Variability of Unit Flexural Bond Strength and its Eect on Strength in Clay Brick Unreinforced Masonry Walls Subject

to Vertical Bending

Leesa Maree Heer

Faculty of Engineering and Built Environment University of Newcastle

A thesis submitted for the degree of Masters of Philosophy 2009

The thesis contains no material which has been accepted for the award of any other degree or diploma in any university or other tertiary institution and, to the best of my knowledge and belief, contains no material previously published or written by another person, except where due reference has been made in the text. I give consent to this copy of my thesis, when deposited in the University Library, being made available for loan and photocopying subject to the provisions of the Copyright Act 1968.

Leesa Maree Heer

I dedicate this thesis to my dearest mother and stepfather, who hoped when all others had lost hope, who believed when all others had lost faith. You are the epitome of goodness, honesty, kindness and unconditional love.

Acknowledgements
I would like thank the Australian Research Council and Think Brick Australia for providing the nancial support for this project.

I would like to thank the sta in the University of Newcastle Civil Engineering laboratory. They were extremely professional and ecient in assisting me conduct the experimental testing and for that, I am extremely grateful. Thanks in particularly goes to Ian, Roger (a.k.a. Q), Goran, Laurie, Shaun, Dan, Tim, Heber Sugo, Josh and Anne in the oce. Similarly, I would like to thank the Sao Carlos Civil Engineering lab sta and Marcio Correa for building and testing the Brazilian walls.

I would like to extend a very, very, very special thank you to Rob Petersen, Andrew Abbo and Gordon Fenton for the seless time they gave me in order to help me understand and get past the numerous technical issues associated with the project. I would like to thank my supervisors Mark Stewart and Mark Masia for their input and support on the project, and Steve Lawrence for his valuable input into the project.

I would like to express appreciation of my oce mates: Ben, Torill, Doris, Khalid, Mohammad and Daniel, and to my closest friends in Newcastle, Bimbi and Moo Moo. Special thanks goes to Ben (a.k.a. Benjamina) for assisting me with technical aspects of Statistics and Matlab.

Finally, Id like to extend a very special thank you to Janet Carr and my mum, for helping me through the dicult times that I had with my body. Their support and strength carried me when I could not carry myself.

May peace be with you all.

List of Publications
Journal paper: Heer, L.M., Stewart, M., Masia, M., Correa, M., 2008 Statistical Analysis and Spatial Correlation of Flexural Bond Strength for Masonry Walls, Masonry International, Volume 21, No. 2, pp. 59-70. Conference Paper: Heer, L.M. Stewart, M., Masia, M., Correa M., 2008 Spatial Correlation of Flexural Bond Strength for Masonry Walls: An Experimental and Statistical Study, 14th International Brick and Block Masonry Conference, 17-20 Feb, Sydney, Australia. Conference Paper: Correa, M., Heer, L., Masia, M., Stewart, M., 2008 Flexural Bond Strength for Masonry Walls: An Experimental and Statistical Analysis, 8th International Seminar on Structural Masonry, 5-7th Nov, Istanbul, Turkey.

iv

Abstract
It has been shown that masonry material properties, in particular, unit exural bond strength (ft ), vary signicantly throughout masonry structures, despite the fact that often only one type of brick and mortar are used. Unit exural bond strength was previously identied as one of the most important material parameters contributing to the strength of clay brick unreinforced masonry (URM) walls in exure. It was the objectives of this research, in the context of clay brick URM walls subject to vertical bending, to examine how unit exural bond strength varied spatially in a clay brick URM wall, determine a best t probability distribution function which can describe expected variability in unit exural bond strength and determine how this variability and other factors aect wall behaviour and failure load using 3D non-linear nite element analysis (FEA). It was hoped that modelling a full sized clay brick URM wall subject to vertical bending using a 3D non-linear FEA model would more accurately predict wall failure load (compared to current analytical methods) and allow the examination of crack pattern development as the wall progresses to failure upon being laterally loaded. The rst part of the research project was to conduct an experimental program to examine unit-to-unit spatial strength correlation within six full sized clay brick URM walls and to characterise a unit exural bond strength probability distribution. It was observed that although weak correlation in unit exural bond strength exists in some courses and between courses, these locations were dicult to predict and didnt follow any particular pattern relating to for example, mortar batch. Therefore, although somewhat counter-intuitive, the results indicate that statistically signicant correlation between adjacent unit exural bond strengths is not likely to be observed. It was also observed that clay brick wall unit exural bond strengths obtained for all of the walls tested best t a truncated Normal probability distribution. Strength of the brick/mortar interface appeared to be governed by factors relating to workmanship (and therefore mortar quality and moisture content), weather (which can aect material characteristics like brick suction rate) and inherent material variability. It would appear that brick suction rate can signicantly aect the overall strength of a URM wall.

