the factory and the remaining fabrication completed on site. Both options were considered undesirable and following discussions with the Engineers it was agreed that the overall diameter of the masts could be reduced in size and the design compensated by increasing the wall thickness of the eight tubes that made up the octagonal cross section of the mast.

PRE-ASSEMBLY OF PYRAMIDS
The four-legged pyramids that support the 12 masts are over 8 metres wide and 10 metres high and therefore had to be constructed on site. The design forces and architectural requirements meant that site welding of the node joints was the only feasible option. The overall construction programme however could not be achieved if the construction of the pyramids was delayed until the permanent foundations were available. The solution was to pre-assembly the pyramids on a temporary foundation in a separate area away from the main construction zone. The completed pyramids were then stored and eventually carried to their permanent location, using a large crane, once the foundations were released. The advantage of pre-assembling all the twelve pyramids in a specific assembly area was that once the assembly jig was set up and checked it could be used twelve times and all the pyramids were sure to be identical and therefore interchangeable. The fact that a large crawler crane in excess of 200 tonnes nominal capacity was required was not a disadvantage because the crane was planned to be on site anyway throughout the period to assemble the mast sections.

Fig 6

May 1997 A completed mast section-.ready for dispatch to the painters

162

This small structural change had a significant effect on the costs since it was now possible to fabricate the masts almost entirely within the factory leaving just five joints along the length to be completed on site in the pre-assembly yard.

The initial objective in developing the scheme for the installation of the cables was to assemble the complete net at ground level and lift it in to position in one operation. This would have the massive advantage of removing almost all the risk of the high level work. It is also many times faster to install a cable in to a net at ground level under zero load, than it is to install it into a existing framework of cables at high level. This initial objective however was found to be impracticable and over ambitious due to the weight, and the many complications, technical difficulties and costs that it introduced. It was next considered splitting the net into two sections and lifting these individually and just mstalling the 72 radial cables between them at high level. Again it was found that the technical issues were too difficult and so the next preferred option of three sections was investigated. This iterative process was continued until, after much debate, the eventual decision was to opt for the pre-assembly of four separate rings and to complete the infill between these rings by lifting one cable at a time at high level. This sort of compromise is necessary and indeed often essential, when developing any complicated erection method. Having to satisfy Safety, programme, and budget considerations inevitably involves compromise. It was possible using this chosen method to assemble over 75% of the cables at ground level under a zero load condition. The difference in terms of man-hours between a cable laid out on the ground and one installed at high level is estimated to be at least six-fold. The saving in terms of cost and time of maximising the pre-assembly of a cable structure is therefore enormous.

Fig 7

A Fully painted section arrives in the site assembly area.

As in the case of the pyramids described in the previous paragraph, the decision was taken to pre-assemble the masts away from their permanent location. This was due to the fact that the foundation works for the masts had to be carried out in parallel with the mast build because of the overall programme constraints. The moving of the completed masts, however, which were 90 metres long and weighed 95 tonnes, was a much more difficult problem. Various options were considered and evaluated. These included using special multi-axial transporters, using a bogey system on a track etc. The solution which was eventually selected was to pick and carry the masts with the large 200 tonne capacity crawler crane. This was only possible if a flat and well-compacted route could be provided and a great deal of investigation was carried out to select and subsequently prove the route. This exercise was further complicated because of limitations to the ground loading pressure that could be applied in the region around the Blackwall tunnel that crossed the site.

The following figures 9,10,11 show the lifting arrangement for the first two nets. This was complicated because the temporary restraint cables for the masts passed over the top of the second net where it was pre-assembled on the ground. The temporary restraint to the masts therefore had to be rediverted via the previously erected central cable truss, once the initial central net was lifted.

Fig 8 A full mast being carried from the assembly area on to the site

PRE-ASSEMBLY OF CABLE NETS


There are over 2600 separate cables that form the 'web' structure of the dome. This presented, perhaps, the biggest challenge to the Watson Steel construction team.
Fig 9

Masts restrained b y t e m p o r a r y forestays with first c a b l e ring r e a d y for lifting

13 6
SITE WELDING
NIERNN. WORMNG F W F O H M

.MMIfW
\
/ I B * O M W \

/\

JMPowWl :

gwyj

||

' 1 H Q W '

Cnrl rg lfe - ms rsri t eta i itd at eta n n ta serd t nw tie dws rnfre o e on
Figure 10

Many engineers tend to avoid site welding wherever possible. This may be due to preconceptions about quality, time or cost. In reality site welding can often be more economic and can provide a better engineering solution than bolting. The difficulty however is that there is no golden rule and the only way to determine which is actually the 'best' method for a particular application is to carry out a detailed comparison on a job by job basis. On the Dome, for example, the original specification was for the mast joints to be site bolted using a pipe flange detail as it was considered to be the more economic solution. The architect however preferred a smooth site welded detail and so an option was included within the tender for the contractor to specify the 'extra-over' costs he would require to site weld, grind and paint the 480 joints in-lieu of bolting. When the actual cost of the options was calculated cost it was cheaper to give the architect and engineer what they preferred and to site weld the complete mast! A good example of a win-win solution!

Tmoay frsas rmvd mss nw e prr oety e oe at o rsri e b tie dws t cnr l r g eta d y o n o e ta i n n
Fig 11

The handling of cables is a key issue that needs careful consideration. It is very easy to cause accidental damage during laying out and handling that may necessitate having to replace the cable. The sequence of the assembly operation has to be planned in-depth to ensure that access routes are maintained and that the site equipment that is being used to handle the cables does not have to run over previously laid cables. The cables used on the Dome were of the spiral strand type, which are highly susceptible to damage caused by kinking or squashing. And any cables that showed signs of distortion had to be replaced.

On the Dome, a method for laying out the cables using a forklift truck and a turntable on a flat wagon was developed. The cables were delivered in coils of a standard inside diameter. The coils were placed on a turntable on the back of a small flat bed wagon. The loose end of the cable was restrained using the fork lift truck while the wagon drove slowly away allowing the cable to unwind on to ground in a predetermined position. Some of the larger cables which ranged up to 90mm diameter also required auxiliary craneage to assist in the laying out.

Fig 12

One of the 480 mast joints being welded on site using a flux cored wire process.

The reason why site welding is sometimes cheaper is because it can dramatically simplify the shop fabrication element of the works. If the site operation is considered in isolation then welding will always be more expensive than bolting but when the savings in fabrication and bolts are taken into account the cost advantage often swings the other way. There are other aspects to consider as well. The site programme will often be extended if welding is involved but in the case of the Dome this was not critical because the welding was taken off the critical path by

164

pre-assembling the masts away from the main site area in parallel with the foundation works. Another factor is the corrosion protection to the welded areas which has to be applied in site conditions and can also effect the cost and programme equation.

JACKS
The attached sketch shows the arrangement of the pulling equipment that was developed for lifting the nets on the Dome. Most of the equipment was designed specifically for this purpose.

Fig 13 The semi-automatic welding equipment used on site

For most site welding applications the preferred process is to use a flux cored wire with a semi-automatic hand held gun. This system is quite robust, can withstand a reasonable amount of draught and has a much higher deposition rate than conventional MMA welding. One of the major advantages that the wire feed processes has is that they do not require the baking and control systems that the MMA electrodes require. The working areas are also a lot cleaner and there is less waste because there are no leftover electrode ends. If site welding is to be considered then it must be well organised with a professional set-up. There is a significant cost to estabUshing a well-controlled site environment and usually there is a minimum scope of work below which it is not usually economic to introduce site welding. Conversely, however, once the decision to site weld has been taken, there are often many other opportunities which present themselves and site welding becomes the preferred solution for that site. The important thing to remember is that there is nothing to be fearful of by introducing site welding. Provided that it is well organised and controlled it can be a major benefit to the project.
Fig 14 Original Sketch of the proposed arrangement for pulling up the cables

TEMPORARY ERCTION GEAR


One of the common elements with wide span structures is that they usually involve complicated and unique erection methods. Where cables are involved the erection method also usually demands special equipment for lifting, jacking, pulling etc. There can be a substantial investment required in such equipment before the construction can commence, in the case of the Dome this was in the order of 0.5m. The major fabricators experienced in such operations often have large stockpiles of specialist equipment that can be adapted for future schemes.

Each mast was equipped with three pull jacks. The jacks were each capable of pulling a six tonne force. The pulling wires were then double reeved which increased the pulling force provided by each jack to almost 12 tonne force. A pull test carried out in site conditions found that the theoretical 12 tonne force at the clamp position had been reduced to 10.5 tonnes due the friction loss in the system. The friction loss is a significant factor that should be allowed for in the design of any lifting arrangement such as the one developed for the Dome. The friction loss would normally vary between 520% however it can be reduced by using special low friction bearings and divertors but this also adds significantly to the cost of the system. The actual design therefore is a trade off between the capacity of the jacks used and the sophistication of the equipment. On the Dome it was found that the most economic solution was to provide enough jacks so that there was plenty of spare capacity and hence the relatively high friction loss did not cause concern. In total 36 jacks were used which generated a combined pulling force at the clamps of 375 tonnes.

165

TEMPORARY CLAMPS
The design of the clamps, which attached to the ends of the permanent cables in order to transfer the pulling force, was an important issue on the Dome. It was expected at the outset of the contract that propriety clamps would be available for each of the 3 different diameters that required pulling. It was found however, that due to the necessary restrictions on the local stresses that could be applied to the spiral strand cables it was not possible to locate clamps 'off the shelf. It was necessary therefore to design and fabricate purpose made clamps. The design was based on limiting the compressive stress to 28 n/mm2 which lead to the clamping length of 500 mm.

Fig 16 The clamps in action at the start of a lifting operation.

STRESSING OF THE CABLES


The final stressing of the cables was carried out at the 72 perimeter adjustment points. Each pair of radial cables incorporated a pair of turnbuckles that were used to take up the adjustment. The cable attachment points were detailed to accommodate a 50 tonne capacity pull jack. A hydraulic pump that had an accurate oil pressure gauge operated the pull jack. The force that was being introduced into the cable was calculated from a calibrated chart based on the hydraulic pressure reading.

Fig 15 The purpose made clamps used to pick up the permanent cables without damage

The clamps also required a lining material to enhance the friction capacity. Various pull tests were carried out during the design period to determine an appropriate lining material. Initially a rubber-based material was used which was found to generate the required friction during the trials. During the first net lift carried out under site conditions however, It was found that the clamps tended to slip in certain circumstances. The subsequent investigation resulted in the conclusion that the friction properties of the rubber material had altered since the initial tests. This was due to the fact that the test was carried out in dry warm conditions and the actual conditions in the middle of winter on site were very different. The problem was resolved by changing the lining material to a type similar to that used in the manufacture of car brake linings. Once the linings had been changed no further problems were experienced.
Fig 17 Arrangement of the stressing equipment

166

EXAMPLES OF OTHER STEELWORK STRUCTURES ON WHICH SIMILAR CONSTRUCTION PRINCIPLES WERE ADAPTED.
TGV INTERCHANGE, CHARLES DE GAULLE AIRPORT, PARIS Site welding was chosen as the preferred method for constructing the trusses primarily for aesthetic reasons but also because of the difficulty achieving the required force transfer between the members. The 50 metre span trusses were pre assembled in an assembly yard some 200 metres away from the construction area and transported by tractor & trailer. The unusual features on this completed structure are the inverted bowstring trusses, which are post tensioned by pulling down the perimeter cable ties. Client Architect Consulting Engineer Aeroports de Paris Aeroports de Paris R.F.R. Partnership, Paris R E E B O K STADIUM, BOLTON The steel roof trusses were pre-assembled by site welding in to sections up to 20 metres x 20 metres. The pre-assembly sizes were determined by the size of the available lifting crane. The trusses were then joined together by insitu welding at heights of up to 50 metres. The complete suspended roof was erected on a series of 72 temporary props. The roof trusses were supported from the propped rafters until all the welding was completed. The props were then struck and the trusses allowed to span the full length of 150 metres. Tie rods from the truss support the front edges of the rafters, which in turn provide lateral support to the top boom of the truss in certain circumstances. Client Architect Steel Designer Bolton Wanderer E C Lobb Partnership Watson Steel

Fig 20 Fig 18 One of the Bow string trusses being assembled in the factory. It was subsequently dismantled for transport to France.

View of the south stand under construction. Note the temporary props to the rafters and the roof truss sections being prepared for site welding

Fig 21

View on the completed stadium

Fig 19 One of the four separate roofs nearing completion.

167

HULME ARCH ROAD BRIDGE, MANCHESTER The 52 metre span bridge was constructed on a series of temporary trestles. The deck sections were preassembled on the adjacent ground and site welded in sections up to 18 metres square. The arch sections were also partially pre-assembled and the remaining joints in the 28 metre high arch were welded insitu and ground smooth afterwards. The cables were installed individually once the welding had been completed and the props removed from the arch. The cables were then tensioned by jacking before the remaining temporary trestles were removed from under the deck. Client. Architect Engineer Manchester City Council Chris Wilkinson Architects Ove Arup & Partners

CHEK LAP KOK AIRPORT, HONG KONG. The 490,000 square metre roof structure was preassembled as 129 large panels up to 36m * 36m square. Each roof panel was fully site welded and painted and then carried over one kilometre to the final location before being lifted and slid in to position. The overall construction programme could only be achieved by pre-assembling the roof in parallel with the concrete substructure. Massive amounts of temporary works were required to assemble, transport and place the roof panels into position. Client Architect Consulting Engineer Hong Kong Airport Authority Sir Norman Foster & Partners Ove Arup & Partners

Fig 24

The first fully welded roof panel in position

Fig 22

The bridge was erected over the busy dual carriageway during a series of road closures

Fig 25 Aerial view during construction. The separate modules have yet to be joined together by site welding.

Fig 23

Note the continuously changing cross section of the plated box section

168

THE GREAT GLASSHOUSE, LLANARTHNE, CARMARTHENSHIRE

SUMMARY
On Wide-span and complicated structures each and every erection scheme will have different priorities and different conditions which have to be taken into account. The one common and essential factor however is detailed planning and attention to detail Also the principles outlined in this paper with regard to pre-assembly, welding, temporary equipment etc. can be applied to most structures and will be equally valid.

The steel and glass roof has a total area of 4300 square metres. The geometry of the complex, doubly curved roof structure is part of a torus. The roof was constructed insitu on temporary trestles by site welding. The curved tubular ribs span up to 55 metres. The site joints were full strength butt welds and were ground smooth.. Client Architect Consulting Engineer National Botanic Garden of Wales Sir Norman Foster & Partners Anthony Hunt Associates

ACKNOWLEDGEMANTS
Client Architects Engineers Construction Managers The New Millennium Experience Ltd Richard Rogers Architects Ltd Buro Happold McAlpine / Laing J.V.

Fig 26 The tubular curved arches were site welded insitu.

Fig 27 The completed Glass house inclined to face the south.

169

CONSTRUCTION OF WIDESPAN ENCLOSURE


David Trench

INTRODUCTION
In March 1997 I was appointed Director of Site, Structures and Infrastructure for the Millennium Dome project. This project is my only experience of constructing a widespan fabric structure. My experiences and conclusions are therefore based on a single unique project. This was a public project using a finite sum of money principally funded from lottery receipts to be completed to an immovable date. From the start I knew if it cost too much I would be flogged but if it did not achieve its completion date I would be hanged.

Fig 1 Interface

Interface (Fig 1)
When considering liability and responsibilities the cabling and installation of masts, posts or towers needed to be under one roof. I here refer to such interface problems of temporary cabling to masts and the changeover of support as the cable net is raised, not to mention setting out and final position of the whole structure. It helps when the fabric is not an intrinsic contributor to structural stability and strength. This allows a split in procurement parties between assembling the structure and installing the fabric although setting out and site measurement of fixing positions becomes an issue.

RISK A S S E S S M E N T A N D OBSERVATIONS
Apart from the "time and cost" risks tackling something that had never been done before, the following conditions were obvious and these will apply to all similar structures; more importantly these features of the project would need to be addressed in the selected procurement strategy.

Safety Design Time


This type of structure lends itself to simultaneous working in Design and Construction. The groundworks will always take time and include the majority of the provisions for distribution of services to avoid the aesthetic and weight problems involved in hanging things on the roof. The Dome also lends itself to parallel working in respect of groundworks and manufacture of a relatively quick fit structure in factories away from the site. CDM (Construction Design & Management Regulations) and safe practice prevent operatives working at roof level and ground level simultaneously so pragmatic flexibility is required in any planning processes.

Foundations
Keeping such structures on the ground requires substantial ballast in the form of concrete. I particularly refer to anchorage points and the ring beam. We used enough concrete at the Dome to cover 25 football pitches in a foot of concrete.

Programmes
Learning curves on new technology require 4 to 6 weeks before optimum speed is established. 12 masts, 72 perimeter poles and 72 inner and outer panels gave good opportunity for repetition and establishing at least two fronts. Planning needs to build in learning curve allowances.

Defects
These types of buildings are subject to continuous and extensive movement both by wind and by temperature. This causes unusual frictional wear on interfaces and in particular seams. Special attention is required to design of vulnerable weatherproofing. Continuous inspection maintenance procedures need to be written into appropriate contracts.

Labour
The industrial relations problems on the Jubilee Line particularly with the electrical labour force were apparent. One of the main stations was in the middle of the exhibition site. Completion bonuses, "work to rule" and strikes on parts of the Jubilee line were commonplace.

Utxad Usa ( C B D . rasidantial. raUil. loisura) Raatdantial (high Uanaity)

Supply Chain
While there are numerous manufacturers and suppliers of PVC polymer coated fabric, the number of PTFE (Teflon coated fibreglass) manufacturers is limited to three, namely: - Chemfab (USA) - Taconic (USA and Ireland) - Verseidag (Germany) It is important to establish what capacity is available and commercial consequences before a procurement strategy is established.

W K U n h l {low donutyi Liiiur* Hotal hraploymant j Pub I K opon i p a c a * Incidental opan i p a c a a R U i l (non-food) RUstl {rood) | hit ton d d Jubiloo U n a | hxiating Uaa M i l l e n n i u m Village , . i*.\ Yacht Club

Public Accountability
NMEC has dual status inasmuch as it is a company incorporated under the Companies Act 1985 and a non-departmental public body by virtue of its share ownership resting with a Minister of the Crown. It is therefore bound by Government rules, guidelines and procedures not least the EC Procurement laws.
Fig 2 What we have left Greenwich

Politics
The project and the Company was high profile and always likely to be subject to political interference in the widest sense, i.e. politicians - media interest Church - professions and trades unions interest. During construction we dealt with as many as 3,000 visitors per month.

Legacy (Fig. 2)
The Prime Minister told the World that Government were looking for legacy out of exhibition. This was widely interpreted that Dome structure would stay and have some future after the year 2000. his the the use

Change Control
Sponsors to be signed up during the project were likely to seek major influence on the design of the exhibition. This influence might be quite late in the design process.

Landlord
The freeholder of the Greenwich Peninsula, English Partnerships, had the overall task of remedition of contamination, establishment of a riverwalk and bringing in the main water, gas, electrical and drainage services to serve the peninsula.

Flexibility
Maximum flexibility in respect of location and content of the exhibition and live events was required. This factor particularly related to capacity and distribution of services and ground bearing capacity for exhibits. Ground bearing capacity was poor until one reached shingle some 11m below surface. Any buildings in the Dome higher than one floor required piling.

Joint Occupation
There was not time to work in sequence. NMEC and EP would need to work contemporaneously on the exhibition site. London Underground would also need to work on the same site completing its station by Christmas 1998 or later.

171

Fabric Envelopes
Fabric structures like the Dome are environmental envelopes that offer protection to a hostile environment. They are not buildings. Heating and cooling is not economically effective beyond 10C of outside temperature. All the usual building codes apply to structures built inside the fabric envelope each on a stand alone basis. Here I particularly refer to insulation, fire protection and means of escape.

advise on any Buro Happold sub-structure design that might penetrate the marker barrier left after EP had delineated its depth of remediation. This way any end purchaser of the property would derive the benefit of a remediation warranty regardless of three previous occupiers and their operations.

Construction M a n a g e m e n t
Construction Management was favoured as the system of procurement. By this method a Construction Management (CM) company (in this case a joint venture of McAlpine and Laing) acts as an agent and not a principal in organising a series of trade contractors to execute the works in direct contract with NMEC. The CM receives a fee for his endeavours in acting as a management consultant. This type of contract with a CM is essentially cost reimbursable and therefore likely to be non-adversirial. Risk is generally shared between the client and trade contractors who are usually on a fixed price. Such arrangements are favoured when the scope of the project is not fully determined and where design and construction overlap. It generally allows more time for refining design and planning different activities before engaging contractors on a fixed price against well defined information describing scope of work.

CONTRACT STRATEGY
My first task in March 1997 was to produce a contract strategy that would address the risks and aspects which I have outlined. In particular I required a set of contract conditions which would address the vital ingredients of a successful project: Clear responsibilities Measured performance Effective communication Early conflict resolution Good information flow Release people from being burdened with processes Strong sense of ownership Co-ordination of energy and resources

The Lease
The Agreement to Lease between NMEC and English Partnerships set out the ground rules in respect of which aspects of the Exhibition project could be deemed to be long term legacy related to a master plan. Legacy has a value to English Partnership, the developer, long after the one year exhibition is over. It also set out a regime that clarified which party should do what and who should fund what element. It was important to us and the Health & Safety Executive that the role of Principal Contractor under CDM regulations should be given to the party with site logistical control delineated by a physical hoarding. The concept of the Agreement was that the 125 acres of the exhibition within the 300 acres of the EP owned peninsula would be remediated from contamination by EP who would leave a platform with full provision of incoming services suitable for an exhibition. EP would also put in infrastructure such as roads and landscaping when it coincided with the Master Plan. This would take two years and therefore phasing was necessary to allow NMEC to construct the exhibition structures and in particular the Dome. The area of the Dome was the first to be released in three segments. W S Atkins were the engineers who organised British Gas' statutory remedial obligations. They were then hired by EP to supervise EP decontamination obligations for suitability of achieving the Master Plan. To preserve the continuity of warranty. WS Atkins were also employed by NMEC to oversee and

+ 000 N o t e : Saviag (feared 50: SO : Ovcraptwl 4mmn4 50:50

+000

-000 Acini
C

-000

-000

-000

* *

<

Fig 3 Pain/Gain - Reward/Penalties Illustration of Adjustment of Free Percentage

Target Cost (Fig. 3)


A Public Accounts Committee in the 1850's examined the disastrously overspent Birmingham to Gloucester rail project. This was the first Government project to be carried out under a cost reimbursable regime. The Committee concluded that "The eye of the master is at its keenest when it is the eye of self interest". To address this issue and create a sense of partnership we introduced a "stakeholder" element into the contract between NMEC and its principal consultants, i.e. designers and construction manager.

172

A Target Cost fee mechanism applied to budget adherence and penalties applied if certain key dates were not met in releasing areas for exhibition construction. The incentive affected fees some V %. The only matters that would lead to an adjustment of the Target Cost were changes in law or a major change in NMEC's brief, i.e. an item in excess of 50,000 increased or decreased cost. This mechanism proved to be self pricing and saved NMEC employing an independent professional quantity surveyor. Thus all the parties shared the cost risk within a 10% band with their client who became a 25% shareholder. The construction manager produced a detailed Cost Plan that matched the NMEC budget and had the responsibility of monitoring and reporting against it.
2

Fig 6 Common communication and control systems

CDM
The Dome was subject to CDM regulations and we were keen that the lead designer Richard Rogers Partnership should have ownership of safety in respect of design. Thus the role of Planning Supervisor was included in the scope of the RRP Agreement, albeit subcontracted to Ove Arup & Partners. We took safety very seriously. I cannot think of a quicker way to destroy morale and sour a contract than endure a major accident. Similarly an accident to a visitor during the exhibition would be catastrophic.

Co-location (Figs. 4 , 5 & 6a)


The design contracts also obligated the designers to colocate with the construction manager in open plan site offices equipped with full IT CAD facilities provided by NMEC. More importantly partners/directors (decision makers) had to also be resident on site on a permanent basis. So in effect the full decision making team including the client was on site, was independent and resourced separately from various head offices. The trade contractors involved with design and production of working drawings were also obligated to set up their design offices within the same establishment adopting the same co-ordinated systems of CAD and document control.

C M Staff Resources
The construction management contract appointed the CM as Principal Contractor under CDM regulations. The fee element of the CM contract also included staff costs. So like the architect and engineer the CM took on the risk of controlling numbers of the staff and what they were paid. An appendix defined staff in name, duties, rates and projected length of employment. This gave NMEC comfort in respect of quality and adequacy of resource. As the CM resources could be affected by trade contractor performance there was an understanding that NMEC would issue a variation order if resources needed to be increased through no fault of the CM.

Fig 4 Co-location

Trade Contracts
The trade contracts were standard but for the following features. We incorporated fairly draconian termination clauses in the event of non-performance and there was an obligation to accelerate if we triggered such an instruction. The contract obligated the trade contractor to provide transparent data in respect to breakdown of costs, detailed resource, programme, estimated man hours on site and in factories. Design warranties called for expertise beyond normal care and diligence but fell short of fitness for purpose.

Fig 5 IT / CAD Systems

173

Internal Audit
Realising that the project would be subject to numerous variations and aware that the Target Cost mechanism would encourage comfortable or conservative cost estimates of variations, we set up an audit role engaging quantity surveyors Currie & Brown. This manifested itself in a senior working partner who would be called in on an ad hoc basis to look at both estimating data and management of specific and random trade contractor accounts by the construction manager. These audit reports then became a focus of discussion and occasional minor changes in practice.

STRATEGIES E M P L O Y E D O N THE PROJECT

Site A u t o n o m y
NMEC successfully isolated the infrastructure construction project from the corporate politics and processes taking place at Victoria which simultaneously involved building up a company and putting together an exhibition. Geographical separation assisted. Jennie Page played a key role by allowing the NMEC site management virtual autonomy within the broad parameters of an agreed scheme and overall budget. As CEO Jennie fielded all the enquiries from politicians and the media allowing the site to get on with its work. As Director I was empowered to sign any contracts or sign off any single payment up to 1 million. Above that figure document went to Chief Executive with my endorsing signature. The CEO's limit was 5 million beyond which it required a Board member's signature and that of our Company Secretary. My two project managers had authority up to 25,000 and the project director of McAlpine Laing JV had authority up to 10,000.
Fig 8 Controls

The trust from Jennie and the NMEC Board was rewarded with transparent weekly and more detailed monthly reports which clearly set out measurable intermediate targets and performance in respect of safety - progress and money. Progress information was provided in summary and detail and highly illustrated in respect of performance and projections. This comprised histograms, trend graphs, data tables, spreadsheets and photographs.

The Site & Structures Team

Fig 6b

Report

Fig 9 The Site and Structures Team

174

N M E C Site M a n a g e m e n t
The NMEC client decision making body resident on site comprised two project managers under myself, as Director. These three client representatives had a combined site experience of 110 years. Not quite the three wise men but I would like to think it was quite hard to pull the wool over our eyes. We kept close to each other and shared a common overall role but with special responsibilities. Peter English focussed on the Dome structure and civil engineering and Dick Coffey concerned himself with Dome core and external buildings and all services. Later this included exhibition structures and central arena which were added to our scope of work. We avoided duplication and I never attended meetings at which either of the other two were present so as not to undermine their authority. I constantly walked the job concentrating on progress and productivity getting first hand experience from working foremen of trade contractors to test how effective we all were as a management team. More importantly, I generally kept myself out of meetings and available for decisions and policy guidance. I would often call ad hoc meetings with no more than six people to tackle whatever problem unexpectedly turned up. I chaired a small cost meeting fortnightly at which we were all present.

Our client team was supplemented by Sue Kidd a construction project experienced and qualified accountant who reported directly to the company finance director although she was very much part of our site team keeping her eye on fees and other expenditure outside the remit of the construction manager. Most importantly she carefully audited and paid all the bills once applications were certified by us. On loan from the Department for Culture Media & Sport she is well versed in public procurement and EC/Government rules and guidelines. So in effect Sue Kidd acted as our compliance officer.

Joint Venture M a n a g e m e n t
The joint venture construction management team mirrored the client team by setting up a flat rather than hierarchical management structure and allowed direct access to people at all levels. Transparency and direct relationship within a whole team led to a good measure of trust and collective responsibility for decision making and performance. Most of all it encourages communication without political interpretation that exists in more hierarchical and structured organisations. A conscious attack on paperwork and procedures forced people to concentrate on management, motivation and progress. "Talk not write" became a byword.

Fig 10 Value Engineering

175

Packaging Strategy
The design team and construction manager instigated a contract packaging strategy that split the major elements in a way that we could retain at least two trade contractors on groundworks, electrical installation, steelwork and fitting out activities. This avoided the employment of agency or casual labour and allowed us to utilise trade contractor core labour not concerned with redundancy at the end of the job. The policy also preserved an element of competition both in terms of performance and bidding for evolving further package works and variations. The lessons of the Jubilee line were well learnt and we had no industrial action on the project. In fact we enjoyed very good relations with all the labour on the project managing to instil a sense of ownership right down to a gang digging up the ground. Although the Target Cost had a built in overall 10% contingency the designers and joint venture agreed to create a further 10% contingency for each package. In effect designers were seeking solutions well within budget leaving an allowance for variations caused by design development and possible costs resulting from site disruption or acceleration. We held a number of value management and engineering meetings involving key players. The twelve external and prominent cylinders housing major plant substituting large spheres with nothing in them are a classic output from one such early "away day". We sometimes held such sessions with potential trade contract bidders prior to awarding a contract in order to capture the best ideas from trade contractors who had spent six weeks tendering and studying specialist aspects of the project. The changes were then included in the contract documents. We chose a route of prefabricating off site as much as we could, almost regardless of the immediate economic benefits. This was to avoid employing a labour force on site above 500 people and to maximise parallel working in order to obtain the fastest programme. Often the factory work included pre-testing and commissioning saving valuable time later. Inaccessibility of the site and the general pull on the London labour market made us strive to build the project on different sites all around the U.K. For instance all our external buildings, some 250,000 sq.ft. were put together

and fitted out in York, Hull and Gilberdyke by three different companies although the buildings appear identical to the visitor. We selected all our trade contractors on a judgement of methodology and a "can do" site management personnel. In some cases the lowest tender bid failed to secure the work. Culture was more important to us than contractual and commercial considerations.

Measurement
The construction manager was rigorous in measuring weekly performance against targets in respect of daily output, resource and trends. This applied to design, procurement and production. At all times we knew exactly where we stood and peer pressure did the rest since everyone submitted to this regime. This also particularly applied to off site activities such as fabrication of rolls of PTFE from Chem Fab in New Hampshire and cutting and preparation of fabric in Buffalo. During this period we stationed a person permanently in the USA who daily faxed back outputs.

Acceleration and B u y i n g O u t Claims


^VVe also strove to identify cases where trade contractors were impeded from performing through no fault of their own. Rather than face a delay in the end date accompanied by a claim for uneconomic working we negotiated specific acceleration agreements whereby some costs were guaranteed and further performance incentive payments were offered if a series of targets to put the element back on track. We encouraged the bulk of the incentive to reach the individual pockets of the labour force involved so everyone had a stake in hitting dates critical to the programme. The change of fabric trade contractor and the re-engineering of the roof from PVC to PTFE meant the first deliveries to site were three months behind our critical programme. An acceleration agreement with Birdair brought the critical completion of the roof back to the original programme allowing the Prime Minister to top out the Dome on 24th June a year after we started the piling on site. This was a date set in stone on a relatively simple programme drawn up in November 1996. At all times we would do anything to achieve some two dozen key milestones.

Cost Meetings
I chaired one regular fortnightly cost meeting with the construction manager's director and commercial manager, my accountant and two project managers. This was a forecasting meeting when we exclusively concentrated on the future. The first rule of project management is that you can only manage what is ahead. Every rumour of an impending cost escalation was thrown into the ring. Often trade contractors initially pricing evolving design was our source. We also looked at identifying uneconomic practices and every cost

Fig 11

Cost Meetings

176

problem in the pipeline. It was generally left for the construction manager to tackle designers on these issues in privacy and without the presence of the client. This is where the peer pressure of three parties locked into a target cost paid dividends. However sacred a design concept it was not in a designers interest to increase construction costs.

O n e Team A p p r o a c h
I cannot emphasise enough the importance of the One Team site co-location approach in a good quality office environment sharing IT systems of document and CAD control. The offices and IT were an investment of some 2 million but it was worth every penny. The bonus is that the buildings are sufficiently well appointed to serve the Production, Operations, Press and Personnel teams in the run up period and during the exhibition.

Conclusion
I have not addressed the one thing that guaranteed success at the Dome. It was people. The project drew together a rare mix of "A" teams determined to make it work and accordingly adopted a common agenda to which the key individuals contributed an astonishing quantum of intellectual rigour. You can have all the right contracts, procedures and systems but without the right people the project will fail. The right people will make anything work. At the Dome I would like to think we got it right on both counts.

SECTION IV Construction and Materials

Lightweight Structures Tensile Space Structures

Materials for the New Millennium Engineering an Integrated

Construction of the Great Court Roof of the British Museum A History of Widespan Structures un the United States

Architecture for Widespan Enclosures

178

LIGHTWEIGHT STRUCTURES
Jorg Schlaich Prof. Dr.-Ing. Drs.h.c. University of Stuttgart, Germany

Any structure designed intelligently and responsibly aspires to be "as light as possible". Its function is to support "live loads". The dead loads of the structure itself are a necessary evil. The smaller the ratio between a structure's dead load and the supported live loads, the "lighter" the structure. We realize immediately that a suspension bridge with knotted cables is obviously lighter than a truss bridge with welded bars which in turn is lighter than a box girder bridge from concrete. This consequently leads us to the question why so few suspension bridges are being built and only for large spans and we intuitively understand that the demand for lightness is not sole criterion when designing structures. Indeed, "natural loads" are the enemy of lightweight structures. These structures tend to deform heavily under snow and temperature changes, they are sensitive towards wind-induced vibrations, they may tear (the structural engineers' trauma of Tacoma), but they literally make light work of earthquakes. Another stern adversary of lightweight structures are today's high labour costs and the imprudent use of natural resources. This promotes the massiveness and hinders the filigree. But before we discuss how to design lightweight structures we need to ask ourselves whether or not lightweight structures today are worth the effort to be promoted and developed. The answer is yes! From an ecological, social and cultural perspective lightweight structures have never been more contemporary and necessary than today. The ecological point of view: Lightweight structures are material-efficient because the materials strengths are optimally used. Thus no resources are wasted. Lightweight structures may usually be disassembled and their elements are recyclable. Lightweight structures curtail the entropy and therefore are superior in meeting the requirement for a sustainable development. The social point of view: Lightweight structures create jobs because filigree structures demand carefully designed labour-intensive details with a great expenditure in planning and above all manufacture. The intellectual effort replaces the physical effort, now time

and craftsmanship supercede the extruding press - the joy of engineering instead of massiveness. But as long as our modern economy equals working hours with costs, we merely pay the mining costs of the raw materials and the overall "external costs" are not even added, lightweight structures will be more expensive than bulky structures with the same function. Therefore lightweight structures might be attributed with an elitist air. It is true that only banks and insurance companies, and sometimes museums, can afford lightweight structures, but nobody in the field of residential or ordinary industrial buildings. And the engineers and architects wallow in this elitist glow (a stark contrast to the pioneers' spirit of lightweight structures: Buckminster Fuller, Vladimir Suchov, Frei Otto). They continue to push this structural exhibitionism and do not even notice that 98 % of the structures around them crave for their attention, thus their actions are highly antisocial - the author know what he is talking about and stands accused. The cultural point of view: Lightweight structures, built responsibly and disciplined, may contribute heavily to an enriched architecture. Light, filigree and soft evokes more pleasant sensations than heavy, bulky and hard. In the typical lightweight structure the flow of forces is visible and the enlightened care to understand what they see. Thus lightweight structures with their rational aesthetics may solicit sympathies for technology, construction and engineers. They may help us to escape the wide-spread monotony and drabness in today's structural engineering which in turn will become again an essential part of the building culture. How to create lightweight structures? When designing lightweight structures we have to: firstly remember a most unfavourable characteristic of the dead loads: The thickness of a girder under bending stress, supporting only itself, increases not only proportional to its span (which is often falsely assumed), but also with the span's square! For example if the girder with a span of 10 m has to be 0.1 m thick, its thickness increases with a span of 100 m not only 1 Of old but 10 x lOfold. Consequently the girder has to be 10 m thick and its total weight increases by the factor 1000!

179

Already Galileo Galilei was aware of the importance of scale. He demonstrated this by comparing the tiny thin bone of a bird with the corresponding big bulky one of a dinosaur (Fig. 1). This teaches us that increasing spans increase the weight of structures, consequently gratuitous large spans are to be avoided.

suspended between two rock faces. But these latter ones are useless, because they deform too much under loads. But in between the upper and lower structures there are the most diverse solutions: arches and suspended cables stiffened by secondary girders in bending and all kinds of fastenings, deck-stiffened arches, strutted frames (left) as well as cable bridges and suspension bridges etc. (right). The further we move down in Fig. 2 the lighter it becomes but also the more critical with respect to windinduced vibrations - and this represents the challenge and the attraction of bridge engineering.

Fig 1 Galilei's demonstration of the scale effect

But this law of nature about scale may be circumvented with some tricks, by secondly avoiding elements stressed by bending in favour of bars stressed purely axial by tension or compression, i. e. dissolving the girder. Basically this is always possible as demonstrated by the truss girder. With struts and ties the entire cross-section is evenly exploited without anything superfluous. Bending completely stresses only the edge fibers while in the center dead bulk has to be dragged along. Here ties in tension act apparently more favourable than struts in compression because they only tear if the material fails, while slender struts fail due to buckling, i. e. a sudden lateral evasive movement. This can easily be tested with a long bamboo stick. We cannot break it with our bare hands, but if we bear down on it, it buckles quickly.
Fig 2

cmrsinesn o pes -ni ot ov


i^I^

^TTTrrrr-ry

Voiiiid/
The evolution of bridges

thirdly these efficient tension stressed elements becomeeven more efficient with increasing tension strength P and decreasing density of the material y, i. e. with increasing rupture length p/y. This clear value represents the length a thread can reach hanging straight down until it tears under its dead load. Wood is more efficient than steel and natural and artificial fibers do even better. These first three approaches to lightweight structures introduce us already to the entire multitude of forms in bridge engineering. We recognize (Fig. 2, starting from the top) the dissolution of the girder into the truss and then (left) the arch structures which carry their loads mainly by compression and their inversion (right) the suspension structures which make use of the especially favourable tensile forces. At the bottom are the most marginal structures, the pure arch or the cable

The keen observer of today's bridge engineering will find that a rather pragmatic attitude prevails, structures are being built "as heavy as justifiable". Solid girders are used up to a span of about 100 m, arches resp. trusses up to approximately 250 m. Dead loads at least five times the live loads are tolerated. Beyond approximately 300 m the dead load becomes so dominant that, as the only alternative, tensile "lightweight structures" remain: up to about 1000 m self-anchored suspension and cable-stayed bridges and for even greater spans back-anchored suspension bridges. The Pont de Normandie in France spanning 856 m and the Tatara-Bridge in Japan spanning 890 m are the world's largest cable-stayed bridges. The largest suspension bridge with a span of 1990 m is the Akashi bridge in Japan. The suspension bridge proposed for the

180

crossing of the Straits of Messina spanning 3500 m is to be suspended from 4 cables each 1.7 m in diameter. These cables consume already half of their loadbearing capacity to support themselves and only one half remains for the actual bridge and the live load which remains insignificant compared to the dead load of the cables and the deck. By definition this is by no means lightweight, but at such span, today's materials do not permit anything lighter - we have reached the limit - unless, steel cables can be replaced by plastic fibres with a significantly greater p/y-value. A strikingly ingenious trick to achieve lightness should be addressed briefly, i. e. fourthly prestress or pretension which permits to transform unfavourable compression stress into favourable tension stress (Fig. 3). The example shows a quadrangle of slats with crossed cables. The diagonal cable receiving compression will not become slack but shares the load because it is prestressed. Initially before applying the outer load this cable was exposed to pretension, thus when compressed it will not experience compression but a reduction of tension which is the static equivalent. This procedure permits the creation of very light cable girders and cable nets which act like ideal structures with tension and compression resistant elements or like membrane shells.

fifthly the use of lightweight spatial structures, or double curved space structures with pure axial stress, called membrane stresses (Fig. 4). These structures are not only extremely light but they also open up a whole new world in architecture, an unsurpassable variety of forms which is not yet exhausted, by no means. Just like bridges, these structures transfer their loads predominantly by compression shells or domes (Fig. 4, left), or by tension cable nets and membranes (right). In between are the plane space structures - the slabs and the space frames.

i il * l -

I i;

Fig 4

The evolution of lightweight spatial structures

Fig 3 The principle of prestress

Top left: Top right:

Bottom left: Bottom right:

Unstiffened kinematic system The diagonal in compression becomes slack and only the diagonal in tension is active. Prestress: before loading the diagonals are shortened i. e. pretensioned In a prestressed system both diagonals share the load

The basic principles of lightweight bridges also apply to buildings such as roofs over large sports arenas or fair pavilions or industrial plants lending an individual character and a human scale to these structures. Since the gap between these cable girders still has to be spanned with transversal girders using bending and thus resulting in semi-heavy or semi-light roofs, the final step is inevitable:

Despite the extremely thin walls of shells and space domes their curved shape stabilizes and prevents them from the dreaded buckling. And applying prestress protects the extraordinarily lightweight nets and membrane from the effects of wind-induced vibrations. The two principal directions of the nets and membranes are mechanically stressed against each other resulting in the typical saddle-shape with an anticlastic curvature, or, if pneumatically stressed by creating an internal air pressure or a vacuum, resulting in a dome shape with synclastic curvature. This can be mastered with modern computers. Manufacture and, as a consequence, costs are more likely to limit the scope of these lightweight spatial structures. Expensive formwork and complicated cutting patterns are required for the manufacture of these doublecurved surfaces (Fig. 5). The details of tensile structures and membranes are complicated and have to be manufactured with extreme precision.

181

But in recent years the textile membrane structures have made a remarkable progress. Since they may be folded they are even used as convertible structures. This marked the beginning of a whole new era in structural engineering completely changing life in our capricious climate. The future is now!

STRUCTURE

MANUFACTURE

GEOMETRY

fe re
SQUARE NET

rsrce etitd
TRIANGULAR NET

fe re
TEXTILE MEMBRANE

Fig 6a The cable net cooling tower at Schmehausen (1974)

rsrtd eti e c
THIN METAL S H E E T MEMBRANE

Fig 5 The geometry and manufacture of typical double-curved lightweight structures

Achieving lightness is a heavy burden, because lightweight structures challenge the boundaries set by the theories of statics and dynamics. The fancy materials put the technologies to the test and the complicated threedimensional structures dare the manufacturing procedures. Lightweight structures tempt the dedicated engineer, because they - exemplary for this profession - equally and simultaneously address his knowledge, his ability and his experience as well as his fantasy and his intuition. With lightweight structures the engineer is able to award the adequate visual expression to an ingenious and efficient structure thus contributing to building culture. Over the years, the author and his colleagues tried to apply these principles of lightweight to all types of structures including bridges and to towers even (Fig. 6), but out of space reasons, this report will be restricted to lightweight roofs, leaving out even the wide and interesting field of concrete shells (Fig. 7).