Stochastic analysis was conducted for walls with and without uncorrelated spatial variability in unit exural bond strength and associated tensile fracture energy (Gf I ). It was found that the TNO DIANA 9.2 FEA package could be used to implement spatial variability of various material parameters and reasonably accurately model failure of clay brick URM walls in vertical bending. From the non-linear FEA model development stage, it was observed that because the brick/mortar bond has signicantly more strength capacity in compression, it appears that the lateral load resistance of the wall comes from a combination of the ability of the brick/mortar bond to tensile soften while providing signicant compressive resistance at the compressive edge. It was found for a spatial stochastic analysis with spatial variability in bond strength (referred to from now on as a spatial stochastic analysis), with COVs of 0.1, 0.3 and 0.5, that COV of wall failure loads were relatively small, being 0.02, 0.04 and 0.06 respectively. For the non-spatially varying stochastic analysis with fully correlated bond strength (now referred to as non-spatial stochastic analysis), with COVs of 0.1, 0.3 and 0.5, COV of wall failure loads were 0.07, 0.20 and 0.32 respectively. For the spatial stochastic analysis, it was found that with a bond strength COV increase from 0.1 to 0.5 the mean wall failure load dropped from 2.25 kPa to 2.0 kPa (an 11% reduction). Despite the relatively small drop in magnitude of the mean wall failure load with increase in bond strength COV, the mean wall failure loads were statistically dierent to one another. For the non-spatial stochastic analysis, mean failure load stayed relatively constant at 2.24-2.25 kPa. These results could be explained by examining the 3D wall progression to failure. For walls with spatial variability in bond strength, it is expected that wall failure load COVs would be smaller because those walls would consistently be composed of smaller valued bond strengths which would consistently contribute to weakness in the wall. For the non-spatial wall simulations, this eect would not occur as failure load is determined by one uniform weak or strong bond strength. It was proposed that failure of a clay brick URM wall is not governed by one course only cracking, but rather, instability in the wall is governed by several courses in the vicinity of locations of large bending moment. It was shown that various current stochastic approximations which employ a unit failure hypotheses in combination with a linear/elastic approximation for rst cracking load all underestimated wall capacity signicantly. The reason for this is suggested as being

vi

because all hypotheses only assume failure is governed by one course and linear/elastic theory only considers the tensile capacity of a joint and neglects strength capacity available as a result of joint tension softening and the resistance to failure provided by compressive strength on the compression side of the wall. The hypotheses also dont take into consideration factors which aect overall wall bond strength mean which result from inuences such as workmanship, weather and material variability factors, such as (for example), variation in brick suction rate due to weather conditions which can make the overall strength of the wall stronger or weaker. Based upon a comparison in wall failure load COV for the spatial and non-spatial stochastic wall analysis results, a more realistic approach for future modelling attempts of spatial variability in masonry material properties is suggested. This would address the issue of external factors such as workmanship and weather on the overall strength of the wall, as well as the inherent bond strength variability due to material variability. For walls with spatial variability in bond strength, upon examination of numerous wall simulation results, several crack patterns were witnessed and are discussed. Keywords: unit exural bond strength, spatial variability, vertical bending, lateral load, unreinforced masonry, clay, single leaf, failure, wall behaviour, failure hypothesis, tensile fracture energy, TNO DIANA

vii

Contents
List of Publications Abstract 1 Introduction 2 Literature Review 2.1 2.2 2.3 2.4 2.5 Understanding URM Wall Behaviour in Terms of Material Parameter Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Factors Aecting the Strength of URM Walls . . . . . . . . . . . . . . . . Random Variation of Material Properties in Brickwork . . . . . . . . . . . Experimental Data of URM Walls in Flexure . . . . . . . . . . . . . . . . Appropriate Numerical Models for Unreinforced Masonry in Flexure . . . 8 10 13 18 20 22 22 24 26 30 39 41 41 iii iv 1 8

3 Experimental Program 3.1 3.2 3.3 3.4 3.5 Wall Materials and Construction . . . . . . . . . . . . . . . . . . . . . . . Method of Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatial Correlation Method . . . . . . . . . . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion of Experimental Results . . . . . . . . . . . . . . . . . . . . . .

4 Developing the 3D Non-Linear FEA Model 4.1 4.2 4.3 4.4 The 3D FEA Modelling Package . . . . . . . . . . . . . . . . . . . . . . .

Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory 43 The Consideration of Two Non-Linear Interface Behaviour Material Models 48 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

viii

CONTENTS

4.5 4.6 4.7 4.8 4.9

An Aside: The Issue of Load Step Size for the 2D Non-Linear FEA Model Examining Failure Load Using Interface Models in the Equivalent 3D NonLinear Wall Model With Self-Weight . . . . . . . . . . . . . . . . . . . . . Modelling Doherty (2000) Experimental Walls . . . . . . . . . . . . . . . . Doherty (2000) Wall 3D Non-Linear FEA Results . . . . . . . . . . . . . . Discussion of 3D FEA Model Development Results . . . . . . . . . . . . .

57 57 63 70 75 76 76 82 88 92 95

5 Stochastic FEA and Wall Strength Variability 5.1 5.2 5.3 5.4 5.5 5.6 5.7 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatial Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Non-Spatial Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . and Upper Bound (Plastic) Analysis . . . . . . . . . . . . . . . . . . . . . Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory . . . . . . . . . . . . . . . . . . . Accounting for the Large Dierence in Magnitude Between First Cracking

Comparing Wall Conguration Failure Load with Lower Bound (Linear/Elastic)

and Failure Load for the the Full Size Wall Structural Conguration . . . 102 Discussion of Stochastic Analysis Results . . . . . . . . . . . . . . . . . . 106 108

6 Unit Failure Progression in a Wall 6.1 6.2 6.3 6.4 6.5 6.6 6.7

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 Expected Behaviour Based Upon Statics . . . . . . . . . . . . . . . . . . . 108 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength109 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 1 Strength - Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Summary of Crack Pattern Observations . . . . . . . . . . . . . . . . . . . 121 Discussion of Unit Failure Progression Results . . . . . . . . . . . . . . . . 122 124

7 Conclusions and Recommendations 7.1

Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

ix

CONTENTS

A Appendix to Chapter 3 A.2 Greyscale Plots for Unit Flexural Bond Strengths in Each of the Newcastle,

133

A.1 Newcastle Wall Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Australia Walls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 A.3 Unit Flexural Bond Strengths Plotted for Each Course in Each Newcastle, Australia Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 A.4 Average Unit Flexural Bond Strength for Each Course in Each Newcastle, Australia Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 A.5 Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls 144 A.6 Autocorrelation function values plotted for each course in all walls . . . . 148 A.7 Autocorrelation function values plotted for between courses in all walls . . 151 B Appendix to Chapter 4 154

B.1 Summary of specications for testing the non-linear 2D and 3D FEA model 154 B.2 Doherty 2000 static push test wall load versus deection results . . . . . . 156 B.3 Summary of specications for modelling Doherty (2000) walls using the 3D non-linear FEA model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 C Appendix to Chapter 5 C.2 Results from Spatially Variable Bond Strength Simulations in the Full Sized Masonry Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Bibliography 165 160

C.1 Summary of FEA specications for the full size wall in vertical bending . 160

List of Figures
1.1 1.2 1.3 Indians using brick construction methods similar to those used thousands of years ago (Internet 2008b) . . . . . . . . . . . . . . . . . . . . . . . . . Inca structures in Peru absent of mortar, opting instead to accurately cut stones and create bond through close packed friction (Internet 2008a) . . Roman aqueducts appear to be absent of mortar, rather utilizing the surface of the accurately cut stones to create bond through joint friction (Internet 2008c) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 Walls SCB1 and SCB2 constructed in Sao Carlos, Brazil . . . . . . . . . . Bond wrench testing units in wall SCB2 . . . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Probability Distribution Fits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, CDF1 . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa) . . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strength for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . 4.1 4.2 2D FEA model test wall set up to compare with linear/elastic theory (no self-weight) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3D FEA model test wall set up to compare with linear/elastic theory . . . 42 43 37 36 35 34 31 32 4 24 26 3 2