Fig 6b

The Killesberg lookout tower in Stuttgart (under design)

Fig 8 cable net roof for the Munich Olympics, Behnisch und Partner and Frei Otto, architects (1972)

Fig 6c The Leipzig Fair tower, Volkwin Marg, architect (1995)

The interesting point is that the arch not only carries the net but that simultaneously the net stabilizes the arch. In order to permit a complete prefabrication of the net and a simple erection but also to separate visually the arch from the net, the trussed arch is triangular in section with the suspenders fixed to its bottom chord (Fig. 9c). With that it was possible to first erect the arch (which had to be temporarily guyed because, as mentioned, alone it was not stable), spread the net underneath and then lift it in its final position below the arch and stretch it by tilting the guyed masts. Whereas the slots between the edge cable underneath the arch were covered with clear acrylic

Fig 7 The 12 ram thick GRC-(Glass fibre Reinforced Concrete) Shell at Stuttgart (1977) (following Candela's Xochimilco design).

It started with a real highlight, the cable net tent for the Munich Olympic Games in 1972 (Fig. 8). This exemplifies the almost unlimited freedom of shapes which the cable net with quadrangular mesh offers. By changing the angles of the original square net it can adapt to almost any surface. Since this roof has been published widely let us proceed to a later cable net over an iceskating rink at the same site in Munich which is elliptic in plan (88 x 67 m) (Fig. 9 a ) . The cable net (mesh width 75 x 75 cm) is suspended from an arch spanning 104 m in the longitudinal direction and 18.5 m high, andstressed at its periphery using guyed masts there.

Fig 9a The cable net roof over an ice-skating rink at Munich, Kurt Ackermann und Partner, architects (1985)

glass, the cable net itself was to be covered with a PVC/polyester membrane, expecting from it not only an economical solution but also a pleasant interior atmosphere. To fix the membrane on the cable net, a wooden grid was used because wood can as well be bolted to the joints of the cable net as the membrane may be attached to it with nails. This on the other side raised the question of fire protection of the arch, which we

183

Fig 9b

Interior view

could overcome by reducing the amount of wood close to the arch to a minimum. There the spacing of the wooden grid is 75 x 75 cm, following the cable net. We could do so because from there the snow would slide down to the flatter region of the roof, thus the reduced snow load would permit the 75 x 75 cm mesh to be spanned by the membrane alone. As a logical consequence the grid in the lower parts along the periphery of the roof had to be closer with additional wooden slats supporting the membrane there. At the end we were very happy in finding that the satisfaction of two technical requirements - fire and snow load - resulted in a beautiful visual appearance, because due to the reduced density of the grid from the periphery towards the arch, the transparency increases from the lower part to the higher part of the roof, thus suggesting more internal height or volume than really exists and thus producing in this iceskating ring a gay and relaxed atmosphere (Fig. 9b). Concerning membrane structures we were for many years reluctant to try our own designs because many of those built in the sixties exposed a surprisingly painful discrepancy between their beautiful overall shape and their nasty details without a tendency towards improvement in the years to follow. Finally, however, we welcomed the invitation of the contractors Hochtief of the Jeddah Airport roofs to advise them in the detailing and during the construction of this SOM/Horst Berger design (completed 1982) as well as of Philipp Holzmann of the Riyadh Stadium roof to do the final design and the construction supervision of this Ian Fraser/Horst Berger conceptual design (completed 1984) for them. These two roofs gave us a chance to develop our own hopefully improved details and to gather overall experience with membrane structures. Besides the large roof over the temporary grandstand for the Munich Olympic swimming hall in 1972 the real chance to design and build original and own membrane

Fig 9c Details of the ridge cable arrangement

structures came only about ten years ago and we even started with their most challenging species, the convertible roofs: The inflated cushion for the Roman Arena in Nimes, France, with an elliptic plan 88 x 57 m is installed there each year in fall and removed in spring (Fig. 10). It goes back to an earlier design for the Roman Arena in Verona, Italy, where we had even proposed to fill it with helium and to fly it to the periphery of the town to serve there as a temporary shed during the summer.

Fig 10 The Roman Arena convertible roof in Nimes, France, with F. Geipel and N. Michelin, architects (1988)

184

It was only last year, when we could apply the same principle to a convertible roof of the bull-fight arena Vista Alegre in Madrid/Spain (Fig. 11). The cushion has a diameter of 50 m. Its upper Polyester/PVC translucent membrane rises 7 m, its inner ET transparent membrane sags 5 m and is reinforced by a cable net with a 1.5/1.5 m mesh of 12 mm cables. The whole cushion can be lifted 11.4 m by winches along 12 vertical columns placed at the inner ring of the permanent cantilevering roof over the grandstand.

Fig 12 The Montreal Olympic Stadium convertible roof as seen from inside during closure, R. Taillibert, architect (1989)

fir

^^7~^zz
II

X X
5000

/
J?
SO.It

Fig 11a Lifting roof, Vista Alegre, Madrid, Spain, with FHECOR engineers, for the fixed outer roof. Cross-section left side cushion in closed position, right side cushion in lifted position (1999)

Fig 13a

The roof over the bull-fight arena at Zaragoza, Spain, total view, convertible inner part not yet installed (1989)

Fig l i b The pneumatic cushion during installation

The huge retractable roof covering 20,000 m_ with a PVC/Kevlar-membrane for the Olympic Stadium in Montreal, Canada (Fig. 12), could not be completed for the 1976 Olympic Games. When, during the mideighties, we were approached by Lavalin Consulting Engineers of Montreal to help them complete it now, the prescribed boundary conditions were so that this turned out to become certainly the most complex assignment we ever accepted. Though the result, as known, was not satisfactory but probably could not be better without accepting major changes of the original design by the architect, we learned a lot, so that we considered it as a pleasure to apply this experience when designing the roof for the bull-fight arena in Zaragoza, Spain, with its fixed circular outer roof and its central convertible part. This roof has the right size for a light and unobtrusive membrane and especially the convertible roof when seen in motion from underneath, resembling an opening and closing beautiful flower, confirms that membrane

Fig 13b From inside during operation of the convertible inner part

structures deserve the attribute 'natural' (Fig. 13). In the case of the four completed large stadium roofs - some more are nearing completion or under design - for the Gottlieb-Daimler-Stadium in Stuttgart (Fig. 14), the Gerry Weber Centre Court in Halle with a translucent and convertible inner roof, the NSC Outdoor Stadium in Kuala Lumpur (Fig. 15) and the Estadio Olimpico de Sevilla (Fig. 16), primary cable structures are applied based on the spoked wheel principle with either two inner tension rings and one outer compression ring or vice versa. In spite of their huge size these make very light structures permitting the membranes really to come forward with their transparency resulting in a friendly and pleasant atmosphere. This is most important in the case of stadia to help calm aggressions which in other

185

places frequently resulted in fights, riots and panic attacks. The atmosphere during the Athletic World Championships 1993 at the Stuttgart stadium was most friendly - thanks to the light and pleasant membrane structure. Unfortunately, older stadia used to be rather ugly and obtrusive concrete monsters and if such an existing grandstand is to be covered - as in the case of Halle - there is no chance to improve the outer appearance even with the lightest membrane roof or with the most carefully designed details.

Fig 16a

Estadio Olimpico de Sevilla. The folded membranes during erection; Cruz y Ortiz, architects (1999)

Fig 14a The Daimler Stadium in Stuttgart; H. Siegel und Partner and Weidleplan, architects (1993)

Fig 16b

Comparison of Stuttgart versus Sevilla membrane arrangement

Fig 14b

Interior view

Whereas for Stuttgart, Kuala Lumpur and Halle we introduced radial cable girders between the inner and outer rings, as a primary structure with the membrane spanning between their lower (Stuttgart, Halle) or upper (Kuala Lumpur) cables and tied arches as a secondary structure, in case of the roof of the Estadio Olimpico de Sevilla (as earlier for the outer roof at Zaragossa), the membrane is an integral part of the primary structure: It is stressed between the upper and lower cables resulting in a folded plate geometry and loadbearing behaviour (Fig. 16 b). To our surprise in this case the architect was not interested in a translucent roof, but insisted to make the membrane opaque. Obviously the quality and success of membrane structures depends on their details, clean and simple details which are in harmony with the structure as a whole. This of course needs constant efforts and includes not only the "hardware" but also the cutting pattern of the membrane as well. If carefully designed, the geometry of the seams can reflect the flow of forces thus improving the visual appearance of membrane roofs. Three recent roofs may exemplify how we tried to follow this idea:

Fig 15

The NSC Outdoor Stadium in Kuala Lumpur; interior view during construction; Weidleplan, architects (1997)

186

The Hamburg-Stellingen Ice-Skating Rink roof covers an existing ice field. Its light appearance results from the fact that it is supported by four masts only with additional cable supported props. As the membrane approaches these singular supports, its seams assume a concentric pattern with additional strengthening strips in order to visualize the concentration of the forces (Fig. 17).

Another effort to solve the difficult singular point supports of membranes resulted in a cloverleaf at the indoor pool adjacent to the Kuala Lumpur stadium (Fig. 19). If a rectangular area is to be covered at minimum cost, as in case of large exhibition halls for fairs, the unidirectional cable girders or hanging roofs usually are

Fig 17a The roof over the Hamburg-Stelling Ice-Skating Rink from outside at night; Silcher, Werner, Redante und Partner, architects (1994)

Fig 19a

Kuala Lumpur Indoor Pool; Weidleplan, architects (1998)

Fig 17b

From inside.

Fig 19b

Cloverleaf, Kuala Lumpur Indoor Pool

The roof over a grandstand at Oldenburg cantilevers from a number of masts using horizontal struts with cable supports and membrane panels in-between, tied down towards the masts. From a distance it looks rather simple and geometrical whereas from underneath it reveals again this clean and joyful appearance, typical for membrane structures (Fig. 18).

1
...

better suited than the double-curved cable nets or membranes. Three examples from our practice may illustrate this type of lightweight tension structure. For an exhibition hall 112 x 184m at the Hannover Fair we added 18 cable supported girders of 122 m span and 9 m depths, resulting not only in a simple and economic but also real light structure (Fig. 20). For another exhibition hall covering 210 x 110m we chose a sequence of three pure stress-ribbons from steel slats covered with wooden panels (Fig. 21). The same system, but much smaller and covered with glass, we applied earlier for a canopy at the Ulm railway station (Fig. 23). Finally most recently for another fair hall at Hannover we applied in a similar way a sequence of 4 full and two half bay stress ribbons at either end, covered with wood and supported in the cross-direction by self-anchored suspension structures: rather complicated, and perhaps at the brink between high-tech and high-effect (Fig. 22)!

Fig. 18: The roof over a grandstand at Oldenburg (1996)

187

Fig 20

Hall 4 at the Hannover Fair at night, Volkwin Marg, architect (1995

Fig 23

Suspended glass canopy in front of the Ulm railway station, H. Gaupp, architect (1992)

Fig 21

Hall 26 at the Hannover Fair inside, Thomas Herzog, architect (1996)

i
4 i
V

This finally brings me to our glass-covered roofs, for which we developed what we call grid-domes. Based again on the structural principle of the square mesh or grid, which adapts by change of angles to a doublecurved surface of any shape, first such a pure grid-dome is built and afterwards stiffened by diagonal cables. The glass-grid dome of the Neckarsulm indoor swimming pool is of pure spherical geometry (Fig. 24). In case of the roof over the courtyard for the Museum of the History of Hamburg two cylinders stiffened by spoked wheels intersect in a free transition surface of the dome shape (Fig. 25). Both roofs, and many more to follow, demonstrate the lightness of such shell-type structures. Double-curved glass covered grid-domes commonly and naturally call for spherically curved glass panes, since the four edges of a grid mesh are not in one plane, a real problem especially if double-glazing is required. In Neckarsulm we really got spherically shaped doubleglazed glass - but never again! - in Hamburg we had simple glazing and enforced the necessary distortion to the glass.

Fig 22a

Hall 8/9 at the Hannover Fair (model), Volkwin Marg, architect (1999)

Fig 22b

Hall 8/9 at the Hannover Fair. The transversal suspension structure during erection

Fig 24

The glass-grid dome of the Neckarsulm indoor swimming pool from inside, K. Bechler, architects (1989)

Fig 27a

Roof over the pools in the hippo-house at the Berlin Zoo, J. Griebl, architect (1999)

Fig 25

The glass-grid dome over the History of Hamburg Museum Courtyard from inside, Volkwin Marg, architect (1989)

Clylindrical shell roofs, stiffened by end diaphragms or by spoked-wheels (Fig. 26), as well as their inversion, the hanging roof (Fig. 23) of course overcome this problem in a trivial manner, as well as our facades, where we fix glass directly to the joints of plane cable nets.

Fig 27b The translational surface of the hippo-house roof

Fig 26

Spandau Railway Station sheds, M. von Gerkan, architect (1999)

obsolete. Now, arbitrary surfaces may be covered with a triangular mesh geometry, where each slat and node is different but in any case resulting in a perfect loadbearing behaviour. If each triangular mesh is formed by three slats joined by nodes with six exits, the glass panes obviously are triangular as well and thus plane in any case, but, of course, with the disadvantage of squandering a lot of glass. Such a computer-designed and -manufactured roof with an arbitrary shape covers the DG-Bank in Berlin (Fig. 28). It demonstrates that the absence of discipline by geometrical manufacturing constraints must not result in chaotic shapes, but can be replaced by self-imposed order resulting in clean and beautiful structures. The author is aware that he owes these rewarding structures to his creative and tireless partners and collaborators, imaginative architects, skillful manufacturers and last but not least trusting clients.

However, only translational surfaces really solve the problem and offer a wide range of double-curved grid shapes with the four corners of each mesh in one plane because as well as for the directrix and for the generatrix any geometry may be chosen. The roof over the pools in the hippo-house at the Berlin Zoo, joining two spheres of different size, is a good example for the feasibility of this approach (Fig. 27). Geodesic domes, grid shells, the grid domes as discussed here or in other words, the work of Bauersfeld, Fuller, Wachsman, Schwedler, Mengeringhausen, Otto and others was characterized by the search for efficient reticulated space lattice structures with as many as possible slats and nodes of equal size and geometry for an economical pre-fabrication and assembly. "Thanks" to computers with CAD, CAM, CNC this all appears to be

Fig 28

DG-Bank, Pariser Platz, Berlin, Frank O. Gehry, architect, 1999

189

TENSILE SPACE STRUCTURES

Wolfgang Renner MERO Systeme GmbH&Co.KG, Max-Mengeringhausen-Str.5, 97084 Wurzburg,/Germany

ABSTRACT SPACE FRAMES


With the rapid advance of technology in engineering, the creation of sound yet innovative construction has evolved into a sophisticated global challenge. For more than fifty years MERO has been a part of world-wide architectural success, in its creation of ingenious modular construction systems, paying particular attention to steel, aluminium and glass structures. MERO's path led from structures made of their classic spaceframe of tubes and nodes to structures of profiles in combination with tensile cables, integrating the cladding as supporting element. In the following this path is described. The traditional MERO System, the Ball Node System (KK) is the first prefabricated space frame system developed by Dr. Ing. Max Mengeringhausen during the second world-war. Loads are applied via the nodes, the members distribute the compression and tension forces. The original idea was to build a space frame with uniform length of members and regular nodes with 18 holes with angels of 45 and 60 degrees. Nowadays the angles between the members may be freely chosen and also the length and diameter of the members is not uniform. The members are round hollow sections, because they have the best resistance against buckling. Diameters from 30-355 mm with different wall thickness are standard. The length differs i.g. 1,5-5,0 m but is not fixed. The nodes are to give a free choice for the connections holes. The standard ball nodes have diameters from 49,5-350 mm.

I. Krolsliohiprofii (KHP) 2 Kegel 3. Gevnndebolzen A. Scnliisseliriutte 5. Knebelkerb&lill

3. Threaded boll 5. Dowoipin

Fig 1

Fig 3

To connect several members to a relatively small node the members have conical ends. The member is connected to the node via a high tensile bolt. The bolt has a dowel pin and is screwed into the node via a sleeve. Compression force is transmitted via the sleeve, tensions via the bolt (fig.3). With this system single and multilayer space frames can be designed. Examples for Ball Node system ( KK ) are the Split Stadium cover and the Music Pavilion in the Grugapark Essen.

Fig 2

190

Fig 4

Fig 7

The MERO Bowl Node System (NK) is a further development of the MERO space frame system developed from the KK-Ball Node-System. The advantage is that cladding or glazing can directly be fixed on the top chord without additional purlins. Thus a higher transparency of glazed structures is achieved. The standard bowl nodes have diameter from 160 to 200m. The members of the top grid are rectangular hollow profiles. The members are pin jointed to the nodes. The screws are fixed invisible via the bolt insertion hole. Members and nodes are flush fitting and torsion proof. The nodes and the members of the bottom grid such as the diagonals correspond to the KK-system.
Fig 5

The biaxial load transfer of space frames often leads to a very low dead weight. They are therefore specially suited for wide span constructions. Loads from cladding or suspended loads should be applied via nodes. Membranes are directly fixed to the nodes, while sheets are fixed to a purlin system. These purlins are fixed on stools of different height to achieve a sloped roof on a horizontal spaceframe.

f
Fig 8

NK Knoten / node / noeud


Fig 6 Fig 9

191

Fig 10

The pyramid shaped greenhouses in Essen (fig. 9) with a lateral length up to 30 m are a perfect example for this system. These new glasshouses were designed as a replacement for the former ones. Three fully glazed pyramid structures of different sizes together with a shed hall and other flat connecting buildings form a square around a totally enclosed garden courtyard. The loads from the glazing are applied directly to the rectangular hollow profiles of the top chord. These members and the circular hollow section diagonals are interconnected via bowl nodes. The steel weight of the structure is 15 kg/sqm surface area. Also the largest hemisphere in the world, the Stockholm Globe Arena, was built with the MERO bowl node system (fig. 11). The Arena designed as a multipurposehall for international events is in the center of the developing district of Hovet. 5.000 to 16.000 seats can be offered according to the different events The geometrical optimization resulted in a dome with 96 meridians and 19 horizontal rings. Thus a very reasonable wide-meshed net with max. field sizes of 3.6 x 4.4m was produced. For reasons of stability a double layered space frame construction with a max. depth of 2.1 m had to be chosen. For cladding an Alucopan

Fig 12

Sandwichpanel was chosen. Due to the requirements of steam tightness a perimeter support of each panel was needed. Thus the MERO Bowl Node System with rectangular hollow profiles in the top grid was used together with an extra secondary member by which the surface was now divided into max. sized panels of 3.6 x 2.2 m. This size was ideal for manufacturing and transport of the panels. With a construction height of 85 m and a globe diameter of 110 m the steel structure has a self-weight of only 32 kg/sqm surface area. Because of their dimensions such large projects are not only a challenge in view of their huge space but also acoustic aspects have to be considered. However the globe shape itself is rather advantageous. It has the highest volume compared with the surface and provides an optimal distribution of temperature as far as the athletes and/or spectators level is concerned and thus creates a comfortable atmosphere.

Fig 11

192

A further development is the MERO-Cylinder Node System ( Z K ) . The cylinder node provides a bend resistant connection. Thus plane and curved single layer constructions can be executed with triangular or quadrangular grid. The quadrangular grid is very economical in respect of the cost for the cladding and/or glazing. Rectangular hollow profiles are connected to the cylinder node by two or four bolts per member (fig. 14). The connection is flush fitting and torsion proof. The standard nodes have diameter of 200 mm, the member sections are from 100x60 - 160x80 mm. The members are screwed invisible.

ZK Knaten

/ node /

noeud

Fig 13

Fig 16

Fig 14

193

Disc Node System ( T K ) The disc node is the most specialized connector. This system can only be used for biaxially curved structures with a triangular grid, ie domes with a diameter up to 50 m covered with prefabricated triangular panels or glass The Rectangular hollow profiles are pin jointed with disc nodes. The glazing is placed directly onto the rectangular tube profiles and sealed on all sides. Fig. 19, shows a dome structure with 30 m Diameter, creating a recreation area for the Rh'n-Klinikum.

CURTAIN WALL SYSTEMS - POINT S U P P O R T E D G L A S S (PSG)


Pursuing the request of architects and owners for more transparency and translucency, glass increasingly becomes the center of MERCTs project. Ongoing research has resulted in the creation of MEROPlus system, developed for modern glass structures, and the advanced MERO PSG which uses point supported glass for maximum transparency in design. At the New Leipzig Fair the multi-functional glass hall as an entrance area, is an expressive example for this development.

TK Knoten / node / noeud


Fig 17

Fig 21

The glazed barrel vault is 244m long with a span of 80 m. The characteristics of the design are (fig.21): clear hierarchy glass, point fixing devices, single layer barrel vault structure, trussed arches welded nodes, axial member connection through hidden bolts, separation of front walls (fig23) and barrel vault, single-layer shell consisting of bending resistant square grid, without wind bracings, stress free, point-fixed suspended glazing, two safety glass panes with a thickness of 8mm each laminated with 1.5mm PVB foil were used for the glazing surface flush sealing, the glass panes are sealed with extremely elastic silicon strips which are glued to the glass edges.

Fig 19

194

The exhibition halls are connected to another via glazed pedestrian bridges (fig.24). The walkways are covered by a point-fixed glazing of 12 mm curved safety glass. A further example for glazed roofs is the project "Underground Station Canary Wharf, London" designed by Foster &Partners. As part of the work on the Jubilee Line extension in London, three Underground stations in the new office district Canary Wharf had to be roofed in. The structures are covered with glass panels curved around one axis, with dimensions up to 3.1 x 1.1 m, which are point fixed with six rotules. The glass panels consist of laminated safety glass. The general layout of the roof is elliptical in shape. This made the geometry of the connecting details a particularly tricky and daunting task

Fig 22

Fig 25

To cover the foyer area and the winter garden of the Musee des Beaux Arts in Montreal the Bowl Node system is combined with stainless steel rods and cables (fig.26+27). The top chord consists of rectangular hollow sections with bending resistant connections to bowl nodes. Rafters divide the grid into dimensions of approx. 1.5 x 3.0 m. Diagonal and bottom chord members are of high tensile stainless steel rods 16-20mm, directly screwed into the nodes. Struts of circular hollow section 88.9 mm connect the bowl node with the bottom chord ball node.

Fig 24

195

TENSILE STRUCTURES
In the development of pure tensile structures, pre stressed structural elements subject to only tensile loads can be used as a plane surface with great deformation and as an extremely stiff structure with counteracting tension members. An example of a pure tensile structure with point supported glass is seen at Tampines Plaza in Singapore (fig28). The challenge was to create a high-tech glass enclosure supported by stainless steel cables and stainless steel fittings.

Fig 27

Stainless steel cables act as tie down cables against wind uplift forces. The double glazing is fixed on the top chord as "Structural glazing".

Fig 29

196

COMBINATION SPACE FRAME, MEMBRANE AND CABLE-NET


The tensile roof to cover the existing Open Air Theatre of the Bilkent University in Ankara is an example of the combination of MERO space frames, membranes and cable net.

Fig 30

The innovative design is based on a tennis raquet, with glass panels supported by pre-stressed cables. The glass is connected to the cables by double hinged connectors which allows movement and rotation in all directions, without breakage to the glass or ruptures to the sealant. An example for a counteracting biaxial curved tensile structure is the cable net structure for the Rhon Klinikum medical facility covering two promenades connecting different buildings. This allows the visitors and patients to move and communicate protected from weather influences. The cladding of the cable net consists of glass shingles connected to the cable net with special clamps (fig.29+30).

Fig 32

The roof is designed to cover a ground plan of approx. 4000 m . The surface of the cable net rain screen is about 400 m .
2 2

The primary steel structure, the MERO space frame, consists of the main trussed arch with a span of about 118 metres and 6 smaller curved trussed girders with variable spans from 34 metres up to 46 meters which are connected to the main arch (fig.32). The trussed girders are rigidly connected with the main arch and pin jointed at the foundation; the main arch is pin jointed at the base points. The membrane roof consists of 7 individual membrane panels spanning between the curved girder trusses. The surface of the membrane is anticlastically curved to take up wind and snow loads.

197

Fig 34

The cable net facade is tensioned between the main arch and the existing building with nearly no curvature. The cable net is covered with glass shingles to protect the stage area from rain but to allow a view as transparent as possible (fig.33). The shapes of the membranes are designed by means of a form-finding process, considering structural and aesthetical aspects, as well as the boundary conditions. The curvature of the membranes is anticlastic (fig 35) with the main load carrying directions following the lines of principal curvature. In this case the wind suction loads are carried by the hogging traverses (fig 35 A-A) and wind pressure and/or snow loads in the sagging direction (fig 35-B-B). All membranes are pretensioned with tension rods to the Mero nodes of the upper chords of the trussed girders (fig.36). The design of the space frame components, tubular members and spherical nodes, was performed by means of a specialized design program, based on the general approval for the MERO space frame system. This program is covers all design steps - beginning with the geometry of the structural model, the analysis and the evaluation of section sizes and diameters of the spherical nodes - and continuing with the evaluation of parts lists and drawings.

Fig 37

198

The key task for the design of space frames in general and especially for the wide span front arch of this structure, is the appropriate choice of the module size and structural height. These design parameters are critical especially for tube sizes for compression members and the angles of the interconnecting members at the nodes which are mainly determining the node diameters and consequently the economical efficiency of the structure. After a number of tests, it was decided to use a section of 4,5 m height and 3,5 m width at the apex, decreasing towards the base support points to 3,0m height and 2,3 m width, resulting in max. member sizes of 300mm diameter with up to 90mm diameter high strength bolts and max. node diameters of 350mm (fig.37)

A further challenge in terms of membrane structures is the Eden Project near St. Austell in Cornwall, designed by Nicholas Grimshaw & Partners. The buildings will contain different climatic conditions to support a diverse range of plant life. The buildings consist of a number of hemispherical domes which are set against the quarry walls. The diameters of the domes vary from 40 m to 120m and are constructed from two layer reticulated steel members. The domes will be covered with inflated three layer ET -foil-cushions which are hexagonal in plan and the edge lengths vary from 2.1m to 5.3 m. The cushions themselves are held with aluminium frames.

CONCLUSION
Using the advance of technology in engineering and the development of innovative materials it is possible to create new sophisticated high tech quality solutions. The integration of cladding materials such as aluminium and/or titanium sheet metal, glass and membranes covers a very wide spectrum. So it becomes more and more important to consider the structure and the cladding as one package.

199

ENGINEERING THE CONSTRUCTION OF THE GREAT COURT ROOF FOR THE BRITISH MUSEUM
Johan Sischka Dipl Ing Managing Director, Waagner Biro, Stahl-Glas-Technik

INTRODUCTION
The British Museum, constructed in the eighteenth century, is arranged as a quadrangle surrounding the Great Court, the centre piece of which is the historic Reading Room. The Great Court, at the very heart of the British Museum was until very recently occupied by the British Library and is now being transformed by Lord Norman Foster into a new and spectacular public square covered by an elegant glazed roof the size of a football pitch. This new space will form the hub of the Museum providing new gallery space retail outlets and a destination restaurant on the upper level.

GEOMETRY
The roof covers the whole area of the Great Court with a length of 95 m and a width of 74m and spans from the edge of the surrounding buildings to the Reading Room in the middle. The whole structure is formed by a net of triangular cells in a shell shape. Because the Reading Room is not exactly in the middle - it is slightly off towards the north - there is only one line of symmetry and therefore the geometry of all nodes and members is different. The final geometry for the roof net and the structural design was developed by Buro Happold using a form-finding process. An essential factor to determine the net geometry was the maximum possible size for the glass panes from the fabrication point of view considering the required glass performance. The final net is comprised of 4878 members and 1566 nodes within the net, all of them different from each other.

Fig 1

Photomontage; British Museum with superimposed computer rendering, Foster and Partners

Fig 2

Work begun on site in March 1998, with the demolition of the book stacks that surrounded the Reading Room fdling the Great Court which hid the original classical Georgian stone facades from view. The erection of the exciting roof structure began in September 1999 and was completed with the de-propping in April 2000. The project will be complete and open to the public in November 2000. It is the construction, glazing and the complex depropping of the unique roof that is the subject of this paper.

Computer model of system axis of roof structure, 3 dimensional, Waagner-Biro AG 1998

Waagner Biro together with other contractors was invited at a very early stage to look at the project in terms of buildability and special fabrication and installation knowledge. The span in the north is 28,8 m the arch height 5,48 m, the span in the east and west 14.4 m the arch height 5,1 m and the span in the south 23,8 m, the arch height 6,4 m.

200

STRUCTURAL SYSTEM
The roof is supported at the Reading Room and at the perimeter of the museum buildings surrounding the Great Court. Because the Reading Room cannot carry additional loads, 20 new composite columns are assembled in an even distribution around it and they sit on new foundations at ground level. On top of these columns there is a circular steel beam to which all incoming roof members are connected. Slightly below this steel beam a big new horizontal concrete ring is positioned. This ring is rigidly fixed horizontally to the columns and forms an important part of the roof in regard to lateral stiffness. No vertical loads are transformed between the ring and roof structure. The ring sits on sliding bearings on top of the Reading Room, therefore, the ring only gives stiffness but no horizontal support to the roof.

Another consequence with an essential impact to the methodology on how to build the roof, is that only once the whole structure is completed does it work in the way it is designed. This means the whole structure has to be temporarily supported until it is finished.

CALCULATIONS
The form finding process and all the structural calculations allowed for the special support conditions and took into consideration the restrictions of height for the roof. At an early stage both a solid and a hollow option for the roof members were considered. The solid version offered slightly slimmer sections but needed twice the steel weight of the hollow option. Because of this huge extra steel weight, forces and moments, deflections and especially the buckling factor were negatively influenced. After detailed investigations and discussions, the Professional Team gave preference to the hollow version. All these calculations were prepared by Buro Happold and in addition in parallel by the Contractor. This independent analysis not only confirmed the results but was mandatory to enable the Contractor to fully understand the structural behaviour and to develop appropriate connection details and the methodology for the installation. In a final run the section sizes were optimised. They have a standard width of 80 mm because of the glass support detail and their height varies from 80 mm near to the Reading Room to 200 mm at the perimeter according to the structural requirements. To realise sharp edges, and to optimise the dead load of the steel, welded box sections were used, which allowed us to define different plate thickness for flange and web. All calculations were carried out using second order theory and including imperfections. These imperfections with a maximum value of 140 mm - span divided by 200 - were applied using the deflection due to the first eigen-form for buckling of the roof structure to cover the worst case. All standard load cases like dead load, patterned snow load and wind load were investigated and combined in the most unfavourable way. To simulate the structure using the computer model, which is as close as possible to the reality, the Reading Room columns including the snow gallery concrete ring and the vertical posts at the perimeter were also modelled. Member stresses were calculated using the forces and moments out of the elastic calculation with factored loads and second order theory and the plastic section values in accordance with the Eurocode. The first eigen frequency is at approximately 1 Hz, the overall buckling factor approximately 2.9 for factored loads.

Fig 3

Cross section East-West through Great Court with final version of roof, Waagner Biro AG 1999

The roof members at the perimeter are also connected to a continuous horizontal steel beam with vertical posts at a distance of approximately 6 meters. These posts are supported on top of a continuous new concrete ring with concrete columns, which have been built onto the existing wall to distribute the loads. The vertical zone between steel and concrete beam is closed with glass in the upper part and operable aluminium panels for ventilation underneath. It became apparent during an early investigation of the existing buildings that no horizontal loads perpendicular to the perimeter could be transformed to the building. To avoid any horizontal load, all posts sit on sliding bearings. The only horizontal support is realised longitudinal to the perimeter beam with cross bracings between adjacent vertical posts behind the four porticos. Additional stiffness is achieved with horizontal lattice beams at the four corners. As a result of the load-bearing capacities of the existing buildings, these exceptional support conditions have a major influence on the structural behaviour of the roof. The usual "arch effect" cannot be realised because of the sliding bearings at the perimeter, therefore the structure gets significant bending moments and shear forces in addition to normal forces. Stiffness and subsequently deflection of the roof are affected.

201

Another important load case, which had to be carefully calculated, was the deflection due to the dead load of the steel structure and the glass. These deflections, which obviously occur once the temporary support of the roof is removed, have to be compensated to achieve the final geometry at the end, which means that the roof has to be built slightly higher.

THE NODE
One of the main challenges, after understanding the geometrical and structural implications, was to develop a node detail which allows the connection of all members and transforms forces, and especially bending moments through the node. The various angles in between the incoming members, minimum 26 degrees maximum 110 degrees, and the fact that the members are twisted relatively to each other, had also to be taken into consideration. The latter is necessary because of the glass support directly on top of the upper flange. The first proposals for the node design, using round and prismatic nodes and starting with triangular frames for each glass pane which had to be connected on site using intermediate elements, were not successful, neither in terms of the visual appearance nor of the economical aspect. After rejecting versions 1 to 7, including some sub-versions and a lot of sketches which were never tabled, the final idea was suddenly born. It was version 8 - the one to be built.

Fig 5

Computer rendering of extreme node with smallest angle between arms, 26.66, Waagner Biro AG 1999

THE WORKSHOP DESIGN


In parallel to the development of the node design and the member end connection, professional ways to automate the process, in order to generate all the necessary data for fabrication, pre-assembly and installation, were investigated. Because of the amount of different members and nodes, it was obvious from the beginning that standard procedures would not succeed. It was therefore essential to fully define all the steps as to how to design the elements. No manual process should be involved. This requirement was achieved by an exact definition of the geometrical and structural coherence and the complete procedure to follow. The starting point was the net geometry we received from the professional team. To this the calculated deflection for dead load was added in the opposite direction as compensation, which then became the "Zero-geometry". The net was understood to be the centre line on the top surface of the upper flange member. All members were positioned in the orientation to equalise the different angles of the adjacent glass panes supported from the member. Because of the various member heights, not all members connected to one node needed to be the same height. Therefore one height was defined per node which was the maximum height required of all incoming members. If the height on the two member ends was different, the member was tapered. The orientation of the node was defined adding the vectors with a standardised length of all members which were to be connected to the node. The next step was to automatically create the node shape following the pre-defined principles such as the width of the legs. Subsequently the end geometry of all members was defined.

Fig 4

Computer rendering of evenly distributed node with welds, Waagner Biro AG 1999

The node is cut out of a thick plate as a star with 6 legs, all of them perpendicular to the plate surface. The incoming members fit in between two legs and have to be prepared with a complex end treatment. The node is recessed both on top and bottom to avoid it penetrating through the surface of the member flanges and clashing with the glass on top. It also allows a fillet or butt weld along the member flange for the necessary structural connection.

202

FABRICATION OF NODES AND MEMBERS


The same comments as for the workshop design are also applicable for the fabrication of members and nodes. Any non-automated process is unacceptable due to economical and programme reasons. It is also mandatory to get all relevant data to the machines without manual intervention.

The positioning of the nodes within the steel plate was also automated and took into consideration the different node heights and the production sequence. Finally the data file for the cutting machine was generated automatically and sent directly to the machine. Together with the cutting procedure the nominal node centre was marked and each node was numbered. The position of this number defined the orientation of the node in the net. Member fabrication also needed to be automated as much as possible, although this was more difficult due to the complex end geometry. The starting point was to preassemble and tack weld box sections consisting of non standard hot rolled flanges and webs. The web for parallel members was also made from rolled sections for the tapered members cut from plates. At this stage all elements were produced slightly longer. The longitudinal welds were done using submerged arc welding with a double head. The member number and the centre were marked at this stage.

Fig 6

Cutting of nodes out of steel plate 180 m m with autogenous cutting, Waagner Biro AG 1999

The node fabrication could be kept simple by flame cutting the elements out of steel plates. The confirmation that this is an appropriate procedure and the nodes are within an acceptable range of tolerances was achieved at* a very early stage in the project by producing some samples.

Fig 7

Automated computer generated nodes of equal thickness and nesting on steel plate, visualised machine file with markings; Waagner Birrj AG 1999

Fig 8

Adapted welding robot for cutting ends of beams with automatic data feed; Waagner Biro AG 1999

The main procedure was the end treatment of the members and therefore the use of a welding robot was adopted. The member was fixed on a turntable and the end in front of the robot was cut using data generated in the design office and directly brought to the machine. In a second step the member was turned 180 degrees and the second end was prepared. This unique procedure needed considerable development and adjustment of detail before it could finish the members to within the

203

required tolerances. For example, the gas pressure to cut the flange and the web depends on the plate thickness and it had to be varied within each process. The plate thickness, therefore, was part of the automatically generated data together with the end geometry.

LOGISTICS
The fabrication sequence for all members and nodes was determined according to the installation progress. In a first step, a total of 8 zones were defined and then each divided into 2 sub-zones afterwards. Without this consequent procedure, the roof would have become a puzzle with more than 6,000 pieces, although it should be noted that a different sequence would have been more favourable with regard to material wastage but unacceptable from a logistics point of view.

The starting point for the ladders was a triangle as the basic unit. Triangles were formed by fitting together three members and three nodes and adjusting them relative to one another. Before final welding each triangle was surveyed and adjusted where necessary. Special attention was paid to the length of the three sides of the triangle in relation to the centre of the node, which was marked during production. This is also essential because the glass panes are produced using theoretical measures. A simple jig arrangement was used for this straight forward pre-assembly process. After a final check, the welding of these triangles was completed.

Fig 10

Smallest spatial element of roof structure: Triangle, ready for assembly to ladders; Waagner Biro AG 1999

For the pre-fabrication of the ladders, a set of special jigs was developed which had to be repositioned and adjusted in height for every single ladder. Firstly, already pre fabricated triangles were set out and followed by loose members which connected the triangles to each other. Before the final welding, the tolerances for the loose members and the overall tolerances were surveyed and adjusted where necessary.
Fig 9 Plan view of roof grid: Preassembled triangles (filled) and full ladders (bold) with installation phases (colours), erection phases; Waagner Biro AG 1999

A tolerance of 3mm for each side of any triangle and 10mm for one ladder was defined and achieved. Even with this tight tolerance we felt comfortable in achieving site installation of the ladders and the glazing of the insulating glass units without any major problems.

A similar procedure was followed for the glass deliveries.

PRE-FABRICATION
To minimise on-site work, the feasibility of off-site prefabrication was investigated. Taking into account the maximum possible transport size, individual ladders were defined which were then pre-assembled and completely welded under workshop conditions. A total of 152 ladders was finally prefabricated and transported to site with more than 50 truck shipments. With this method, approximately 3,000 out of 4,878 members and all the 1,500 nodes were formed into elements and only the balance of the members was shipped loose to site.

Fig 11

Ladder assembled out of individual triangles; Waagner Biro AG 1999

204

After the completion of the ladder welding another survey was carried out to check whether shrinkage of the welds has caused deformation. Finally, the ladders were sandblasted, and the first and intermediate coats were applied with the node areas being masked, because of the missing connection of loose members which had to be subsequently installed on site.

The whole process was started in the north and continued clockwise and anti-clockwise towards east and west. The working platform was installed in several steps ahead of this procedure. All activities were checked by an independent surveyor, and finally the structure was closed in the south without any problems of tolerance. The paint completion followed the steel welding. The node areas were treated at the beginning, with the rest of the members getting their final coat later.

SITE INSTALLATION OF STEEL


Because from the beginning the roof structure was on the critical path for the completion of the whole project, it would have been beneficial to complete sections earlier and enable following works, such as the paving of the Great Court, to also start earlier. However, the complex structural behaviour of the roof structure did not allow for various advanced options for the installation method. The roof is only stable once the entire steel is completed and therefore has to be supported temporarily during the entire installation process. To allow such a temporary support and at the same time to continue with some work underneath, a watertight working platform was installed at the snow gallery level. This platform was a standard structure using scaffold towers, steel beams and sheeting for the platform itself. On top of this platform, a total number of approximately 600 props were set out to support the pre-fabricated ladders. Spreader beams transformed the prop loads to the steel grid of the platform. The props were standard elements for the construction industry, adjustable in height, with a custom-made head.

DE-PROPPING
The de-propping could only start after the completion of the steel. As mentioned earlier, the whole roof structure was built in a higher position using the "zero-geometry" to compensate the calculated deflection of the structure once it is set free and has to carry its load and that of the glass. The main purpose of the de-propping procedure was to lower the structure in a controlled manner and load it step by step. Because the maximum theoretical deflection is 142 mm, it was not possible to lower sections of the roof completely in one step without overstressing parts of the structure. The conclusion after a very detailed investigation of the behaviour of the roof was finally to lower it in four equal phases. The roof area was divided into eight zones, four diagonal ones, which are much stiffer compared with the rest of the structure, and two zones in front of each portico. The four phases and twelve zones together define 48 individual steps. Detailed calculations proved that after each of the steps no member or node connection was over-stressed. After phase two pre-defined props, which are not loaded any more, can be removed. The amount of roof glazing which is already installed at the start of de-propping influences which props, especially in the south, can be removed earlier and this played an important part in our deliberations. The glass installation itself can be continued during, and finished after, the de-propping.

ROOF GLAZING
Fig 13 Ladders jigged on site with isolated ladder in front of connected net prior to connection to net with individual members; Waagner Biro AG 1999

After setting out the props, the ladders were lifted in and fine-tuned using the adjustable props. The next step was to fit in the loose members between the ladders, check the tolerances of these triangles and finally weld the connections.
1

The roof glazing is comprised of 3312 glass panes, each a different size. They are supported directly by the main steel structure comprising a total area of 6.000 m . Because the Great Court is covered completely with the roof, high performance glass was required to ensure comfortable interior conditions throughout the year. This was achieved by using insulated glass units (IGU) with fully toughened 10 mm glass outside, an 16 mm air cavity and on the inside annealed laminated glass of 2 x
2

6 mm. Additionally the outside pane is body tinted green and screen printed on the inside with a density of 57 %. The inside pane at the face to the cavity is low-E coated. This combination results in the following performance data: r light transmission of 29 % outside reflection of 23 % total energy transmission 23 % shading co-efficient / G-value 0,26 U-value 1.9 W/m K
2

bubble gasket were designed around the edges to cover part of the cavity in between the panes and to minimise the amount of mastic silicone applied on site. This solution allows the glass panes including the gaskets to be finish in the workshop and therefore the main parts of the sealing works are carried out under workshop conditions. Additionally the gaskets worked as edge protection during transport.

The IGU with the above performance data is sufficient to achieve the required conditions inside the forum without any additional shading devices.

One of the major design criteria was to have a flat surface on the outside without any up-stands for the glass fixation along the edges and on the corners. This was achieved by resessing the upper glass pane within the IGU and fixing the glass with a clamp which sits on top of the annealed glass. This clamp is fixed with a bolt which is screwed into a threaded stud welded to the members at regular intervals of 500 mm. These fixations are hidden within the joint between adjacent glass panes. This solution does not mechanically fix the outside glass pane, but uses a structural silicone connection to secure the outer glass pane against uplift.