xi

LIST OF FIGURES

4.3 4.4 4.5 4.6 4.7 4.8 4.9

Mesh Densities for the 2D and 3D brick/mortar bodies in the FEA model Non linear tension softening model for interface elements (Cornelissen et al. 1986) and (Hordijk 1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . Diagram used for linear/elastic theory calculation (no self-weight) . . . . The composite interface model two-dimensional interface yield function (Hendriks and Wolters 2007) . . . . . . . . . . . . . . . . . . . . . . . . . Typical behaviour of the mortar joint between two bricks which are uniaxially loaded (taken from (Lourenco 1996a)) . . . . . . . . . . . . . . . . . Tensile fracture energy material data from Lourenco (1996b) report . . . . Tensile bond strength versus tensile fracture energy for clay brick masonry arranged by mortar type (Van Der Pluijm 1992) . . . . . . . . . . . . . .

45 45 46 48 49 50 51

4.10 Tensile bond strength versus tensile fracture energy for clay brick masonry arranged by brick type [Joosten: wire cut clay bricks, Vijf Eiken: soft mud clay bricks] Van Der Pluijm (1992) . . . . . . . . . . . . . . . . . . . . . . 4.11 Tensile bond strength versus tensile fracture energy for clay brick masonry with general purpose mortar (1:1:6 + air-entrainer) from (Van Der Pluijm 1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.12 Stress displacement curves for tests which used clay bricks and 1:1:6 mortar. Average theoretical descending branches according to the non-linear tension softening model (Hordijk 1991) and the composite interface model (Lourenco 1996a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.13 Fitting a linear relationship between tensile bond strength versus tensile fracture energy for clay brick masonry with general purpose mortar (1:1:6 + air-entrainer) using Figure 4.11 from Van Der Pluijm (1997) . . . . . . 4.14 Load step versus deection curve for the 3D non-linear model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.2, peak load = 1.8kN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.15 Load step versus deection curve at point of maximum deection for the 3D non-linear model shown in Figure 4.2 using the composite interface elements, ft = 0.2 MPa, peak load = 1.92 kN . . . . . . . . . . . . . . . . 59 58 56 53 52 51

xii

LIST OF FIGURES

4.16 Load step versus deection curve at point of maximum deection for the 3D non-linear test wall model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.8, COV = 0.5, peak load = 3.85 kN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.17 Load step versus deection curve for Dohertys 3D non-linear model shown in Figure 4.2 using the composite interface elements, ft = 0.8, COV = 0.5, peak load = 4.35 kN . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.18 Final deected wall shape for a wall using non-linear tension softening interface elements, with variable bond strength according to ft = 0.8, COV = 0.5, peak load = 3.85 kN . . . . . . . . . . . . . . . . . . . . . . . 4.19 Final deected wall shape for a wall using composite interface elements, with variable bond strength according to ft = 0.8, COV = 0.5, peak load = 4.35 kN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.20 Static push test rig setup from (Doherty 2000) . . . . . . . . . . . . . . . 4.21 Load versus deection curve for the 3D non-linear FEA model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.328, peak load = 2.5 kN . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.22 Load versus deection curve for the 3D non-linear FEA model shown in Figure 4.2 using the composite interface elements, ft = 0.328, peak load = 2.7 kN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.23 Stress distribution across the thickness of the wall at the location of peak deection (course 10) for the 3D non-linear wall model shown in Figure 4.2 using the composite interface elements, ft = 0.328 . . . . . . . . . . . . . 73 74 4.24 Limiting stress distribution across the thickness of the wall at the location of peak deection (course 10) for the wall shown in Figure 4.2 . . . . . . . 5.1 5.2 5.3 5.4 5.5 Convergence of mean failure load and COV with increased number of simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary conditions for the 3D full sized wall . . . . . . . . . . . . . . . . Change in mean failure load (kPa) and COV with unit exural bond strength COV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.1 Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.3 83 84 85 78 80 72 72 62 64 61 61 60

xiii

LIST OF FIGURES

5.6 5.7 5.8 5.9

Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.5 Cubic spline interpolation t to 3D non-linear FEA wall peak loads for uniform bond strengths ranging from 0.01 MPa to 1.4 MPa . . . . . . . . Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . .

85 88 90 90 91

5.10 Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.11 Initially the full sized wall acts as a propped cantilever until stresses due to the moment at the base exceeds bond strength at the extreme tensile bers of the brick/mortar bond . . . . . . . . . . . . . . . . . . . . . . . . 5.12 Plastic limit denition of behaviour obtained from observing the behaviour of the brick/mortar interface subject to loading as shown in Figure 6.5, Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.13 Structural conguration used for analysis of the wall using the three failure hypotheses and linear beam theory . . . . . . . . . . . . . . . . . . . . . . 5.14 Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5.15 Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5.16 Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 5.17 Stress distribution across the thickness of the brick/mortar joint at the base of the wall at peak load for the wall with uniform ft = 0.4 MPa . . . 103 5.18 Modied linear beam conguration to account for restoring moment at the base of the wall due to tensile softening of the bottom brick/mortar joint 6.1 For wall with uniform ft = 0.4 MPa, despite rst crack (rst crack load = 1.3 kPa) being reached at the base of the wall, it can be seen from the reverse curvature at the base, that the base remains relatively bonded to ground by the mortar joint at a load of 1.4 kPa (load step 8) . . . . . . . 110 104 98 94 93

xiv

LIST OF FIGURES

6.2

For wall with uniform ft = 0.4 MPa, with continued loading, bottom mortar joint cracks and tension softening occurs, Failure (Peak) Load = 2.25 kPa (load step 16) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.3

For wall with uniform ft = 0.4 MPa, after failure (peak) load, cracks developed over courses 15 to 18 of the pressure loaded wall, Load = 2.08 kPa (load step 18) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

6.4 6.5 6.6 6.7 6.8

For wall with uniform ft = 0.4 MPa, load versus deection curve for a node located at course 16, where peak out-of-plane deection occurs . . . . . . 113 The stress distribution in the bottom course across the thickness of the joint at various load steps (uniform ft = 0.4 MPa) . . . . . . . . . . . . . 113 (The crack opening thickness in the bottom course across the thickness of the joint at various load steps (uniform ft = 0.4 MPa) . . . . . . . . . . . 114 The stress distribution at the course of maximum deection (course 16) across the thickness of the joint at various load steps (uniform ft = 0.4 MPa)114 The crack opening thickness at the course of maximum deection (course 16) across the thickness of the joint at various load steps (uniform ft = 0.4 MPa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.9

Initially, crack opening occurs in the weaker joints in courses close to the base of the wall, (ft = 0.4 MPa, COV = 0.5) . . . . . . . . . . . . . . . . 116

6.10 At failure, full crack development occurs at a course near mid-height of the wall, (ft = 0.4 MPa, COV = 0.5) . . . . . . . . . . . . . . . . . . . . . . 117 6.11 Initially, crack opening occurs in the weaker joints in courses close to the base of the wall (ft = 0.4 MPa, COV = 0.5) . . . . . . . . . . . . . . . . 118 6.12 As more load is applied, cracking develops rst at a location of extremely weak bond strengths in course 16 (ft = 0.4 MPa, COV = 0.5) . . . . . . 119 6.13 At failure, crack has continued propogation in course 17 (ft = 0.4 MPa, COV = 0.5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 A.1 Newcastle Wall 1, 1:1:6, High Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa) . . . . . . . . . . . . . . . . . . . . . . . . 138 A.2 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa) . . . . . . . . . . . . . . . . . . . 138