Fig 15

Application of extruded silicone profiles to glass edges with special edge protection closed cell silicone foam profile; Waagner Biro AG 1999

60 mm 10 mm I

A big challenge was to handle the logistics to produce and deliver the glass panes in the necessary sequence and to mark and identify the number and orientation of the panes. After the first parts of the steelwork for the roof were done the glazing operation started and followed the sequence of the steel. The glass installation itself started with installation of the carrier gasket followed by the positioning of the glass panes and finished with the mastic seals between the glass panes. For handling and installation of the glass several options were investigated. One was a specially designed platform penetrating the roof and being supported from the working platform with a jib crane on top that has a reach of 7 m. This platform was repositioned using the existing tower cranes. Another option was to install the panes directly using the tower cranes, and the third option was to use a specially designed gantry system which was positioned in rows and relocated to install complete rows running from the perimeter to the reading room. Finally all three methods were used to achieve the best results.
Fig 16 Sealing of Nodes with solid silicone plate in cross point, joints closed with final silicone seal. Waagner Biro AG 1999

Fig 14

General layout of joint between IGU after installation on the net. Edge gasket preapplied before dilivery to site; Waagner Biro AG 1999

To keep the detail tight and compact, only a carrier gasket was applied onto the upper flange of the steel members and the glass sits directly on top of the carrier gasket. This carrier gasket detail was designed to cover the various angles between the glass panes. A flip and a

206

Because of the special shape of the roof a standard manual cleaning operation was preferred rather than the use of machines or special access equipment. For the outside of the roof access is provided by gutters that run around the perimeter beam and by a service trolley at the snow gallery. Additionally a Latchway system was introduced which runs along the apex of the roof. The inside cleaning is planned to be done using cherry pickers operating from ground level.

Around the perimeter of the roof a gutter is incorporated into the vertical facade. This allows for the drainage for the outside part of the roof between the apex and the perimeter. A special down pipes system connects the gutters with the water management system of the building.

FACADE TESTING
To certify all theoretical considerations and to verify the sufficiency of the proposed technical solution a mock up consisting of nine triangular roof bays and the respective part of the facade including vertical glazing, aluminium panels and the gutter was built and tested at the Arsenal, a specialised Viennese laboratory for building technology. A series of tests including water and air tightness in combination with wind suction and pressure and structural safety tests were conducted. The mock up finally passed all tests successfully, the proposed design was proven by the tests. Besides standard tests during fabrication and installation another site water tightness test for agreed sectors of the roof was carried out after the completion of the structures.

VERTICAL GLAZING AND ALUMINIUM PANELS


The gap of 1,5 m to 2,0 m between the concrete beam on top of the existing buildings and the perimeter beam of the roof itself is closed with a combinati6n of glass and aluminium panels. The aluminium panels are positioned in the lower part up to the top of the concrete columns and are openable to allow for natural air ventilation. To open these panels electrically driven actuators are employed which are operated by the building control system. To avoid bird infestation a stainless steel mesh is installed in front of the panels. The upper part of the vertical facade - the celestery glazing - is closed with fixed single laminated glass. A similar glazing system is used around the reading room.

Fig 18

9 Bay Mock-Up with openable panels in vertical facade at the independent testing institute arsenal research after testing; Waagner Biro AG 1999

Fig 17

Cross section of vertical facade under Perimeter Beam; Waagner Biro AG 1999

207

SUMMARY
With the support of computers, it is now possible to realise such complex projects as the Great Court roof within acceptable cost parameters which are not too far away from comparable costs for more standard solutions. Of course, software is the ultimate basis of the design but custom-made programme modules also have to be developed and therefore highly qualified personnel are required for this purpose. It is also necessary not only to handle the geometry from the design point of view but also to have a continuous information chain through the entire project, including fabrication and installation. The method of fabrication, pre-assembly and installation was specially developed for this project and is without precedent. The combination of all these elements, together with the commitment of a highly motivated team, is mandatory for the successful realisation of innovative architectural concepts. Last, but not least, a close working relationship and co-operation with the Professional Team is of vital importance.

Fig 19

British Museum Great Court Roof, steel structure completed 15.3.2000, glass installation app. 6 0 % ; Waagner Biro AG 2000

208

A HISTORY OF WIDESPAN STRUCTURES IN THE UNITED STATES


Robert Silman President, Robert Silman Associates, New York and Adjunct Professor of Architecture, Columbia University

INTRODUCTION
The history of widespan structures in any country is a reflection of its needs and requirements as well as its industrial capacity. Thus one would expect the origin of widespan structures in the USA to have lagged behind those of the more developed European countries of the 18 and 19 centuries. Indeed this is true we in America did not have the great cathedrals or palaces or exhibition halls to rival those of Britain and France. Our needs were quite different and in fulfilling our somewhat questionably formulated doctrine of manifest destiny, the demands were to provide easy access to the enormous expanses of the country in addition to facilitating local connections for the growing cities, most of which were founded adjacent to rivers, lakes or the sea. The means for meeting these needs was the construction of a large number of bridges to carry the road and rail traffic in all parts of the country. It is my hypothesis that almost every great breakthrough in the history of widespan structures in the USA was first demonstrated in bridge > construction, and only later in buildings. The one exception to this statement is fabric structures.
th th

stress as the loads moved across the span, to the use of wrought iron posts and diagonals and to sophisticated connections. In building design, timber gable and gambrel roof trusses were developed to parallel the original appearance of various truss types in bridges. The joinery skills required to make the truss joints were considerable and this was usually the limiting factor in their use. Medieval European truss construction of large barns as well as the cathedrals set a valuable precedent for timber trusses. It was realized that triangular panels were the most stable components and roof trusses were usually composed of an aggregation of these triangles. Of course most of the early widespan buildings were roof structures, not floors. The pinnacle of 18 century timber construction was an unusual array of timber girders used in the floor of Philadelphia's Independence Hall [1732-1748]. The gambrel trusses roof of Independence Hall are a masterpiece of ingenuity, utilizing tie rods down to the attic floor to engage it as the bottom floor of the truss. However, the greatest challenge in timber structures was to create widespan floors. In most cases where the depth was limited, the design and construction of floors became a reflection of the availability of timber beams and girders and the ability of tradesmen to handle them on site. The two major column-free spaces in Independence Hall are 12.20 m (40 feet) square, framed with four massive 280 x 305 mm (11" x 12") girders in a pinwheel fashion. This "fireman's basket carry" arrangement is difficult to justify theoretically, but of course it has performed successfully for 260 years. 19 century timber construction was dominated in the USA by the development of the balloon frame house. This was anything but widespan construction, utilizing small dimension members closely spaced and fastened with multiple nailed connections. There was no need for skilled joiners in building these houses as evidenced by their proliferation. However widespan timber/iron composite structures were being constructed as train sheds for the burgeoning railroads. Truss spans of approximately 30 m (100 ft.) became commonplace, first at Lowell, Massachusetts (1835) and later in Chicago and other cities. Arched vaulted timber trusses with wrought iron tie rods reached spans of as much as 46 m (150 ft.) in Philadelphia (1852) and ultimately 50.6 m (166 ft.) in
th th

TIMBER
The early use of timber in widespan bridges was mostly in the configuration of a truss, or a truss in combination with an arch. The mathematical theory of truss design, developed in France during the first half of the 19 century, was not understood by the men who built bridges in the United States. Their empirical designs of multi-panel trusses were often highly redundant, imbued with the notion that some members would deteriorate due to weathering or insect damage and that others would pick up their load. It was also expected that the bridge would often require repairs or upgrading. Early Palladian arched trusses built in the years 1792 -1806 by Timothy Palmer achieved spans of up to 195 feet. Combinations of trusses with arches bolted to their sides (Theodore Burr) achieved spans of over 200 feet as early as 1806. And large laminated timber arches by Lewis Wernwag spanned up to 340 feet in Philadelphia in 1812. Experimentation in the design and construction of these timber bridges led to trusses with single diagonals, then with two diagonals (X-braced) to allow for reversal of
th

209

Chicago's Grand Central Station (1856), making it the second widest span in the world, after New Street Station, Birmingham, England at 64.3 m (211 ft.). One of the greatest timber roofs ever built is the Mormon Tabernacle in Salt Lake City (1868) whose overall dimensions are 45.7 x 76.2 m (150 x 250 ft.). This roof is framed with segmented vault lattice trusses 2.74 m (9 ft.) deep. There are two pairs of top and bottom ribs, two intermediate pairs of ribs and innumerable lattice diagonals to create this rigid, highly indeterminate structure. Of course with the advent of steel, widespan timber structures in the 20 century became uneconomical and quickly dropped out of fashion. Engineered wood products such as glue laminated timber and sophisticated timber connectors have created a renaissance in modest spans where the visual appearance of wood is a desirable characteristic. Parabolic and Tudor arches as well as lamella type vaults and domes are now common.
th

ring was completed they knelt upon it and laid up the next one. Their trademark was the use of herringbone pattern groined vaults, such as the ceiling of the Registry Room at Ellis Island Immigration Station in New York.

IRON AND STEEL


Clearly American engineers have made their greatest contribution to widespan structures using iron and steel. Although we have nothing as romantic as the cast iron of Iron Bridge, there is a long history of early uses of wrought iron in the USA. At first it was combined with timber in bridge trusses, where members carrying the highest stresses were made or iron while the more lightly stressed were timber. Then truss bridges built completely of wrought iron began to appear, especially when the locomotive loadings became too great for timber members to handle. In the eastern US, where the iron industry was developed, truss bridges appeared commonly by the middle of the 19 century. However in the western US where timber was plentiful, iron and steel did not replace the huge timber trestles until the end of the century. Patented trusses by engineers such as Pratt, Whipple and B oilman were constructed in the period J850-1875 with spans up to 38 m (125 ft.). Shortly after 1875 the need for longer spans became apparent when waterways such as the Ohio River needed to be bridged. Truss spans suddenly increased to more than 150 m (500 ft.) as mathematical analysis became readily available.
th

MASONRY
Masonry arch bridges were quite common from the 17 to the 19 century. However there was nothing terribly dramatic about their use in the USA. Both brick and stone were utilized and as the quality of the mortar improved over the years, so did the span length and the durability of the bridges.
th th

By the time that the need for widespan building enclosures was created in the Americas, materials other than masonry were available to create these structures. Besides, there were few masons skilled enough to construct large scale arches, domes or vaults. Among the monumental masonry vaults of note is the unusual floor of the rotunda in the U.S. Capitol by Benjamin Latrobe and Charles Bullfinch, built during the first half of the 19 century. Here a system of radial groin vaults and annular vaults forming double concentric rings supports the central floor. Later the SubTreasury Building on Wall Street in New York (1842) was patterned after this floor. The Sub-Treasury also had a handsome 18.3 m (60 ft.) marble dome roof.
th

One of the two most dramatic bridge designs in the 19 century was the great tubular steel arch bridge spanning the Mississippi River in St. Louis, designed by James Eads and completed in 1874. There are three trussstiffened arch spans of 153, 158.5, 153 m (502, 520, 502 ft.) built without hinges. The arch ribs are trussed together with wrought iron braces, however the main 400 mm (16 in.) diameter tubular sections are composed of carbon steel and chrome alloy steel staves riveted together. Because the steel industry was so newly established, Eads specified that every piece of steel be tested for mechanical properties before it was allowed to be installed. This bridge employed the first use of steel in a major structure in the USA. The other imposing bridge structure to be remembered from the 19 century is John Roebling's Brooklyn Bridge. Parallel wrought iron wire suspension bridges were pioneered by Charles Ellet, who built a 109 m (358 ft) main span in Philadelphia in 1842 and a 308 m (1010 ft) main span in Wheeling, West Virginia over the Ohio River in 1849 (later rebuilt and stiffened by Roebling in 1854 after suffering wind damage). Ellet designed a number of other significant spans up to 366 m (1200 ft) that were never built. Roebling received a formal engineering education in Germany prior to emigrating to the US in 1831. He founded a wire cable manufacturing company in 1849 and combined this commercial venture with an interest in suspension bridges that dated to his
th

th

Perhaps the most unique structural contribution to widespan enclosures in masonry from the USA came from the Guastavino family, father and son, Rafael Guastavino Sr. and Jr. Rafael Sr. emigrated to the USA from Spain, carrying with him the traditional skills of Catalan masonry vaulting. During a 74 year period from 1889 to 1963 thencompany built more than 1000 structures, including some significant widespans. Using clay tile of their own manufacture and high-strength, quick-setting mortar, they were able to build some significant vaulted structures very competitively and very quickly. In fact some of their domes, such as the 20.1 m (66 ft.) diameter Cathedral of St. John the Divine in New York, were constructed without the use of centering or formwork. Masons built rings of overlapping tiles, usually three wythes thick, and when one

210

thesis as a student. His first five suspension bridges carried aqueducts; the one at Lackawaxen, Pennsylvania (1849) is a four-span continuous structure that has been preserved as a national landmark. He went on to build a 250 m (821 ft) span at Niagara, a 322 m (1057 ft) span at Cincinnati (1867) and finally, what most agree to be the greatest technological feat of the 19 century in the western hemisphere, the Brooklyn Bridge in 1883 at 486 m (1596 ft). The Brooklyn Bridge eclipsed the record for long spans by more than 50% in one leap and successfully solved the problem of aerodynamic instability. Moreover it was the first use of steel wire in suspension bridge cables.
th

thrusts at the base. The mammoth roof of the St. Louis Union Station (1894) covered an area of 183 x 213 m (601 x 700 ft.) but was divided into five spans. These huge vaults were built only over a thirty year period as the coal-fired locomotive gave way to diesel-electric generation starting around 1910. In England and France the great exhibitions provided the venue for innovations in widespan buildings. From the Crystal Palace to the Galerie des Machines engineers were able to demonstrate their achievements. Although several opportunities were available in the USA, not much of significance came from exhibition structures. The Columbian Exposition of 1893 in Chicago was an architectural victory for the Beaux Arts, suppressing any notions of modernism or exposed structures. Louis Sullivan said, "the damage wrought to this country by the Chicago World's Fair will last half a century," unfortunately an accurate prophesy. It did contain, however, the vast Liberal Arts Building with a span of some 100 m (328 ft,) and a height of 61 m (200 ft.), making its volume second only to the Galerie des Machines. Widespan trusses were used in several prominent buildings. The Produce Exchange in New York (1884), an early skeleton frame by George B. Post, utilized wrought iron trusses above the open trading floor at the second level, carrying four floors of offices above over a 17.7 m (58 ft.) span. The great Auditorium Building of Adler and Sullivan in Chicago (1889) used both flat and elliptical trusses to span the 4,000 seat theater which was 36 m (117 ft.) wide. The dome of the U. S. Capitol, finished in 1864 after the Civil War, was a great work of cast iron meridional trusses. The top of this dome is 87 m (287 ft.) above the main rotunda floor. These 19 century widespan buildings paved the way for the numerous, and sometimes fantastic, structures of the 20 century. The great train sheds were replaced by airline terminal and hangar buildings. As cities grew and land prices soared, it became expedient to build on top of existing buildings or services, requiring serious widespan structures. Mass entertainment created the need for large theaters, auditoriums and stadiums. Expositions and fairs offered the chance for the USA to show off its industrial prowess and real estate developers often needed no excuse to create showy widespan structures. It would be nothing but a catalog to list many of the widespan structures built in the last half of the 20 century. Instead, this paper will select some of the more unusual efforts and concentrate on their techniques.
th th th

The 20 century saw many longer bridge spans in trusses, arches and suspension bridges. However the theory had been established by the 19 century pioneers. Trusses reached spans of 244 m (800 ft) and cantilever trussed bridges, patterned after the Forth Bridge, spanned up to 480 m (1575 ft). The great steel arches of Hell Gate, Bayonne and New River gorge near Charleston, West Virginia went from 298 m (977 ft) to 518 m (1700 ft). In suspension bridges, the spans increased from the George Washington Bridges of the 1930's at 1100 m (3600 ft) to the Golden Gate, the Mackinac Straits and the Verrazzano Narrows at more than 1280 m (4200 ft).
th

th

How did each of these advances in steel bridge design and technology find their way into widespan buildings? Of course steel was the material that architects and engineers were waiting for to enable their large scale projects to become reality. Its performance per pound was ideal. Train sheds were the first group of significant widespan iron and steel structures in the USA, much as they were in Britain. The boom in railroads started after the Civil War, such that by 1900 there was 150,000 miles of line. Whereas widespans were not a particular requirement, great height was necessary to dissipate the smoke from the coal-fired locomotives. To make the proportions of these sheds reasonable, they were designed with widespans to compliment the required heights. Gable trusses of wrought iron bearing on masonry walls in Washington and Jersey City spanned up to 43 m (142 ft.). But with the vision of St. Pancras to inspire them, American engineers soon tried the balloon shed, where the trusses or arches sprang directly from the ground or were integrally designed with iron or steel columns. In 1871 the first Grand Central Station in New York opened with a train shed spanning 61 m (200 ft.). The bottom chords of the wrought iron, latticed, fixed Howe trusses sprang from the ground and were tied across at their bases by iron tie rods extending under the tracks. In Philadelphia in the 1890's, two great three-hinged wrought iron arched train sheds were constructed. The largest was the Pennsylvania Railroad's Broad Street Station with a span of 91.7 m (300.7 ft.), making it the third largest span in the world in 1893. These two sheds utilized steel beams beneath the tracks to resolve the arch

Perhaps the most interesting contribution of the American engineers was the use of cables to achieve long span roof structures. Taking off from where the bridge builders had left off, and often using their hardware and technical know-how, the cable supported roof was able to achieve interesting stylistic characteristics as well as offering reasonable economy. Certainly the

211

"performance per pound" of cable roofs was unrivaled. The first significant cable roof was Matthew Nowicki's design for the Raleigh Arena. Because of his untimely death in a car crash, the building was finished by Fred Severud in 1953. (It was only seven years after this date when I worked together with Ted Happold in Severud*s office.) The cables are suspended from two intersecting arches, with one set serving as the main supports, spanning a maximum of 99 m (325 ft.), and the second set place orthotropically and curving downward to prestress the roof and eliminate any tendency to flutter. The supports for the cables are two giant intersecting canted concrete arches. Fred Severud was fond of comparing the structural action to two men holding hands and leaning backwards, who counteracted the pull on their arms by transferring the loads through their bodies down their legs and into the ground. Severud went on to design the hockey rink at Yale University using a single concrete spine with cables draped to both sides. Another, but more symmetrical structure was the Utica (NY) Memorial Auditorium by Lev Zetlin in 1959. Spanning 73 m (240 ft.), the circular roof was composed of a conventional compression ring around the outside, tension ring in the middle, suspended cables sagging downward to carry the load, but with the unusual design of a second set of cables given an upward curvature by separator struts off of the lower cables. This second set of cables was prestressed to take out the flutter. A few years later, Severud completed the Madison Square Garden arena in New York using the dead load of the mechanical equipment room placed in the center of the round cable roof over the tension ring, to eliminate any aerodynamic instability. At Dulles Airport (1962) outside Washington, DC, the stability of the single curvature roof cables was assured by engineers Ammann and Whitney by using relatively heavy precast concrete panels for the roof deck. Perhaps the most unique widespan structure to emanate from the USA was the geodesic dome, designed by a truly free frontier spirit Buckminster Fuller. Patented in 1947, Bucky Fuller invented a truly original form, not indebted to any structural precedents. The component elements are all straight pieces, mostly alike, capable of easy mass production, light in weight, inexpensive to fabricate and erect. Of the many geodesic domes actually built, we will look at only two. The first is the huge volume of the Union Tank Car repair shop in Baton Rouge, LA, built in 1958. With a span of 117 m (384 ft.) and a height of 39 m (128 ft.), it was the largest enclosed space in the world when finished, yet the total weight of structure is only 567 tons or 48 kg/m (9.8 psf) of covered floor area. The other prominent publicity that Fuller garnered was for his designs of USA pavilions at world's fairs, most notably in Moscow in 1959 and Montreal in 1968. By using plexiglass glazing, the buildings had unlimited natural light by day and they glowed by night. Their immense scale, achieved by using tiny components, was awesome to most visitors. Other steel domes were notable in serving as coverings
2

for sports stadiums. The Astrodome in Houston was the first major covered baseball stadium in the USA. Many roofed, climate controlled stadiums followed, with the largest being the mammoth Superdome in New Orleans, a six layered lamella steel dome spanning more that 213 m (700 ft.). More recently the Georgia Dome, built for the Atlanta Olympic Games in 1992 by Weidlinger Associates, is a fine tuned cable-strut structure covering an elliptical area 235 x 186 m (770 x 610 ft.).

CONCRETE
Engineering design of widespan concrete structures in the USA clearly has lacked the elegance demonstrated in other parts of the world. With a concentration on speed of construction for conventional flat plate buildings because of high labor costs, the use of concrete for widespans has been relatively limited. We have had no Nervis or Torrojas or Candelas. One of the earliest concrete widespans in the USA was the Hershey (PA) Arena (1936) by Anton Tedesko of Roberts and Schaefer. The 69 m (225 ft.) wide concrete shell was stiffened with heavy reinforced concrete arches spaced quite closely. The architect Eero Saarinen became interested in qoncrete shell structures and two of his early examples are the Kresge Auditiorium at MIT and the TWA Terminal at JFK in New York. Neither can be called a "thin" shell; both are quite "clunky" in terms of engineering elegance but they are rather dramatic structures. Kresge is a truncated sphere, supported only on three points. A more free form structure is the TWA Terminal, designed by Saarinen shortly after his return from Australia where he served as a juror on the Sydney Opera House competition. TWA's bird shape is deceiving, for two bands of skylight glass really separate the roof into three parts, a concession to the indeterminate analysis problems of designing a free form in the early 1960's.

FABRIC A N D AIR-SUPPORTED STRUCTURES


One of the great contributions of American engineers was the pioneering of fabric and air-supported structures. However the history of these structures and their development, not only in the USA but throughout the world, will be the subject of a number of other papers at this symposium. So it will be left to others to discuss.

212

SUMMARY
As the USA became the largest industrial nation in the world in the 2 0 century, the need for widespan structures developed quickly and the ability to fulfill the demand was available, both conceptually and practically. Many unusual forms were developed or refined by American engineers. However as the 20 century wore on, and now in the 2 1 century, no one country is in the ascendancy with regard to widespan structures. They are happily well distributed throughout the developed world and the ideas of their designers are available for all to use. We hope that we will continue to have our share of interesting widespan structures in the future, even as we keep in mind other more pressing needs such as sustainability and quality of urban life.
th th st

REFERENCES
Condit, Carl W. American Building* Second Edition, The University of Chicago Press, Chicago, Illinois, 1982.

DeLony, Eric. Landmark American Bridges* American Society of Civil Engineers, New York, New York, 1992.

213

MATERIALS FOR THE NEW MILLENNIUM


Nick Goldsmith

The rich Chinese businessman leaves his hotel in Montreal and gets into his 72 kilogram Russian chauffeur driven 2148 kilogram Mercedes 300 SE whilst carrying his French ultra light carbon fiber reinforced plastic briefcase. Unknowingly he demonstrates that we can observe strange paradoxical clashes between the formerly accepted idea that there is no end to the earth's resources, and the seeds of insight that lightness may help improve the quality of life. Slowly but inevitably we are becoming aware of the fact that the price we pay for energy is unjustifiably low and part of the slash and burn culture of that has become too familiar to be noticed. In the near future lightness will, however, once again turn into an accepted starting point for the way we construct things." This quote from the book, Lightness by Adriaan Beukers and Ed van Hinte, underlies the notion that the history of architecture has been a slow and steady path from mass to membrane. Today using lightweight materials is no longer a visual aesthetic, it is becoming an economic reality. Over the past hundred years, the development of many new materials such as steel, aluminum, titanium and synthetic ceramics have changed the way we build. Woven architectural fabrics have been around longer than most other materials. The Romans used woven leather tents in their campaigns across Europe and nomadic cultures have used fabrics as cover for thousands of years, be it the North American Teepee, the black goat hair tent of the Bedouins, or the Mongolian yurt. But all these traditional tent structures relied on materials that lasted only a few years. With the advent of industrialization, fabrics did not translate into an acceptable building material because of ephemeral qualities. Fabrics and foils as building materials were revisited in the nineteen sixties due to a cultural interest in inflatables, instant cities and alternative lifestyles. From the inspirational drawings of Archigram in the UK, the pioneering lightweight structures of Frei Otto in Germany, the influences of Bucky Fuller and his tensegrity structures, the inflatable structures of Walter Bird in the States, 20th century fabric structures were born. Industrial designers started to design inflatable chairs, pneumatic structures and one-piece injection

molded plastic furniture. Otto used cotton and polyester fabrics for his early garden show structures and Bird used urethane-coated polyester for his early radome enclosures. Eventually PVC coated polyester fabric was adopted by both, which is the ancestor of today's polyester fabrics, and today the most widely used membrane structure material. But the evolution of woven membranes continued. From space exploration, we have seen the introduction of Teflon coated glass fibers, which can withstand tremendous temperature differences and harsh outdoor climates. The material was adapted to be used in baking oven conveyor belts and during the past twenty years has 'become a permanent structural membrane for many larger architectural projects including the new Millennium Dome in Greenwich. From the yachting industry, we have seen the introduction of Kevlar with a greater strength-to-weight ratio than steel. From the airplane industry, Spectra fibers have been introduced into an architectural parlance. Vectran, a high performance thermoplastic yarn spun from liquid crystal polymers is beginning to be used in architectural applications. In addition, many materials that we are familiar with such as stainless steel can now be woven into textiles that embody tremendous structural and architectural potential.

How as designers should we use these new materials? What are the intrinsic qualities in these wovens that we should appreciate? What are their spatial implications? Unlike other materials, fabrics and cables can only have strength by being placed in tension (the opposite of bricks or stone); this applied tension is called prestress. Since the thickness of the fabric remains constant throughout the structure, it is desirable that the tension placed on the fabric is similar in both woven directions, that is the warp and the weft. To achieve this state of equal prestress, a design process starting with physical models, proceeding to computer "formfinding" models, progressing to load analysis, developing details based on this geometry and finally generating a series of cutting patterns, has been developed.

214

In addition to the notion of prestress, lies a second crucial maxim, which is necessary in understanding woven structures. Due to external pressures such as wind or snow, which affect these structures, designers must use the notion of opposing curvatures to gain strength in the material. Upward curves resist snow and downward pressures; downward curves resist uplift winds. Together they describe a saddle or cone shaped surface which are the base building blocks of this material. The tectonic quality of these elements describes radically different spaces. Hyperbolic, parabolic, and conoid forms all feel dramatically different from the box. In addition, the translucency of the woven material adds a diffuse luminous quality to the spaces; it retains the exquisite beauty of being under a tent. The least expensive fabric material for architectural applications is the woven polyolefins such as polyethelene and polypropylene. They are coated, flame retardant, UV resistant and are fabricated up to lOoz/square yard weight. Their lifespan however is relatively short (5 years or less). Our experience with this fabric is the Titan series for Coverall, which we designed. This wide span product is a steel truss frame system, which reaches spans of up to 180 feet. Its installed cost in North America is under $15 per square foot (fig. 1).

Fig 2

Tension tent for Armbruster (fig. 2). These products generally use a PVC laminated fabric since they are designed for lifespans of five years or less and relative cost is a driver in this market. In the field of deployable and relocatable structures, the PVC coated polyester fabric is widely used. Reasonably priced, this fabric can easily fold up. Among our numerous deployable structures we have designed over the years, the Cadillac Exceleration Center is a typical example of the use of the PVC polyester fabric. The Cadillac Exceleration Center is a traveling theater facility specifically designed for easy one-day erection and dismantling (fig. 3). Using easily transportable fabric, steel and aluminum components which unfold from the back of two semi-trailers, the facility transforms into a main theater space of approximately 3,000 net square feet and consists of a free-form fabric enclosure, the envelope of which approximates a 70 foot diameter cone

Fig 1

PVC coated polyester fabrics are probably the most common architectural fabric used today. It is either PVC coated or laminated onto a polyester base fabric. Dozens of different manufacturers provide such a material, which range from the laminated fabrics for party rental tents to heavy-coated fabrics for permanent (15-year replacement cycle) architectural installations. The fabric comes in numerous colors, has three different topcoatings (Tedlar, Fluorotop T, Acrylic) and is considered as a fire resistive material. As part of our product design work, we have designed and engineered several tent systems in the party tent field, including the Genesis and elite series for Eureka!, the Anchor Mod and Century series for Anchor, the

Fig 3

supported at a height of 60 feet by a steel truss tripod. Perimeter masts and tiebacks secure the boundary shape and are anchored into the ground to maintain the geometry under anticipated wind and weight bearing loads. The fabric divides into three 1000 square foot sections, which store on standard palettes. Seating approximately 125 people with stage capacity for one automobile in full view and one automobile on each side of the stage and off ramps, the facility is used to train

215

and educate dealers on the merits of new Cadillac autos. Two adjacent ancillary structures for dining and seminars have a capacity of 30 to 40 people. A smaller centralized reception/lobby structure combine the three structures. Designed to be installed on-site at temporary locations primarily large parking lots within twelve hours, each of the two "Exceleration Centers" that were constructed covered five cities on the East and West Coasts during a seven week U.S. summer tour. Another interesting lightweight building material is Kevlar fiber Air Tubes. These high pressure air tubes can take on the support function of a beam, an arch, or a grid becoming a type of frame structure. When this type of support is joined with a tensile membrane structure, a symbiosis occurs creating a stable structure greater than the sum of its parts. Since the air pressure in the tubes is inflated using a compressor and is capped off after inflation, the development of the material technology for the actual air beams becomes critical. Tubes made of welded seams loose pressure and stability over long periods of time. New materials and joining processes needed to be developed. The FTL Happold design for a series of innovative structures for the US Army in collaboration with Natick Research and Development, allowed the exploration of this nascent technology. The Transportable Maintenance Enclosure (TME) and Light Area Night Maintenance Shelter (LANMAS) (fig. 4) projects are ongoing explorations, which have developed the air tube technology. We used Kevlar (para-aramid) fibers which are braided into curved forms and bonded to an inner urethane membrane to create seamless inflatable arches of approximately 30 psi spanning 30 feet without support. The Kevlar fabric is enclosed with a PVC cover to protect the fibers from ultra-violet degradation. Kevlar was used as a material in this design because of its excellent tensile strength to weight ratio and its low elongation values. These Kevlar fibers consist of long molecular chains, which give excellent dimensional stability.

We have also used Kevlar rope as edge rope for polyester membrane structures. Based on technologies developed initially for the yachting industries, Kevlar rope is used as an excellent edge detail for tensile structures. Unlike traditional steel cables, which are relatively stiff at large diameters, Kevlar ropes maintain flexibility at any scale. This allows membrane structures to be designed as deployable and relocatable structures where they are folded up and moved from site to site. The Carlos Moseley Music Pavilion in New York by FTL Happold was designed for the Department of Cultural Affairs, the Metropolitan Opera and the New York Philharmonic as a traveling performance pavilion which sets itself up in only three hours (fig. 5). Six semi trailers with concrete foundations poured in the trailers constitute the fleet of trucks. The pavilion travels from park to park in New York and can facilitate performances every 24 hours at a different site. Kevlar rope jacketed in a polyester braid was used both in the edge cuff of the fabric and the exposed catenary cables to allow for quick installation.

wkw

AMWV^

m m

HER
Fig 5

Teflon (PTFE) Coated fiberglass fabric is the most permanent of the coated architectural fabrics. First employed as a roof in 1974 for the La Verne College Student Center in California, it has a lifespan of over thirty years. It can be used only for permanent applications and is not relocatable. The fabric is considered non-combustible and as such meets the most stringent building codes worldwide. Off the roll it has an oatmeal appearance which bleaches out to white in the Sun after a couple of months. With translucencies of up to 20%, it has been used on such projects as the Georgia Dome in Atlanta, the Denver Airport and is currently being used for the new Millennium Dome in Greenwich, England. It has a low coefficient of friction, is chemically resistant, and a high UV resistance. A recent application of this material is the new Cirque de Soleil theater at Disney World in Orlando (fig.6). The design brief was to make a circus theater that was a transitional step between the big tops of traditional

Fig 4

Fig 6

circus tents and a contemporary theater. Initially we looked at making the roof of the theater drum a tensile membrane, but due to performance criteria concerning machinery and systems in the roof, we were requested to work on the facade only. Our approach to the facade was to create an illusion that the theater was clad in fabric. Joining onto the drum shaped theater required that we develop the curvature by moving in and out with the tie down points on the surrounding plaza. The fabric was Teflon glass fabric supported by painted steelwork. Glass bottles are rapidly being replaced by plastic ones and soon water will be shipped across oceans in large plastic bags. In a similar way glass roofs will be replaced in many cases by foil roofs. Unlike woven industrial fabrics, which consist of a substrate of woven fibers' sandwiched between film or coatings, polymer films consist of thin extruded sheets of plastic. On their own these large sheets are too flexible, however when two or three layers are sealed and inflated like a pillow, an efficient structural unit is created. Depending on the strength and thickness of the unit, these cushion units can be made up to 12' widths and unlimited lengths. Supported by a primary structural system such as a cable net or frame, air pillows provide a cladding system that has advantages over glass and fabric structures. Translucencies range between 25% and 95% of light transmission. The Eastleigh Tennis Center located near Southampton, U.K., and designed by the Architects, Euan Borland Associates and Buro Happold Consulting Engineers, is an example of this technology (fig.7). A key factor in the design of the roofs was that they needed to provide lighting comparable to daylight at a cost less than a traditional glass roof. The building houses both a swimming pool and tennis courts using both double and triple layer membranes for thermal insulation. The foil cushions were transparent and printed with a dot matrix to reduce solar gain inside, similar to fritting on glass panels.

Fig 8

The idea that metal can be structured in the same manner as textiles is not new. It has existed for centuries in different places and cultures. Today with the use of stainless steel wire rope, woven metal mesh fabrics are available to be used in architectural applications both as interior textiles and as exterior shells. Stainless Steel can be woven as wire rope to form membrane fields or butt welding them together can fabricate crimped wire cloth in large areas. New developments in woven stainless steel technology using slit film and "melt o f f technologies are moving from the laboratory and art studio to architectural applications. (Structure and Surface, Contemporary Japanese Textiles by Cara McCarthy and Mathilda McQuaid, Museum of Modern Art, New York 1998).

217

The Russell B. Aitken Seabird Aviary that we designed and engineered for the Bronx Zoo is an example of a woven wire rope net (fig. 8). Using a series of galvanized parabolic steel arches spanning 90 feet and covering 30,000 sq. ft of area, a one inch stainless steel mesh net enclosed an aviary environment. The net acts in unison with the supporting arches utilizing double curved shapes to provide a rigid structural system. Joining the stainless steel mesh to the ground is a sinuous perimeter wall that in parts is submerged below ground and other times emerges to accommodate cave-like entrances. The woven steel mesh was chosen as an invisible roof for the birds. Looking up from inside the structure, the mesh appears to disappear entirely. Tenara is a woven Teflon fabric probably most known to the public as dental floss fibers. However woven Teflon has many other applications including architectural fabric structures. Unlike the glass fibers, which are stiff and unbending, the woven Teflon has the drape and feel of a fine silk. Manufactured by GoreTex, the material is considered a non-combustible material, meeting all building codes. However, its unique characteristic is its excellent ability to fold and unfold continuously without fatiguing the yarns. The Prophet's Mosque in Medina Saudi Arabia by Bodo Rasch Architects and Buro Happold Engineers is an example of this fabric applied to architecture (fig 9). The courtyard of the mosque required a shaded area during parts of the day. Designing a series of twelve retractable umbrella structures of approximately 60' square made of the Tenara fabric allowed the umbrellas to open up in just a few seconds. The opening is regulated by a computerized system, which responds to seasonal climatic requirements. When closed the umbrellas take on the image of a domed minaret and when opened, they create a lightweight vault. Other types of fabric would be too stiff and require bulkier folding systems. Several shade mesh cloths exist from nylon shading for agricultural uses to PVC coated polyester shade mesh is used in architectural applications where intense solar gain needs to be shaded in roof or facade conditions. The shade cloth is made of an open mesh fabric, which comes in opening sizes from a sixteenth of an inch to 5/16" openings. The fabric comes in many colors and is considered flame resistant. Its lifespan is approximately 12 to 15 years, before it needs to be replaced. The New Central Library in Phoenix AZ designed by Wil Bmder Architects was designed with a north glass facade, which accepted direct early morning sun. FTL Happold was asked to design and engineer a series of 24 seventy five foot long tensile fabric brise- soleil which acted to shade the glass wall, yet at the same time allow visual transparency through the mesh fabric. A series of horizontal cables which were attached to the building and a series of sliding aluminum struts joining the glass facade tensioned the fabric. Incorporating photovoltaics into fabrics can now be achieved by laminating thin film amorphous silicon cells onto the coated fabric. The result is a shade cloth that also acts as a

Fig 9

power station using renewable, non polluting energy. The cells produce about 2 watts per square foot, generating about 5 - 8% efficiencies. The top surface of the PV thin film is encapsulated with a Tefzel laminate. The Tensile Pavilion in the Under the Sun exhibition at the Cooper-Hewitt, National Design Museum in 1998 was the first prototype of this application (fig. 10). Using the garden of the museum, we designed the exhibition, which included a solar canopy that incorporates PV technology. The fabric was a PVC mesh Ferrari fabric which had an amorphous silicon PV film laminated onto it. It generated approximately 1 watt per square foot of electricity with an efficiency of about 7%. This technology has tremendous potential in developing countries with high solar gain. A small tensile structure would not only provide shade but also generate enough electricity to support communication and limited power without any grid connections. We must remember that woven architectural fabric is one more material, with its own language and details, to be used in conjunction with those we are more familiar with: steel, wood, concrete, glass, and masonry. The challenge and the excitement today is to be part of the development of its intrinsic architectural language. With pin connected masts which look like large cocktail sticks and want to be at just about any angle other than 90 degrees, with catenary edges that don't join easily to rigid walling systems, we have discovered that the structural accomplishments have outstripped the architectural qualities of this technology. Now is the time to take all these new materials and develop their integration into mainstream building.

218

ENGINEERING AN INTEGRATED ARCHITECTURE FOR WIDE SPAN ENCLOSURES


Horst Berger Light Structures Design Consultants, White Plains, N Y, USA Professor, School of Architecture and Environmental Studies The City College of the City University of New York

ABSTRACT
This paper deals predominantly with tensile architecture whose application for permanent buildings has occupied this writer for the more than 30 years. In tensile architecture the historic unity of structure and architecture is maintained and many building functions are integrated. The fabric membrane acts as structure and enclosure; reflector and transmitter of light, heat, and sound; generator of the interior space and the exterior sculpture. Using the Denver Airport terminal and other structures in whose design and engineering the writer played a critical role, this paper mainly presents principal tensile structure forms and their impact on function and construction of the building. The examples include the Hajj Terminal of the Jeddah Airport, Riyadh Stadium, Canada Place in Vancouver, and the San Diego Convention Center. Their dramatic forms and spaces consist primarily of minimal surfaces deriving from their structural tensile order. Weight of construction material is drastically reduced, construction time shortened, energy saved, maintenance simplified, and life cycle cost improved. Raising technology to an art form lets tensile architecture add a softer tone to a new vocabulary of architectural design. The paper ends with the new UniDome roof structure, which replaced the 25 year old air-supported roof with a combination of an opaque griddome and a translucent fabric structure in its center.

INTRODUCTION: STRUCTURAL FORM IN ARCHITECTURE


Architecture has the purpose of creating and enriching space for human activities. Structure is the means by which space is spanned and enclosed. Structure, then, is an integral and inevitable part of architecture, its form, its function, its economy, and its spirit. Today this simple relationship is often lost, since, for smaller buildings, contemporary structural technology can support almost any chosen form. For large spans structural form is still important, for tensile solutions it is critical. Yet this is not always obvious to architectural designers at a time when new technologies are evolving rapidly and design tools are not yet user friendly. We live in a period of transition from the relatively settled world of the Middle Ages to a New Age whose outlines are only beginning to become apparent.The last two centuries were marked by a huge population growth (six times world wide, three times in my own life time) and by drastic changes in the way people live as a result of the innovations of the industrial age and the electronic age. The evolving built environment is a critical part of this changing world in which human activity puts a burden on the resources of our planet and exerts pressure on the delicate balance which maintains an environment friendly to human existence. The consequences could be

Fig 1 The Jeppeson Terminal, Denver International Airport

Fig 2 American Indian Wigwam Frame

Fig 3 UniDome air-supported roof structure, 1975

disastrous. Therefore, to survive on this planet may make it necessary to select order systems in which visual form and structural form are congruent and which respect the natural balance of the natural environment. It is my belief that our ideas and images of what constitutes architecture were first formed long before the tiny fraction of the human evolution which we call 'history'. There is evidence that human dwellings of substantial size and grouped in community settings date back over 400 000 years. More significantly, the form and structure of these dwellings was most likely similar to village houses found in Africa and Asia reaching into the last century and to the American Indian wigwams encountered by the European settlers. Their shape derived from the process of building the shelters using available natural means. Flexible saplings, would be set in the ground in a circular or oblong floor pattern. Bending opposite members inward, lacing them together, and adding horizontal rings, domes were formed. Two principle patterns emerged: radial and orthogonal grids. They are identical with the two principal engineered dome forms we have today. Thatched with grass, leaves, or reed, they provided protection against rain and wind, produced ventilation and modified temperature. These enclosures were minimal surface lightweight structures forming comfortable interior spaces and gracious exterior building forms. The similarity of their geometric order (Fig.2) to recent air-supported fabric domes (Fig.3) and the most recent grid domes is amazing.

foundations, required less mechanical equipment and simplified the drainage. It saves energy because of the use of daylight, the reflection of heat from the sun, and the outward night radiation. And there is less general maintenance. Therefore, its life cycle cost is lower than that of any comparable roof system. Above all, the bright interior (Fig.4 ), with its sweeping tensile shapes offers a great space for the traveler. And the exterior sculpture is powerful and distinctive (Fig.l). Architectural form is identical with structural form. And the structural form I kept as pure and direct as possible. It is one of a number of significant public buildings using tensile structure as the dominating architectural feature. The roofs of the San Diego Convention Center and of Canada Place in Vancouver have become landmarks for these two cities. The roof structure for the King Fadh Stadium in Riadh is still the largest stadium cover(despite its large central opening). The Haj Terminal of the Jeddah International Airport, now almost 20 years old, is still by far the world's largest roof cover. Amphitheaters, indoor sports facilities, malls, stores, and industrial structures are among the other frequent areas of application. These and many structures by other designers indicate the successful entrance of fabric tensile technology into the world of permanent architecture and the potential of a larger role in the future when fabric properties will advance and their cost will reduce, and when architects

FABRIC T E N S I L E S T R U C T U R E S F O R PERMANENT BUILDINGS


Tensile structures satisfy at least part of this challenge. The terminal building of the new Denver International Airport, completed in 1994, illustrates most of the significant features of a fabric tensile structure. It took less time to build than a conventional roof system and provided protection during construction of the spaces below. It weighs one tenth of any conventional roof system. Using Teflon coated fiberglass, it cost more than a conventional opaque roof, but less than any roof with similar translucency. It reduced the cost of supports and

Fig 4

Jeppeson Terminal, Denver International Airport, Interior View

220

and engineers will be more familiar with their design, and when this technology and its forms become more
acceptable to both design professionals and the general

outside the plane defined by the other three to achieve the double curved surface which gives the structure its
stability and its capacity to carry load. The alternative

public.

PRINCIPAL CONSIDERATIONS: THE DENVER EXAMPLE


As a structural category fabric tensile structures are a special form of lightweight surface structures which include shells, grid-domes and cable nets. In each of these the continuous spatially curved surface is a critical and integral structural element. In tensile structures the surface elements, consisting of structural fabric and high strength cables, can carry load in tension only. The primary advantage of tensile members over compression members is that they can be as thin and as light as their tensile strength permits. Consequently the weight of tensile structures is almost . The weight of the Denver roof, for instance, is 10 kg/m2, which is one tenth the weight of a light steel truss roof, one thirtieth the weight of the most intense snow accumulation which this roof is designed to carry. The fabric skin is not only part of the structure but also the building's enclosure, requiring no additional dead load for cladding. A further advantage of thin, lightweight tensile components is that they are easy to ship and erect. Their flexibility allows them to be coiled, rolled or folded into small packages. Cables can be a few hundred meters long, requiring no splices or internal connections. They can be raised and connected to their end supports by* cranes, winches or helicopters, requiring no scaffolds. In fact, the erection time for a fabric structure is much shorter than that for a conventional structure. Form and prestress, rather than gravity and rigidity, are the basic means of providing the stability and the strength to carry load. Structural form becomes a critical determinant of architectural form. To make a tensile surface structure work, requires a minimum of four support points, one more than needed for a rigid structural system. The most basic form, therefore, is a four point structure. (Fig.5). If an orthogonal grid is used, this is the basic module. One of the four points has to be

geometry is a radial tent. As long as these surfaces are in tension the structure is stable. Under external loads part of the surface can be permitted to go slack in one direction as long as the stability of the support system is not lost in this state. The pattern of surface stresses which is required for the stability and load carrying capacity of the structure results in horizontal forces at the anchors in addition to the customary vertical forces. This is the price to be paid for the advantages of a tensile structure. The skill and efficiency with which these horizontal forces are

A U-ClE2U
Fig 6

Denver Section, showing ridge and valley cables, and the building's horizontal anchor elements

anchored or balanced has a large impact on the economy of the structural system. The Denver roof, for instance, is anchored to the conventional sub-structure by supplementing the existing structural frame with diagonals to balance the horizontal forces along the shortest possible path.(Fig.6). Because of the lack of structural weight, there need to be elements which resist upward loads from wind suction in addition to the elements which carry downward loads. In order to generate the structural surface grid which satisfies all these requirements there have to be supports at the high points of the surface, others at the low points, and still others located around all sides of the periphery. The choice of these support points defines the shape of the structure. Their geometry combined with the stress pattern assigned to the surface leads to the form of the structural surface. New forms can be explored with the help of stretch fabric models which simulate the actual shape rather well and are easy to make. The final shape is determined with the help of a formfinding computer program. It puts all the tensile forces in all the elements in equilibrium. For one given configuration of supports and one internal stress pattern there is only one equilibrium shape. Form clearly follows structural function. Since the surface which is generated in this way is also the enclosure, the structural form defines the sculptural shape of the building on the outside and the form of the space on the inside. There is no longer any distinction between engineering and architecture.

Fig 5

Four Point Structure

221

The shape of the Denver roof consists of fabric spanning between alternating ridge and valley cables, with the periphery defined by edge catenaries. Fig. 7 shows the entire form of the 320 m long roof. This image is based on thee writer's iterative geodesic formfinding system.

Fig 9 Aerial View of Denver terminal roof

Fig 7 Denver membrane grid

The photo of Fig.9 shows the completed structure. My initial proposal for the shape was to keep all interior fabric units identical . The concern was the simplicity and economy of the structure. The visual impact would be naturally enriched by the deep perspective caused by the large scale, an effect seen in medieval cathedrals. The architects' desire to emphasize the two main entrance points which also divide the terminal into three functional sections, led to the use of four larger units with higher masts.( See Fig.l, Fig.7, and Fig.9). This resulted, of course, in a tremendous variation of shapes due to the continuity of the stress pattern. The impact on cost was considerable but probably worth it. Fabric as the surface element in a tensile structure is critical in maintaining the hierarchy of materials which makes the system compatible. Fabric stretches more than cables, they stretch more than rigid structural elements. Rigid surface elements instead of fabric cause compatibility problems unless frequent expansion joints are provided or the surface is regarded a rigid shell and included as such in the analysis. There is no expansion joint in the 320m length of the Denver roof.

Translucent fabrics further define the character of the spaces they enclose by bringing in daylight. High reflectivity and low absorption of heat greatly moderate the interior climate. And the surface geometry, together with characteristics of the fabric or of an inner liner control the acoustics in the space. The sound dissipating geometry of tent shapes combined with the sound absorbing surface of the inner liner acts as a "black hole" for internal sound. Users of the Denver airport, which has an acoustic inner liner, comment on the quiet atmosphere inside this busy terminal. A feature of critical importance in a permanent building .with a fabric structure roof is the treatment of the connection between the flexible membrane and the rigid periphery wall. Clamping the components of the roof structure directly to the top of the wall requires the wall to be designed for substantial horizontal forces. If the membrane forces are anchored separately, a connection has to be found which allows for the substantial differential movement between fabric and roof membrane. In the case of the Denver roof with its high, cable supported cantilevering glass walls and the big fabric roof overhangs, a workable solution was the introduction of an inflated fabric tube which allows roof movements in the order of 0.65 m at the clerestory windows (Fig.8). and around 1 m at the south and north walls. Simple spring operated valves let the air escape and the tube flatten out or elongate. A small pump keeps the tube inflated. The inner fabric liner, connected directly to the top of the periphery glass walls, hides the tubes from the inside. Fig.8 shows the tube before installing the inner liner.

Fig 8 Clerestory with inflated tube closure.

222

An interesting and integrated part of the Denver enclosure are the cable supported glass walls around the entire periphery of the terminal space. The south wall itself is one of the largest glass walls built, being up to 20 m high and 67m long. The upper edge anchors the inner liner. The deflection of the top of this wall under wind load is only 8 cm.

Fig 10 Construction of Denver roof

MAST SUPPORTED STRUCTURES


The example of the Denver terminal building shows the principle structural features of a mast supported tensile structure. The upper support points are formed by pairs of masts which are spaced 46 m apart. Ridge cables are draped over these masts and anchored to the adjacent lower roofs similar to the main cables of a suspension bridge. They occur every 18.3 m along the length of the building and are designed to carry the downward loads,

Fig 12 Shoreline Amphitheater, during constrution

A few notes on a number of other mast supported structures, pointing out features of special interest: The roof of the Shoreline Amphitheater shows a mast supported structure in its simplest form and largest scale. The two masts are 45 m high, spaced 61 m apart, supporting a roof with 8,000 m of plan area. The front edge catenary spans 140 m between two pile supported abutments. The fabric spans between ridge cables and edge catenaries with only a few internal cables placed within the fabric surface for sectionalizing the membrane and reinforcing it along a few critical lines. The fabric was stressed by jacking the masts at the ground level.
2

Fig 11 Denver, main fabric, stressed.

mainly snow in the case of Denver. Valley cables are placed between any two ridge cables and run parallel, taking on the form of an arch. They carry the upward load from wind suction and are tied to lower roof anchors. The edges of the roof are formed by edge catenaries outside the window walls which are anchored against the building. Construction progressed linear (Fig. 10), a bay at a time, starting at the north end , and ending at the south, where external anchors complete the structure. The exterior fabric was stressed by pulling down on the main connectors right outside the clerestory walls. (Visible in Fig. 11 at the far end of each valley cable). This photo shows the main fabric, stressed and before installation of the inner liner. The cables running parallel to the fabric seams are redundancy cables which act as rip stops and as replacement of fabric stresses in case of a rip or during replacement.

In the roof design for Canada Place (Fig. 13) in Vancouver the masts are placed at the ends and are anchored back with external tie-down cables. The tent units have a 45o skew in plan, orienting them parallel to the city streets. This arrangement made the patterning

.1Fig 13 Canada Place

223

complex. But it gives the building the sail-like character for which it has become known. The large external moments created by the position of the high masts at the ends was balanced by engaging two floor levels of the building and utilizing the building's structural components. Pairs of cables are used for the external anchorage to provide for redundancy and to make it possible to replace them. In the earlier design for the Haj Terminal of the Jeddah Airport, completed in 1982, central mast supports were avoided by suspending the 46 m span square tent units at their peaks. Eight suspension cables carry the load of each unit up to the top of the 46 m high pylons, which consist of single masts in the interior and of rigid frame double pylon structures along the periphery of each module as well as between modules. The roof covers a total of 420,000 m2 or 105 acres of plan area, by far the largest roof cover in the world. The roof's purpose is to moderate the climate by simulating the functions of a forest in the desert. The translucent roof provides shade and reduces the effect of the heat and light from the sun to about 10%. It avoids the heat storage in the ground and its subsequent radiating back into the space. It allows warm air to rise up and escape through the center ring openings. The construction of this very large project made use of its repetitive design, which becomes visible in Fig. 14. The 210 tent units are arranged in 10 modules, each three units wide and seven units long. The 21 units of one module were assembled close to the ground. The support ring in the center of each ring was split in a top and bottom section. The top ring, hanging from the main support cables, contained winches and jacks, which could be operated from one central control space on the site. The winches lifted all 21 units simultaneously within about one meter of the top ring. Four screw jacks each were then installed. Again, simultaneously jacking all 21 units the rings were docked, the structure fully stressed and the rings bolted to each other. In the photo the five modules of one side of the structure (Modules A to E) are completed. The first module of the other side

(Module F) has been raised and is being stressed. Module G, next to it, is being installed near the ground, soon to be raised. It should be noted that this process was tested on two full scale test modules which were also instrumented with stress sensors to check the accuracy of the computer analysis. The test results deviated from the analysis output by less than 5%, giving us confidence in the reliability of our analysis process. Because of the tremendous scale of this nearly 20 year old structure it is becoming a test for the reliability of fabric tensile construction. The Riyadh Stadium extends the concept of mast supported tent units to create the largest span roof structure to date. (The design could have been adjusted without difficulty to cover the area formed by the central opening which is only one quarter of the total plan area. Functionally this was not desired). This 247 m diameter

Fig 15 Riyadh Stadium Roof

span is achieved by arranging 24 units in a circle with an outer diameter of 290 m, covering an area of 49,000 m2. In each unit a main vertical mast and a smaller sloping mast combine with triangulated peripheral tie downs to provide the rigid supports which hold the structure out and up. On the interior the horizontal forces are balanced by a large ring cable with 130 m. diameter. Again, ridge

'

* :

Fig 14

Jeddah Airport Roof: Construction

Fig 16 Riyadh Stadium : start of fabric erection

224

and valley cables form the main elements to which the fabric membrane is attached with the valleys forming the downward anchors. The ring cable, suspension and stabilizing cables provide redundancy and make a simple erection feasible. Fig. 16 shows one step in the erection process. The entire cable system is in place. Fabric is laid out on the ground, ready to slide into position. Note in both photos that only two fabric panel shapes were required to make up the entire roof and give it its dramatic shape.

In 1997, Light Structures!Horst Berger were engaged to provide an enclosure design for for the area under this roof. The schematic design proposed a convertible enclosure to include a free standing, cable supported glass wall at the 91.5 m long open end similar to the south wall at the Denver airport. Movable wall panels were to convert the space from naturally ventilated to fully air-conditioned, curtains and fabric baffles from bright daylight to a shading level permitting video presentations to 6,500 people. A different scheme by a design/construct team is presently under construction.

Fig 19 Mitchell Amphitheater, near Houston

Fig 17 San Diego Convention Center, exterior

A-FRAME SUPPORTED STRUCTURES


The roof of the San Diego Convention Center provides a 91.5 m clear span by suspending the masts. They rest on the main suspension cables placed 18.3 m apart, which carry the load to triangular concrete buttresses whose dominant forms give the building its character. The roof structure is again formed by stretching the fabric between ridge cables, valley cables, and edge catenaries. A special feature of this roof design is a horizontal flying pole with forked ends which has the purpose of resisting the tensile forces of the two open ends. (Fig. 18) This makes it possible to keep the end openings totally free of supports, giving the roof its sense of floating weightlessness. A visually delightful feature is the socalled rain-fly, a closure structure on top of the main roof which covers the ventilation openings of the main roof. Tent shapes require a support at the peak of each tent unit. Architectural spaces most often need to be free of interior supports. Of the examples above, at Canada Place this was resolved by moving the supports to the edge. The result is a space which is high at the ends and low in the center, and a structure which is not very efficient. At Jeddah the masts were placed at the corners and extended upward to be able to suspend the tent units from them, again a structurally inefficient solution. At San Diego the masts ride on support cable which transfer the load to the perimeter requiring heavy anchors there. One way to resolve this problem is to replace the mast by an A-frame. One of several such structures is the roof of the Cynthia Woods Mitchell Center of the Performing Arts at the Woodlands outside of Houston, Texas. It covers 3000 fixed seats. Three A-frames form the support system together with the stage house structure. Horizontal anchors are avoided by introducing compression struts which link the support columns and edge cable anchors to the stage house, thereby balancing the horizontal components of the membrane forces. The supports of the A-frames form low points of the membrane which function as drainage locations for the rain water. The trussed columns supporting the A-frames contain the rain leaders and support platforms for the follow spot lighting of the theater.

Fig 18 San Diego Convention Center, interior

225

Fig 20

Mc Clain Practice Facility

Fig 21

Bayamon Baseball Stadium Roof Design

ARCH-SUPPORTED STRUCTURES
For spans of rectilinear structures of up to 100 m arch supported fabric roof systems can be highly efficient. For domes with circular, elliptic or super-elliptic edge shapes spans of more than 200 m can be an efficient solution, as long as the arch components remain within dimensions which are shippable by trucks. A number of structures have been built using prefabricated steel sections, often with a triangular cross section. The largest one using such prefabricated steel arches is the McClain Indoor Practice Facility of the University of Wisconsin in Madison. This building covers a football practice field. Arches of 67 m length, spaced 18.3 m apart, span the the full width between rigid concrete abutments. They are 2.1 m deep. Shop fabricated in 12 m long sections they were bolted together in the field to form half-arches. These were lifted by cranes, pinned in the center and braced against the adjacent arch, requiring no temporary support elements. It took 10 days to assemble the entire arch system. The outer quarters of the roof are covered by standing seam, stainless steel roofing. Only the middle half is covered by fabric membrane which spans between the arches and is held down by valley cables. This arrangement provides excellent natural lighting conditions for sports by concentrating vertical light in the center. Also the combination of the insulated opaque roof sections with the translucent, reflective fabric roof help reduce thermal energy consumption. Up-lighting against the reflective underside of the roof make for good lighting conditions in the night. One of the many other arch supported designs was for the tennis practice facilities of the AELTC in Wimbledon. It uses exterior, exposed precast concrete arches from which the fabric is suspended. This provides a neutral geometry of the translucent ceiling which is essential for playing tennis. It was completed in 1988.

STADIUM DOMES
A single arch spanning 168 m was proposed to support the cover for an existing baseball stadium in Puerto Rico. This dramatic design illustrates one of many ways of spanning a full size stadium facility. The arch, rising over the middle of the field, supports two cable reinforced fabric membranes, one anchored to a horizontal edge beam behind the outfield, one connected to two cable stayed masts located in front of the stadium. Fabric structures entered the world of permanent buildings with large and super-large spans. Geiger Berger's low profile air-supported roof design for the US Pavilion at the 1970 World's Fair in Osaka led to eight stadium-size roofs built in the United States and Japan between 1973 and 1985. All followed David Geiger's special geometry, consisting of a superelliptic ring and a cable net with cable lines parallel to the diagonals of the superscribed rectangle. The economy and speed of erection of these domes together with the attraction of high levels of daylight of the new Teflon coated fiberglass fabric made them win out over conventional structural systems. They became the engine that drove the new train of fabric structure technology. Problems with snow melting and removal, the cost and inconvenience of operating mechanical devices to maintain the stability of the roof structures, and the limitation and expense of a highly pressurized building led owners to return to static structural systems. This writer's first opportunity to respond to this development with a fabric tensile roof came in 1983 with his initial design for the St. Petersburg Sundome, for which he was the partner in charge. He called the system cable dome. The main principle of this patented system came from the idea of spanning suspension cables from opposite points of the ring beam and supporting sets of

Fig 22

Original Cable Dome system developed by author for Sundome, 1983

Fig 24

Hybrid Cable Dome system with arches as top chord members, author, 1985

flying poles on them, similar to the basic arrangement of the San Diego roof. Integrating these elements leads to this simplest of all cable dome systems, where each cable carries two poles, each pole is supported by two intersecting cables. Again, one, two, or several layers can be used, whereby each layer is added like a cantilever. Erection needs no temporary supports.

a peripheral ring. The cable system below the arches becomes very light as its function is reduced to carrying part of the unbalanced roof loads, stabilizing the arches and allowing the roof to be erected without a scaffold and a minimum of interference with the space below. In studying the replacement of the air-supported UniDome roof at the University of Northern Iowa, a cable-dome proved to be impractical. It was not possible to adapt its radial configuration to the existing orthogonal geometry and the first row of flying struts interfered with the sight lines from the upper seats which is a common shortcoming of all cable dome structures. It was also not economical for Iowa's heavy snow loads. The answer evolved from taking advantage of the special nature of the existing geometry in which the horizontal forces from the cable grid are in perfect funicular balance with the shape of the ring beam. The initial concept was a grid of compression elements following the same plan configuration as the existing cable net but located above it. The compression members were assembled from shop fabricated, three dimensional truss sections which were connected by vertical ties to the old cable net re-installed below. This combination offered the most direct force flow for downward or upward loads for the 15 000 m2

Fig 23

Cable Dome for Sundome by David Geiger, built 1986

In the final design, carried out by David Geiger (after the dissolution of Geiger Berger Assoc. in 1983), the configuration was changed to a system consisting of concentric rings and radial cables. ( Fig. 23). There fabrication and erection is difficult. A number of other cable dome structures have been executed, most prominently the roof of the Georgia dome in Atlanta, designed by Weidlinger Associates, using a triangulated configuration. Cable domes of this type are not efficient in heavy snow areas because of the multiplying effect which this geometry has in transferring loads from the center to the periphery. This leads to very high cable quantities accompanied by very large deflections. To avoid these problems this writer's cable dome patent includes a version with arch-shaped compression members at the top. These carry gravity loads in the most direct way to

Fig 25

New UniDome, Arch and cable grid. The cable net is that of the former air structure

Fig 26 New UniDome Hybrid roof, computer image superimposed on photograph

Fig 28

Erection of steel grid dome members.

dome spanning 140 m across the diagonal. The cable net stabilizes the grid dome and provides sufficient bending capacity to accommodate eccentric load cases for snow and wind. In the final version of the design the center section was replaced by an arch-supported fabric tensile roof which reduced the dead load where it is most critical and provided translucency where it is most desired. The rest of the roof surface is enclosed with a stainless steel standing seam roofing on metal deck and bar joists. Fig.26 shows the roof design in a computer generated image superimposed on an existing aerial photograph. The concrete ring beam which on this structure was made of rather thin precast sections was prestressed with tendons rapped around its exterior face to give it the capacity to become a tension ring.

Fig 29 Beginning of steel erection

Though a support free erection was studied, the use of four construction masts under the intersection points of the four continuous arches proved most practical and economical. (See Figs. 28 and 29). The arch sections (1.2m X 1.8m ) were shop fabricated in up to 17m straight lengths, bolted on site into sections ready for installation. The four long arch sections were strengthened by tie cables. Bar joists, spanning between arch ribs, and metal deck, spanning between joists, followed. Insulation and stainless steel roofing was installed parallel to the center fabric structure and the cable net below.

Fig 27

Prestressing tendons applied to the out side of the existing ring beam

The construction began with the prestressing process in the winter of 1997/98, while the stadium was in full use. (The air-supported roof had failed in a sudden snow fall two winters before and had been repaired with PVC coated polyester fabric, a process which took Birdair only weeks to complete). Parallel to prestressing, shop fabrication of structural components took place.The stadium remained in use until the middle of March 1998. The new roof was completed and the first football game took place in the stadium in October of 1998.

Fig 30 The UniDome with its new roof

228

Under the new hybrid roof system, the natural light level remains approximately the same as in the air roof. Over sixty percent of the roof surface is insulated. Air pressure is no longer required. The resulting energy and operating savings are sufficient to finance the difference in cost of the new roof as compared with simply replacing the fabric in the existing air supported system. Above all, the risk of failure under snow load is eliminated. The construction cost was under $11 million. In 14 years time the life cycle cost will be below the cost of simply re-installing the air-supported roof.

visual excellence, fabric structures will be able to prove themselves as highly desirable, when looked at by the slightly adjust ancient standards of usefulness, stability, economy, environmental desirability and visual delight. The buildings shown in this paper are, hopefully, part of a development in this direction. It should be mentioned at this point that for each of these buildings there was a design team, generally led by an architect, sometimes assisted by other engineers. This short paper covers too many projects for comprehensive credits. They can be found in the the writers book (Horst Berger: Light Structures - Structures of Light), which also covers additional information on the subjects above. Tony Robbin's book, Engineering a New Architecture, tries to give an overview of the potential offered by new engineering advances and ideas towards a new architecture. And, finally, an attempt to set up a broad scope of objectives for architecture and of the built environment in general can be found in the new issue of American Building.

CONCLUDING

NOTES

This paper looks at a variety of fabric tensile structures and a large hybrid grid dome as examples of surface structures which each form the dominating architectural feature of a permanent building. They are major projects out of over 40 designs built over the last 25 years. Emphasis is on the integrating impact, which does not only extend to the unity of structure and enclosure, but also to the building's main functions as control of light, heat, and sound. Construction is regarded an integral aspect of the structure. While a substantial number of fabric tensile structures have been built world wide, this art and technology is still on the fringes of architecture and building construction. And as the example of the UniDome roof replacement demonstrates, even for substantial spans (the span is of the size of Madison Square Garden) conventional materials in a grid dome configuration may* at this time, be more cost- effective than a pure tensile fabric structure. Several things need to happen to make the art and technology of fabric structures a common component of the new built environment: design tools need to become more user friendly so that architects and engineers will be willing to use them. the cost of construction needs to be reduced. This requires a cheaper, more translucent, longer lasting fabric which is easier to handle; simpler, less labor intensive detailing and construction; and more use of prefabrication. a broader acceptance of these new forms by the general public.

LITERATURE
Light Structures - Structures of Light The Art and Engineering of Tensile Architecture by Horst Berger, Birkhauser, 1996 Engineering a New Architecture by Tony Robbin, Yale University Press 1996 American Building The environment forces that shape it by James Marston Fitch and William Bobenhausen Oxford University Press, 1999

The Denver airport terminal, one of the very few tensile enclosures of a regular 24-hour public building has had a very positive reaction. The Millennia Dome can be expected to have an important impact. Further, by measuring a building's qualities by standards beyond

SECTION V Sports Stadia


The Millennium Stadium, Cardiff Development of the New Wembley Stadium Roof Keeping the Doors Open: The Olympc Stadium, Sydney Widespan Enclosures: Cost Considerations Lightweight Enclosures in Japan (and the World Cup Stadia, 2002)

230

THE MILLENNIUM STADIUM, CARDIFF


Mike Otlet Director of Engineering Design WS Atkins - Oxford

INTRODUCTION
The Millennium Stadium is located on the site of the original Cardiff Arms Park stadium in the heart of Cardiff the capital City of Wales. Conceived as a prominent and attractive landmark, it received 46 million of lottery money from the Millennium Commission and became one of the major projects to mark the new Millennium (fig 1).

BACKGROUND
The new stadium, which seats 72,500, was built by John Laing Construction, over a three year period on the restricted inner city site of the original Cardiff Arms Park rugby ground. It has close neighbours on all sides, including the River Taff. In order that the stadium can host significant events besides rugby or football, two sections of the roof can be moved across to completely cover the spectator and pitch areas and form a weathertight arena. This closing roof is the first of its kind in the United Kingdom and the largest in Europe. The quality of the acoustics ensures that noise breakout is reduced to a minimum, neighbours are disrupted as little as possible and there is, within the stadium, an atmosphere that will attract top performers and large audiences to the venue. The architect HOK Lobb have balanced a series of factors to achieve the optimum configuration that will ensure that the spectators are close to the pitch and have excellent sight lines, seating comfort and safety. High quality facilities for all the family have been provided, including restaurants, shops, bars and fast food outlets. Behind the scenes, below the entrance concourse level there are changing rooms with state of the art physiotherapy and medical facilities, offices, kitchens, storage and parking. Unique to the project is a fully palletised system of interlocking turf modules which can easily be lifted out and replaced when worn or damaged (fig. 2). The whole system can also be completely removed to create one of the largest covered arenas in Europe, capable of hosting almost any indoor event.

Fig 1

It is the first opening roof stadium in the United Kingdom and took four and a half years from conception to completion. In order to hone the design and refine the details to suit the Arms Park site, budget and programme, many structural forms were considered. The Rugby World Cup was to be hosted by Wales in October 1999 and this event, provided both a catalyst and a completion date for the project. This paper reviews some of the key stages in the work of the design office and fabrication workshops, which led to the final spectacular solution. Nowadays, the design process relies heavily on the use of computers and, in this, the Millennium Stadium was no exception. They were used extensively throughout the design process for analysis purposes and to express the design proposals.

Fig 2

231

In these respects as an advanced technological building and as a focus of urban activity and renewal, the new Millennium Stadium can be considered to be one of the first of the "Fourth Generation" stadia - a stadium for the new Millennium.

ROOF DESIGN DEVELOPMENT


The stadium needed to be about 50 metres larger than the pitch in all directions to accommodate the 72,500 seats and the opening had to be at least the size of the pitch. This gave roof dimensions in the order of 220 metres long and 180 metres wide with an opening of approximately 120 metres x 80 metres. At the outset, following discussions with the various members of the team, a number of design criteria were decided upon; 1. To keep the roof as low as possible to reduce the stadium's impact on adjoining buildings e.g. Westgate Street flats. 2. To keep the edge of the opening as low as possible to reduce the extent of shading on the pitch bearing in mind the requirement for roof falls for rainwater drainage 3. To make any structure around the edge of the opening as small as possible, also to reduce the effects of shadows on the pitch. 4. To make the track for the retractable roof to move along, as near to flat as possible, again bearing in mind the roof falls for water run-off and drainage, and also to assist with making the retractable roof mechanism simple and therefore less problematic. 5. It must be a quality design.

STANDS
In order to hold the required seating capacity and comply with the space restrictions around the site, the stands rake outwards as they rise. The interesting structural solution needed to achieve this, led in turn, to a dramatic architectural form. The structure above the entrance, which is at concourse level, is constructed from 6,500 tonnes of steelwork in CHS, RHS, open sections and plate girders. It comprises a series of frames at typically 7.3 m centres. The frames are stabilised radially by concrete shear walls and, although there are only two basic frame types with shear walls, either close to the pitch or remote from the pitch, the shape of the stadium means that virtually every one of the 76 frames is different. The steel frames are supported by a reinforced concrete substructure and piled foundation system (fig. 3).

THE RETRACTABLE ROOF


The direction and form of the moving roof was an initial concern. The drive systems however were not considered to be a significant factor in this decision and have not unduly affected the structural form since.
Fig 3

Pre-cast concrete stepping units sit on raking steel plate girders around the bowl to form the seating areas. At the back of the stands, these girders carry not only the seats but also some of the roof weight and, by means of tie rod hangers, the extensive level 6 upper concourse. Tubular steel props assist in limiting bending moments and deflections in these girders. Level 5 (Box and Restaurant level) and level 4 below (Club level) are of pre-cast concrete slabs and are supported by deep plate girders on steel columns. Holes are provided in all the horizontal plate girders for services penetrations. A horizontally propped raking plate girder supports the seats for the dramatic middle tier. This cantilevers 14 metres out from the floors at levels 4 and 5.

Due to the plan shape of the stadium seating bowl, and the aim to create a roof as flat as possible, dome forms were dropped in favour of linear "sliding door" style systems running on straight rails. Most schemes have involved two sets of 5 similar sections combined in some manner to form a total unit at each end of the opening. Initial ideas centred around methods for concertinaing sections so that they could be stored in a shorter length, than the area to be covered, clear of the pitch. One of the original sketches produced at the time of the studies is shown (fig. 4). The third scheme (fig 5) was pursued in the greatest detail and certainly could have been made to operate successfully but the cost was

232

ho..,__

. | . - V Ktl

Fig 6

Fig 4

prohibitively expensive. Instead, the efforts were concentrated on creating two 55 metre x 76 metre "doors" to cover the 110m long opening.

An initial idea produced in the first two weeks of the design process, in April 1995, was eventually to bear a surprising resemblance to the final form. The first scheme was developed over the following weeks, ready for the first submission for Millennium funding, which was made in May 1995. This, unfortunately, was not successful. Scheme 2 Following lengthy discussions and the consideration of alternative sites for the stadium through the summer of 1995, a new location, partly on the existing Arms Park site and partly on the site of an existing BT building and TA centre to the south, looked to be feasible. This had the advantage of improved access from Park Street. Again we opted for two masts to support the main structure and retractable roof track, but this time to the south of the stadium (fig 7). Effectively, it was the same as the first scheme but turned through 180. To avoid the road, the masts were moved towards the centre line of the Stadium and transfer structures were incorporated. This second scheme was submitted for Millennium funding and following close scrutiny by the Millennium Commission and its representatives, received 46 million of lottery money on 23 February 1996.

Fig 5

FIXED ROOF AND SUPPORTING STRUCTURES


Design Evolution There was insufficient space on the site both at the ends and each side to allow any arch forms starting at ground level and it was decided not to follow the tied arch and deep truss route used on the Ajax stadium in Amsterdam, due, again, to the shadows created by such a high structure. Instead the schemes investigated all made use of masts and tension systems in an effort to improve structural efficiency. Scheme 1 Over the first weekend of the project we sketched some ideas and started putting rough numbers to the member sizes and depths, for a two mast solution, picking up 2 large lattice trusses for the retractable roof track to sit-on (fig. 6). From this we started to get a "feel" for the scale of the problem and the magnitude of the various elements involved.

Fig 7

Scheme 3 Through early 1996 we had been having increasing difficulties with the foundations and buried services that would have been too costly to move elsewhere. When these problems were combined with uncertainty

regarding the availability of the Empire Pool site to the south, we started to investigate alternative mast arrangements that did not involve such a large site. By going back to the beginning again and considering the options available it became clear that four masts could be successfully employed, one in each corner at 45, to lift the corners of the opening. Being symmetrically loaded, the ability to offer a more efficient design also became possible (fig 8). After lengthy discussions with the client, the architect et al, the four mast scheme was eventually adopted by all in the summer of 1996 and developed in conjunction with the contractor John Laing Construction, through to the signing of a Guaranteed Maximum Price, in March 1997.

THE FINAL SOLUTION


The Roof Covering Both the fixed and retractable roofs are clad in a standing seam aluminium sheet with about 120 mm of insulation which is supported by a 128 mm deep profiled aluminium sheet (fig 9). This was all manufactured by Hoogevans and installed by Kelsey Roofing Industries Ltd on-site. This type of make-up and weight is unusual for a stadium, but was necessary to comply with the acoustic criteria noted earlier and allow more concerts to be held annually.

Fig 9

Fig 8

Scheme 4 One or two adjustments in early 1997 lead to the final arrangement we have today. These were: i. The seating bowl that originally varied in its row numbers to the sides of the pitch, and was deeper and therefore higher to the long sides than at the ends of the pitch, was rationalised to a constant level. This deepened the radiused corners and pushed the masts further outboard at this point requiring large diameter columns externally to transfer the loads to ground. ii. A section of the original North Stand was retained, cutting into the roof zone adjacent to the Cardiff Rugby Club. This required structure to spread the loads onto the existing concrete stand and an adaptable solution to allow the roof to be extended at some time in the future if required. iii.The masts to the north were rotated by approximately 22 to ensure they did not encroach on adjoining properties' land. Unfortunately the masts to the south could not be similarly rotated and so a less efficient asymmetric structure was the only solution.

The top sheet continues from the roof opening out to a perimeter gutter, which runs practically all the way around the perimeter of the bowl. A syphonic drainage system, made by Fullflow, then takes the water away from the gutters to the ground. Roof Services Because the roof is closed completely for special events which require protection from inclement weather, there are a greater number of services suspended from the roof than would otherwise be necessary. There are two rings of walkways running around the stadium to access these. The first is located back from the edge of the opening and the second in the middle of the fixed roof, 24 metres back from the edge of the opening. Both walkways support heavy pitch lights and speakers weighing up to 165kg each, together with cabling (fig 10).

Fig 10

234

MAIN STRUCTURES
Purlins The roof cladding is supported by 14 lines of purlins that run circumferencially around the roof at 4.0 metres centres. The surface created is very much like that of an egg with varying radii in both directions. As a consequence, the purlins twist from one bay to the next as they pass over the structure below. The roof deck provides lateral restraint at top boom level and small CHS tubes provide lateral restraint at bottom boom level. These tubes also provide support to the metal ductwork suspended from the roof.

Primary Trusses Two major pieces of structure, known as the Primary trusses, are located on each side of the pitch in a north/south orientation. Rising 35 metres above the pitch, these are continuous over the full 220 metre length of the stadium (fig 12). Support is provided at two intermediate positions (at the corners of the opening) via cables up to the corner masts which are then tied down to anchors outside the stadium. With a 1067m diameter top and bottom boom, the trusses range in depth from 4 metres at each end to 13 metres in the centre.

Tertiary Trusses The Tertiary trusses support the roof deck purlins and walkways for the fixed area of roof (fig 11). There are 44 in total generally at 14.6m centres around the stadium. With a span up to 50 metres, they are supported at one end by the back of the stands and at the other by the Primary/Secondary trusses which surround the opening. To achieve good sight lines the trusses reduce from 4.3 metres deep at the junction with the Primary/Secondary Trusses next to the opening, to only 400 mm deep at the back of the stands. Here, the trusses sit, via individual sliding bearings, on a perimeter truss (fig 11). The perimeter truss spreads the end weight of the Tertiary Truss uniformly onto two adjacent stand frames.

Fig 12

A 778 dia. middle boom, four metres above the bottom boom, provides a connection point for the tertiary truss top booms and resists high compression loads from the mast structures, (ref. analysis) On one side the trusses provide the support and rigidity for the continuous runway beam which support the moving roof. On the other side they provide support for the fixed roofs on the east and west of the opening.

Fig 11

The bearings ensure that differential horizontal movements between the roof and the stands will not have an adverse effect on either element, e.g. under wind loads and thermal expansion/contraction.

235

Secondary Trusses The secondary trusses run in an East-West orientation and trim the North and South edges of the opening. They traverse the full 180 metres width of the stadium and are formed from a 915 diameter top boom and 550 diameter bottom boom (fig 13). Support is provided at each end by the stand structures, and at the intersection points with the Primary Trusses, by the corner mast and cable assemblies. They principally provide support to the pitch end of the North and South Tertiary Trusses and also by a lesser extent, to an area of roofing to the corners.

The corner tertiaries are restrained back to the adjacent parallel roof section (either east-west or north-south). The total roof is trimmed by a 4060 CHS which supports an eccentrically applied cladding load and holds the shallow Tertiary trusses vertical at the bearing positions on the perimeter trusses.

THE CORNER MASTS


Four corner mast structures are key to both the vertical support and horizontal stability of the roofs. Each mast structure is made up of a pair of lower columns (concrete filled steel tubes 12190) which sit upon a 16000 fabricated steel tensioning chamber which, in turn, rest on reinforced concrete foundations (fig 15). The tensioning chamber is connected to an 8m deep reinforced concrete shear wall via 10 no. 750 Mac Alloy bars cast into the wall. On top of the pair of lower columns is a complex series of connections commonly known as the elbow and knuckle. The elbows form the link between the roof and the stand structures providing total stability horizontally to the roof via the eight elbows, in 4 pairs, and the cross-bracing between them.

Fig 13

Bracing and Lateral Restraints The fixed roof is connected together to perform structurally as one homogeneous unit. The straight, rectangular, roof areas are braced in both directions on plan at top and bottom boom level for stability and lateral restraint purposes (fig 14).

FIXED P O S I T K J N S <RJ.

Fig 15

Fig

14

The A-frame mast rests on the knuckle and is held down by the high tensile forces in the cables which on one side lift the main roof and, on the other, are tied down to the tensioning chamber at the base of the pair of concrete filled steel columns. The high tensile forces in the cables generate compression in the two horizontal structures. On the pitch side these are known as the Mast Tertiaries. These are fabricated units 2.6m deep and are made of 60mm thick plate to form a Tee-shape section which were then welded to a 6600 CHS tube at the bottom. Pairs of Mast

The purlins perform the role of lateral restraint, at top boom level, with CHS tubes at bottom boom level. In addition the bracing holds the track for the moving roof in position laterally. This requires both levels of bracing to resist the torsion effects of the moving roof loads being applied eccentrically to the Primary Truss.

236

Tertiaries are braced together for stiffness and buckling resistance. On the other side of the knuckle outside of the stadium, is an A-frame outrigger made from 9150 CHS tube. The tubes are restrained by a tensegrity structure to stop the outrigger from bowing under its own weight. This also provides buckling resistance. The A-frame masts rise 40 metres above the edge of the roof and 70 metres above the surrounding ground level (74 metres above the pitch). Each leg is a fabricated oval section 915 x 1415 overall, tapering down to 9150 at the knuckle. Again, as with the outriggers, the mast A-frame is restrained by a tensegrity system of McCalls tie rods and struts.
Fig 17

FTOM iro

or

mmm

r r r r c s (70.

tuxi

The tension system, although loosely described as cables, is in fact a group of 15 mm diameter high tensile steel strands by PSC Freyssinet inside 6 No. 2730 HDPE sleeves.

MOVING ROOF
There are two moving roof sections that are generally located one to the north and one to the south of the opening over the fixed roofs. Both sections are 76 metres wide and 55 metres long and made up of 5 individual units, each 11 metres wide (fig 16). The units are linked together, principally at the ends with vertically orientated sliding bearings. Each unit is prismatic in cross section and 8 metres deep at the centre. The truss curved in elevation has a single CHS top boom and two CHS bottom booms. The flat roof deck sits on purlins above the bottom boom with all diagonals and the top boom* exposed to the elements. The units are allowed to move differentially horizontally (fig 17) to accommodate curvature of the track on plan. Sliding bearings are provided above the wheels to cater for variations in the distance between the two retractable roof runway rails. The retractable roof units were assembled at ground level and lifted onto the roof in 76 metre sections. Since positioned and connected, it has functioned well with few problems.

THE MECHANISM
The moving roof sections have no power connection to them whatsoever. The actuating mechanisms are mounted on the fixed roof between the track and the Primary Truss (fig 18).

Fig 18

ttCTXX

THROUOM WfSt S K *

Fig 16

237

These are on each side of the moving roof located at the four comers of the opening to pull each half of the roof open or closed via a cable loop (fig 19). This operation takes 20 minutes in each direction, a speed of only 2.6 metres a minute.

Fig 21

Wind Tunnel testing was carried out prior to the commencement of the design to ascertain the most suitable wind loads to be applied to the roof analysis model. This also provided an opportunity to check the effects of the new stadium on the surrounding buildings. The 3D model was loaded with self-weight, dead load, live load and wind load cases individually and then with combined load cases with the retractable roof units in open, closed and partly closed positions. The output was then sifted to obtain the worst combination of axial and biaxial bending stresses for each roof member. The model was refined to improve structural efficiency and uniformity in truss boom sizes. The Primary Trusses were pre-tensioned by varying amounts until the optimum location and pretension were determined. This showed that maintaining the theoretical Primary/Secondary truss intersection node "level" following installation of all the dead load and retractable roof trusses was the best solution. This meant that the cable system had to be shortened by approximately 500 mm at ground level to achieve the required position. During the construction period, sections of the model were deleted to reflect the partially constructed state. Further analysis was undertaken to ensure that the partial stability and strength of the roof structure and its various components during the erection period were satisfactory. These figures were then used during the tensioning sequence for comparison purposes with the actual figures measured on site.

Fig 19

A system of hydraulic motors and gearboxes power the drums back and forth. Substantial brakes are provided to each drum. Hydraulic buffers are located at the centre of the travel and where the units rest at the outer ends of the track, as end stops (fig 20). The speed and actuation are all computer controlled such that both sections move at the same time and at the same rate.

Fig 20

CONSTRUCTION AND CONNECTIONS


All the steelwork for the stadium was manufactured in the UK by British Steel (now Corus) and shipped to Italy for fabrication by Costrusioni Cimolai Armando SpA. Due to the large size of all the trusses it was necessary to subdivide them into transportable sections no bigger that 5 metres by 17 metres. It is interesting to note that seventy five percent of the roof structure is there purely to support its own selfweight.

ANALYSIS
The roof was initially broken down into simplified 2D frames before the assembly of a complete 4349 member 3D roof model (fig 21). The model was developed over three months in order to reduce the highly loaded and high displacement/deflection points to acceptable levels.

238

With this in mind it is obviously important to keep the selfweight to a minimum. Given that the steel was being transported from Italy to Wales after fabrication it was important that the connections between each piece were both small and efficient. During the fabrication drawing period the Primary truss which was previously a 3Dimensional prismatic form was redesigned as a 2Dimensional element for ease of transportation and in particular shipping. Additional lateral restraints were required as a consequence. From an early stage it was decided to follow the simple principle for the connections of: i. The ends of complete trusses or members would be emphasised architecturally within the practical constraints of tolerance, fit-up and economy.

from a single 100mm thick high grade steel plate cut to the external profile of the overall connection, (fig 23) The plate is orientated vertically in the direction of the cables to enable the tube and cable termination housings to be welded directly to each side. Short stubs for the incoming truss members were then welded at the appropriate angle onto the central plate to give sufficient space for bolted splice connections to be made. Where forces were prohibitively high in-line butt welds were made on site, but these were rare.

ii. All intermediate (splice) connections would be hidden to give the impression of being a continuous monolithic piece. In the majority of cases such as the tertiary truss ends and lateral restraint/bracing member end connections, a simple tapered tube detail was developed with single plates protruding. A plate each side with multiple bolts in a circular arrangement then linked the pieces together. Tolerances for length and direction were achieved via four interfaces each with 3mm oversize holes for M27 bolts (fig 22).

Fig 23

The mast top cable termination, outrigger end cable termination and base tensioning chamber all followed the same principle of the single central plate cut to the external profile of the connection. All other plates and tubes were then welded to the sides of this (fig 24). Cover plates (bent plates) welded outside these have ensured that the external appearance is as smooth flowing as possible, within' the budget constraints. The use of a central plate has ensured that the forces flow more evenly across the connection and high local bending moments and shear forces were kept to an absolute minimum.

Fig 22

The connections for the mast assembles were considered individually due to their varying requirements of movement, tolerance and adjustments The Primary node which connects the Primary, Secondary and Mast tertiaries together as well as numerous smaller bracing and lateral tubes is formed

Fig 24

239

The knuckles which are located at eaves level on 1219 A tubular columns form the focus and connection point for the A - frame masts, Outriggers and Mast tertiaries (fig 25). This central knuckle (or hub) has to resist approximately 40000 kN axial compression from each incoming member. During the cable tensioning process and when the retractable roof sections close, or when it snows, the forces in the cables (which join the outer ends of those members) increase, causing them to elongate significantly. This in turn causes a rotation at the knuckle necessitating a pivot at the same point.

The splice connections between the Primary and Secondary trusses required the greatest development time. Various options were considered (some of considerable weight and complexity) before the final detail was found. Bolted connections were considered to be essential for speed and ease of construction. Due to the large diameter of the tubes involved (770 diameter and above) it was possible to climb inside to make a hidden bolted connection(fig 27).