xv

LIST OF FIGURES

A.3 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa) . . . . . . . . . . . . . . . . . . . . . . . . 139 A.4 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa) . . . . . . . . . . . . . . . . . . . . . . . . 139 A.5 Newcastle Wall 1, 1:1:6, High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . . . . . 140 A.6 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa) . . . . . . . . . . . . . . . . . . 140 A.7 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . . . . . . . 141 A.8 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . . . . . 141 A.9 Newcastle Wall 1, 1:1:6, High Quality Mason, Average Unit Flexural Bond Strength for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . . . . 142 A.10 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strenght for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . 142 A.11 Newcastle Wall 1, 1:1:6, High Quality Mason, Average Unit Flexural Bond Strength for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . . . . 143 A.12 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strenght for Each Course (MPa) . . . . . . . . . . . . . . . . . . . . 143 A.13 Newcastle Wall 1, 1:1:6, High Quality Mason, Probability Distribution Fits 144 A.14 Newcastle Wall 1, 1:1:6, High Quality Mason, CDF1 . . . . . . . . . . . 144 A.15 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Probability Distribution Fits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 A.16 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, CDF1 . . . . . . . 145 A.17 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Probability Distribution Fits 146 A.18 Newcastle Wall 3, 1:1:6, Fair Quality Mason, CDF1 . . . . . . . . . . . . 146 A.19 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Probability Distribution Fits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 A.20 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, CDF1 . . . . . . . 147 A.21 Newcastle Wall 1, 1:1:6, High Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . . . . . . 148

xvi

LIST OF FIGURES

A.22 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . 148 A.23 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . . . . . . 149 A.24 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . . . . 149 A.25 Sao Carlos Wall 1,1:2:9, High Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . . . . . . 150 A.26 Sao Carlos Wall 2,1:2:9 , Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall . . . . . . . . . . . . . . . . . . . . . . 150 A.27 Newcastle Wall 1, 1:1:6, High Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall . . . . . . . . . . . . . . . . . 151 A.28 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for Between All Courses in the Wall . . . . . . . . . . . . 151 A.29 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall . . . . . . . . . . . . . . . . . 152 A.30 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Autocorrelation Function Values for Between All Courses in the Wall . . . . . . . . . . . . . . . 152 A.31 Sao Carlos Wall 1,1:2:9, High Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall . . . . . . . . . . . . . . . . . 153 A.32 Sao Carlos Wall 2,1:2:9 , Fair Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall . . . . . . . . . . . . . . . . . 153 B.1 Load Deection Curve for Static Push Test Wall Specimen 3 from (Doherty 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 B.2 Load Deection Curve for Static Push Test Wall Specimen 12 from (Doherty 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 B.3 Load Deection Curve for Static Push Test Wall Specimen 13 from (Doherty 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 C.1 Peak load CDF1 for a ft = 0.4MPa and COV = 0.1 . . . . . . . . . . . 163 C.2 Peak load CDF1 for a ft = 0.4MPa and COV = 0.3 . . . . . . . . . . . 163 C.3 Peak load CDF1 for a ft = 0.4MPa and COV = 0.5 . . . . . . . . . . . 164

xvii

List of Tables
3.1 3.2 3.3 3.4 3.5 Summary of mortar type and mason quality for the experimental walls . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Location of Mortar Batches in the Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Bond Strength Values (MPa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Statistical parameters for NA wall data sets . . . . . . . . . . . . . . . . . Within and between course statistics for adjacent unit correlation (i.e when k=1 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 4.2 4.3 Summary of dimensions used in the 2D and 3D FEA model test wall . . . Summary of 2D and 3D FEA element type and mesh selection for the test wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of linear/elastic approximation (rst cracking load) compared with 2D and 3D FEA predictions using the structural conguration shown in 4.1 and 4.2, (taking a wall length of 960 mm for the 3D model)(no self-weight) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Summary of results for the two interface material models examined, with a tensile fracture energy Gf I of 0.012 N/mm (suggested from Lourenco (1996b)), using the 2D non-linear FEA model shown in Figure 4.1 (no self-weight) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Varying tensile fracture energy for specic unit exural bond strengths to examine if there is an upper limit on peak load with increased tensile fracture energy (no self-weight) . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Summary of dimensions used in the 3D non-linear FEA to model the three Doherty (2000) walls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 55 53 47 44 38 42 28 33 27 25

xviii

LIST OF TABLES

4.7 4.8 4.9

Summary of results from (Doherty 2000) for the uncracked static push tested walls (P indicates results are from prism bond wrench tests) . . . . Summary of FEA element selection for the walls in Doherty (2000) . . . . Summary of failure loads for Dohertys uncracked static push tests and the 3D FEA model with two dierent interface material models . . . . . . . . 71 75 65 69

4.10 Predicted failure loads using various approximations . . . . . . . . . . . . 5.1 5.2 5.3 5.4 5.5 Statistical parameters to be used to obtain failure load distributions for the 3D FEA model of the full wall . . . . . . . . . . . . . . . . . . . . . . Summary of wall dimensions and details used for the 3D FEA model for the full sized wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of wall failure load means and COVs for the full wall with spatially varying unit exural bond strength . . . . . . . . . . . . . . . . . .