Fig 25

Fig 27

A 2.4m diameter cylinder 2.4 metres long orientated horizontally, like the centre of a bicycle wheel has around it removable steel plates with PTFE and stainless steel contact surfaces to allow the small rotation to occur (fig 26). The drum was sized to accommodate the very large circular end plates from each of the three incoming members with the low working stresses in the PTFE material

The majority of tubes are highly stressed and an innovative detail was required to solve the problem. The axial loads are transferred by 4 flat plates bolted to fabricated tees welded to the inside of the tubes (fig 29). The tees are of sufficient depth to allow the splice plates to pass over an internal flange on the ends of each section. The flanges are bolted together to transfer shear and torsion. Where tubes were too small to climb inside, port holes were formed to gain access to the bolts and flange plates.

Figure 26

Figure 28

240

The 355 diameter bottom booms of the tertiary trusses were too small for this detail and instead a steel cruciform was welded into each transportable section (fig 29). Flat splice plates were used to link the pieces together. Cover plates flush with the outside of the tube hide the splice.

Client
Welsh Rugby Union in conjunction with South Glamorgan County Council (later formed into Millennium Stadium Ltd)

Funding
46 million Millennium Funding

Architect
Lobb Sport (now HOK Lobb)

Civil and Structural Engineers


WS Atkins

Main Contractor
John Laing Construction

Mechanical & Electrical Engineers


Hoare Lea & Partners (Detail design by Ove Arup and Partners for Drake and Scull)

Steelwork Fabrication
Costrusioni Cimolai Armando SpA

Roof Covering
Kelsey Roofing Ind Ltd Fig 29

SUMMARY
The 8000 tonnes of roof steelwork was erected and clad in only 10 months. This was made possible by the design of practical, workable site connections and the assembly of large sections (generally 50 metres long) at ground level lifted into the air with an 1800 tonnes crane. This avoided the use of scaffolding and minimised work at high level. To design the first retractable roof stadium in the UK has been a long and at times difficult process due to the varying interests of the people involved. Structurally it has been a wonderful challenge and opportunity to design to a scale rarely experienced by most engineers, (fig 30) The deadlines were achieved and I believe the results speak for themselves. The next one will be easy!

Figure 30

241

DEVELOPMENT OF THE NEW WEMBLEY STADIUM ROOF

Michael J Barker Director Mott MacDonald Ltd

ABSTRACT
This paper describes the current development of the new Wembley Stadium Roof. The final design has not yet been completed at the time of writing, however the principles of the scheme are established. Once constructed, this roof will be one of the largest in the world. The paper describes the background to the project and the reasons behind the evolution of the current scheme for this large area, long span roof.

would be used to purchase the existing stadium and it's business. The finance would be raised by a development company, Wembley National Stadium Ltd (WNSL) a wholly owned subsidiary of the FA. This finance would be secured on the basis that the FA would continue to hold it's flagship events at the new stadium. After negotiations with Wembley pic the existing Stadium and business was purchased on 15 March 1999.

THE NEW STADIUM BACKGROUND


The New Wembley Stadium was conceived from the sense that whilst the current stadium could boast a glorious past it had little, if any future. As the centrepiece of the 1924 British Empire Exhibition, it has staged many celebrated football matches, since the famous 1923 "White Horse" FA Cup Final. The most notable football match ever staged however, was the 1966 World Cup Final. The stadium was also the venue for the track and field events during the 1948 Olympics. It has also become the traditional home of the Rugby League Challenge Cup Final. By default therefore it has become the National Stadium. All agree that the current stadium has clearly passed the point at which it can usefully continue to serve the as the National Stadium. Faced with this position and the prospect of a number of applications for major new lottery funded stadia, Sport England established with the Football Association, (FA) the Football League, The FA Premier league, the Rugby Football Union and the British Athletic Federation, a competition for the development of a New National Stadium. In July 1995 bids were received with Wembley emerging as the preferred location. It was intended that the owners of the Stadium, Wembley pic. would take this forward with a body especially established for the purpose, The English National Stadium Trust. The Trust would lease the site from Wembley pic and be responsible for the design, construction and finance for the new stadium. This arrangement ultimately could not be made to work, and in the face of bids from private competitors such as Arsenal FC, the lottery grant assigned to the project The configuration of the new Stadium was primarily ^developed through the brief complied for the competition, and expanded to suit the requirements for expected use. The new Stadium is to be capable of hosting events similar to that of the existing. Together with the primary sporting events of football and rugby, major athletics events were to be capable of being held. Recent national publicity on the topic of athletics at Wembley has shown this to be a very emotive subject, one that thankfully is not for this paper. Other events that complete the portfolio are concerts, pageants and exhibitions. The brief called for a "world class" Stadium which would continue to uphold the traditions of the existing Stadium. A full "bowl" arrangement for the terracing is utilised as the whole of the stadium was being reconstructed. This provides the opportunity to ensure all the are spectators are wrapped around the action creating an intimate environment, full of atmosphere. With this arrangement, the Wembley "Roar" would continue! A spectator capacity of 90,000 was eventually chosen. During the development of the business case for the operation of the stadium a 200 bed hotel, interactive museum, 100,000 ft2 of offices a 5,000 seat (the largest in London) banqueting suite and full spectator hospitality and corporate facilities are to be provided. These facilities could not all be fitted into the concourse areas under the spectator bowl structure and are housed mainly on the North side of the building, facing Olympic Way. The roof of the Stadium is designed to cover in one sweep the both the spectators and the other facilities housed in the building. This has resulted in a very large roof, one of the largest in the world, containing long clear internal spans.

242

The roof of a stadium is a very highly visible structure and which dominates all views of the building, both internally and externally. It therefore demands special consideration from the design team. The existing stadium's roof boasted the "Twin Towers". This very emotive and powerful, national icon was the subject of much debate. To keep the Towers or not? It was not a viable option to retain the Towers in their original positions. The new Stadium needed to be moved to the North to create a piazza around the whole of the building. It is possible to relocate the Towers, however it was felt that a new image should be provided for the Stadium. The roof for the new Stadium was the subject of a very intense and detailed option study. It was necessary to provide a column free space within the spectator bowl and to cover the additional facilities housed in the North of the building. The brief required all the spectators to be covered. There was not however a requirement to have a completely retractable roof, as the primary sporting events for the Stadium had to be held in the open air. However, could the capability for a closing roof be excluded and the roof not "future proofed"? It would be very difficult and prohibitively expensive to try and retro fit a fully closing roof once the Stadium was constructed. Future proofing is discussed later in this paper. The configuration of the spectator bowl adopted with it's much shorter sight distances and a roof covering all the spectators also generates problems of shadow lines on the pitch and lack of air movement over the playing surface. Both of the above cause problems with grass growth. Healthy grass needs both direct sunlight and fresh air. The current Stadium configuration with it's low set back roof, running track between the spectators and the pitch, and the shallow seating tiers allows easy air movement over the pitch. Even at 4.45pm in mid May (FA Cup final) there is only a small portion of the pitch covering the South Western corner flag in shadow. Apart from pitch health, this is very important to the television

companies as their cameras cannot cope very easily with moving in and out of shadow. Given this existing condition it is imperative that the new configuration gives a similar performance for both pitch health and shadow lines. The solution for the roof needs to be able to address these issues satisfactorily.

SCHEME DEVELOPMENT
The starting position for the scheme was that there should be no columns in the spectator bowl. The perimeter of the roof therefore needed to be supported at the high back edge of the bowl. The saddle shape of the bowl edge reflects the capacity requirements together with maintaining the required standards for the pitch and scoreboard sightlines. The edge of the roof will be supported on the perimeter truss. This element of structure is formed from the extension of the bowl primary raking beams triangulated with V props to form a continuous perimeter truss on which the roof edge elegantly rests. It was immediately apparent that the North side of the building would be the dominant area and would need to contain the major supporting elements of the roof. The roof structure could then align primarily North South. If a line of support could be gained at the internal leading edge of the North Roof, main trusses could be utilised to span from this edge to the Southern edge of the bowl. Further advantages to the overall solution could be gained by following this concept. As discussed, the problems of shadow lines and air movement over the pitch needed to be addressed. These problems could be largely eliminated if the roof along the Southern side of the Stadium could be retracted back to allow sunlight onto the pitch. They could be closed before or during a match if the weather deteriorated. The main North South trusses could support retractable roof panels running along the truss top chord. The requirement to cover all spectators would still be achievable, albeit only during inclement weather.

Fig 1 Schematic Section

243

To enable this option to work, a major supporting element needed to be introduced which could provide support to the internal leading edge of the North roof and could be supported outside the footprint of the building, ensuring still a column free spectator bowl. The initial public scheme utilised 4 masts at the front (North) of the stadium with twinned forestay cables attached to internal edge of the roof. Twin backspans cables were anchored back to foundation blocks to the North of the masts. This scheme worked well as a structure, however it was felt that the intrusion of the masts directly in front of the building, together with their cables, and the same mast type solution with other similar structures would not be special or unique enough for the new National Stadium. Subsequently, a solution that found favour with all was produced. This utilised a massive arch, which was positioned over the Northern roof and spanned the whole building East to West. This solution provided an efficient and elegant solution to both problems of roof support whilst giving the necessary icon to the Stadium. A series of forestay cables is attached to the arch supporting the internal leading edge of the Northern roof, and backstay cables anchor to the edge of the bowl, neatly eliminating all the externally anchored cables of the mast solution. Refer to Figure 1. The arch quickly became the accepted image of the new Stadium, replacing the existing Twin Towers as the icon for the new National Stadium. Once basic agreement on the arch solution was reached working models of the Stadium were constructed, Figure 2. A full wind tunnel test model was prepared tested to confirm existing design data, determine accurate wind loads and highlight and quantify any special effects on the roof and arch for this configuration.

THE ARCH
The arch takes the form of a 7m dia. 138m high, 315m span open "basket weave" unclad lattice structure. It is formed of 457 dia CHS longitudinal chords with diaphragms at approximately 20m centres. Alternate diaphragms are primary and support the stays. Steel grades are S355 JO or J2 to BS EN 10025. Rolled Hollow Sections S355 J2H to BS EN 10210. Protection is 400 dft micron epxoy primer / buildcoat and a 75dft micron finish coat, over a blast clean surface to Sa 2.5 of BS 7079, giving the period to first maintenance of 30 years. Access to the arch will need to be undertaken for the following reasons: Structural Inspection Lighting maintenance / replacement Repainting (30 year interval) Festivity / celebration (eg pyrotechnics) Dressing the arch with flags or banners.

It is anticipated that maintenance will be through the centre of the arch. For the more thorough maintenance tasks a pre fabricated platform that is launched from the North Roof would be winched up under the section requiring attention, providing a safe working platform for the maintenance crew. The position of the arch and it's inclination have been the subject of an intensive iterative analysis process. This to try and ensure that under all load combinations the arch acts as far as possible in it's most efficient state, direct compression. The treatment of the springing points of the arch will be important as it is expected that these points will be used by the visiting spectators as favourite spots for

Fig 2 Architectural Model

244

photographs. In the design of the knuckle springing points, there is also the security and safety issues to be considered from people potentially being able to scale the arch. Figure 3 below.

Soffit treatment (lining) will be provided in certain areas to hide walkways and services. This lining will also have acoustic benefits during concerts. Further to this there will also be areas treated with specific acoustic lining panels to both absorb sound and modify the general acoustic properties of the spectator bowl. 350 mm cold rolled nested purlins at 3m centres running East West, support the roofing spanning between the main roof trusses. The centre 10 bays are at 13.5m reducing to 10.5m for the end 6 bays. On the North roof, the primary fink type trusses are 6m deep spanning up to 75m. These are supported from the arch stays and the Northern edge of the bowl. The top chord is a box section generally 750 x 500 mm deep with a cable bottom chords and CHS V struts at third points. A longitudinal truss T i l is provided to support the alternate secondary North South trusses and to provide in plane rigidity to the roof plate when considering assymetric and dynamic loading. The main T3 and T4 trusses span to the Southern edge of the bowl. T3 spans 155m and T4 129m. These trusses, together with the edge T5 trusses support the main Southern roof and carry the rails for the moving roof panels. Lateral stability of these main trusses is provided by a series of horizontal cable ties. At the leading edge of the Southern roof truss T13 is located. This spans 135 m between the T3 trusses to support the central section of the Southern roof. As the truss is curved in plan, diagonal cable ties are introduced back into the roof plate to counteract the bottom chord kick out. The main North South 6 m deep fink trusses are utilised as runway beams for the sliding roof panels. Fabricated box sections top chords are provided, as in the North roof. Refer to Figure 5 for a complete isometric view of the roof structure.

Fig 3 Arch Support Detail

The stays are spiral strand galvanised wires grade 1570. Cables. Cable interior and exterior corrosion protected. First inspection within 5 years of initial coating and a major inspection after 15 years from initial coating. Touch up will be required at the time of the initial inspection, with complete re coating a likely option at the time of the major inspection. Under the approved' maintenance regime the cables have a guaranteed life of 60 years. Generally the forestays cables range between 110m and 135 mm dia. The backstays between 55 and 95 mm dia. There are 8 support points provided along the North roof leading edge. The first 2 supports at the East and West ends are primary picking up the main North South T3 and T4 trusses. These trusses span to the South edge of the bowl and are the main elements which support the whole of the Southern roof area.

SLIDING ROOF
In order to meet the requirement to provide maximum covered seating whilst still allowing daylight for pitch health a moving roof over the whole of the Southern side of the stadium is required. The roof panels are nested over the static section of the roof and at each end double stacked The permanent roof structure running North South provides the runway beams supporting the track for the panels. The area of roof that moves is split into 5 bays, one middle section extending the length of the pitch (135m between trusses T3), and two bays at each end, covering the end stands. These panels are supported of the T3,4 and 5 trusses. The end bay panels are subdivided to allow them to double stack on top of the fixed roof without projecting over the Southern edge of the building.

THE ROOF PLATE


The roof plate main structure runs North South. Refer to figure 4 for the member references. The surface is profiled to fall from the North South centreline away to the East and West. The roofing material is to be a mixture of standing seam aluminium (eg Kalzip) and 30% translucent polycarbonate sheeting (eg Lexan). The polycarbonate sheeting is introduced to allow diffused light through the roof towards the leading edges. This provides a light gradient of open to fully solid roof which improves hard shadow lines.

245

ARCH O E VR

ARCH OVER

Fig 4

Roof Plan

During operation, in order to maintain a positive wheel friction to the rails, it is proposed that a wind speed limit of 20 m/s (approximately 50 mph) is imposed. This is in line with common practice at other operable roof stadia. This limitation to operation pertains to operational reasons only. The panels themselves are designed to withstand the design wind loading at any location, fully closed, part open or fully open. The panels are generally framed by fabricated box sections (up to 3m deep for the central large cantilever panel) which are connected to the running bogies. Secondary framing UB sections are utilised with full diagonal rod bracing for each panel to ensure racking of the panel does not occur. A full cycle for the roof to open or close will take 20 minutes.

the T3 trusses. To cope with this additional load the T3 trusses, the supporting arch stays, the arch and the arch foundations would all need to be strengthened. It is considered that if this strengthening was not carried out at the time of construction, any retro fit would be virtually impossible and expensive to carry out. The technical problems could be overcome however any future fully independent structure over the top of the roof would be visually intrusive and would be unlikely to gain planning permission. This would effectively deny the retro fit route for this option.

SUMMARY
The roof, especially the arch of the New National Stadium provides the icon that will take over from the Twin Towers and continue the tradition of Wembley for the next 50 years. The stadium will be a truly magnificent stage for the major events that will be held there. It will elevate England into a country capable of hosting almost any international sporting event in style. As the kingpin for bids for the World Cup, and the Olympic Games it is considered that any such bid would be very strong built around the New Wembley Stadium.

FUTURE PROOFING
Whilst there is no current requirement for a fully closing roof for the Stadium, there may be, in the future cause to want this facility. The existing design is able to be adapted to cater for this by the strengthening of certain key elements. The fully closing roof would take the form of two additional central overlapping panels clear spanning the length of the pitch (135m). These would be supported on

Fig 5

Roof Isometric

NOTE
The structural engineering for the New National Stadium is being undertaken by the Mott Stadium Consortium.* This consortium is lead by Mott MacDonald and contains Connell Wagner Pty, Modus Consulting Engineers and Weidlinger Associates Inc.

REFERENCES
1 Wembley National Stadium, House of Commons Culture, Media and Sport - Fourth Report 2 March 2000 The New English National Stadium, World Stadium Team February 2000 Key Development Criteria, Technical Development Criteria, English National Stadium Development Company Limited March 1999

247

KEEPING THE DOORS OPEN: THE OLYMPIC STADIUM, SYDNEY


S Morley Principal Modus Sinclair Knight Merz

Fig 1 Aerial view of Stadium Australia

Sports facilities, where shelter is provided, invariably fall into the category of widespan enclosures. Stadia and arenas tie up vast amounts of a client's capital and the planet's resources and therefore should be designed to be suitable for many uses as far as practicable. This may mean providing a range of environments through altering their widespan enclosures. Where field sports played on natural turf form some of those uses the enclosure needs to be open centred to promote grass growth and comply with current

regulations for 'outdoor' sports. Having the ability to close this central opening can greatly increase the possible uses and also provides event surety at least in the face of a temperamental climate. It is, to a degree, possible to quantify the benefits of this flexibility of extent of enclosure by assessing the revenue from additional usage and avoidance of lost revenue from cancelled events (although the latter is difficult to assess as past events at other facilities statistically represent such a small sample of total available event

248

days that little guidance can be drawn from "I've never known an event cancelled in thirty years"). However, this judgement will necessarily be made based on current knowledge of the forthcoming developments in the event markets in which the new facility is intended to compete and the only certain thing about such markets is that they change, often unpredictably. Hence it can be of benefit to provide for flexibility of enclosure or at least for adaptability to be able to introduce such flexibility at a later date. Two of Australia's most recent multipurpose sports and entertainment facilities - Stadium Australia at the Olympic site in Homebush Sydney, and Colonial Stadium, Docklands, Melbourne, are good examples of the application of this strategy. Stadium Australia, the host venue for the 2000 Olympics, is designed to operate in two distinct modes. In Olympic mode the North and South sections of the bowl are unroofed to allow space for large temporary grandstands, each accommodating 15,000 people. After the Olympics these end tiers are due to be removed an the perimeter enclosure completed with the addition of North and South Roofs. At the same time the lower tier of seating will be moved inwards by nearly 16m on the sides and 20m on the ends to greatly improve their proximity to field sports. This strategy was part of an innovative financial package which helped secure the project for the Multiplex led team in a B.O.O.T (build, own, operate and transfer) competition. Under the

auspices of such a procurement method the development team was acutely aware of the need for long term financial viability. Whilst, based on 'current' understanding of how the stadium might be used, there as insufficient justification for providing the flexibility of full as well as partial enclosure, it was considered important to 'keep this door open' by allowing adaptability for this flexibility in the future. The roof of Stadium Australia follows , and in fact was borne out of the geometry of the seating bowl. In the longer term post Olympic mode, from the lowest point on the end stands to the highest point on the side stands there was a height differential of perhaps 40 metres and the natural shape to fit this saddle perimeter was the hyperbolic paraboloid. With this geometry the roof over the side stands curves gently downward maximising weather protection for every dollar spent on the roof whilst hugging the higher sightlines. There was also potentially a construction advantage in that this iconic doubly curved form is generated from two sets of straight lines parallel to but progressively rotated from two principal generators at 45 degrees to the main axes of the stadium. Structurally such a surface can be very materially efficient as loads can be transferred by in plane or membrane forces. However this geometric and structural purity is rudely interrupted by the roof plan form in Olympic mode when just crescents of roof over the side stands are required. In plane action across the stadium is not possible in this

Fig 2 Architectural Image of the Stadium with End Roofs in Place

Fig 4 Architectural Image of fully roofed arena

mode and instead the surface would have to act primarily in bending to some form of edge stiffening along the front edge of each crescent. Nonetheless the other advantages of the hyperbolic parabaloid remained and it was decided to pursue this geometry. True to this surface, the front edge of the crescents were parabolic which suggested a form of arch as edge stiffening. Once provided the arch stiffening would attract a large proportion of the load in both Olympic and post Olympic modes and to increase the efficiency of this system still further the main arch was lifted above the HP surface by up to 12 meters at its apex. The concentrated lateral load components generated by arch forms are best taken directly to foundations on a continuation of the line of thrust (curve of the arch). However, the arch line had to remain above the HP surface to the edge of the roofed area which left it more than 20 metres above precinct level. There was simply not enough site area to continue on this line and therefore some form of cantilever thrust block was required. Interestingly a degree of mitigation was provided by allowing the arch line to change in response to areas of concentrated gravity loads. Because of the geometry of the opening such a concentration occurs where the roof ends meet the arch (here there was also a requirement to support 50 tonne video screens) and the arch line was therefore deliberately diverted inward and downward at the ends like the profile of a crab. This lowered the thrust blocks slightly such that the arches could spring from a point 17 metres above the precinct. The resulting roof geometry therefore was the hyberbolic paraboloid surface curtailed to a vertical cylinder defining the back of the seating and with two crabbed arches stiffening the edges of the central rectangular opening in post Olympic mode and providing full edge support in Olympic mode. Early on in the development of this design the client instructed the designers (Architect Bligh Lobb Sports Architect and Engineers Modus with Sinclair Knight Merz) to consider how a fully closing roof could be provided and what steps might sensibly need to be taken

now to ensure this remained a possibility for the future. It was found that geometrically the hyberbolic parabaloid could readily accommodate a fully closing roof. The arched edge reinforcement running East West provided natural trackways for a simple sliding system albeit on a constant radius curved track. Furthermore, the fact that the top surface of the end roofs dishes down following the HP surface between these trackways provided additional space for the depth needed for the moving roof panels to clear span the 100 metres plus between the rails. This property was particularly useful as it proved necessary to stack a pair of moving leaves above arch and fixed roof within the curtilage of the stadium. Inclusion of a closing roof would therefore increase loads on the arches and their supporting thrust blocks and foundations requiring commensurate strengthening. Also as this additional loading is to one side of the arch centreline, the roof diagrid supported off the arch which restrains the arch horizontally also would require strengthening locally. The design work instructed allowed informed decisions to be taken on which aspect of their strengthening could and should be carried out during initial construction to 'keep the door open' whilst minimising the amount of capital tied up in steel and concrete for a potentially lengthy period before any revenue is gained from this benefit. The strategy for

Fig 5

End segment of arch truss

Fig 6

Computer model of west stand

strengthening the arches was to add a new central compression element within the arch trusses. This could conceivably be carried out at a later stage by unloading the arches using the same temporary towers employed for their construction and feeding in the new compression member in pieces small enough for assembly by site welding (used extensively for initial construction of the arches). However, strengthening of the arch supports would be altogether more involved to the point of being impractical. Fortunately it was found that the cost of strengthening the pins at the ends of the arches, the pin plates, thrust blocks and foundations was reasonable and therefore these measures were included in the initial development. Also it was found that the diagrid strengthening local to the arches could be achieved simply and with minimal cost by deepening the diagrid in this region. By this means the option of installing the future closing roof has been kept open in a logical way and with the minimum of additional expenditure and construction beforehand. With Colonial Stadium, in Melbourne's Docklands, the decision was taken that the roof should be closeable, subject to the additional costs involved being supported

by the business plan. Thereby a stadium designed as an open stadium complete with an AFL sized natural turf pitch could become fully enclosed to operate as a multipurpose venue to attract a greater number or events. Although having a moderate seating capacity of 52,000, the infield is exceptionally large by European or US standards, to accommodate the AFL pitch and closing the 1.7 hectare retractable roof creates an enclosed arena of vast proportions. Several forms of retractions were explored and a simple sliding mechanism with two 50 x 165 m span doors selected for cost efficiency and design efficiency. Such an arrangement naturally concentrates loads on two distinct lines at the ends of the doors and recognising this the Architectural team of Daryl Jackson Architects Pty Ltd and Bligh Lobb Sports Architects, proposed that the upper tier be divided into quadrants allowing four corner supports to be 'pulled in' under these loads. This arrangement could readily be accommodated as the moderate capacity compared with the long infield perimeter meant seats would be redistributed without detriment to viewing quality or efficiency. The Engineering Team of Connell Wagner with Modus used this opportunity to the full to produce a very efficient structure which, even with the 165m span doors

Fig 7 Colonial Stadium main roof support arch

and supporting tracks weighed less than 100kg/m2 overall. In detail the doors comprise shallow I sections suspended off pairs of asymmetric tied arch prismatic trusses. This asymmetry was borne out of a desire to minimise shadows on the pitch and to reduce the visual impact of the roof by chamfering the volume of the inner truss to the pitch side and outer truss to the precinct side respectively. By making the I section roof beams deliberately flexible the doors are better able to accommodate differential movements and slopes of their supporting tracks. The tracks themselves comprise 4m deep prismatic trusses which are carried over the 120m span between corner supports again by tied arch trusses with chamfers to the pitch side. Great care was taken in developing active positioning systems both of the door relative to the tracks, to control racking of the bogie sets relative to the door trusses, to minimise the forces induced in the structure should racking occur. This was particularly important here as racking forces would have the benefit of a 165 metre lever with which to annoy the fixed supporting structure. Constructors, Bauderstone Hornibrook have now completed the stadium and it is hoped its flexibility of enclosure will contribute to the long term success of the venue. In conclusion, the sports and entertainment market appears increasingly competitive. In this market it can be an advantage to have flexibility of enclosure or at least the ability to add this flexibility later. Both approaches can be integrated into Structural and Architectural designs seamlessly if included into the design scope early enough.

252

WIDESPAN ENCLOSURES AND STRUCTURES COST CONSIDERATIONS WITH ILLUSTRATED CASE STUDIES
Alan S. Willby Partner Davis Langdon & Everest Construction Cost Consultants

INTRODUCTION
DLE have had the privilege over the years of being involved with some highly innovative schemes which fall, for the purposes of this paper, into the category of "Wide span Enclosures / Structures". In reaching this definition we have considered any clear span that is in excess of 30m at its minimum. Rather than preparing a paper that purely compares the relative costs of different structural solutions to the problems of spanning such large distances we have also included detail of the actual costs of a selection of schemes with which we have been involved and have reliable cost data. The reason for doing this is that, in our experience, the decision process followed in arriving at the eventual solution very rarely emanates from a full analysis of all available structural solutions and material choices. One good reason, of course, is that the design team are rarely paid a fee to complete a number of designs to the level where a complete and total analysis can be performed that involves Substructures, Main Superstructure Frame, the Roof Structure itself and other elements such as external cladding and M&E Services Installations. We normally find that the requirements of the structure and the architectural form tend to suggest a system which, is then developed in terms of cost. Of course different structural solutions often lead to alternative architectural solutions, which again would need to be worked up fully if a full comparative cost analysis were required. Often wide span structures are extremely complex and display a cost sensitivity that demands that specialist contractors are required for their expert advice. This again tends to reinforce the favoured design solution. One of the biggest problem for anyone providing cost advice is that of the large quantity that often can only have one rate applied to it. There is often a lack of detail, at concept stage, in respect of those areas that have a cost impact. DLE have an advantage in this field as we do have experience of all structural forms and, consequently, have a detailed and reliable database and can advise upon the likely costs of alternative solutions. When we first became involved in Arenas, a new building type in the UK, in the mid 80's the general

principle was to look to reduce weight. This of course was counterproductive if the costs of fabrication and erection are not considered. Furthermore there is often a different point of view between the designing engineer and the specialist contractor who will carry out the work as to the relative costs of certain aspects of the design. Over the years we have therefore taken a more balanced view to the cost of such structures and have resisted the temptation to merely consider weight, but rather concentrate on a system being designed within the context of the function and architectural form required. Of course for a low cost distribution warehouse utilizing a simple truss solution we will revert to type eventually, but we are QSs after all! For the purposes of this paper we have arbitrarily sub divided wide span buildings into three basic types, namely: Sport / Entertainment (includes Arenas and Stadia) Exhibition / Museums / Galleries Sheds

(includes Distribution Centers, Warehouses, Manufacturing Facilities, Retail Outlets, Airports and Railway Stations) We have done this, as it seems that the first category deals with uninterrupted spectator views as its primary need, the second often have a high architectural influence and quite specific needs associated with the use of the building and the third is the remainder. It goes without saying that all require maximum flexibility of use. We will now discuss the cost drivers for wide span enclosures / structures.

253

COST DRIVERS
The cost drivers for wide span enclosures and structures can be split into two quite specific areas. Firstly there are those that are fundamental in the decision for the use of a wide span enclosure / structure, namely:

made clear to the Client. For example does the building need to be column free. What is the cost of that decision and does it fit with the Business Case. Secondly are those that are specific to the structure itself, the necessity for a wide span enclosure / structure having been already made under the first criteria, namely: Span Environment Self Loads Working Loads This is normally the more technical aspect of the analysis where the costs of a particular project are compared and benchmarked to other similar schemes. This is an area where accurate, known, cost analysis and advice is so important At this stage we will consider different systems such as: Trusses Spaceframes Shells Cable Nets Air Supported Moving Roofs

Function
This changes from one building type to another. For Sports and Entertainment buildings the main purpose is to provide uninterrupted views and spectator comfort. For Exhibition / Museum / Gallery buildings the purpose is the same, but with exhibit flexibility and regard to the particular needs of the exhibited material. For Sheds the main purpose is a completely column free space to allow maximum flexibility and ease of use.

Of course some large buildings are only temporary! These are special case indeed.

Utilising differing materials such as:

Architectural F o r m
This is a very important aspect for wide span enclosures. By definition these buildings are large, often very large indeed. In some locations the need for an aesthetic solution may not be that important, either from a planning or "context" point of view. Good examples are buildings such as the Boeing 747 Assembly facilities in Seattle (worlds largest building by volume at 7.77 million m3, with a 220,000m roof area) and the Vertical Assembly Building in Cape Kennedy Space Centre, Florida (measuring 218* 158* 160m high which is the second largest building by volume at 5.51 million m3).
2

Concrete Timber Steel Aluminium Fabric With consideration, where necessary to innovative materials, such as: Kevlar Carbon Fibre Once these basic decisions are made almost the most important process of all begins, that of procurement.

However many buildings require a sympathetic treatment and often the structural form is the primary method by which this is delivered. Some buildings are fortunate in that particular aspects of their function almost dictates their form, and therefore aesthetic. Such occasions are not as rare as may be expected. Good examples are the Eden Project and the American Aircraft Collection building, both of which are subject of case studies in this paper.

Cost
Given all of the above the cost needs to balance the previous two functions, to a level that is acceptable to the Client and represents value for money. It is most important that all the basic assumptions are queried in depth and the cost implications of the decisions reached

254

PROCUREMENT
The procurement of large span structures is absolutely critical. These structures always represent: A significant percentage of the overall cost of the project and corresponding cost risk. A significant design programme implication. A significant construction programme implication. A significant construction methodolgy implication. A significant overall project risk.

SPORTS AND ENTERTAINMENT


Case Study - National Indoor Arena, Birmingham

Even in the case of simple structures, such as steel trusses, it is important to obtain specialist advice at the earliest stage. One should follow a route that gives : Design certainty Programme certainty Cost certainty Buildability Speed Competition We have detailed some of the different routes used in the Case Studies that follow.

Architect: Engineer:

HOK Geiger Associates / Mero

CASE STUDIES
We have presented here a number of Case Studies and the results of cost analysis carried out by DLE into different building types. The Case Studies are intended to give an indication of cost and also some background as to why the particular structure was adopted and how it was procured. The Case Studies cover Sport and Entertainment Buildings, Exhibition / Museum / Gallery Buildings and Sheds as follows:Sport and Entertainment: National Indoor Arena - Birmingham Odyssey Arena - Belfast Sheffield Events Centre - Sheffield Manchester Velodrome - Manchester City of Manchester Stadium - Manchester Exhibition I Museum I Gallery: Eden Project - Cornwall American Aircraft Collection, Imperial War Museum - Duxford Sheds: Waterloo International Station

Mero Space Frame. 600 Tonnes I27m*90m clear spans 50kg/m 12,000m roof area 3,433,000.00, including roof coverings, in 1990 equating to 286.00/m Steel cost approximately 2,500,000 in 1990 equating to 4,200/tonne
2 2 2

This scheme is a good example of a project where the eventual design solution is completely different from that originally envisaged. In this case the design for the roof structure had to be as light as possible, furthermore the initial design also envisaged the main loads coming down in four locations. This was due to site constraints, where the building straddles a railway line (the floor of the Arena is the tunnel roof!) and the Arena floor is 14m above general site level. A design was developed by the Structural Engineer which involved two major trusses forming a diagonal cross over the building. Of the 1,000 tonnes in the roof some 70% of the weight was in these two structures which, in themselves were highly complex, utilising cambers, "crucifix" bottom chords, changing sections every 9m, etc. When tenders were returned the cost of these elements was in excess of 6,000/tonne, resulting in an overall roof rate of 5,000/tonne, at 1990 prices. When the contractors were interviewed it became apparent that the complexities and risks foreseen in fabricating these two major elements had led to the level of pricing. In particular the risk factor was a major cost driver with the overall price making allowance, effectively, for a "prototype" truss. At the time there was major pressure for the contract to be signed. The only viable alternative on offer was an allin structure and coverings price for a space frame

255

solution. This was an expensive solution, still over budget, but was cheaper and required minimal re-design to the structure below. Further it was calculated that savings would accrue from the deletion of the four major roof supports, loads being transferred instead via the outer bowl columns.

Case Study - Sheffield Events Centre, Sheffield

Case Study - Odyssey Arena, Belfast

Architect: Engineer:

HOK/Marshall Haines + Barrow Oscar Faber

Built for the World Student Games Truss Solution. Part Hollow Section 118m* 89m spans 800 tonnes which equates to 80 kg/m 2,200,000 in 1990 which equates to 2,750/tonne 10,000m roof area
2 2

Architect: Engineer:

Marshall Haines + Barrow / Consarc Design Group Oscar Faber

This building was procured using the Management Contracting Form of Contract, and the Steelwork Contractor was thus a specialist Package Contractor who tendered on a competitive basis.

100m*70m clear spans 700 tonnes which equates to 100 kg/m 1,600,000 at 1998 which equates to 2,285.00/tonne 7,000m roof area Arched Truss Design As a point of interest this building contains a Bow String Truss internally at one end to support approximately 40m of concourse where it oversails the Indoor Running Track.
2 2

Case Study - Manchester Velodrome, Manchester

This building was procured as a Two Stage Design and Build. As part of the Stage One Tender Submission the Contractor was asked to confirm fixed rates for the roof structure and then to work with the Clients Design Team in Stage Two to develop the design within the fixed price. This proved to be a successful strategy.

Architect:

HOK

120m Main Span 40m Trusses 615 Tonnes Weight 75kg/m 1,200,000 8,200m plan area
2 2

Design and Build Procurement Route.

256

Case Study - The City of Manchester Stadium

Generally: As can be seen extensive experience of Arenas and Stadia has been gained. This has allowed us to analyse cost drivers in some detail as follows:

A R E N A R O O F S - K E Y ISSUES

Trussed Solutions: The key issues related to an Arena Roof Structure concern the economics of large clear spans and the design for progressive collapse. Of the structural alternatives available to designers the most used are trusses with some use of space frames and glulam beams. Architect: Engineer: Arup Associates Ove Arup and Partners / Arup Associates Steel solutions, as well as being relatively inexpensive, have a structural depth that is useful in accommodating mechanical services, PA Systems, Service Gantries and Walkways, etc. Clear spans of more than 60m incur significant cost penalties related to the tonnage of steel required and the complexity of the fabrication and erection. The average tonnage of a 60m clear span roof is 85-90kg/m , increasing to an average of 100-110kg/m for a roof spanning 100m.
2 2

Ground anchored cable stay roof with tension ring. 1,900 tonnes steel purlins, struts, stays, etc. 200 tonnes cable 90 kg/m overall Cost cannot be disclosed as yet. Rates vary from 2000/tonne to in excess of 6,000/tonne for some elements 23,000m area
2 2

The current design solution is the last in a series that have been developed over the years. For a long time a moving roof solution was being considered, which led to a solution with a pair of arched leading edge trusses (similar to the Hong Kong Stadium illustrated later in this paper). When the moving roof was eventually deleted for financial reasons, the leading edge arches remained. However the Architect was unhappy with a straight roof edge over a curved bowl and this led to the current solution, where a cable forms the leading edge. The main contract is progressing as a Two Stage Design and Build. However in this instance we took the slightly unusual step of utilising a separate two stage approach to tendering the steel package. In fact the steel tenders were invited before the main contract tenders. This was to ensure that this important, and cost and programme critical, element of the design was tested early and that a specialist would be working with the design team from the earliest time possible. We asked the steel tenderers to provide a detailed price breakdown that split every element into: Material Shop Fabrication Erection

Arena roof structures are designed as working platforms providing a load bearing element for scoreboards and concert staging. Central scoreboards weigh at least 10 tonnes, and a total suspended load allowance of 40 tonnes should be provided above stage areas. One can query why there are not more lightweight solutions given the systems available today. For example the Louisiana Superdome was constructed with a steel lamella domed roof with a diameter of 210m and a total steel weight of 125kg/m .
2

Technically there is no limit to the length of span or area of enclosure that can be achieved. The limitations, as always, are financial in measuring capital cost against cost benefit, i.e. 'value for money'. Lightweight structures with minimum use of material can be erected quickly for minimum cost.

This was to enable a full analysis to allow costs to be saved and the design to be developed in the most important areas. To date this has been a successful strategy.

Space Frames: These are essentially three dimensional structures with loads being spread in all directions and balanced out. Consequently peak loads are diminished, inner stresses reduced and cross sections of members reduced. Theoretically therefore these structural solutions should be more economical for long spans, and a number of proprietary systems have been developed over the years. One of the best known is the Mero System where steel tubes are screwed into forged connectors capable of accepting up to eighteen members. As detailed earlier this system was used in the NIA Birmingham.

257

Space frames can also utilize aluminium when an even lighter weight is required. The Exhibition Hall in Anhembi Park, Sao Paulo, Brazil measures 260m*260m with a span of 60m. Aluminium has the advantage of weight, ease of maintenance and transportation, but typically tends to be more expensive than steel for the same spans and loadings. The weight of a space frame is typically 65-70% that of an equivalent truss solution. The use of a space frame can generate additional savings, related to frame and foundation design, that offset the higher cost. However, the small, proprietary, market for space frames leads to procurement risks at an early stage, and most designs adopt a trussed solution for this reason at the outset of the design process.

Cantilevered: These have either a bulky "hockey stick" rear truss, which requires additional space and can impact adversely on surrounding areas, or prop, or tree, cantilevers where the geometry of the cantilevers affects the structure beneath. Cantilever trusses are usually more expensive than king truss structures. Further costs can be incurred because access for cranes is often limited to the rear or end of a stand.

Reebok Stadium, Bolton, England

STADIA R O O F S - K E Y I S S U E S
There are three main types of roof structure. King Truss: This is similar to beam-and-post constructions, but no intermediate columns are used between the end posts. Although frequently regarded as the cheapest solution, it is not appropriate for schemes where the stands are swept around the corners, and this now represent the preferred stadia design solution. Main truss depths often equate to 1/12th of the length of the span (usually the stand length), and access for future painting and repair is always a problem. Trusses that use hollow sections can be considerably lighter than their solid counterparts, but the unit costs are usually higher - 3000/tonne compared with 1500/tonne. A more recent development has been the use of large leading edge arches. Hong Kong Stadium, Hong Kong

Tottenham Hotspur FC, London

Mm$&

^SKtmhi

^^^^^

^3^H

\m\m\Wmar Tension Structures: These are generally the most expensive, but can give greater architectural expression. (See The City of Manchester Stadium Case Study).

258

Simple Cost Comparison: Most isolated stands are designed using either king post trusses or cantilevers. The price of both systems is dependent on the square of their length. This means that short, fat stands are always cheaper using the king truss method, whereas long thin stands are cheaper cantilevered. Guide to roof types for various stand sizes:

The probable extra over cost was in the region of 15 million. We did also consider the cost of a "retro-fit" moving roof which had a cost in the region of 20 million.

Stand Length

Stand Width

Recommended Roof Design Cantilever Either King Truss Cantilever Either King Truss

Air Supported Structures These are convex or cylindrical membranes which are prestressed through an internal air pressure. Although the ideal shape is spherical unusual forms can be attained if desired by the introduction of varying stiffness. Avoidance of wind buffeting is critical, but most other aspects have high tolerances.

110m (typical side stand)

22m 33m 50m

70m (typical end stand)

14m 21m 31m

Weight and cost of steel work (typical stands - actual costs):


Scheme 1 2 3 4 5 Structure Type King Truss King Truss King Truss Cantilever Cantilever Width m 37 37 57 29 40 Weight kg/m
2

Cost /tonne 1214 1250 1230 1500 1100

157 185 207 163 208

Prestressed Membranes The more curved the surface, the more effective the prestress as a means of providing surface stiffness. Large variations in curvature across the surface can lead to substantially different properties in different parts of the surface. For very large surfaces (enclosures) the use of rigid cables supported on masts is the norm, although stiff arches, portal frames and lattice shells can also be used. This form of roof is very effective in areas where earthquakes or wind buffeting are a potential problem due to their low mass and high frequency. The decision to use a fabric structure can be due to: Image Cost Special Material Properties (i.e. transluscency, radio interference)

Base date July 1995 - South East England

Location.

Moving I Closing

Solutions

These clearly are the discussion point of the moment in the UK, whereas they have been common in the USA and Canada for many years. It is interesting to note that the USA has never found a problem in justifying these expensive solutions from a Business Plan view, but there is always a problem in this country. Millennium Stadium, Cardiff (Illustration Only)

The air supported roofs at Vancouver, Pontiac and Minneapolis have clear spans in the range of 200m with surface areas in the region of 25,000m and are reported to have been cheaper to construct than rigid structures.
2

When considering such roof types greater emphasis than normal is placed upon matters such as: Life Durability Maintenance Fire

If high translucency were to be combined with improved insulation then one can move towards the concept of zero energy structures. The problems associated with these roof forms, when in direst competition to more traditional forms, appears to be in the area of procurement risk. These systems prevail when technical or architectural considerations dictate such a solution.

A recent stadium scheme with which DLE have been involved has an overall cost for a moving roof of approximately 26 million. This means an additional cost over a normal roof of approximately 17 million. As stated earlier The City of Manchester Stadium originally had a proposal for a moving roof. This solution would have used a pair of leading edge arches with the moving roof panels "stacked" at the ends of the stadium.

259

EXHIBITION / MUSEUMS / GALLERIES


Case Study - American Aircraft Collection, Imperial War Museum, Duxford

SHEDS
Case Study - Waterloo International Station, London

Architect: Engineer:

Foster and Partners Ove Arup and Partners

Architect: Engineer:

Nicholas Grimshaw and Partners Limited Anthony Hunt Associates Limited

Single span vault, 18.5 m high. Arched geometric shape, torus. 90m span Twin shells of concrete used. The mass of 7,500 tonnes is also used to control internal humidity. 333no. llm*2m precast Tshaped panels 6000m area in total Cost 2,528,000 Hybrid Hanger/Museum 405/m cost over area of roof structure. Add cost of frame and this rises to 510/m . Roof membrane adds another 50/m
2 2 2 2

480m long with max 49m span tapering to 33m Steel FA 4. l m 203/m in 1990 56kg/m 3630/t Total roof cost 685/m Three Pinned Bow String Arches Maximum standardization in fabrication
2 2 2

Construction Management Contract.