76 81 84 89 99

Summary of peak wall failure load means and COVs for the full wall with non-spatially generated unit exural bond strengths . . . . . . . . . . . . Summary of wall failure load using the weakest link, averaging and load redistribution hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . .

6.1

Unit exural bond strengths (in MPa) for course 16 and 17 in Figure 6.13 119

A.1 Newcastle Wall 1, 1:1:6, High Quality Mason, Location of Mortar Batches in the Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 A.2 Newcastle Wall 1, 1:1:6, High Quality Mason, Bond Strength Values (MPa) 134 A.3 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Location of Mortar Batches in the Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 A.4 Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Bond Strength Values (MPa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 A.5 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Location of Mortar Batches in the Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 A.6 Newcastle Wall 3, 1:1:6, Fair Quality Mason, Bond Strength Values (MPa) 136 A.7 Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Location of Mortar Batches in the Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 A.8 Newcastle Wall 4, 1:1.5:6 + a.e. , Fair Quality Mason, Bond Strength Values (MPa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

xix

LIST OF TABLES

B.1 Summary of material parameter values to be used for the FEA wall analysis using the non-linear tension softening interface elements . . . . . . . . . . 154 B.2 Summary of material parameter values to be used for the FEA wall analysis using the composite interface elements . . . . . . . . . . . . . . . . . . . . 155 B.3 Summary of material parameter values to be used in the 3D FEA analysis of the walls tested by Doherty (2000) . . . . . . . . . . . . . . . . . . . . . 158 B.4 Wall loading and analysis specications for uncracked static-push tests from Doherty (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 C.1 Summary of FEA element type and mesh selection for the full wall . . . . 160 C.2 Summary of material parameter values to be used in the 3D FEA analysis of the full wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 C.3 Boundary conditions for the full wall . . . . . . . . . . . . . . . . . . . . . 162 C.4 Wall analysis specications for the full wall . . . . . . . . . . . . . . . . . 162

xx

Chapter 1

Introduction
Masonry, or more simply, bricks with or without mortar has been one of the most widespread used construction materials all over the world for thousands of years. From the pyramids in Egypt to the Great Wall of China, many large and complex structures, lasting thousands of years, were built based upon the simple idea of placing two stones or bricks beside of, or on top of one another and binding them together with a suitable material (or using friction between the two adjoining surfaces) to produce structures which could provide long term protection from the elements. Sun baked clay bricks with additions of chopped straw or grass, which were used in building construction thousands of years ago, are still used today in several developing regions of the world. Examples of this can be found in parts of Africa, Pakistan and India (see Figure 1.1). Thousands of years ago, some civilizations didnt use mortar beds. This can be seen in some of the building ruins still standing today in Cuzco, and some of the Roman aqueducts, where it can be seen that builders and masons instead opted to shape the stones in such a way so that they t intricately together, the bond being formed by the friction between the two adjacent stone faces (see Figure 1.2). Some of the advantages of clay brick masonry include its longevity, ease of construction, insulation characteristics and aesthetic appeal. Masonry is inherently strong in compression. This characteristic was used by architects and engineers of the past to build amazing structures such as Roman domes and aqueducts (see Figure 1.3) and Gothic cathedrals (Bassetti 2000). However, masonry is inherently weak when subject to forces creating

Figure 1.1: Indians using brick construction methods similar to those used thousands of years ago (Internet 2008b)

tension in the structure, such as those forces created by placing lateral loads on the face of a masonry wall. Tensile properties of masonry can be improved by reinforcing the structure, but this is generally not as simple to do when clay bricks are used and is therefore often limited to use in concrete block masonry construction. Reinforcing masonry walls also signicantly adds to the cost of building with masonry. The research herein examines spatial variability in unit exural bond strength and its eect on failure load variability and behaviour of modern day clay brick unreinforced masonry walls (URM) subject to lateral loads which cause vertical bending. Due to the complex nature of the structural interaction between an array of bricks and mortar and its relatively unpredictable overall behaviour, in the past, it has proven difcult to develop a rational based mathematical formulation to reliably predict the behaviour and strength of masonry, especially for masonry walls required to resist lateral loads. Therefore, although this has been been achieved quite well for other building ma-