Case Study - Eden Project, St. Austell, Cornwall

Other Sheds by Type Expected typical weights for: Airports: Hong Kong Airport Span 36m Weight 70kg/m Aircraft Hangers: Span 100m Weight 150kg/m
2

Architect: Engineer:

Nicholas Grimshaw and Partners Limited Anthony Hunt Associates Limited

Clinker Store: Span 95m Weight 100kg/m

Mero Space Frame Biome facts (one element only) Plan Area 15,590m Surface Area 20,000m Volume 330,110m Maximum Internal Height 43m Maximum Structure Height 55m Steel Weight 465 Tonnes Frame Weight / Plan Area 30kg/m Span 100m diameter Cost 3,800,000.00 for steelwork which equates to 8,000.00/tonne or 190.00/m of surface area Cost 4,200,000.00 for the Foiltec Coverings which equates to 210.00/m of surface area
2 2 3 2 2 2

Retail Superstore (Food): Span 46m Weight 90kg/m


2

Distribution Centre: Span 30m Weight 38kg/m


2

Warehouse: 30m span @ 7.5 centres - 20kg/m 40m span @ 7.5 centres - 30kg/m 50m span @ 7.5 centres - 40kg/m

Costs are in the range of 1,250 - 2,500/tonne

The NEC form of contract has been used within a target Cost framework.

260

SUMMARY
It seems clear that the design of wide span enclosures has demanded, with great success, innovative thinking and solutions. The Cost Consultant's role when dealing with innovative design is to ensure that it is not stifled but that it is cost effective, represents value for money and that the risks of implementation are minimised to measurable levels. DLE continue to be at the forefront in being capable of providing such advice as the range of successful schemes presented in this short paper illustrates. A far cry from our early "keep it simple, keep it light" approach.

REFERENCES
Coxeter-Smith, J.A., Stadia Cost Model, Building Magazine, July 1995

Forster, B., The engineered use of coated fabrics in long span roofs, Arup Journal, 1985

Wilkinson, C , Supersheds, 1996

Willby, A.S., Arena Cost Model, Building Magazine, October 1999

261

LIGHTWEIGHT ENCLOSURES IN JAPAN


Kazuo Ishii Professor emeritus of Architectural Engineering Yokohama National University

ABSTRACT
Lightweight structures which are used in large-span structures can offer a sense of lightness to people, giving new direction to architectural design compared with that of traditional structures. In Japan, technologies for lightweight structures are being developed, with consideration given to characteristics of construction and earthquake resistance. In 2002, the Soccer World Cup will be held cooperatively in Japan and Korea, and 20 stadiums are to be newly built or renovated. In many of them, a hybrid structure consisting of cables and membranes will be adopted for the roof above the spectators' stands; this type of roof has unique characteristics which are not realized by other materials, such as brightly lit audience seats due to the translucency of membrane materials, its lightweight, ease of formation of curved surfaces, and ease of construction. In this article, outlines of lightweight membrane structures used for the roofs of these new soccer stadiums and large span enclosures in Japan are presented, and their characteristics and problems are discussed.

q = 120 W H

kgf/m

(q = 1176

WH

Pa)

H: Height of the building (16 m or higher)


q = 60x>/H kgf/m
2

(q = 588 -JH

Pa)

H: Height of the building (16 m or lower) There is a difference between the legally established design wind velocity and design velocity based on actual past records of wind velocity at a meteorological observatory located close to the construction site. At present, it is determined that designers should survey the wind velocity at the construction site with a return period of 100 years using past records of wind velocity obtained at a nearby meteorological observatory. The value obtained should be compared with the legally determined value, and the higher value should be used as the design wind velocity. Changes in the value of design wind velocity in the direction of height due to the conditions of the site, location and surroundings should be appropriately evaluated, and the design wind velocity should be set in accordance with the height of the building. Matching of the difference between the legally established value and values which take into account site and environmental conditions is an issue to be resolved. Wind pressure coefficient The distribution of wind-pressure coefficients on a curved surface is complex; it should basically be obtained by wind-tunnel experiments, excluding cases of roofs with simple forms such as spherical and cylindrical shapes. Peak pressure coefficient -1.5

1. L O A D A N D E X T E R N A L F O R C E F O R LIGHTWEIGHT STRUCTURES 1.1 Wind load


Membrane and cable structures used in large-span structures are greatly influenced by wind. Consideration of the problems of deformation and dynamic vibration induced by wind is important in the design of the structures; therefore, they are discussed as follows. Wind velocity Japan is located in the path of typhoons, and each year, several typhoons pass through Japan during early autumn. The wind load is set in the design, based on the assumption of the wind velocity of typhoons. The standard wind velocity is legally defined as the maximum instantaneous wind velocity of 62 m/s 10 m above the ground, and structures are designed on the basis of this value. For this value, the following windvelocity pressure is legally used in accordance with the height of buildings in the actual design.

Figl Peak wind pressure

262

Particularly high peak wind pressure generated at the edges and angular parts of the roof should be considered. The peak wind pressure affects the roof material and the points at which it is attached. In the wind-tunnel experiments, measurements should be carried out with consideration given to peak pressure. Legally, the peak wind-pressure coefficient on the roof surface and at the angular parts (within 3 m) of the roof should be -1.5 or above in the design. Evaluation of indoor internal pressure (consideration of the increase in the internal pressure due to breakage of the glass of windows or doors) When a door or window breaks due to strong wind, the indoor internal pressure increases, which affects the roof surface. A membrane material used for the roof is generally subjected to wind load in the upward direction; therefore, increase in the internal pressure increases this wind load. The wind-velocity pressure q for wind velocity V at the entrance point of wind is generally given as follows. Increase in the indoor internal pressure ; q = V/16 kgf/m* (=W 1.63 Pa) The occurrence of an incident in which the glass of windows and/or doors significantly breaks greatly affects the indoor internal pressure. An accident caused by such the case occurred in Japan. To reduce the level of damage, it is necessary to predict the internal pressures during such incidents and design the structure accordingly. However, judgement on whether or not a design should, or is allowed to, assume the breakage of exterior materials including glass is difficult; this is currently determined independently by each designer. Flying objects on membrane surface A membrane surface is easily damaged by flying objects with sharp edges. Flying glass chips can directly penetrate a membrane surface and cause damage. The membrane moves significantly under strong wind, which will easily propagate tears caused by such damage. Examples of accidents caused by such incidents have been reported in Japan.

wind load is strong enough to cause its curvature to be reversed, the direction of wind load changes; accordingly, the wind-load direction for the shape of the membrane surface after deformation should be taken into consideration to obtain the stress applied to the membrane. Vibration of membrane surface due to wind Vibration is a part of the behavior of the membrane surface under wind. The low rigidity of the membrane material leads to large deformation. When the reversal of the positive/negative curvature of the membrane occurs due to wind, vibration which has a 0-tension region, i.e., fluttering, occurs. This fluttering must be prevented in consideration of the durability of the material. In particular, fluttering is problematic in glass-fiber fabrics. Similar consideration is required for cables. Damping of membrane surface vibration The forced vibration on a membrane surface is fairly easily damped experimentally, when the curved surface has curvature in two directions. In addition, the closer the curved surface approaches to a flat plane, the smaller the clamping coefficient. This may cause fatigue of the membrane material, due to the occurrence of traveling waves on the membrane surface when the roof membrane is extended to a flat shape.

1
. 1 j

1
j _

1
i

Photo 1

Membrane surface damaged by a flying object

j
~"i

Consideration of changes in the wind-load direction due to deformation of membrane surface by load When the deformation of a membrane surface due to

j
1.2 1.4
1.6

0.0

0.2

0.4

0.6

0.8

1.0

Time (sec)

Fig 2 Damping of membrane surface vibration*!

263

In Japan, which is located in a region characterized by strong winds, the issue of wind load is a critical problem for secondary materials including exterior materials and glass materials only when building with regular reinforced concrete (RC) structure or steel framed reinforced concrete (SRC) structure. For such buildings, earthquake loads are more critical in Japan. However, wind loads become an important issue for lightweight structures. At present, the number of examples of the measurement of the wind-pressure coefficients for large roofs with curved surfaces is increasing, and time-history response analyses are carried out for structures with large spaces; therefore, behaviors of roof surfaces under strong wind are being analyzed more accurately. However, with respect to local peak wind pressure at angular parts of roofs, actual values are predicted to differ from those obtained in wind-tunnel experiments. It is difficult to set an accurate value of local peak wind pressure; therefore, designers are required to set a value with a sufficient margin of safety. There are several actual examples of damage to large membrane structures due to typhoons. However, since the history of membrane and cable structures is short compared to that of traditional structures, they have not yet been subjected to a sufficient variety of conditions, and unexpected accidents may occur. Even without serious accidents, lightweight structures are often subjected to critical conditions. It is particularly important to clarify the behavior of lightweight structures in a typhoon to ensure the safety of the structure. This situation has significantly influenced the design of lightweight structures in Japan, and has lead to the adoption of rigid frames.

Design snow load for general areas: short-term snow load only, 60 kgf/m2 (588 Pa), (depth of accumulated snow: 30 cm)

For a roof with a membrane surface, accumulated snow easily slides off compared with roofs made of other materials. If the roof membrane has a shape which allows easy sliding of snow, a decrease in the gradient of the roof, which is permitted only for membrane structures, is possible based on snow-sliding experiments; therefore, when there are no objects which prevent the sliding of snow off a roof surface, the design can be carried out using the decreased gradient.
Table 1 Membrane roof slope (?) and multiplier for snow load Slope (0) Multiplier Snow load 1 1/2 1/3 0 0<25 25<0< 30 3O0<4O 4Os0

Photo 2 Snow sliding off experiment

1.2 S N O W L O A D
Snow loads differ depending on the area in Japan. A number of areas along the Japan sea coast and in the north of Japan are subject to heavy snow. Snow loads are classified into short-term snow load and long-term snow load in the design of structures, and different allowable unit stresses are set. Design snow loads for areas in the north and on the Japan Sea side: Short-term load: 450 kgf/m (4410 Pa) (depth of accumulated snow: 150 cm, specific gravity of accumulated snow: 0.3). Long-term snow load: 70 % of the short-term snow load value. (Short-term snow load is defined as the snow load after 3 days of snowfall)
2

Points to be considered in design of membrane)'cable structures in terms of snow accumulation Problems concerning the snow-accumulation mode on a curved roof surface : The state of snow accumulation on a curved surface differs depending on conditions such as temperature, gradient and shape of the roof, effects of wind and the quality of snow. In particular, snow accumulation on a concave part of the roof produces a high local load due to sliding snow from surrounding parts. Local snow load : Local snow load may be produced due to the shape of the roof surface and sliding snow. When the snow load concentrates at a certain region of the membrane surface due to sliding snow and melt-water, the concentrated load should be appropriately evaluated with consideration given to deformation of the membrane surface. Regarding "ponding", snowfall is the primary issue in Japan. In cases of fairly high atmospheric temperature, accumulated snow contains a large amount of water and becomes heavy. "Ponding" should be considered when determining the shapes of the membrane surface and cables, and deformation characteristics should be verified in terms of local load.

264

Single-side load on a curved roof should be considered. In lightweight structures, the degree of deformation due to loads is fairly high. The change in the snow load due to the deformation of the roof surface should be considered. Structural details of the attachment parts of the roof membrane (projections on the roof surface to prevent snow sliding, ice, cables, steps, details of eaves) may prevent snow sliding, which should be considered. Snow-melting system: In heavy-snow areas, the use of snow-melting systems is required to reduce the snow-accumulation load on a membrane surface. There are several methods of melting snow, (i) blowing hot-air along the inner surface of the membrane, (ii) applying infrared rays to the inner surface of the membrane. The snow-melting characteristics vary depending on the reliability of the snow-melting system, as well as conditions such as temperature, snow quality, gradient of the roof and effects of wind; therefore, the performance of the snow-melting system should be verified based on accurate experiments. Failures in the snow-melting functions directly lead to tearing of membrane surfaces and ponding.

ratio of tensile stiffness Et in the warp direction to that in the weft direction coincides with the ratio of Poisson's ratio v of the warp direction to that of the weft direction).
%
_
F
V

wf

Et

v ^

For this reason, uniaxial and biaxial tensile tests are performed on the membrane material, and the design elastic constant of the material is determined based on the least squares method using the results obtained from these tests. Standards for material constant test methods were established by the Membrane Structures Association of Japan, and elastic constants are currently determined based on this standard. In addition, for determination of the design in-plane shearing stiffness of membrane materials, the following test method is used on the basis of the test-method standards established by the Membrane Structures Association of Japan.
F

Snow sliding by forced vibration of a membrane surface: Accumulated snow can be slid off the roof by vibrating the membrane surface. The installation of a low-frequency vibration motor can aid in the sliding of snow off the roof surface. Although the number of actual examples is still small, some effectiveness has been observed, and there is a case in which snow load has been reduced through the use of such motors; this is a new method.

2. A N A L Y S I S
To clarify the behavior of a membrane structure against loads and external force, a finite element method is currently used.

2.1 Elastic constants of m e m b r a n e materials


Using the current analysis method, it must be assumed that membrane materials are elastic. The elastic constants of membrane materials are based on the results of tensile tests of the membrane materials. However, the visco-elastic characteristics and woven fabric properties of the membrane materials lead to difficulties in determining the elastic constants. It is generally difficult to satisfy the reciprocal theorem with membrane materials (according to this theorem, the
Fig 3 Testing method for in -plane shear stiffness

Material nonlinearity In general, material constants of membrane fabrics are determined under the assumption of linearity. However, load-extension curves for membrane fabrics obtained from tests exhibit obvious nonlinearity. The application of analyses wherein the material constant of the fabrics

265

with the above characteristic is determined from a simple linear approximation is inappropriate for cases in which a more accurate load-strain relationship is required to elucidate the actual behavior of the material or in which the material constant in the low-load region for introduction of pretension is required.

1) The shape of a membrane surface formed by the design initial tension (in cases of air-supported structures, internal pressure during normal operation) should be appropriately obtained in the analysis. 2) Based on the curved surface obtained from shapeanalysis results, the curved surface without tension should be designated as the manufactured curved surface, using appropriate cutting patterns. 3) With respect to the manufactured surface, a stressdeformation analysis should be carried out as follows: considering the self-weight of a membrane and applying initial tension in accordance with the tension-introduction method, a membrane surface determined by forced-deformation analysis at the edge boundary is used as a standard curved surface, and load and external force are applied to the standard surface. 4) Before the re-introduction of tension to the membrane surface, creep of the membrane materials should be considered and the tension of the membrane surface should be reanalyzed in accordance with the planned method of re-introduction.

Fig 4 Example of biaxial tensile test results

Therefore, more accurate analysis will be required using material constants which can be used to approximate the load-strain curve obtained from the biaxial tensile test. In this case, the material constants vary depending on stress or strain level. Wrinkling analysis In the analysis of membrane surfaces, the behavior of wrinkles of membrane surfaces must be considered. The analysis of wrinkles generated locally on a membrane surface is not so difficult; however, when wrinkles are generated over a large area, analysis becomes difficult, particularly when initial tension is eliminated from the membrane surface.

5) For the reinforcement parts using the membrane material on a membrane surface, an analysis in which the direction of warp yarns of the membrane surface and the direction of weave yarns of the membrane material used as a reinforcement is considered should be carried out. 6) With respect to load and external force applied to the membrane surface, which may induce large deformation such as reversed radius of curvature, an analysis should be performed under the assumption that the loads and external forces react on the membrane surface after deformation.

2.2 Standards for analysis of lightweight structures in J a p a n


The analysis of membrane surfaces requires special inspection because the requirements that must be satisfied are very strict. The legal standards for cable and membrane structures in terms of structural analysis are as follows. Analysis of membranes In the analysis of membrane structures, shape, deformation, stress and stress distribution under the design load should be determined to verify the strength, rigidity and serviceability of the structure. A partial or complete verification can be obtained by means of experiments. If necessary, analytical results can be verified experimentally.
Fig 5a Shape analysis

Fig 5b Zooming analysis

(a) Undeformed model

Fig 6 Time historical dynamic analysis for wind

X. ^
F^3
-|

j1
...ij/ A

1
. !

..
\

.,c\ .

\_..\'

y-

0
J;

^/l - - \\\/" 1
pj Time (sec)

(a) Undeformed friction element cable

Dnm bhv u o ec h t ya i e ai r f ah u c o

Fig 7 Nonlinear time historical dynamic analysis

7) At parts where the concentration of stress is predicted on a membrane surface, detailed analysis should be carried out. 8) Vibration of membrane surface and cable materials due to wind should be examined in accordance with actual situations. 9) In cable structures, analysis of cables under construction should be performed in accordance with a method to introduce initial tension and sequence of introduction, when necessary. 10)In sections where the lap joints of a reinforced" material and membrane material, as well as lap joints of only membrane materials, are concentrated, analysis should be performed with appropriate evaluation of the elastic stiffness of the parts. Modeling and analytical method of membranes 1) In the structural calculation of membrane surfaces and their boundary structures, an appropriate model should be used in terms of joint conditions, attachment conditions, shape, elemental division, warp-yarn direction and load conditions for structures and membrane surfaces, in accordance with actual situations. 2) When the woven fabric structure of a membrane material is assumed as a plane-elastic body, an analysis in which the material anisotropy is considered should be performed, and the relationship between tensile stiffness in the weave yarn direction and Poisson's ratio should precisely satisfy the reciprocal relationship. 3) The definition of the stress and the nonlinear analytical method used in the analysis should be appropriate to the membrane material being analyzed.

Fig 8 Concept of bendable elementand friction element *2

4) A method of following the wrinkles generated on a membrane surface and of loosening of cable materials in the analysis should be properly presented and verified. 5) In cases where cables are reinforced, the analysis should be performed in terms of a hybrid structure consisting of cable materials and membrane materials. When cables slide over a membrane surface and do not support the stress of the membrane surface, the stress should be borne only by the membrane material being analyzed. 6) When the analysis is performed using a bar model for membrane materials, the analytical results should be compared with experimental results in terms of the relationship between biaxial tensile stresses and the strains of the membrane materials, as well as in-plane shear stress, to verify their agreement.

Fbc ar i

267

3.1 Problems in the performance required of m e m b r a n e materials


At present, the following test items are required in the performance tests for membrane materials in Japan.
weft direction

Coating adhesion strength of the coating layer Weight of the coating material Lap-joint strength: when a certain angle is formed between the yarn directions of a membrane material to be joined, the lap-joint strength is expected to decrease. Peel strength of lap-joint Creep resistance characteristics: when a membrane material is repeatedly subjected to stress, the membrane material creeps, leading to loosening of the initial tension, and reintroduction of tension is required. Fatigue resistance due to repeated load: for glass fiber fabrics, single-swing tests are repeated 300,000 times. When the fatigue test containing 0 tension is repeated, deterioration of the strength becomes clear. Flex resistance: glass fiber fabrics in particular have problems associated with flex resistance; a decrease in the strength is observed at the folds of membrane materials in actual use. There are problems in terms of construction, and problems of stress decrease in membrane materials which are subjected to large deformation including wrinkling. Abrasion resistance Weather-ability: there are problems associated with the performance of weather-meters used in the analysis of membrane materials; currently, weatherability is determined based not on the exposure time, but on the amount of energy received by the membrane. Tensile strength under high-temperature (60, 150 and 260C): the tensile strength of a joint decreases under high-temperature. In particular, when a fire occurs, the joint separates. Water resistance Low-temperature resistance Chemical resistance Resistance to fire: incombustibility tests and flame resistance tests are performed to determine the range of use of materials according to their grade. When a mesh-like fabric is used to increase the translucency of the membrane material, flames may penetrate the fabric during a fire; therefore, in Japan, mesh-like materials are not legally defined as incombustible materials.

Fig 9 Bar model of fabric *3

Reliability of the solutions obtained based on results of analysis 1) In cases when the stress-deformation analysis is performed using the finite-element method, the unbalanced force at each node of elements after deformation should be specified; in addition, the total load and total reaction force should be specified. In the shape analysis, with respect to the assumed tension in the direction of weave yarns and to the inplane shear stress, unbalanced forces should be specified at each node of elements for the final shape. In addition, these unbalanced forces should be within the respective allowable ranges. 2) The program used in the stress-deformation analysis should be verified for its efficacy and accuracy in terms of load, under which the obtained shapes are theoretically possible with respect to membrane and cable materials. 3) The method of following the wrinkles generated on a membrane surface and loosening of cable materials in the analysis should be properly presented and verified. Recognition of the need to ensure the safety of membrane and cable structures in terms of the abovementioned items is gradually taking root among designers in Japan, and many high-level structural calculation reports have been submitted.

3. M E M B R A N E M A T E R I A L S A N D CABLE MATERIALS
Membrane materials and cable materials are important materials for lightweight structures. In Japan, PTFEcoated glass fiber fabrics are used for lightweight structures, as well as in other permanent buildings with membrane structures, because of their resistance to fire, durability and self-cleaning characteristics. Currently, since the performance of PVC-coated polyester fiber fabrics, including durability and the antisoiling characteristic, is being improved, these fabrics are used as roof materials when incombustibility is not required.

These tests for membrane materials for structural use are compiled in the Membrane Structures Association of Japan: Testing Method; Standard for Architectural Use. Membrane materials are tested in accordance with these test methods, and the standards for membrane material performance are defined based on the test results.

268

3.2 Allowable unit stress of m e m b r a n e a n d cable


The allowable unit stress and allowable yield strength of cable and membrane materials used as structural materials for lightweight structures are set as follows, for general use.
Allowable unit stress for short-term load Cable materials Membrane materials Fc/2.2 Ff/4.0 Allowable unit stress for long-term load Fc/3.0 Ff/8.0

The allowable unit tensile stress of membrane materials subjected to folding is listed in the table below under the general usage conditions of 3-4 times of opening and closing per day. However, when the frequency of opening and closing is high, these values should be increased appropriately in accordance with load conditions, environmental conditions, usage conditions and generation of wrinkles during folding.

Allowable unit stress for short-term load Membrane Fm/5.0

Allowable unit stress for long-term load Fm/8.0

Fc: Breaking strength of cable materials. Ff : Breaking strength of membrane materials.

Short-term load: wind load, snow load in general areas, earthquake load, temperature load. Long-term load: self-load, live load, snow load in heavy-snow areas. When cable and membrane materials are used for retractable roofs: 1) Allowable tensile strength of wire ropes The allowable tensile strength of wire ropes used in a retractable roof which is subjected to opening and closing 3-4 times per day (general-use condition) is set as listed in the table below. However, when the frequency of opening and closing of the roof is high, these values should be increased appropriately, in accordance with load conditions, environmental conditions and usage conditions.

Fm: Breaking strength Using the allowable unit stress design method, the safety factor of membrane materials is determined based on experiments regarding the following items. Scatter of material strength (number of test samples: 20 or more, the scatter of material strength among lots should be considered) Decrease in strength due to outdoor exposure Dynamic and static creep characteristics Decrease in strength at joints of membranes Others (analysis accuracy, fabrication accuracy) By multiplying the decreasing percentage of each strength, the design safety factor of the material is set to be 4.0. This value is by no means satisfactory, but it takes into account the possible deterioration in the strength of the membrane material. In view of the low cost of membrane materials, this safety factor may be the lowest society will accept. Past examples of accidents in membrane structures show that damage is caused by tears spreading in the membrane material. Tears are generated by localized incidents that are not accounted for in structural calculations. The safety of membrane structures therefore depends a great deal on the design of details.

Allowable tensile strength for short-term load

Allowable tensile strength for long-term load

Main-cable Rope for traction driving Cases of winding Cases of repeated bending Cases of hanging + winding

Fc/3.0 Fc/3.3

Fc/4.5 Fc/5.0

Fc/3.3

Fc/5.0

Fc/4.0

Fc/6.0

(Note) Fc: tensile strength. *) The main-cable ropes are used to hang the roof; they receive moving load via pulleys, etc. *) The rope for traction driving is subjected to winding or repeated bending. The loads applied to wire ropes during traction are longterm loads.

4. A P P R O A C H T O L I G H T W E I G H T CABLE/MEMBRANE STRUCTURES
There are many possibilities with respect to the shape of a lightweight structure, a hybrid structure made of cable and membrane. These structures can be used for buildings with various shapes and structural characteristics. This light-weight, large-spatial structure is currently being studied in Japan, and the possibilities of various structures are being sought systematically.

2) Allowable tensile strength of membrane materials for retractable roofs For membrane materials used at retractable parts subjected to folding where wrinkles are generated, glass fiber fabrics should not be used.

Fig 11 Approach to lightweight cable/membrane structure

Roofs for stadium for the W-Cup Soccer in Japan Lightweight structures have been adopted in stadiums for Soccer World Cup in Japan even though the adoption of lightweight structures may not be essential in their design. In the future, lightweight structures are expected to draw the attention of more designers and developers in Japan.

For the spectator stands of these soccer stadiums, steel structures are mainly adopted, and the number of hybrid structures with cable materials is currently increasing. Examples include: cantilevered girder structures, support of the entire structure by system space trusses, steel structures. Membrane materials are often used for the roof material. Two of the stadiums have retractable roofs, and one a movable floor.

5. C O N C L U S I O N
A number of accidents have occurred as construction of membrane structures has increased. In addition, many accidents have occurred in small-scale structures. Accidents caused by failure to control the inflation pressure are especially frequent in the case of air domes. These are the first structures to depend on the satisfactory control of inflation pressure for their success and to rely on control or supervision for their safety. Compared to traditional structures, membrane structures have not yet encountered a wide variety of accidents. Consequently, sufficient consideration of all possibilities is necessary in their design. Underestimates of loads and the safety factor of materials, insufficient recognition of structural characteristics, subjection of materials to unreasonable use, and inappropriate connection and attachment of membrane surfaces are still conspicuous. The careless structural use of membrane materials can lead to accidents. Membrane structures must be made as reliable as other, traditional structures if they are to be accepted by society.

6. L I S T O F W - C U P S O C C E R S T A D I U M S A N D R E C E N T W I D E SPAN L I G H T - W E I G H T S T R U C T U R E S IN J A P A N

Summary
Location : Architects : Engineers : Ohita City, Kyushu Kisho Kurokawa Architect & Associates Takenaka Corp. K. Kurokawa Architect & Associates Takenaka Corp. soccer & events Steel frame, RC concrete Frame membrane structure (Retractable) 52,234 m
2

W-Cup Soccer Ohita Stadium (Retractable)


Prioncipal use : Building structure : Roof structure : Buiding area : Roof material : Dimension : Seating capacity : Completion :

PTFE coated glass fiber fabric + Titanium plate Dia.274m, Height 57.5m 42,000 2000

271 Summary
Location : Fukuroi, Shizuoka Pref. Architects : SatoSogo & M. Saitoh Engineers : Plus One & M. Saitoh Principal use : Soccer, events Building area : 26,967m Roof area : 28,500m> Building structure : R.C. Roof structure : Steel pipes, cable and membrane structure PTFE coated glass fiber fabric Roof material : 260.8 x 215m Height 43.4m Dimension : Seating capacity : 45,700 2001 Completion :
2

Name W-Cup Soccer Shizuoka Stadium

W-Cup Soccer Saitama Stadium

Location : Architects : Engineers : Principal use : Building area : Roof area : Building structure Roof structure : Roof material : Dimension : Seating capacity : Completion :

Urawa City, Saitama Pref. Azusa Design Office Azusa Design Office ^jUk

Soccer
53,000m 24000m
2 2

^t((^^^^^

S.R.C. rahmen structure Steel pipes space truss + membrane structure PTFE coated glass fiber fabric Dia. 260m Height 58m 63,000 2001

W-Cup Soccer Kajima Soccer Stadium

Location : Architects : Engineers: Principal use : Building area : Roof area : Building structure Roof structure : Roof material : Dimension : Seating capacity : Completion :

Kajima City, Ibaragi Pref. NIKKENN SEKKEI NIKKEN SEKKEI Soccer 36,000m 25,000m
2 2

S.R.C, R.C., S. Steel frame + membrane structure PTFE coated glass fiber fabric + Polycarbonates 231 x l 9 5 m Roof Height 44.6m Mast Height 49.5m 42,000 2001

W-Cup Soccer Nagai Athletic Stadium

Location : Architects : Engineers: Principal use : Building area : Roof area : Building structure : Roof structure : Roof material : Dimension : Seating capacity : Completion :

Ohsaka City, Ohsaka Pref. SHOWA SEKKEI SHOWA SEKKEI Soccer, Athletics 26,700m 19,000m
2 2

R.C. PS concrete rahmen structure Steel pipes space truss + membrane structure PTFE coated glass fiber fabric 267 x 2 1 7 m Height 31m 50,246 1996

W-Cup Soccer Niigata Stadium

Location : Architects : Engineers: Principal use : Building area : Building structure Roof structure : Roof material : Dimension : Seating capacity : Completion :

Niigata City, Niigata Pref. NIKKEN SEKKEI NIKKEN SEKKEI Soccer, Athletics 15,262m
2

S.R.C. structure Keel truss steel structure + membrane structure PTFE coated glass fiber fabric 285 x 270m 43,000 2001

272

Name
Park Dome Kumamoto
Location :

Summary
Kumamoto City, Kumamoto Pref. Daichi Kobo & Fujita Corp. Daichi Kobo & Fujita Corp. Sports and events 26,325m 23,000m
2 2

Architects: Engineers: Principal use : Building area : Roof area: Building structure : Roof structure : Roof material : Dimension : Seating capacity : Completion :

R.C. & Steel structure Air-inflated membrane structure PTFE coated glass fiber fabric 234 x 184 m Height 48.2m Roof dia.: 107m 2,000 1997

Athletic Stadium Kumamoto

Location : Principal Use : Architects : Engineers : Building structure : Roof structure :

Kumamoto City, Kummoto Pref. Athletics, Soccer Nikken Sekkei Design Office Nikken Sekkei Design Office R . C , Precast, Steel Cable +
2

Suspended arch membrane structure 20,240m Building area : Membrane covered Area : 8500m 250 x 210 m Height: 59.7m Dimension : 33,471 Seating capacity :
2

Membrane materia! Completion :

PTFE coated glass fiber fabric 1998

Toyota Stadium (Retractable roof)

Location : Architects : Engineers: Principal use : Building area : Roof area : Dimension : Building structure : Roof structure : Retractable roof: Roof material: Seating capacity : Completion:

Toyota City, Aichi Prefecture Kisho Kurokawa Architect & Associates OveArup and Partners Japan Soccer 40,734m 28,500m Dia. 223.6 m Mast Height 92.7m Roof height 68.6m
2 2

Steel rahmen structure Suspension roof structure Air-jnflated membrane structure PVC-Polyester fabric 45,700 2001

Yamaguchi Dome

Location : Architects : Engineers: Principal use : Building area : Roof area : Building structure : Roof structure : Roof material : Dimension : Seating capacity : Completion :

Yoshiki, Yamaguchi Pref. Nihon SekkeiInc. Nihon Sekkei Inc. Sports and events 27,666m 27,666m
2 !

R.C. structure Steel pipes space truss + membrane structure PTFE coated glass fiber fabric 172 x 167.5m Height 53.4m 2,058 2001

Ohdate Dome

Location : Architects : Engineers: Principal use : Building area : Roof area : Wall structure : Design snow load : Roof structure: Roof material: Dimension : Seating capacity : Completion :

Ohdate City, Akita Pref. Toyoo Ito & Associates + Takenaka Corp. Toyoo Ito & Associates + Takenaka Corp. Baseball and Multi-purpose stadium 21,900m 21,900m R.C. structure
2 2

450 kgf/m Laminated timber + membrane structure PTFE glass fabric 175 x 160 m Height 52m

j^KQ ^i^SsBfeP'

jSSS
y ^ / \

10,000
1997

273

REFERENCES

1.

Masaoka, N., Ishii, K., Vibration Characteristics of Suspension Membrane Structures, Proc. of IASS+ASCE Symposium on Spatial, Lattice and Tension Structures, Atlanta, April, 1994.

2.

Matsumura,T.,Oda,K., and Tachibana,E., Finite Element Analysis of Cable Reinforced Membrane Structures with the Use of Bendable-Finite Element, Proc. of IASS International Symposium '97 on Shell and Spatial Structures, Singapore, Nov. 1997.

3.

Kato, S., Takeda, F , Formulation of Nonlinear-Continuum Constitutive Equations for Fabric Membrane based on Schock's Model, Proc. of IASS+ISMES+REDESCO International Symposium on Spatial Structures, Milano, June, 1995.

4.

Kato.S., Minami, H., Yoshino, T. and Namita, T., Analysis of Membrane Structures Based on Fabric Lattice Model Considering Viscous Characteristics,. Proc. of IASS International Symposium '97 on Shell and Spatial Structures, Singapore, Nov. 1997.

Shizuoka Stadium

274

SECTION VI Glass and Fabric Atria

The British Museum Great Court, and the Music Centre, Gateshead Engineering the Great Court Roof From Schlumberger to the Dynamic Earth, a Sequence of Membrane Roofs Tensioned Braced Ribs in Architectural Projects Three Widespan Enclosures in the USA A Philosophy of Widespan as applied to Smaller Enclosures Design and Construction of Two Large Span Buildings in Germany

276

THE DESIGN OF THE ROOFS OF THE BRITISH MUSEUM GREAT COURT AND THE MUSIC CENTRE AT GATESHEAD
Spencer de Grey Architect Foster and Partners, Architects and Designers

ABSTRACT
This paper provides the background to the design of two current projects by Foster and Partners. The British Museum Great Court in London and the Music Centre at Gateshead have a single major architectural feature in common - both are covered by lightweight, large-span roof structures. The complex geometries of the roofs have required cutting-edge computer models and parametric modelling software to assist in the design process. Technically, the two roof structures are amongst the most advanced of their kind. But the two projects share themes that have more farreaching cultural and philosophical ramifications than the technical virtuosity of their structures. Both create new democratic urban spaces - public spaces that serve not only as circulation areas for their respective buildings but also as internal piazzas for their respective cities at large. Each project involves 'repairing' its site - the British Museum Great Court reclaims a space that has been lost to the public for more than 150 years and the Gateshead Music Centre makes a major contribution to the cultural redevelopment of the derelict south bank of the River Tyne. The Great Court will be completed in November 2000, and the Music Centre at Gateshead, which is still in design development, will be completed in late 2002.

Figl

I will begin by outlining the historical background to each of the projects, focusing on the ways that each of the buildings deals with broader issues of urban planning. This will be followed by a description of their sites and the functional requirements that led to the development of wide-span structures. I will conclude with a brief technical description of the structures and their energy strategies.

Fig 2

277

HISTORICAL BACKGROUND TO THE PROJECTS


The Music Centre at Gateshead is a central part of the regeneration of the Tyne riverside. Other key projects include the new Baltic Centre for Contemporary Art and a new pedestrian bridge by Chris Wilkinson. The area will be further enlivened by shops, a hotel and leisure facilities. The Music Centre complex will provide accommodation for three auditoria and the Regional Music School. Each of the auditoria has been designed to provide acoustic excellence in relation to the number of seats required. As such, the exact forms of the three spaces have essentially generated themselves along functional lines. The largest hall will seat 1650 people. The second hall is intended for folk, jazz and blues concerts, including those by the resident Folkworks, and will have an informal and flexible seating arrangement with a maximum capacity of 400 seats. The third hall will be used as a rehearsal space for the Northern Sinfonia and a major performance space for the music school. The three auditoria are conceived as separate enclosures placed alongside each other on the riverbank.

It would have been possible to leave the auditoria as three discrete buildings, housing their own foyers and auxiliary spaces. However, the specific characteristics of the site and the future development of the quayside suggested an alternative solution - a large roof structure enveloping the auditoria. Firstly, the windswept nature of the site required some kind of common shelter for the three buildings. Secondly, the building complex needed to supply its own access routes and infrastructure on what was a totally derelict site. Both issues suggested the need for a concourse, in the form of a covered 'street' on the riverfront, beneath a large roof structure. Below the concourse is the music school. This concourse becomes a major public space - a shared foyer for the three auditoria, a common room for the music school, and a sheltered environment from which to enjoy the river. It symbolises the ethos of cultural fusion inherent in the establishment of the Music Centre - a complex shared by musicians and audiences of a range of different music, and a meeting point for students, professional performers and the public. This integration has been encouraged by reducing the back-of-house hospitality areas for performers, so that visiting musicians will meet with students and their audience in the concourse bars. Lastly, the roof gives visual cohesion to the project, and provides the waterfront with a landmark structure that formally echoes the great arch of the neighbouring Tyne Bridge.

Fig 5

Fig 4

The Great Court project at the British Museum is, at one level, a solution to the problems of welcoming visitors to one of the world's busiest museums and providing a clear primary circulation route from which they can visit the many galleries. But it also rescues from obscurity one of the most impressive public spaces in the capital - a courtyard the size of the football pitch at Wembley Stadium.

278

The present British Museum building was designed by Sir Robert Smirke to house the King's Library and act as a permanent home for the collections of the museum founded in 1753. Completed in 1847, Smirke's design was conceived as four wings of galleries arranged around a central quadrangle. Measuring 96 x 72m, this courtyard was to be used as a breathing space at the heart of the museum - a space to perambulate, relax, talk and think about the museum's extraordinary collections. Unfortunately this dramatic space existed for no more than five years. Almost as soon as the building was completed, it became clear that there was insufficient space for the museum's growing collections. The solution, conceived by the then Keeper of Printed Books, Antonio Panizzi, was to construct the great circular Reading Room within the central courtyard. Sydney Smirke, who had succeeded his brother as the museum's architect, began construction of the Reading Room in 1852. Completed in 1857, it is undoubtedly one of the most impressive and beautiful interiors in London.

The remaining space between the facades of the museum quadrangle and the drum of the Reading Room was gradually filled in with buildings to house the ever growing collection of books that now constitutes the British Library. These book stacks were extended between the wars, partly damaged in the Second World War and subsequently re-built. As a result, not only was the central courtyard lost to the public for over 150 years, but the museum was robbed of a primary circulation route. This problem became more acute as the museum's popularity grew. Today it has a worldwide reputation for the scope, quality and rarity of its collections and for its role as a centre of education and scholarship. Every year the museum attracts 5.4 million visitors compared to the Louvre's 5.7 million and the New York Metropolitan Museum's 5.2 million. The museum's entrance hall is a magnificent space but its plan dimensions are small and contained. It now has to accommodate seventy times more people than allowed for by the original design, so that it is constantly packed with visitors and is a frustrating and disorienting space from which to move on to the galleries. The removal of the British Library to a dedicated building at St Pancras has left the accommodation in the courtyard empty, freeing approximately 40 per cent of the museum's area. This has provided the perfect opportunity to establish the Great Court as the museum's central orientation space. The undistinguished post-war buildings that served as bookstacks have been demolished to recreate the courtyard at the heart of the Museum. The Reading Room is retained, serving as a reference library and a multi-media information centre about the museum's collections. Upon completion of the project, the Reading Room will be open to the general public for the first time in its history.

Fig 8

Fig 7

In order that the Great Court can be used by visitors all year round, it is being covered with a lightweight roof that spans the space between the facades of Sir Robert Smirke's original quadrangle and the drum of Sydney Smirke's Reading Room. The lightweight roof is designed to let in light and keep out rainwater. This creates an indoor piazza - the largest of its kind in Europe - that will be open outside normal museum hours, providing London with a dramatic space for evening events. The courtyard links the main museum

279

entrance on Great Russell Street, via the new gallery in the North Library, to the rear entrance on Montague Place, establishing a public thoroughfare directly through the centre of the museum. In this respect, the project can be seen in a much wider context. It offers the opportunity of establishing a new diagonal route - a cultural route - across London, perpendicular to the River Thames. This route starts in the north with the new British Library and the three major railway stations - Euston, St Pancras and King's Cross. It continues through Russell Square, leading to the extensive area occupied by London University. Opposite Senate House, the British Museum and Great Court with its through-route and covered public space is a focal point. The route then moves south to Covent Garden, which attracts in excess of 10 million visitors each year. The improved pedestrian walkways on Hungerford Bridge, currently under construction, link Covent Garden with the revitalised South Bank and the international terminal at Waterloo Station.

Fig 11

Fig 9

280

THE DESIGN OF THE ROOFS

The Gateshead Roof The starting point for the Gateshead roof was to design a structure that would shelter the auditoria, the concourse and the music school beneath the concourse in the most efficient manner, closely hugging the buildings, and generating a form that would unify the complex. Initially a tensile structure was considered, but was abandoned in favour of a more permanent solution. Three adjacent shell-forms were generated. Initial these were entirely free-form shapes - not governed by any geometry. However, it was clear that it would be necessary to make the roof conform to geometric rules in order to rationalise the setting out and the manufacture and construction of the building components. Parametric modelling was employed to do this.

The roof has an area of 10,200 m2, spanning a distance of 100 metres north to south and 115 metres east to west. The three shell forms are cut at the rear and cantilever at the east and west edges to provide entry canopies. As it swoops down to the riverfront a portion of the roof is glazed. At the mid-point, the roof varies in height from 22 to 37 metres. The majority of the roof is clad in 2mm rain-screen stainless steel. This sits 600mm above a waterproof membrane. The glazed area on the riverside is 1,700 m2, with a 20m2 free area of high-level opening glass. The roofing system ensures that all panels, whether solid, glazed or louvered are interchangeable. The faceted roofing panels vary in length, but have been rationalised to only twelve different widths.

GcwMion of COM Section

erntntion of Spiral AA*H

Fig 14 Enclosure Go er e mty

Fig 12

In long section, east to west, the roof is a series of arcs that meet tangentially. These arcs are rotated longitudinally to create a toroidal geometry. The parametric model allowed the architects to alter the radii of any of the arcs and immediately generate a new roof form. This meant that recalculating the information each time a change was made, which would have taken hours or days if done conventionally, could be done within seconds. In response to structural, financial and aesthetic issues, the design team generated more than 100 alternative roof designs, sometimes mocking up 4 or 5 schemes per day. Such a degree of responsiveness would have been impossible without the parametric model.

The steel structure consists of four primary arches running north to south, which are 838mm universal beams. These are supported by sixteen props, which are 457mm circular hollow sections. There are an additional four props, which are 323 circular hollow sections, for each of the two cantilevered entrance canopies at the eastern and western edges. The props are set out radially. The secondary arches run east to west and are 406mm universal beams. The tertiary members running north to south are 168mm circular hollow sections. This integrated structural system is further braced with diagonal rods of 32mm diameter. The three main sets of structural elements are fixed with bolted connections, while the diagonal bracing is pinned. The whole forms a continuous shell structure.

Fig 13

281

The Great Court Roof The key element of the design for the Great Court is the glazed roof. The underlying strategy is to produce a canopy that is delicate and unobtrusive, avoiding the need for columns within the court, which would obscure the handsome internal facades of Smirke's building. Geometrically the roof has to negotiate the space between the Reading Room and the surrounding facades and is constrained by planning requirements, which limit its height relative to existing structures. The roof had to be constructed of components that would be small enough to be lifted into position by crane, there being no other access to the construction site. This has resulted in a geometrical form, generated by a complex mathematical model, in which, despite its apparent simplicity, every single triangular glazing panel is unique.

Fig 16

The double-glazed units are assembled with an outer 'monolithic' lOmm-thick, toughened-glass panel; a 16mm air-filled cavity and an inner laminate glass, comprising two panes of clear-float glass and two clear PVB interlayers. The total thickness of the glazing unit is 38.76mm. The roof allows daylight to filter through and illuminate the court, passing into the Reading Room and, in very controlled quantities, into the surrounding galleries. In order to reduce solar heat gain the glazing units combine body-tinted glass with a white dot-matrix fritting pattern- over 7 5 % of the sun's heat is prevented from entering the court - while a high proportion of the visible spectrum is transmitted. The glazing panels are supported on a fine lattice made up of 5162 purpose-made steel box beams that intersect at 1826 structural six-way nodes, each totally unique in its x, y and z co-ordinates and rotation angles. The 80mm-wide roof members are both the primary structure and the supporting frame for the triangular glazing units. The structure consists of 10 km of steel.

Fig 15

The roof is 6100m2 and comprises 3312 triangular glass panels. Only the north-south axis represents a line of symmetry for the roof because the Reading Room is offcentre within the Great Court by 5m towards the north facade. The structure spans lengths varying between 14 and 40 metres. The varying lengths result in the mid point heights of the roof varying from 3 to 7 metres in relationship to the horizontal boundaries. The maximum distance from the floor level of the Great Court to the highest point of the roof is approximately 26m. The triangular glass panels vary in size from 800mm x 1500mm to 2200mm x 3300mm; the average area of the glass panels is approximately 1.85m2.

Fig 17

An extruded silicone gasket provides the interface between the supporting steel frame and the glass panels. This 15mm-high gasket is not only shaped to cater for the angles at which each of the panels meet - varying between nearly 0 and 30 - but also to respond to the combined system's tolerances. As the steel roof members and nodes are fabricated through computercontrolled machining, precise tolerances can be achieved in the steelwork fabrication.

282

The glazing panels are mechanically restrained by means of stainless steel bolts and cleats, fixed to the steelwork at approximately 500mm centres around the double glazing units' perimeters. The double glazing units are manufactured with stepped edges, which provide the beaming surface for the fixing cleat. At its junction with the Reading Room the roof is supported on a ring of 20 composite steel and concrete columns which align with the structural form of the original cast-iron frame of the Reading Room. These columns will be concealed by a new skin of limestone surrounding the entire drum of the Reading Room, the exterior of which was not designed to be seen from within the museum. This skin also provides space for vertical services. At its perimeter the roof is supported by Smirke's original load-bearing masonry walls. It is connected to the walls by a sliding bearing carried by a concrete ring beam surmounting the existing walls. The roof's glazing system has been designed to be walked on for cleaning and maintenance. To ensure operatives' safety 200 harness attachment points, linked by continuous cables, have been provided in strategic locations across the roof. Both glass and steel have been designed, fabricated and installed with fully tried and tested technology and rigorously tested before assembly.

At the British museum it was important to integrate modern services with minimal alteration to the building's historical structure. Having sealed the Great Court in order to keep the weather out, it is necessary to bring fresh air into the new spaces and the Reading Room at a rate of 45m3/ second. This is achieved by the construction of four new primary plant rooms in the basement of the existing buildings to the north-east, south-east, south-west and north-west of the court. These perform the initial filtering of the incoming air before it is passed to four secondary plant rooms beneath the court. Within these, full conditioning of the air takes place before it is distributed to the education centre, gallery spaces and the restored Reading Room. In the Reading Room the new systems follow, in broad principle, the original strategy of Smirke's design by using the existing 'spider' - a series of brick air ducts to carry insulated ductwork beneath the floor to supply air through the reading desks. The extract system will also use the original routes in the structure of the dome. The first level of environmental control is provided by passive, natural ventilation. Air is drawn in through highlevel openable louvres around the perimeter of the Great Court. These, combined with a direct fresh-air feed to the floor-recessed displacement louvres, produce a large stack effect and wind effect to self-ventilate any internal heat gains. The passive system can also be used to 'purge' the entire volume at night when outside air is much cooler. This prevents the 'heat soak' from which many large structures can suffer if they are not allowed to 'breathe' at night. During the winter months, the Great Court is heated by an underfloor heating system with a network of pipes in the screed. To enhance cooling in the galleries and auditoria during the summer, the same pipes are used for a chilled water system. At night, when the galleries are closed, the central chiller plant is redundant. It is therefore possible to run this plant outside of occupied periods, using off-peak electricity to feed cold water to the slab. This then pre-cools the large floor area to approximately 18C in preparation for the following day. The chilled slab encourages fresh air to remain at floor level rather than being drawn into the higher, unoccupied volume of the space. The resultant scheme allows the Great Court to be maintained between a minimum temperature of 18C in winter and a maximum temperature of 25C in the summer.

Fig 18

Heating and Ventilation Stategies With both projects we have attempted to rely as much as possible on passive systems of cooling. The aerodynamic form of the Gateshead roof assists in a system of natural ventilation. The south-west wind is drawn over the roof, creating an area of low pressure at the building's riverside facade. This encourages air to be drawn in through low-level opening vents. A natural stack effect is created and air is exhausted through highlevel opening glazed panels. This system is augmented with mechanical ventilation that supplies air and warmed air as necessary. Heating to the concourse is provided by an under-floor system using hot-water pipes. The auditoria are fully air-conditioned.

283

Engineering the British Museum Great Court Roof

Stephen Brown Parner, Buro Happold

The British Museum is one of England's most popular venues, visited by millions of tourists, students and academic researchers every year. To create more space for the Museum's continuing expansion and modernisation of its visitor facilities, it is witnessing change on a scale never before experienced on this tightly populated site in Bloomsbury.

THE DESIGN
The architectural scheme proposes spanning the Great Court, and encircling the grade one listed Reading Room, with a graceful streamlined glass roof enclosing the court below, providing a sunlit, comfortable space for visitors and museum staff. To meet the requirements of planning consent, the height of the new roof construction is restricted and the support of the outer perimeter on the quadrangle buildings does not visually intrude on, or structurally disturbing the classical Georgian facades that face into the Great Court. The roof is a fine lattice shell structure spanning in three directions from the four sides of the quadrangle on to a ring of 20 columns that will surround the Reading Room. The Reading Room is actually not located at the centre of the courtyard, but some 5m towards the North facade. These columns carry the roof load down to the foundations ensuring that no additional load is applied to the Reading Room. They will be of structural steel composite construction to achieve the required fire rating and stiffness to span from floor level to the snow gallery while remaining slender enough to be hidden behind a new stone cladding of the Reading Room. The columns designed in accordance with Eurocode 4 will be fabricated using tubular steel, an outer 457mm diameter reinforced with an inner 250mm square and filled with concrete.

Around the Reading Room, of the roof will be prevented from spreading laterally by the Snow Gallery, which acts as stiff diaphragm balancing the thrusts from opposite sides of the roof. To achieve this the existing brick arched snow gallery will be demolished and replaced with a new reinforced concrete construction which will also house the main extract fans. On the other hand, around the outer perimeter of the roof, to avoid applying any lateral load to the quadrangle buildings, the roof is supported on sliding bearings. These bearings allow the roof to spread laterally under load , normal to the relevant facade, independent of the buildings. This freedom means that for the roof to hold its form, the outer radial members near the perimeter quadrangle must work in bending and

284

The torodail framing of the roof has been generated to provide an easy transition from the circular form of the Reading Room to the quadrangle of the surrounding Museum buildings. The geometry has been defined using customised form generating computer programme resolving both the architectural and structural requirements. Forming a smooth flowing roof that adheres to the height restrictions while curving over the stone porticoes in the centre of each of the quadrangle facades. The high points in the roof are located such that the lateral forces exerted on the Snow Gallery from opposing sides of the roof are generally balanced, minimising the risk of any nett force being applied to the Reading Rooms iron frame. As a further precaution the new reinforced concrete snow gallery will be supported on sliding bearings, so that the stiff ring floats above the historic frame.

THE STRUCTURAL GRID


The roofs structural grid follows that of the glazing supporting each panel along its edges minimising the complexity of the glass fixing. Therefore, the maximum size of glass available set the final structural grid size. The grid is formed by radial elements spanning between the Reading Room and the quadrangle buildings, that are inter-connected by two opposing spirals so that the roof works as a shell. While rectangular fabricated hollow sections are the preferred structural solution for the structural elements, a alternative slightly finer option using solid sections has been prepared. For both options the elements taper to smoothly accommodate their increasing depth towards the Quadrangle buildings. This reflects the architecture maintaining the sharp flowing lines of the structural elements dividing the individual glass panels. With the roof having only one line of symmetry, there are individual 1826 structural nodes where six elements are connected. All connections must fixed to transfer the forces and bending moments between the structural elements. Design of the roof evolved using a three way lattice of steel members which add in plane stiffness, creating a very efficient form. The roof shape itself is curved to a tight radius of approximately 50m, which means it can act much in the same way as a dome, while imposing minimal loads onto the existing surrounding structures. The curvature of the roof has allowed Buro Happold to develop a light weight construction relying on arch compressions. The curvatures of a perfect torodial are usually steep so that it acts in an arching fashion, converting vertical loads into compression in radial members. In this project, the great Court roof is restricted in height and the outer perimeter is unrestrained laterally. Wind tunnel tests carried out by Bristol University provided information on the external and internal pressures which will influence internal ventilation and air movement of the great Court once it has been covered over. The results showed that wind flow separates at the outer perimeter of the museum, and does not re-attach over the new steel and glass roof in the great Court. This means

Roof Plan colours show how the stress corres[ponding element size varies.

TYPICAL SECTION NEAR READING ROOM:

TYPICAL SECTION NEAR QUADRANGLE:

Section sizes increase from 80 x 80m around the reading Room to 80 x 180mm deep at the extremeties of the Perimeter.

compression. These effects must pass through the joints in all directions. The size of the steel members therefore are smallest adjacent to the Reading Room and increase in size towards the perimeter, being largest at the corners. The forces generated by the abrupt change in direction at the corners are large and the structure is further stiffened in these areas with a tension cable across each corner.

285

is commonly selected for long span structures for many reasons, particularly because it provides high strength and stiffness at low cost. It is easily connected by bolting, or welding, and with a surface coating, has excellent weathering characteristics. By suitable selection of different components to form the whole cross section of the beam elements, the amount of fabrication can be kept to a minimum and the efficiency of the section can be maximised. The successful connection of the some 6000 individual members is critical to the integrity of the roof structure. The high stresses and slenderness of the steel elements lends itself to welded connections. To minimise the risks of weld failure, Grade D steel, more often used for marine, or petro-chemical applications rather than construction, is to be used. With such a precise project, it was felt that the impurities present in lower grade steel may allow too much margin welding error. Buro Happold sought the advice of The Welding Institute (TWI) when preparing the structural welding specifications to ensure that the welded joints will have sufficient ductility to prevent brittle failure. The specification included a stringent testing program to ensure that the quality of the steel and welding will allow the structure to behave as predicted. The architects are keen that from the ground, the double glazed roof has as light and clear an appearance as possible. This has led to the use of fabricated steel box beams, with sufficient selfweight to resist any wind induced uplift, and with enough strength to carry the roof and its cladding. The steel weight for the entire roof is approximately 420 tonnes, or 75 kg/sqm. The double glazed cladding system will add another 60 kg/sqm. This light weight form of roof minimises additional loads imposed onto the existing facades.

that the wind pressures on the roof will be small and consistently negative (uplift). On this basis, the net once in fifty year uplift force does not exceed 0.3kN/m2. This is well below the total dead weight of the roof with double glazed cladding. The roof's outer perimeter is supported at every other nodal point by a short steel column down to the new reinforced concrete parapet beam system around the top of the existing facades. The roof is laterally stabilised around the perimeter with cross bracing situated behind each of the porticoes working parallel to the relevant facade. At the centre of each side of the roof, behind the porticoes, the lateral spreading movement of the roof is one directional, normal to the line of the facade. At these locations the roof can be laterally restrained parallel to the facades sitting the stub columns on one directional sliding bearings without inducing awkward secondary effects. A wide range of materials was considered for the construction of the structural support for the roof grill before steel was selected as the most appropriate. Steel

286

the ladder sections will be craned over the museum buildings, that will remain open to the public throughout the construction process, on to a precise system of temporary props. Adjacent ladders will then be stitched together using on site welding techniques. The installation of the glazing will follow the steel erection. Only when the structural lattice is complete and vast majority of the glass has been installed will the temporary props be systematically removed. During this process the roof will be carefully monitored to ensure that it is behaving as predicted to achieve the defined final shape. Installation of the steel roof structure finished at the Museum in early 2000, with the project due for completion this Autumn.

CONSTRUCTION
It is proposed that the roof will be constructed in a series of prefabricated ladder beams erected off a crash deck that will cover the entire court. As there will be only a very limited degree of repetition in the node types, the use of steel castings to form the nodes would be uneconomic. As a result, the star shaped nodes, some 200mm deep, may be cut from single thickness of plate, the points of each star shape set at the appropriate relative angles for each and every node. Members between nodes will be made as a series of straight elements meeting at the nodes.The prefabricated ladder beams will be assembled using precise jigs in the steel fabricators workshop. Because of the cramped congested site there is no area for storage, the ladder beams will be trucked to site to meet immediate requirements. On site

Stephen Brown BE (Civil) CEng MIStructE is Group Director at Buro Happold.

CREDITS
Architect: Sir Norman Foster & Partners Structural and Building Services Engineers, Fire Engineers and PlanningSupervisors: Buro Happold Construction Managers: Mace

287

FROM SCHLUMBERGER TO THE DYNAMIC EARTH, A SEQUENCE OF MEMBRANE ROOFS


Bill Taylor - Director, Michael Hopkins & Partners John Thornton - Director, Ove Arup & Partners

Saga Group Headquarters, Folkestone

288

For fifteen years or so we have been working together to develop a membrane architecture within the context of our wider practices. The intent of this paper is to illustrate, by example, the evolution of this work through built projects and to highlight the key parameters that inform our design decisions. The genesis of this work was not primarily technical. Through our discussions with our clients we initially seized upon innate opportunities within their briefs to create unforeseen types of space, spaces which didn't need full environmental control, where activity could be carried out under cover, or where the social networking of business could be promoted and encouraged. Clear spans were necessary, whether at Schlumberger to facilitate the passage of gantry cranes or at the Inland Revenue for playing sport. We considered all the options, but what captivated us as architects was the spatial dynamism and excitement that architectural membranes can bring. Standing beneath the taut and undulating canopies high overhead, one is aware of the perfect diagram of forces - a pure expression of the inseparability of architectural form and structural function - the dream of every modern architect. For us, membranes would not follow the free form, organic shapes generated by Frei Otto and the others, beautiful and brilliant as they might be. They needed to be controlled, to have a logic and rationale that was based in and reinforced by a 'bigger' architectural picture. From this, our roofs all possess a scale and rhythm that responds to the sub riffs of the architectural and sub structural systems, which themselves grow out of the* program and the human activity. It cannot be claimed that any of our membranes are 'wide span' structures, the largest being a relatively modest 50m clear. Our rule of thumb is that a single membrane field spans no less than 9m and no more than 15m without an event such as a pick up, a lens, a ladder at these centres. For the architect this provides interest , relief and scale For the engineer it provides the opportunity to introduce additional curvature and form. We can though claim that our membranes are the overall unifying element, architecturally, structurally, environmentally and, perhaps most importantly, socially.

Schlumberger Cambridge Research Centre

Schlumberger Cambridge Research Centre View of Winter Garden

2X4

We realised immediately the benefits of combining transparency and transluscency. From an architectural perspective, the combination of diffuse and direct light is fundamental in realising internal modelling and external awareness. Established knowledge and experience enables us to use the same basic ingredient of silica in either slab or cloth form - one in compression and bending, the other in tension, to such effect. The Schlumberger Research Building at Cambridge (1985) was, at the time, probably the most significant membrane structure in the UK. Its dramatic form was very photogenic and there was clarity in the distinction between the central space, with its PTFE glass fibre roof covering the test house and restaurant, and the conventional side blocks with offices. The simplicity of this diagram combined with rigour in the detailing avoided the clumsiness often seen when membranes meet rigid structures. The Mound Stand at Lord's (1987) was another very successful project winning numerous awards. The visual bulk was reduced by layering the building, which was supported on 6 central columns. The effect of lightness was reinforced by the PVC polyester membrane canopy covering the debenture seats, bar and restaurant. The canopy, supported on a continuation of the central columns, is completely integrated into the overall design both as a form and in the detailing and the objective of creating a summer space redolent of marquees was achieved. It was the twentieth century response to the oversailing canopy of the Main Pavilion. The New Inland Revenue Centre at Nottingham (1994) consists of 6 office buildings and an amenity building. It soon became clear that the space planning requirements of the amenity building led to a similar configuration to Schlumberger, albeit 2 storeys, with a membrane roof covering a central sports hall. The translucency of the membrane was of great benefit while the lack of insulation was not a problem in these circumstances. 5i
"'

i
3 J
Mimwi "S___

_J

iffj i
' '
I h_f US fir

- I'

i
li.

ll. _.

Schlumberger Cambridge Research Centre

Schlumberger Cambridge Research Centre Detail

Mound Stand, Lords Cricket Ground

290

In the pre-concept stage we carried out a critical analysis of the earlier projects which led to three broad conclusions. The first was that the absolute distinction at Schlumberger between the parts of membrane covered central area and the side blocks, although logical, was too severe. The relationship was somewhat disconnected since the side blocks, apparently, could be placed anywhere. The second was that the external mast structure formed a cage which contained the membrane while the third was that a major factor in the elegance of the Mound Stand roof was that the membrane floated free of the structure and its edge could be seen. Seeing the edge of the fabric and both surfaces immediately reveals it is a membrane not a solid, this is further refined by the catenary cables and fittings. Without this free edge the membrane creates an interesting form but its lightness is lost.
Mound Stand, Lords Cricket Ground

With these lessons in mind the objective at the Inland Revenue was for the roof to link and enclose all parts of the building with a form which would also reveal free edges of the membrane. Early schemes, rectangular in plan with a single PTFE glass fibre membrane over the whole building, were unsatisfactory, the forms were uninteresting and the large expanse of fabric had a blandness which did not suit the context. The solution was to develop a boat shaped plan and break the roof so that it became a central principal roof with two subsidiary flanking roofs over the side blocks and canopies front and rear. This also helped us reduce the height of the building. Lens shaped windows were introduced at the boundary between the principal and flanking roofs. The structural' frames were supported by the points of the A shaped push-outs for the main roof and their bottom edges formed the upper boundary of the flanking roofs. These windows introduced light and a visual link between inside and outside. They also allowed the upper fabric to oversail, revealing the edge of the membrane and creating shadows emphasising the form. Another device to reduce height, add interest to the membrane and introduce more light was the suspension system for the main roof. The glazed ladder trusses which form the pick ups not only give a better stress regime than the rings which were first considered but also, importantly, follow and accentuate the vaulted form of the roof. The connection with the sky is important in emphasising the qualities of the fabric. Elongated versions of the ladder trusses were used where the masts pass through the roof. Not only did this avoid the clumsiness of masts passing through the fabric, it also provided a rigid support for the glazed end walls and the front and rear canopies. The inclined masts, cables, ladders and the form of the fabric all express the tension in the roof and avoid the containment of Schlumburger.
Inland Revenue Centre, Nottingham Roof Detail

Mound Stand, Lords Cricket Ground View looking above to the canopy

291

The structural principal of the roof is simple. The main roof is picked up by 5 transverse ladder trusses suspended from 4 masts. The edges of the canopy are pushed out by the points of inclined A frames which spring from the inner edge of the side blocks. The lens window frames are attached to these points as are the suspension rods for the lightweight internal roofs of the side blocks. These roofs are glazed in places to reveal the fabric and suspension system above. A tubular steel grid over the side blocks is the springing point for the booms which push out the lower edges of the flanking roofs. The reality, however, is complex. Linking the components in this way creates a sense of coherence but the interdependence of the various elements and the variety of load combinations present technical difficulties. A consequence, for example, of suspending the side block roofs from the main roof structure is that the movement criteria for these become significantly more onerous. The geometry of the principal ladder suspension inevitably leads to significant vertical movements as loads change. These movements when transmitted to the A frames led to unacceptable deflections of the side block roofs. This problem was solved by prestressing the suspension system by the addition of additional internal cables from the side block edges to the points of the ladder trusses. The geometry of these cables was such that they were very effective in absorbing changes in load on the roof with greatly reduced deflections. An incidental advantage, which was seen after completion, was that they accentuated and gave scale to the vaulted form of the internal space. Similarly, although the overall arrangement now appears natural and obvious it was only as a result of much work reconciling many requirements. This is as true for the solution of the ladders linking membrane, columns, canopies, end and flanking roofs as it is for the detail at the points of the A frames where so many diverse elements meet. The variety of components, the magnitude of some of the forces and, of course, the demands of fabric and cable geometries created a considerable challenge. One particular detail which caused difficulty was the seal where the flanking roofs oversail the side blocks. This seal has to accept the vertical movement of the canopy as the loads vary but we did not want a ridge to appear in the fabric and spoil the form. This detail was left to be resolved with the contractor and although the inflated transparent tube functions perfectly well its appearance is less satisfactory and slightly incongruous.

Inland Revenue Centre, Nottingham Model

Inland Revenue Centre, Nottingham

Inland Revenue Centre, Nottingham

292

The next project in this series is the amenity building for the Saga Group headquarters at Sandgate near Folkestone (1999). The functional requirements led to a similar form to that at Nottingham but the difference here is that the site, on the side of a valley, looks out to the English Channel. The special qualities of the location led us to develop a design which, unlike the earlier projects, was orientated in a single direction. The idea of having a design which clearly opened out to the sea was satisfied by the metaphor of the perambulator hood pulled up to give protection while giving an open view to the front. The structural concept is that of a PTFE glass fibre membrane barrel vault formed by pairs of inclined tubular arches. Between each pair of arches a third tubular arches rises to pull the membrane up out of the vault into a hood. The lower arches are braced against in plane buckling by fans of rods which radiate out from the side blocks while the hood arches are braced by radial rods which form a bicycle wheel truss. Further sets of rods which pull the hoods forward give a very clear expression of the tension in the roof. An incidental economic benefit of this roof form was that the side block could be rectangular, instead of curved as it was at Inland Revenue. In this roof the fabric edges are not free, they are connected to the tubular arches. However, because both surfaces can be seen the visual quality of the free edge is not lost, the tensile qualities strongly emphasised by the way the visitor approaches from below and enters the sheltering hood.

1 |
1

1 1 IJ
Saga Group Headquarters, Folkestone View of Pavilion from Offices

Saga Group Headquarters, Folkestone

The form of the Inland Revenue roof was very successful at uniting side blocks and central space but there was obviously a cost premium associated with taking the fabric beyond the perimeter walls, both in terms of area of membrane and the detailing, particularly the seal. We concluded that sail-like shade structures over clerestory windows would provide the necessary visual links, create the all important free membrane edges and avoid complicated sealing details. These were a continuation of the main roof but the seal against the windows was at a valley point. It was also cheaper to enhance the specification of the side block roof than to provide a separate, enclosing, membrane roof. The contracts for the earlier roofs had been let on the basis of a full tender on detailed drawings. In this case we had to get early cost security because of budget and time constraints so the roof was procured on a two stage basis whereby the contractor, Koch, was appointed on the basis of outline information which gave the basic form, member sizes and key details. Koch then worked with the design team as the final design was developed. This was very successful. A small building of a different type was the temporary demountable ticket office for Buckingham Palace. The requirement was for a building to be located in Green Park close to Buckingham Palace during August and September from 1994 to 1999 to sell tickets for the summer opening of Buckingham Palace and thereby raise funds for the restoration of the fire damage at Windsor Castle. The concept was a gently curved rectangular wooden cabin which would accommodate the ticket sellers while the queuing customers would be sheltered by a membrane structure which would cover both them and the cabin. The scale of the structure, its location and its transient nature called for a "relaxed" structure. We did not want the hard, tense, rigid appearances of the typical engineered structure; the ideal would have been cotton or canvas. The sheltered location and the summer use allowed us to reduce wind loads and be less concerned that usual about movements and durability. An early scheme with pick up rings similar to the Mound Stand was dropped because there were insufficient rings to create the right visual rhythm. Instead the design has two gently inclined masts which support a series of 6 "coathangers" which pick up the fabric from above. The coat hanger is a development of the thinking at Inland Revenue where the linear supports were seen to enhance and add crispness to the form of the roof. Booms on the mast and coat-hanger lines push out the edges. The proportion of the roof, its rise and the angle of the masts were carefully chosen to create the desired balance of lightness and tension.

The masts and booms all spring from the front edge of the cabin and are of laminated timber for continuity with the cabin and for "softness" while the fabric is acrylic, whose appearance is more like a natural fabric. The programme for design and construction was tight and to short-cut as much as possible Landrell were appointed early and the tasks shared in the most expedient way. The final project is the Dynamic Earth building Edinburgh (1999). Design on this building, which contains an exhibition explaining the history of the earth, started soon after the Mound Stand was completed. The exhibition is enclosed in a three storey reinforced concrete structure whose roof forms an open deck for cafe, bar and circulation. A dome over Part of the upper exhibition level appears in the upper deck where it establishes the presence of the exhibition. The upper deck is enclosed with a PTFE glass fibre membrane. Like the Mound Stand the lower level of the Dynamic Earth incorporates existing masonry construction. Another link with the Mound Stand was that the first scheme consisted of two lines of pick-up rings suspended from pairs of masts. After initial development the scheme was postponed for some years and early concerns that the form was rather lifeless were reinforced by the success of the Inland Revenue Roof. Consequently, the design was reappraised when lottery funds enabled the project to proceed. The layout of the building was determined by the site constraints and the exhibition design and planning permission so this could not be altered. The key changes were to change the rings to ladder trusses similar to those used at Inland Revenue, to recognise the entrance on one of the long sides byextending out a canopy which lifted up to welcome the visitor, and to lift the two ends. Another aspect of the design which was given considerable attention was the relationship between the show dome and the membrane; there had to be enough space between the two so that the continuity of the membrane could be seen but not so much that the show dome would be insignificant. The Dynamic Earth has been interesting precisely because it was developed in two distinct phases which has enabled us to revisit some of the design in the light of later experience. It was particularly interesting to note the transformation in appearance created by relatively small changes, which would not have been so evident if the design had developed in the normal way. This reinforces the authors' belief that the design of a successful membrane structure requires careful attention at every level, even the most apparently insignificant details can have an enormous effect. The principal designers must have involvement and capability in all aspects from concept, through analysis to detailed design.

Buckingham Palace Ticket Office

Buckingham Palace Ticket Office Detail of column head, mast junction

295

Perhaps the most attention was given to the edge condition. The edge is pulled out by a series of raking struts and tie-downs and over-sails the perimeter glass wall sufficiently to create a sheltered walkway. The further the glass is from the fabric edge the taller, and more expensive, it is and the greater the membrane movements to be accommodated. Unlike Inland Revenue, where the closure detail is tucked away above the side block roofs, this closure detail is highly visible throughout its entire length and was given careful attention from the very beginning of the second design period. The tensioned folded flap closure developed by Birdair has proved both functional and elegant. There is a thread of development which connects these projects at both conceptual and detailed levels and shows the importance, in technically complex work, of continuity of working relationships and a shared base of experience.

William Younger Centre, Edinburgh Detail of perimeter condition

F7

William Younger Centre, Edinburgh View by night

CREDITS Schlumberger Cambridge Research Mound Stand, Lord's Cricket Ground Inland Revenue, Nottingham Saga Headquarters, Folkestone Buckingham Palace Ticket Office Dynamic Earth, Edinburgh Stromeyer Koch Koch Koch Landrell Birdair

William Younger Centre, Edinburgh View by day

296

Inland Revenue Centre, Nottingham Amenity Building at Night

297

TENSIONED BRACED RIBS IN ARCHITECTURAL PROJECTS

Brian Forster, Ove Arup & Partners, London

During the last 12 years several of us in Ove Arup & Partners, stimulated by the late Peter Rice, have pursued through various projects an interest in tensile systems which provide restraint to slender arches. The most recent of these is the Saga project which John Thornton has presented in a separate paper - "From Schlumberger to the Dynamic Earth - A Sequence of Membrane Roofs" [1]. This paper therefore looks at some of Saga's antecedents. Some of these have cable nets or textile membranes prestressed onto them, others are deadweight glass roofs. In each case it is the geometric stiffness of the tie restraint system which is significant rather than the levels of prestress applied. Each project has relied on the use of non-linear analysis software to develop and justify structural behaviour. The software is based upon the Dynamic Relaxation technique originated by Alistair Day [2] and it explicitly models the following effects: Large displacements Change of stiffness of beam elements due to the axial forces and moments developing within them Tie and membrane element forces go to zero if they attract zero strain.

THOMSON LGT, CONFLANS STEH O N O R I N E , F R A N C E , 1985


Here it was required to cover a 100m long x 18m wide internal "street" with a PTFE/glass membrane roof. The buildings on either side of the street have quite different stiffnesses. One is a 2 storey insitu concrete frame, the other a tall single storey steel shed. Directly coupling them together with a purely tensile membrane was simply not sensible because of the scale of the displacement that would be imposed upon the membrane, by the relative movement of the buildings. The preferred architectural solution was to use a closed framework across which spanned tensioned braced arch ribs support panels of membrane (fig. 1). The whole framework was fixed to one building and released from the other. However for reasons of cost the truss was realised as fig. 2 with slender struts forming the bottom member of a 3 dimensional truss with tensile shear bracing.

'

50.8kN

'

Day [3] has described the software and its use in the design development of the first of the following projects:
N o n - L i n e a r Analysis

Futt

n d

Ties in compression

Linear Analysis S t e p 1 Fig 1

Linear Analysis S t e p Fig 2 Thomson, France Fig 3 Thomson, France

Fig 4 Bukit Jalil Station, KL

Day's paper [3] compares the results of using both N-L and linear analyses. The latter involved multiple steps and was awkward and slower to carry out. This was because under wind load loading some of the diagonal ties go slack with consequent changes in the geometric stiffness of the structure. This means when using a linear analysis method the structure has to be re-analysed with those members removed as fig. 3. A larger version of the Thomson roof covers the STAR LRT station at Bukit Jalil in Kuala Lumpur. The station serves the National Stadium and was built for the 1998 Commonwealth Games (fig. 4).

299

SAN NICOLA STADIUM ROOF, BARI, ITALY, 1989


This project was built for the 1990 Football World Cup held in Italy. The architect, Renzo Piano, conceived the stadium superstructure tier as a flying saucer hovering above the arena (fig. 8). So the superstructure and the roof have a simple rounded shape and it was important that the roof be composed as a series of simple calm sunshades. There are 26 of these cantilevering up to 27m from the back of the concrete upper tier. The infilling structure within each canopy is minimal through the use of slender tubular ribs curved to follow the profile of the roof (fig. 9). Each is braced with a system of "chord-ties" - a set of 3 tie rods springing from each end of the rib and joining to the 1/5th points of each arc (fig. 10). Under the downward load the chord-ties act to reduce the buckling length of the rib by resisting its in-plane deformation. Straight lateral tubes, used to stabilise the ribs out-ofplane, complete the framework producing an architectural effect similar to that of a Japanese screen.

Fig 6 Aviary, HK

"itJIf.

"^C -*^jjafiuB 9' '_h


1 i

Fig 5 Aviary, HK

Fig 7 Aviary, HK

300

Y O U D E AVIARY, H O N G K O N G , 1990
This project was named after the eminent botanist and is a sub-tropical aviary situated in an urban park in Central district on Hong Kong island (fig. 5). The aviary straddles a steep sided twisting valley on the lower slopes of the Peak. It was made tall enough to accommodate existing mature trees and consequently has a maximum clear internal height of 30m (fig. 6). Woven stainless steel mesh forms the enclosing skin. This is suspended from a cable net system prestressed against 3 tubular arches. The largest of these is 560mm diameter and spans 62 metres. This equates to a span/depth ratio of 110 indicating the stiffening effect that the cable net provides to the arch (fig. 7). It was found that the bending stiffness of the arch elements in relation to the axial stiffness of the cables influenced the size of bending moments developing in each arch. Simply increasing the arch size attracted more moment. The trick was to use thick wall tubes giving the highest bending capacity (section modulus) but with the lowest bending stiffness (moment of inertia) for cable net stiffness (area). Stability analyses performed on a full model showed that the nets sufficiently constrained the 60m arch to buckle in its second mode.
Fig 10

Fig 9 Bari Stadium. Italy

25m

1 3 9 m m dia. CHS rib

Fig 8

Bari Stadium. Italy

L O U V R E , PARIS
This project, completed in 1994, involved covering the three open courtyards of the Richelieu wing of the museum, to give enlarged gallery space for sculpture (fig. 11). The dimensions of the structural elements were an important consideration in the roof design, not only to reduce the visual 'weight' but also to avoid casting deep shadows at floor level within the new sculpture courts. The courtyards taper in plan and spans vary from 2841m. The principal means of support are tied arches with radial ties providing restraint to in-plane buckling of each arch member (typically a 139 dia. CHS). A spine truss mediates between the stiffness of the hipped ends and the interior arches.

Fig 11 Louvre, France

302

C H U R STATION R O O F , S W I T Z E R L A N D
This project was the outcome of a public competition won by architects Richard Brosi and Robert Obrist. The roof is a fully glazed vault covering both railway and bus stations in a single span of 52m (fig. 12). The primary structure is a 10m deep tied arch with intermediate radial ties providing restraint to buckling of the principal compression members. Each "arch" is a pair of 460mm dia. CHS slightly inclined to one another as they pass over the vault but converging elegantly onto a common springing point (fig. 13). The inclination of the ribs when combined with the longitudinal purlins obviated the need for any other bracing. The main support columns occur at 15m spacing and are composed of a pair of 406 dia. CHS and cantilever from below (fig. 14). They were chosen so as to be strong and stiff enough to resist lateral wind but, importantly, not so stiff as to generate untoward resistance to arch spread under snow load or thermal expansion. For a tie braced arch of this type it is not possible to verify stability using a conventional "Code" approach. As with the previous structures the Arup programme FABLON was used to simulate elastic buckling of the framework and capacity checks were performed using a Merchant-Rankine approach in a manner consistent with Code requirements. This project received the ECCS Steel Award of 1993.

Fig 13 Chur Station, Switzerland

Fig 12 Chur Station, Switzerland Fig 14 Chur Station, Switzerland

ACKNOWLEDGEMENTS:
The work of the above projects was carried out by a number of people. Those working on the particular topics discussed were: Thomson John Hewitt, Brian Forster, Alistair Day Bukit Jalil Stadium : Tristan Simmonds, Andrew Trotman, Brian Forster Youde Aviary : John White, Amanda Gibney, Brian Forster San Nicola Stadium .Tristram Carfrae, Brian Forster, Peter Rice Richelieu Wing : Alexandre Cot, Alistair Lenczner, Peter Rice Chur Station : Alistair Hughes, Matthew Lovell, Peter Rice

References:
1. Taylor W., Thornton J: From Schlumberger to the Dynamic Earth, a Sequence of Membrane Roofs , International Symposium on Widespan Enclosures, University of Bath, April 2000. Day A.S: An Introduction to Dynamic Relaxation, 219, 1965. The Engineer

2.

3.

Day A.S., Haslett T., Carfrae T., Rice P: Buckling and Non-Linear Behaviour of Space Frames, First International Conference on Lightweight Structures in Architecture, Sydney, Australia 1986.

304

THREE WIDESPAN SPACE ENCLOSURES


Tim Macfarlane and Damian Murphy, Dewhurst Macfarlane and Partners

INTRODUCTION
We are currently constructing three projects in the USA with roofs spanning between 180' and 420'. The design approach to each roof is fundamentally different and this paper will discuss in broad terms the different approaches.

The brief called for a completely glazed roof surface with transparent end walls to the barrel to achieve maximum transparency. The major axis of the vault is 350 ft long and the diameter of the vault is 174 ft. From the outset studies were carried out to find a structural system that would align with the glazing bars to ensure that only the primary system would be visible, to achieve the greatest transparency. A folded plate barrel vault constructed of vierendeel frames was adopted, which allowed for simple flat glass panels measuring 3'2" x 7 ' 1 " to be framed onto the vierendeel members (Fig 2). The vierendeel frames are fabricated from 5"x4" and 5"x5" tubes with wall thickness altered to reflect the changes in force between the crown and the springing

PHILADELPHIA REGIONAL PERFORMING ARTS CENTRE


The roof geometry is a classical barrel vault within which two major performance spaces sit as self-contained structures (fig 1).

305

End Wall Design The height of the end walls (Fig 3) varies from 84 ft at the crown to 0 at the springing point. Laminated glass panels _" thick measuring 5' 9"x 4 ' 2 " are attached to 7/8" diameter steel cables which are suspended from a steel arch and tensioned by attaching cast iron weights of up to 12 tons to each cable. The cables deflect horizontally under wind load up to 2 ' 8 " in the centre and the weights which are attached to the adjacent roof structure via a linked arm move up to 3 " vertically to accommodate the deflection of the wall. This arrangement ensures that the load on the steel arch remains constant and that vertical deflection of the arch structure remains unchanged during variable wind pressure. The weights are linked together to ensure that should a cable fail, the weight will be retained in place by the adjacent cables. Wind tunnel studies were carried out to ensure that the load variation in pressure which could occur would not cause unpredictable relative movement between adjacent cables. The project will be completed in December 2001.
Fig 2

point. The fabrication process involves constructing 4 segments 1_ folds wide, welded and jigged in the workshop and bolted on site at the member mid points. The sections will all be painted with the finish coat applied prior to erection. Fire protection which was required for the lower 20 ft. of the trusses was achieved using a combination of sprinklers and intumescent paint. The glazed panels will be installed in bands of 3 units on a minimal aluminium frame and the joints between panels will be sealed with site-applied silicon. The total weight of steel is 825 tons which is equivalent to 27.0 lbs/ft2 of plan area. The vierendeel frames provide stiffness in the longitudinal direction as well as the span direction, therefore no additional wind bracing elements are required, resulting in a clean uniform structure.

BOSTON CONVENTION CENTRE


The Boston Convention Centre (Figures 4 and 5) will occupy a site some 1800 feet by 500 feet wide with its major entrance addressing the revitalized old dock area and its rear end abutting the finer scale of residential South Boston. Architecturally the 3 storey service buildings and hotels were designed as access strips running down each side of the 5 single storey visitor halls. Each hall measured 300 feet square on plan and with the front end of the building the final 300 x 300 space was a three storey volume with reception and public function suites. The roof to the convention halls and reception area was conceived as a continuous surface 300 feet wide x 1800 feet long, changing in profile from an arched section at the entrance end with its crown 120 feet above ground, to a flat section at the residential end with an elevation of 40 feet. At first sight the roof surface looks like a section of a regular solid. The geometry however was more complex than this and each section through the roof had a different radius, crown and springing dimension. Initially a strategy was developed which separated the roof from the adjacent service buildings and from the bridge elements which connected these structures. This resulted in a number of solutions which involved supporting each individual hall roof with 300 x 300 foot spans from up to eight columns set 60 foot in from the

306

Fig 6

free edges. The columns and steel roof trusses for this structure were inevitably large (each one unique), resulting in average weights of 25 lbs/sq. ft. for the roof steelwork. Value engineering came close to reducing the architectural gesture of a continuous roof plane to a series of flat stepped surfaces. At the eleventh hour a solution was found which simplified the erection and fabrication process and reduced the overall weight to less than 20 lbs/sq. ft.. The isolation of the roof from the adjacent service buildings and bridge links was reconsidered. It was established that the roof could be supported from and braced by these elements, by introducing a sliding joint that could accommodate thermal movement but was capable of supporting vertical and horizontal wind and earthquake forces. Working with a nominal 60' long W14 section, a surface was created by building out from the corners of the 300 sq. ft plan to each hall. The 60' length was established as the ideal length, both for normal transportation and because at the steepest curved section of the roof, a chord of 60' would result in a maximum difference of only 2" between the actual and ideal position of the beam, which could be easily accommodated in the purlin connections.

The construction involved lifting each steel beam into place, as a planar single 60' long member or as a 120' long trussed element fabricated on the ground from 60' long straight sections. The only exception to this rule was the far corner pieces. These were lifted as welded cruciform assemblies supported at three nodes by the boundary walls. They offered a positional location at the fourth node for the 4 internal support columns which were then swung into place. The remaining members were then lifted into place, building out from the corners using simply supported connections toward the centre defining the roof surface at the connection points, as construction progressed. This configuration resulted in a 2-way spanning diagonal co-planar configuration, offering both vertical and horizontal stiffness due to the natural triangulation inherent in its constructional logic. Metal deck on purlins or pre-fabricated triangular panels were then erected to complete the surface. The 120' long fabricated king post trusses provided an excellent opportunity to incorporate the service walkways designed to run at 60' centres above the exhibition hall.

308

Fig 7

DAVID L L A W R E N C E C O N V E N T I O N CENTRE, PITTSBURGH


The initial architecture proposal to create a curved roof form to cover the new 300,000 ft2 convention hall in Pittsburgh took its formal inspiration from the three suspension bridges which cross the Allegheny River just downtown from the building's riverside site (Fig 6). To accommodate site geometry, a trapezoidal plan shape generated three halls 300' long by widths varying from 430' to 320'. The cross section of the building resembled a suspension cable bridge cut at mid-span (Fig 7). The principal carrying cable started at a height of 74' above ground level and climbed to an elevation of 180'. It was supported by a steel mast and by its continuation from the mast to a connection 54' above ground back down to the building frame on its south elevation. To coordinate with the ideal grid of 60' at the lower conference hall levels, cables and their supporting frames on the north and south, or Bow and Stern frames as they become known, were set with this spacing. The triangulated bow and stern frames transmitted the tension forces from the main cables to the foundation and to the foundation level/conference hall level couple, which countered the large horizontal forces generated by the suspension cable.

Tie down cables were introduced to control wind uplift and overall deflection. At the principal bridge connection and end wall, the tie down cables were connected directly to the building structure to limit the cable live load deflection to 8 inches. This was necessary as free live load deflection of up to 3 feet would otherwise have been encountered, which would have made the attachment of the vertical walls exceptionally difficult and awkward looking. The curved metal roof deck was supported by fabricated metal steel trusses spanning 60 feet onto the main cables. The cables comprising 7 no. 3 " strands were designed to carry a maximum tension load of 3,600 kips. The mast section was 40" diameter steel and the maximum steel truss sections were fabricated from steel plate weighing 1000 lb/ft.

309

A PHILOSOPHY OF "WIDESPAN" AS APPLIED TO SMALLER ENCLOSURES


Bernard Vaudeville & Mitsu Edwards RFR Engineers, Paris

If "wide" is defined in absolute terms, the projects presented here - which do not exceed 50m span - cannot strictly be classified as "widespans". However, we hope to justify this diversion by means of the following arguments, which attempt to address the philosophy hidden behind the phrase. Structures, in the 18 Century understanding of the term, are by nature hollow, therefore constituting an enclosure. The steel constructions of the 19 century revealed this fact with brilliance. Siegfried Giedion, the famous theorist of Modernity, evokes this new experience. He draws attention to the way in which our world view is transformed by this possibility to penetrate and move through the structures: In the air-flooded stairs or the Eiffel Tower, better yet, in the steel limbs of a transporter bridge [at Marseille], we confront the basic aesthetic experience of today's building : through the delicate iron net suspended in midair stream things, ships, sea, houses, masts, landscape, and harbor. They lose their delimited form : as one descends, they circle into each other and intermingle simultaneously (S Giedion-Building in France, Building in Iron, Building in Ferro-concrete Santa Monica : Getty Center 1995, p91)
th th

presented during this conference. However, when the structural spans are small enough to not interfere with the functionality of the place, the optimisation or structural efficiency become secondary. In general, the small- or medium-sized constructions do not invite a discussion of their structure, this being conceived in an automatic fashion, just like an assemblage of standard or repetitive elements. RFR's approach clearly distinguishes itself from this negation of structural design, in particular with respect to glass roofs or facades where the structure is on show. RFR is an agency which was founded in the 1980s by Peter Rice together with Martin Francis and Ian Ritchie. From its inception, RFR has been a point of convergence between a 'British' approach, where the structure valorises and often determines architecture (i.e. the hitech) and a French tradition where the architect and engineer each lay claim to their freedom and independence, often leading to mutual incomprehension. To caricaturise the situation in France, the engineers (that is to say, the large state bodies and construction companies) take responsibility for public works (bridges, nuclear stations, the underground), leaving no place for the architect except a role as embellisher. In return the French architects jealously guard their prerogatives as regards small and medium-sized projects. The latter are above all constructed according to aesthetic, political or urban principles, or in relation to the usage and functionality of spaces. Engineering, especially as applied to structure, is, to a far lesser degree than in Britain, the motor force behind the architectural project. RFR's particular contribution, permitted by its cultural blend, consists in reintroducing the structural question into architectural form, even when it is small. To be sure, in this case the structure or the technology used does not determine the shape. But the philosophy and the design input which Peter Rice developed at RFR goes far beyond simply making the form stand up. Its dialectical relationship with French architects has consisted in searching for a certain transcendence of architectural form through structure and engineering. Since the death of Peter Rice and the departure of Ian Richie a few years previously RFR continues to work in this vein whilst maintaining its tradition of between-two-cultures.

Structures are also always widespan inasmuch as they distinguish themselves from inert, superfluous mass. They, to varying degrees, are characterised by their slenderness, and for a slender structure, a span is always wide. This is corroborated by the scalelessness of the abstract tools used for structural analysis. This is the case with graphs, algebraic calculation, technical drawings, computer models, where only the unit (m, KN, Pa,...) indicates the relation between the object or the physical phenomenon and their abstract representation.

One could suppose that the theme of widespan enclosure is a way of obliging engineers and architects to address the specifically structural stakes of construction. When the span is large in comparison with the mechanical capacities of a material, with the mode of construction, the conception of the structure becomes a crucial element of the project's design. Such is the case of the Millennium Dome and of most of the projects

310

What are the principles of this 'adaptation of structure to a form' that is defined in advance and for which there is no structural imperative? Firstly, there is the somewhat perverse desire to reintroduce the wide span into a small building. It is the means deliberately to dramatise an element of the project by playing with the effects of scale. This is what we did for the Fingal facade, in Ireland, where horizontal wind trusses span 30m.. This big span allows us to give to the large glass screen of the atrium a exceptional presence and an autonomy in relation to the rest of the building. The economic impact of this choice of aesthetic order is nevertheless minimal as the structure of this facade only represents an insignificant part of the overall budget of the building. A given architectural shape always harbours structural potentialities which can, if put to good use, contribute to giving meaning and substance to it. The projects of the Peninsula and the Lentille provide good illustrations of how, as engineers, we have tried to use the structure to accompany the architectural form. In both cases, the shapes were predetermined in the main in accordance with urban and functional constraints, or simply for their beauty. We then tried to exploit their intrinsic geometric strength. By making use of their curves and folds, we designed structures which are primarily axially loaded. As a consequence they are finer and tauter than structures designed for bending strength. When the form precedes the structure in this way, it is generally not possible to have recourse to a homogenous structural * scheme. Instead, hybrid structures are generated which superpose and bring together diverse structural modes of functioning (e.g. the Peninsula is the superposition of bicycle-wheel-like flanks and a three-pinned arch). In order that this combination of structural schemas results nevertheless in one unique object, it is necessary to blur the articulations between these different schemes. This translates as a greater complexity of the structural behaviour, less predictable, more indeterminate, sometimes non reversible. This structural hybridity also necessitates the creation of continuities in the geometry and the design of the details. This implies a less mechanical, more organic conception of structures. Often it is the case that the geometry of a structural element is not fixed in advance, but that the physical context in which it will find itself is most specific and limiting. This was true of our glass roof at Diisseldorf whose perimeter is determined by the U-shape configuration of the building and by the urban regulations. Our approach was to look for the geometry that offered the best compromise between these constraints and the structural functioning. Here again the structure is not necessarily optimal as such, it is instead characterised by its adaptation to its context.

It is important to note that all our projects search for maximum slenderness, which naturally permits a refinement of the structure. However, this is not our sole aim. Beyond the lightness, there is an aesthetic pertaining to the structure, which comes partly from the thrill it inspires (the enthusiastic rush of emotions provoked by observing a grid shell traversing a void), partly from the challenge that it represents to the irrepressible tendency of the matter to return to a heap-like state. This aesthetic is not reserved to the widest spans, it can equally be aroused by projects with more conventional dimensions like these presented here below.

Thus, our work on structures can appear far removed from the criteria of efficiency, rationality and economy usually associated with engineers. It is indeed a question of work of a technical order, with all the rigor that implies, but which also necessitates a recognition of the fact that the motivation is for a large part aesthetic, philosophical, architectural. The structures possess a symbolic value with which we play and which, all too often, finds itself suffocated by a narrow vision of economy.

G L A S S P E N I N S U L A E , T E R M I N A L 2F, ROISSY CHARLES-DE-GAULLE A I R P O R T , PARIS


The two glass Peninsulae which serve as gateways to France via Roissy airport's new terminal 2F (fig.l) continue a trend of innovation and theatricality which was started there by architect Paul Andreu in the 1970's. The project, begun in 1993, has been a long time in gestation and is threaded through with the influence of Peter Rice. After defining the overall form of a typical peninsula, Andreu and Aeroports de Paris handed over to RFR the responsability for technical design of structure

Fig 1 Aerial view of Peninsula, Terminal 2F Fig 2 Interior view from Departure hall

and envelope; our aims were to break down structural hierarchy as far as possible and to promote sleekness and visual simplicity by putting the glass "skin" in direct contact with the steel "bones", creating a breathing composite assembly (fig.2).

311

Fig 3 Plan and elevation of the Peninsula

The 200m-long hall, whose tapered shape resembles that of an upturned boat hull, increases in span from 13 to 50m and reaches a maximum height of 22m (fig.3). In the manner of a 19 century glasshouse, it was conceived as a huge, transparent entity which, despite its dimensions, has been constructed as a single piece with neither joint nor cut. The formal unity of the design would have been greatly compromised by the presence of intermediate expansion joints, and so we sought an alternative solution to the thermal stress problem. The framework is therefore fixed to the Terminal's main concourse building by a single point at its ridge (fig.4), and the sides of the peninsula are supported by a hundred articulating arms which can each move six centimetres to the front or back, allowing the structure to expand and contract freely along its full length.
,h

The structure is built up from fifty one-way spanning steel trusses - the hull's ribs - linked together by longitudinal stability bracing. The ribs are hybrid structures whose slenderness is achieved by a fine-tuned assembly of different sub-systems, designed to minimise bending stresses in the individual elements (fig.5). The truss profile was chosen to reduce deflections by allowing the central portion to act as a three-pinned arch under downloads, but to respond to wind uplift as a catenary. This non-reversible structure increased the complexity of the calculation process, by requiring non linear analysis, but enabled us to design the lower chords as fine tension-only rods. Curved flanks transfer roof loads to the lower floor slabs; these are stiffened and stabilised by the addition of fan-shaped arrays of tension rods, bicycle-wheel fashion (fig.6).

Fig 5 Typical section, Peninsula

Fig 4

Connection of "blade" to Main Concourse"blade" to Main Concourse

312

We developed our initial ideas by physical modelling of the diverse structural elements on a week-by-week basis (fig.7). The tactile nature of this process helped to provoke a rich dialogue within the design team, enabled us to get a "feel" for efficiency and allowed us to revise problem areas quickly as they arose. One interesting debate during this phase concerned the presence of the longitudinal ridge beam, nicknamed the "blade". Our initial structural concept, a pure one-way spanning system, relied on cross-bracing for longitudinal stability and we had no need for such a big element in this location (fig. 8); Andreu, however, considered it an essential part of the overall architectural composition. Rather than accept it as a purely formal element, we decided to eliminate the longitudinal cross-bracing and redesign the blade as a 200-m long strut which could transmit all longitudinal forces - the loads from the wind and the thrusts arising from the chevron-like setting-out of the final transverse arches - back to its point of anchorage (the concrete roof of the Terminal's main concourse). This lightened the structure as a whole by retaining the thrust at ridge level, rather than spreading it through the rest of the framework. The metal caisson therefore functions the way it reads, as a type of spinal column to ensure global structural cohesion. These two design principles - a hybrid, non-reversible structure for the transverse trusses and a longitudinal spine beam without expansion joints - led us to carry out the final calculations using a global ten-thousand element computer model, and this became the keystone of all

Fig 6

Detail, curved flank

p
Fig 7 Detail, curved flank Fig 8 Study model, without blade

Fig 9 Model analysis, deflections under critical loads

Fig 10 Model analysis, internal stress distributions

Fig 11

Roof detail, Peninsula

313

detailed design development. The behaviour of a building such as this remains nevertheless difficult to grasp through calculation because of its high degree of structural indeterminacy. We were greatly helped in this intuitive task by the use of colour-coded stress visualisation which allowed us to explore the global and local effects of all sorts of different loadcases, and also proved useful in explaining our work to other members of the design team, or negotiating structural improvements to the overall form (figs.9 & 10). The global model enabled us to refine element sizing to a very high degree in order to achieve the filigree appearance of the final structure (fig. 11).

The architects, Arte Charpentier, came to us with a fullydeveloped transparent form whose soft curves resemble those of a bubble surfacing or a lens for viewing into or out of the underground world (figs. 12 & 13). Its seemingly informal surface is in fact strictly controlled, described by the juxtaposition of sphere and toroid. The illusion of free form is achieved by the siting of the ground plane along a bias slice through the torus and also by the superposition of a bisymmetric cartesian grid on the skin surface, instead of the more usual axisymmetric radial grid. A second diagonal slice through the torus creates a fault line to mark the boundary between the protective enclosure and the sliding doors. Although its development was geometrically pure, the shape and gridding of the lens (or "Lentille") was far from efficient in structural terms, and the low level of symmetry posed potential difficulties with the fabrication process. Faced with a similar problem and a free rein, the engineer in search of efficiency would typically prefer an arch or shell form which allows the structural members to transfer load in compression rather than in bending, and would rationalise the structural grid into a repeatable module, either by segmentation or tessellation. Here, the geometric manipulation had flattened the shell's arches, given them bow legs and crippled the shell on one side! The span of the Lentille, at only 15m, would have allowed us to get away with a straightforward "beam-inbending" design, but this would have led to a heavy grid, lacking the "elan" of the initial architectural concept. We decided instead to work towards visual lightness by recapturing the potential structural efficiency of the central cupola and by concentrating bending stiffness where it was most needed, in the bow legs and the arched door frame. In this sense, we treated the structure as a "small wide-span" rather then a "classic medium-span". We needed to minimise spreading of the cupola zone under load in order to achieve the compression-rich force transfer which would keep the size of the grid members as small as possible. We looked initially at achieving this by means of a circumferential tension ring assisted by grid joints which were stiff in-plane; this approach was rejected by the architects, however, who felt that the source geometry was re-emerging too strongly. The alternative was to limit joint rotation and mobilise shelltype interaction by introducing a lightweight mesh of cables into the grid plane (fig. 14). The cables were judged to be sufficiently unobtrusive and were accepted; they also played a major role in reducing the risk of global buckling, as the two-way shell action eliminated the first,"arch" buckling mode, thus reducing the out-ofplane effective lengths of the grid members and allowing us to shrink their size further (fig. 15). The intrinsically flexible "bow legs", which could not have easily developed the perimeter thrust necessary to contain an unstiffened cupola, could now be designed for lower rigidity, as bent beams under essentially vertical

METRO ENTRANCE, SAINT-LAZARE STATION, PARIS


Recent years have seen significant construction or extension of underground urban transit systems in Europe and further afield. The street-level entrances to the invisible subsurface network take on great symbolic importance and present clear opportunities to experiment with architectural, geometric and structural ideas in a relatively pure fashion. This project is a case in point.

Fig 12

Exterior view, Metro entrance

Fig 13

Interior view, Metro entrance

314

INTERNAL ATRIUM FOR KBB OFFICES, DUSSELDORF


This project for a large internal atrium, 40m square, displays an unusually-shaped surface skin despite its simple footprint. The transparent envelope is nevertheless a rational expression of the roof support conditions, which allowed us to bear on three sides only. Our challenge was to combine this specific structural constraint with the architectural requirement to produce a structure that was as lightweight as possible and the economic need to find a flat-plane geometry for the glass panels. The architects, Kohn Pederson Fox's London agency, had designed a prestigious building for their client and wanted the atrium roof to be a sleek, modern symbol of the German bank's identity (fig. 17).

Fig 14

Magnified deflections, vertical loads : tension ring c.f. in-plane bracing

% #
Fig 15 First buckling mode

loads. The door arch was also spared the difficult task of resisting a significant out-of-plane thrust by this bracing of the cupola. The form of the structural members, an extruded "delta" section, was also designed to promote visual lightness. The section minimises its presence when viewed from below by reflecting as much light as possible, and provides a recessed upper channel in which to seat the glass fittings (fig. 16). The section is sized for the cupola zone, but adapted for use throughout the framework by extending the side flanks of the bow legs with extra plate where additional out-of-plane bending stiffness is required.

Fig 17

Interior view, atrium

The upper levels of the office block took the form of an extruded U-shape, to be constructed as a monolithic concrete framework without internal movement joints. This high frame stiffness gave us ideal support conditions for a gently curving "soft" gridshell which would have a recognisable form and at the same time keep within the strict height limit of Dusseldorf's zoning regulations (figs. 18 & 19).

j
Fig 16 Delta profile, detail

We developed a curved shell by taking a surface of revolution and slicing it into successive conical sections which could be subdivided into planar facets. This virtual object was then rotated and bias-cut to fit the available roof opening, undergoing successive optimisations to squeeze as much curvature as possible out of a tight space (fig.20). The circular arc segments created by the slicing process would become not only the ribs of the grid but also the top boom of a bowstring truss used to

315

project, but here the "free" volume underneath the atrium roof allowed us to adopt a more efficient method whereby the roof was stabilised, parachute-like, by sets of cables gathered together and anchored against the building wall at lower level. Under unsymmetrical loading conditions, the tension resistance of the cables in the unloaded zone stops the development of the first mode of buckling failure, which is typified by a "push in/pop out" waveform, and the equivalent stiffness of the shell is increased. The model was checked by non-linear analysis, and was further developed to allow for initial imperfections of the global geometry which could arise due to construction tolerances. This allowed direct buckling analysis of each individual member, thus providing a more direct validation process for each critical load combination. The project is currently on hold.

Fig 18

Study model, view from side

FACADE, FINGAL COUNTY HALL OFFICES, DUBLIN


The glass facades of the City of Science and Industry at La Villette, Paris were the first of a series of projects by RFR using structural glass. These projects all use the glass to contribute to several different load paths within a strong but brittle structural screen, which requires the movements and stresses to be precisely defined. These structural systems contain a series of evolving leitmotifs such as the patented, articulated point fixings and the sparse cable trusses. This atrium facade extends the tradition to wider span structures and reinterprets the familiar vocabulary in a new context, with new materials. The newly completed Fingal County Hall in Swords, north of Dublin, defines the rear edge of a small tree filled park. The architects, Bucholz-McEvoy, had replied to this context by curving their building to embrace the oval park and the 150-year-old Himalayan Cedar at its centre .The entrance atrium is recessed into the building and is closed to the park by a curved glass curtain 32m long by 18m high. We proposed to structure this facade with horizontal plane cable trusses lined up with each floor slab, to allow uninterrupted horizontal views. The trusses span the full 32m length of the atrium, thus reinforcing its volumetric unity and generating cable trajectories of the same harmonic as the curved glass facade (figs.21 & 22). Dealing with lateral deflection under wind loads was a challenge, given the particularly large span, which we countered by the design of a progressive prestressing sequence.The stiffness of the truss was further tuned by modifying the relative stiffness of the pressure and suction cables. In order to reduce vertical deflections, we twinned the heavier suction cables and formed them into a vertical-plane catenary.

Fig 19

Study model, view from rear

Fig 20

Roof geometry : Plan, section, elevation

bridge across the fourth, open edge of the roof, with the inclined slope of the truss allowing it to carry the lateral thrusts from the radial ribs of the grid in addition to the vertical loads. The accumulation of these design decisions enabled us to keep the depth of the grid members down to 120mm. The advantage of a surface of revolution is that it can be facetted with quadrilaterals, and thus easily adapted to a stiff, planar cladding material like glass. However, fitted to a square template, it is impossible to get the edges to lie flat, so the side supports must accommodate this variable geometry. The edge zone was glazed to explicitly describe this constraint. Stable under uniform loading, the shallow shell was like all arched structures - particularly sensitive to unbalanced pattern loads and the project became in many ways a study of slenderness limits. In order to conserve the filigree nature of the grid, we needed to eliminate the lowest modes of buckling failure. This could have been done by in-plane bracing, as typified by the Saint-Lazare

316

Fig 21

Plan of atrium, showing facade and cable truss

Fig 23

Interior view, facade and roof

Fig 24

Detail of flying spars, view towards park

Fig 22

Facade, vertical section

The vertical grid of the project, with a floor-to-floor height of 3.5m, was too large a span for a single sheet of toughened glass. This specific location encouraged us to develop an element in timber for hanging and supporting the glass panels. These flying spars, in laminated American oak, are pivot-connected to the cable trusses and suspended from small-diameter cables which are prestressed between foundation and roof (fig.23). Their stability mechanism, which was inspired by our previous work on similar systems (Channel 4), uses the stiff length of the spars to mobilise the tension in the vertical cables as a counterbalance to horizontal-axis moments created by wind buffeting on the glazed surface. The spars "float" in the glass plane, their strong form and material presence emphasising the ethereal nature of the facade (fig.24).

This paper is dedicated to our friend and former colleague, Mark Richard, who was tragically killed in a recent climbing accident in Russia. The Dusseldorf project owes a lot to his inspiration.

317

STRUCTURAL DESIGN OF TWO LARGE SPAN BUILDINGS IN GERMANY


CARGOLIFTER AIRHSIP HANGER, BRAND - NEUE MESSEHALLE 3, FRANKFURT

Dipl.-Ing. Riidiger Lutz Arup GmbH MalkastenstraBe 2 D-40211 Dusseldorf Germany

Michele M. Janner, M.Sc. Ove Arup & Partners 13, Fitzroy Street London W1P 6BQ England, U.K.

ABSTRACT
Currently, a new generation of airships, the CargoLifter CL 160, is being developed. Therefore a new hangar for two airships is going to be built. With a span of 220 m, a height of 107 m and a length of 360 m it will be the largest volume in Germany. Messehalle 3A will be the new hall for the Fare in Frankfurt. This 40,000 m_ two-storey exhibition building is covered by a doubly curved roof structure spanning 165 m clear. The paper reports loading, analysis and design of both structures.

The whole project comprises the hangar - the "Werfthalle" - itself, a number of subsidiary buildings for the production of components as well as a visitors centre. (Fig. 2)

1.0

CARGOLIFTER AIRSHIP HANGAR


Fig 2 Site plan with hangar, subsidiary buildings and position of anchor masts (SIAT).

In the middle of 1997 CargoLifter AG commissioned the design for an airship hangar in Germany to house two new airships. The hangar is needed for production and maintenance of a new generation of airships. These CL160 airships are so-called "blimps" - airships filled with helium without skeleton. The airships are designed to lift and transport goods of up to 160 tons on long distance haul. The site is an airfield in Brand, approx. 50 km south of Berlin, formerly used by Soviet troops.

1.1

BUILDING CONCEPT AND ARCHITECTURE

Despite the fact that it will be an industrial building, the architecture of the CargoLifter hangar is of considerable importance, both in functional and visual terms, as the architects SIAT from Munich have designed it. The airship hangar has to fulfil various very specific requirements: it has to house two airships at a time, it must provide appropriate working conditions (min. internal temperature 17C 2m above ground, avoid drafts, guarantee sufficient lighting) and allow for light cranage under the roof construction. The total building is 360m long, 220m wide and 107m high. A total covered area of 63.000m2 - completely free of columns - would allow for eight football pitches. The shape is following closely the clearance diagram for two airships (Fig. 3 and 4).

Fig 1 Map of Central Europe.

318

arches. There are no additional tension member tying the bases. The steel arches have a clear glass roof between their top cords at 3.4m centres to allow daylight into the building

Fig 3 Plan of the CargoLifter Airship hangar (SIAT)

Fig 5

Steel trusses are supported on large concrete bases, which form the entrances of the hangar

The steel arches have cross bracing between them internally and props on the outside to avoid overall torsion buckling of the arches. Thus, the steel arches provide a very stiff framework against horizontal thrust from the sliding doors. At both ends of the building the sliding doors are located which consist of two fixed and six moving elements each. Both doors form a semi-circle in plan and a quartersegment of a circle in elevation. Three moving elements slide under one fixed element. Each shell-shaped element is fixed to a hinge at the top of the end arch and guided horizontally by rails, both in tangential and radial direction, at the bottom. Each sliding door element has two motor drives at both ends at ground level (Fig. 6). The floor is constructed as a concrete slab, using road

Fig 4

Longitudinal section (SIAT)

Sliding doors, which rotate at both ends of the building, allow the airships to be hauled in and out under appropriate wind conditions. Two-storey high concrete "naves" along both sides of the production floor provide staff facilities for 250 employees, labs and offices for a further 75 staff. Under-floor heating as well as radiant panels hanging from the steel arches in the side areas provide heating.

1.2

STRUCTURAL CONCEPT

The structural concept, developed by ARUP GmbH in the Dusseldorf office, distinguishes two main parts of the building. The central part is of a cylindrical shape consisting of five steel arches at 35m centres - each of the four bays being covered with a textile fabric - and the sliding doors at both ends. The arches spring off concrete plinths, which also act as a covered entrances, providing safety against snowavalanches coming from the fabric roof (Fig. 5). These bases are founded on large concrete pad footings designed to limit the settlement to 30mm and to avoid sliding due to horizontal wind loads and thrusts from the

Fig 6

Clamshell doors, opening (SIAT)

construction techniques. The side buildings are simple flat-slab-on-column concrete structures with an 8.1 m x 5.6 m grid.

319

1.3

LOADING ASSUMPTIONS

Initially all the known and assumed loading data has been assembled and gathered in one document, which has been sent to and checked by the entire project team. After receiving approval for all the loading the "Loading Assumptions Report" formed a very important base for the whole engineering team carrying out the calculations. The snow load considered is based on the German standards using a base value of 0.75 kN/m_, which can be reduced depending on the slope of the surface. The IFI Institute of the Fachhochschule Aachen carried out a wind tunnel test in order to determine the wind loading. The crosswind (0) Cp-values along the middle arch are shown in Fig. 7 (left) and for comparison those according to Cook [1] for a cylindrical building (right). Wind tunnel tests were carried out for the closed-door

cylindrical part of the building, these arches are not perfectly circular but polygonal. Seventeen straight segments, each with a length of about 18m, form one arch. The wind bracing is connected to the bottom chords at each intersection of two arch-segments. At the same intersection between the top chords, external props restrain any torsion in the arches, induced by the eccentrically connected membrane. At their ridge the arches are longitudinally connected by

Fig 8

Main steel arches, internal bracing, external props, typical system

Fig 7

Wind profile derived from wind tunnel test and according to Cook for a cylindrical building.

case only. This is justified, as the doors of the hangar will only be opened at wind velocities of maximum 10 m/s. Hand-calculations were sufficient to show that the open door case would not be a design load case. As special loads a 45C temperature variation was applied to the external steelwork and a 10C to the internal steelwork. Furthermore, a load due to ice of 30 mm thickness (0.21 kN/m_) was applied onto the external steel. The influence of foundation settlements (50 mm) has also been analysed. For this size of structure, the effects of foundation settlements were negligible, as would be reasonably expected.

a four chord, 8m deep truss, similar to the structure of the arches. This ridge beam enables the connection of the membrane and the valley cable at the top, and takes up the large compression force between the two end arches, generated by the doors. Tubular hollow sections were used as structural elements for the arches because of their high torsion resistance and their good buckling performance. The chords have an outside diameter of 559mm, the diagonals and the bottom straight of 355mm and the side and top straight of 273mm. An iterative analysis allowed to optimise the steel

1.4

THE MAIN STEEL ARCHES

The cylindrical part of the building is formed by 5 steel truss-arches, placed at 35m centres. The arches have a structural height of 8m and span over 225m. The top chords are at 3.441m centres and the bottom chords at 2.0m centre. The chords of the truss-arch are braceconnected to each other, with the exception of the two bottom-chords; only straight members, forming a Vierendeel-system connect these. Although the arches are always referred to as being the

Fig 9

The cylindrical part of the building covers an area of about 31.500m^ and has a total steel weight of approx. 4,200 tonnes

320

utilisation over the total structure in a sensible way. The cylindrical part has been designed by means of the Ove Arup non-linear structural program FABLON, also using the buckling option of GSA to determine the mode shapes.

1.5

FABRIC ROOF

The building enclosure is achieved using a stressed membrane spanning between the trussed tubular arches in the warp direction and between the ridge truss and the edge cable attached to the arch bases in the fill direction. Initial concern was raised about the ability to generate enough curvature in the membrane to limit stresses and deflection due to the large radius of the arches and their relatively small spacing. Initial analysis based on a form found surface with equal pre-stress in the warp and fill directions confirmed this. An additional constraint on the surface shape was to limit the deflection of the membrane under wind uplift so that it did not clash with the external bracing of the main structure. This was achieved by adopting a valley cable midway between the arches. Early on in the project, coated polyester was chosen as the preferred membrane material over the stiffer PTFE/glass chiefly on the basis of supply and cost. After the initial analysis, the polyester was shown to be the more onerous of the two with respect to deflection of the membrane but at the greater benefit of producing much lower localised stresses. High-localised stresses produced by PTFE/glass would have proven more complex to remove.

The development of robust details that are reasonably insensitive to construction tolerance is essential to the ease and speed of construction and pre-stressing of a structure of this size [2]. Proposed is an extruded aluminium luff groove. By doing so the details around the tube, that included glazing system and weather closure details, are greatly simplified. The luff groove is kept at a distance of approximately 700mm from the main structure by swinging links. After stressing diagonal links are installed at the ends of each arch segment to remove shear forces.

1.6

CLAMSHELL DOORS

The clamshell doors, as shown in Fig. 6, are each supported at 3 points. A common support for all moveable doors is the cantilevered "ridge point".

Fig 11

Hingeable connection for the door segments

The different door radii, as well as the different support height for each door (necessary for the opening mechanism), required a complex detailing of the ridge point and the upper door parts (Fig. 11). The 2 supports at the bottom of each door contain the driving mechanism.

Fig 10

Membrane connection to the steel arches

Fig 12

The concrete strip foundations for the doors include the necessary rail details to allow the doors to move in the desired direction during opening or closing and to restrain the doors in the tangential and radial directions in the closed position

321

1.7

DOOR STRUCTURE AND DESIGN

The enormous dimensions of the doors (arch length 168m, bottom width 42m) result in large weights. The door weight strongly influences the hangar costs: the upper door support reactions are collected in the ridge point, adding up to a point loading on the end arches. This point load has a large effect on the section dimensions in the arches. The costs of the foundations and the driving mechanism depend on the lower door support reactions which are mainly a result of the steel weight of the doors. Minimising the door tonnage was a main goal.

movement of the ridge points, several hundreds of load combinations were generated. With the ARUP (linear) structural calculation program GSA, a first impression of governing load combinations was achieved. A distinction of 2 loading cases, producing the typical deformation shapes of an arch, could be made. The appropriate loading factors were determined by making use of those shapes.

Fig 14

Two governing wind profiles with their accompanying deflection shapes

The shape on the right shows a typical "gravity dominated" deflection, whereas the shape on the left shows a typical "wind pressure dominated" deflection. In the latter situation the loadings acting approximately in gravity direction are multiplied by a loading factor 0.9. The German steel design code (DIN 18800, part 2) stipulates a nonlinear analysis for structures with critical loading factors _cr less than 10, as linear analyses do not produce realistic section forces for such structures. The procedure is to superimpose scaled buckling shape deflections (see Fig. 13) and the original geometry and perform the nonlinear analysis on the thus obtained imperfect structure. With the ARUP nonlinear calculation program FABLON the imperfect structures were nonlinearly analysed. FABLON recognises eventual instability of the structure (the development of the collapsing mechanism can be monitored on the screen), so that the derived stresses in a "survived structure" can be directly used for comparison with allowable steel stresses (without slenderness limitations). Additionally, however, a lateral torsional buckling check for the individual elements in the structure was required as FABLON does not account for this instability phenomenon. The deformations of the lower part of the shell, relative to the edge beams, were in the order of 300 mm. Even though the horizontally oriented elements are very flat, the presence of an initial arch height of about 2 m does not give rise to a snap-through occurrence.

Fig 13

The structure of a moveable door segment and one of its first buckling mode shapes

The lightest structure was achieved by adopting a shell principle: The inner part of a door segment consists of a spherically shaped grid structure, realised by identical horizontal, vertical and diagonal elements (DIN HE240A-sections), which are rigidly connected in their joints. The cladding (corrugated metal sheets) spans between the horizontal elements. The side beam dimensions are d x w = 3000 x 800 [mm]. The shell is eccentrically connected to the side beams: at one side to the top flange and at the other side to the lower flange. Hence, a horizontal section through a door shows an approximate Z-shape, allowing for the sliding of door underneath the other. At the bottom the shell joins the lower beam (d x w = 2300 x 800 [mm]) concentrically. Due to the hangar plan-symmetry the investigations could be restricted to 4 different doors (including the fixed door). The wind tunnel test produced 8 different wind profiles for each of the 4 doors. Together with dead weight of steel frame and cladding, snow loading (according to DIN 1055), temperature loading and

322

1.8

CONCRETE ELEMENTS

In addition to the large amount of steel, large amounts of concrete are also used for the Airship Hangar, about 20.000 m_ in total. The main concrete elements are the foundations of the arches and the doors, and the floor slab. Between the concrete bases of the arches a twostorey high office area is under construction.

A strip foundation, consisting of a one metre thick reinforced concrete slab and an upstanding block of 2 m height was used for the sliding doors. The foundation geometry was mainly governed by the required rail pattern for moving the doors. The slab width varies from 6 m to 10 m (from the centre towards the fixed door segment). The large total length of the strips necessitated the use of doweled temperature/shrinkage joints. These joints are located in the centre of each segment (covering one door segment). Single slab foundations are used for the arches in the cylindrical part. The bottom level of the 2 m thick foundation slab is at -5 m. Two different slab dimensions can be distinguished, depending on the position in the "cylinder". For the end arches, where the large door loads occur, foundations of 12x26 m are necessary. At the inner arches, the foundation sizes could be limited to 9x16 m. The arches are based at a level of +8.85 m and connected to the foundation slab at - 5.0 m through 800 mm thick concrete walls, which form the short side of the office zone.

Fig 15

In front of the offices, between the concrete arch-bases, large "pools" have been incorporated for gathering the rain water from the membrane Fig 17 Connection of the top chord of the arches to the concrete base. The connecting plates for the membrane and its edge cables are clearly visible

The fairly good soil conditions at the site did not require complicated foundations. A maximum allowable soil pressure of 500 kN/m2, with the assumption of a maximum settlement of 40 mm, was used for the calculations of the door foundations. The groundwater level at -15 m was not of importance for the structural design so that sliding, stability, occurrence of tension and maximum soil pressure were used as design criteria.

In order to minimise the foundation moment about the hangar's longitudinal axis, the foundation slab is placed eccentrically und the arch. Use is made of the fact that the horizontal arch forces act in the outward direction: putting the slab perimeter outside of the hangar results in counteracting moments from horizontal and vertical arch forces. The decks of the office area are designed as flat-slab-oncolumn. The 300 mm thick concrete slabs span between the concrete bases of the arches over 3 rows of columns, with a maximum clear span of 8.5 m.

Fig 16 Cross-section through the arch-bases (SIAT)

323

1.10

CREDITS

Fig 18

Concrete bases and steel arches. Between the concrete bases the steel structure for the facade of the offices is already in place

CargoLifter AG. Wiesbaden SIAT Architektur + Technik, Miinchcn Building Services: Kloffel, Bruchkobel Client's representative: Connert + Wolfram, Dusseldorf Structural Design: Arup GmbH, Dusseldorf Road & Landscape Design: Cordes + Partner, Miinchen Fire engineering: Halfkann & Kirchner, Erkelenz Wind tunnel tests: IFI, Institut fur Industrieaerodynamik, FH Aachen Photographer: M. Lindner

Client: Planer and Architects:

The slabs act as diaphragm between the concrete bases, restraining the fassade against horizontal loading. The slabs are only connected to one base. An expansion joint connection to the other base allows for temperature movement of the slab.

1.9

ORGANISATION

CargoLifter AG as the client commissioned SIAT Bauplanung und Ingenieurleistungen GmbH & Co. OHG with the planning and architectural design for the whole site in early 1997. The structural design by Arup was started in October 1997 and finished in June 1998. Site works started in May 1998 with the steel construction due in October 1998. The building is due to be completed in October 2000. Total building cost for the hangar is estimated at DM 140 million.

Fig 19

The CargoLifter airship hangar under construction, with Joey, the 1 to 8 model of the CL160 on the foreground

324

2.0

M E S S E H A L L E 3A

The fair in Frankfurt, Messe Frankfurt, is expanding and therefore it is building a new fair hall towards the south side of their grounds: Neue Messehalle 3.

building units are working together to fulfil the functional as well as the esthetical needs. The main elements are clearly defined (Fig. 21): The The The The roof structure, with their A-frames middle deck, often referred to as the "sandwich" side buildings foundations

The roof structure, spanning in the long direction of the hall, covers the 20,000 m upper hall completely free of intermediate columns. Along its east and west side the roof is supported on 12 A-frames, which have been incorporated as separate elements within the concrete side buildings. The A-frames transfer all the vertical and horizontal roof loading down to the foundations.
2

Fig 20

Plan of the Fare area with the existing halls and the new faire hall, Neue Messehalle 3 (Rendering by Nicholas Grimshaw & Partners)

The 20,000 m deck, separating the lower from the upper hall is formed by pre-cast concrete units on 32 m long steel trusses, carrying 15 kN/m live load. In order to use the volume generated by the 3.4 m deep trusses sensibly, the larger parts of the technical plant rooms were located in this zone, resulting in the conception of a double deck: the sandwich. At the north side the sandwich cantilevers, creating a space for the Via Mobile: the people mover or the travelator network of the Messe Frankfurt.
2

The Messehalle consists of a lower hall with columns in a 32 m x 17.5 m and 6 m grid, an upper hall, which is completely free of columns over the area of 165 m x 120 m. On the east, west and south side of the exhibition area smaller concrete side-buildings enclose the Messehalle.

2.1

BUILDING CONCEPT AND ARCHITECTURE

*ni|jj|i m i

imoflgmnfrmm

The design of the Messehalle is characterized by the unity of Architecture and Technology. Individual

-! 1 1 1 *
W

1
HJ Fig 22 A cross-section through the Neue Messehalle 3. The sandwich separates the upper from the lower hall and houses the technical plant rooms

Fig 21

Isometric view on the Neue Messehalle 3 (Rendering by Nicholas Grimshaw & Partners)

The side buildings are located on the east, west and south side of the main exhibition area. They include offices, restaurants, toilets and other support functions. Five concrete "boxes", positioned between the roofs' steel Aframes, and a flat-slab-column system form the load bearing structure of the side buildings east and west. The concrete boxes also restrain the sandwich horizontally. A special doweled connection allows in plane expansion of the sandwich due to temperature movement (Fig. 23). The pre-cast concrete units of the sandwich are

325

connected to each other by a 120 mm concrete top layer, causing the sandwich deck to act as a stiff plane. This way the sandwich restrains the steel trusses against buckling and is able to transfer the wind loading from the north facade to the horizontal load bearing structure in the side buildings.

Dowels

.^-Dowels

Fig 23 The side buildings restrain the sandwich slab horizontally, yet allow the deck to expand within its plane

Fig 25

The geometry of the roof reflects the need for structural height (sketches by Nicholas Grimshaw & partners).

Sc^f-iTr IS.

/ A-NticHr

HON rV*<x ^ m S t x r ^ N

The second step in the development of the roof structure is the translation of the roof geometry into a roof structure. To do this the load paths within the geometry were analysed and a corresponding stick model developed. The roof structure is composed of 5 compression arches and six tension zones (Fig. 26-27).

Fig 24 A front elevation of the concrete boxes showing the A-frames between the boxes and the large openings required for the ductwork from the plant rooms (Sketch by BGS)

The boxes of the side buildings share foundations with the A-frames of the roof. A combination of a flat slab and piled foundation proved to be the most economic solution for the foundations. The columns of the lower hall have single pad foundations.
Fig 26 The compression zones are curved in two directions and carry the load mainly by axial forces to its supports, where the horizontal thrust force is partially taken up by the tension zone (see Fig. 26) and partially by the A-Frames, which also support the roof vertically (sketch by Nicholas Grimshaw & Partners)

2.2 THE ROOF STRUCTURE The design of the roof structure finds its roots in the principle of a folded material. Like with corrugated metal decking, the geometry is determined by the loading and the way the loads are carried. The roof geometry was adapted to react in an optimal way to the forces within the roof structure, generated by the loading. At mid span the structural height of the folding is at its maximum, reducing towards the supports (Fig. 25).

326

Fig 27

The tension zone is doubly curved as well. Its geometry was developed using computerised form-finding methods (sketch by Nicholas Grimshaw & Partners)

During the initial design stage, the concept was developed to take up all of the horizontal thrust force by the tension zone. On one side the arches would be connected to the A-frames with a horizontal sliding connection. As can been derived from Figure 28, the movements of the roof structure in longitudinal direction would mount up to 250 mm under snow load and concentrate only on one side.

Fig 29

The A-frames, positioned as individual elements within the side buildings, support the side buildings' roof and restrain the main roof structure horizontally (rendering by Nicholas Grimshaw & Partners)

During the early design stage, the tension zone was developed by having the main ties running from A-frame to A-frame, taking up the horizontal thrust forces only at the supports. The corresponding shape was generated using computerised form-finding methods (Fig. 30).

GEOMETRY AFTER 0 ITERATIONS

Fig 28

Longitudinal roof movements using the concept of horizontal sliding supports. Here, the red colour stands for horizontal movements up to 250 mm

For infrastructure projects, like bridges, movements of 250 mm can easily be taken up in a normal movement joint. However, when facades are considered, the development of simple robust connections resulted in locking the roof structure by the A-frames, reducing the longitudinal movement to about 1/10. Apart from the main roof, the A-frames also support the roof structure of the side buildings, as an extension of the main roof (Fig. 29).

GEOMETRY AFTER 100 ITERATIONS

GEOMETRY AFTER 791 ITERATIONS NATURAL OPTIMISED SHAPE

Fig 30

The form finding process

327

After the competition stage, the tension zone was again investigated, aiming to optimise the structure in weight and stiffness. To do this various alternatives were investigated. The most promising alternative is shown in Figure 31 underneath the original concept.

Fig 32

The axial force diagrams for the loadcase "dead load plus snow load" for the two different structural systems as shown in Figure 31

Fig 31

Two concepts for the tension zone. Top, concept A: the main tension ties run from support to support. Bottom, concept B: the main tension ties, run in a contra curvature to the main arches, linking up to the arches at intermediate points

The design method used for analysing the Frankfurt roof structure is similar to the design method applied for the Airship Hangar. Imperfections were applied to the perfect geometry after which the geometry was fully non-linearly analysed. Figure 33 shows one of the buckling mode shapes of the roof.

In Figure 32 the corresponding force diagrams are shown. For ease of understanding these figures are placed next to structures. Although the maximum compression forces in the arches are clearly smaller for concept B and they do not built up towards the supports as with concept A, the geometry of A was chosen as it turned out to have the most simple geometry with respect to the roof cladding. For the design of the roof structure a wind tunnel test was carried out to define the design wind load and in fact the snow load as well. The Messe Frankfurt specified their requirements with respect to exhibition hanging loads. Foundation settlements have been agreed with the geotechnical engineer and limited to 20 mm. In Comparison to the CargoLifter Airship hangar, this roof structure is much more sensitive to foundation settlements due to the very low rise of the roof arch. The roof structure with A-frames has been analysed by means of the Ove Arup non-linear structural program FABLON, also using the buckling option of GSA to determine the buckling and the dynamic mode shapes.

Seals 1 1 064E+3 Isometric Scale 1:1 303E*3 Deformation magnification: 1260 Case: A1 Al, mode I, Load factor. 5 913

Fig 33

One of the roof structures' lateral sway modes

328

2.3 O R G A N I S A T I O N
In January 2000 the Messe Frankfurt as client commissioned Hochtief AG from Frankfurt with the planning and construction of the Neue Messehalle 3. The Architectural design by Nicholas Grimshaw & Partners started in summer 1999 in close collaboration with Ove Arup & Partners in London as structural engineers and Kuehn Bauer Partner from Munich for services. The building is due to be finished by June 2001 before the international automobiles show IAA. The total building cost for the Neue Messehalle 3 is estimated at DM 230 million.

3.0

CONCLUSION

The very specific requirements of the airship-production for the CargoLifter AG have challenged SIAT-Architects and ARUP-Engineers to design an interesting lightweight building to high aesthetic standards. The building, due to its size, is of course a huge challenge, however by using proven technologies and finding a tailor-made shape for its use, an economically viable solution has been found to suit the client's requirements. Totally different and with its own specific requirements the new Neue Messehalle 3 has evolved as challenging and innovative applying the same calculation method as for the CargoLifter Airship Hangar. Due to the very strong concept, the design process has been experienced as a true development rather than an agglomeration of changes.

3.1
1

REFERENCES
Cook: The Designer's guide to wind loading of building structures, Butterworths, London

2 Fig 34 Rendering by Nicholas Grimshaw & Partners of the roofs' steel structure

Bubner: membrane construction, connection details, Wehlmann, Essen, 1997

2.4

CREDITS
Messe Frankfurt AG Hochtief AG, Frankfurt WPV, Frankfurt Nicholas Grimshaw & Partners London ABB, Frankfurt Ove Arup & Partners, London Arup GmbH, Dusseldorf BGS, Berlin Kuehn Bauer Partner, Munich BPK, Dusseldorf Ingenieursozietat Prof.Dr.Ing. Katzenbach und Dipl. Ing. Quick IFI, Institut fur Industrieaerodynamik, FH Aachen + Wacker Ingenieure

Client: Contractor: Client's representative: Architects:

Structural Engineers:

Building Services: Fire Engineers: Soil Engineers:

Wind tunnel tests:

Anda mungkin juga menyukai