Figure 1.2: Inca structures in Peru absent of mortar, opting instead to accurately cut stones and create bond through close packed friction (Internet 2008a)

terials such as steel and concrete, current analytical formulations for URM walls subject to lateral loading still possess limitations and shortcomings. Some of the most notable insight into the behaviour and formulations for clay brick URM walls in exure are works by (Lawrence 1983), Willis (2004), Baker (1981), Candy (1988) and Han (2007). Design methods currently used to design unreinforced masonry walls subject to lateral loading have not been developed from reliability-based calibration methods but rather calibrated to past experience. Although it is commonly believed that current design methods are conservative, the actual level of safety of masonry structures is unknown. It is therefore believed that designing and building with structural masonry in developed nations using current design specications often leads to more material and therefore a higher relative cost required to build with masonry, than say, with other construction materials like concrete. If one could gain a better understanding and more accurate and reliable techniques for assessing the performance, safety and reliability of masonry, the cost of construction with

Figure 1.3: Roman aqueducts appear to be absent of mortar, rather utilizing the surface of the accurately cut stones to create bond through joint friction (Internet 2008c)

masonry may reduce. Also the overall impact that masonry material manufacturing and transportation of masonry products has on the environment could reduce. Unless of course the demand for masonry increases as a result of a more ecient use of the material, in which case brick and mortar companies stand to increase prots signicantly. On top of mechanics based complexities of modelling and predicting masonry strength in exure, another added complexity is that masonry components, the bricks and mortar, are known to be highly variable in their strength properties, even amongst one type of brick and mortar. For example, unit exural bond strength, one of the most important local mechanisms contributing to the moment capacity of clay brick URM walls in exure (Willis et al. 2004b), is known to possess coecients of variation (COV)s ranging from 0.1 to 0.5 (for example see McNeilly et al. (1996)). Therefore, despite the fact that often only one type of brick and mortar is used in the construction of a masonry structure, it

has also been dicult to be condent about the actual strength of the overall structure, because of the degree of spatial strength variability in masonry materials. Another issue which has been a limitation to the ability to accurately predict strength of URM walls in exure is that these highly variable strength properties of bricks and mortar have proven dicult to measure accurately and reliably. Small specimen test results such as those used to measure material modulus of elasticity, unit bond strength and mortar compressive strength which are often used to dene material parameters used in the analysis of masonry are thrawt with inconsistencies and rarely represent the strengths exibited in situ (for example, see Correa et al. (2008)). However, an encouraging insight made by Ellingwood (1981) noted that large size masonry structures have been found to behave more uniformly than prisms and wallettes and are less aected by small variations in workmanship and materials. With recent advances in computer technology and numerical methods which aim to deal with complex mechanical interactions between composite materials such as those found in masonry, it is now possible to examine the issue of both wall behaviour, combined with spatial variability in material parameters and their eect on overall wall strength. In particular, it is hoped that using the latest developments in 3D non-linear nite element analysis (FEA) which can be used to model masonry reasonably well, will provide fruitful insight into issue of masonry strength variability due to material variability. Previous methods which attempted to examine the eect of spatial variability in material properties on wall failure load did not attempt to, or could not accurately capture the non-linear aspect of the behaviour of clay brick URM walls combined with the spatial variability of material properties. Although this thesis also hopes to provide insight into the meso-level (local stress) behaviour of clay brick URM walls in vertical bending, the main focus of the thesis is to examine how variability of the most signicant material parameter contributing to the strength of walls in exure aects the variability in wall strength. Through a comprehensive literature review, this material parameter is the unit exural bond strength (ft ).

Previously, various analytical studies using a variety of approaches were conducted to investigate the issue of spatial variability of masonry material parameters and their aect on the strength of clay brick URM walls in exure. Notable were those conducted by Lawrence and Cao (1988) and Stewart and Lawrence (2002). In both cases, linear approximations using various methods were employed to predict rst cracking load of walls in exure with spatial variability of certain material parameters. These results indicate that spatial variability in unit exural bond strength in particular, can have a signicant eect on the strength and reliability of a wall. Lawrence and Cao (1988) commented, that based upon the ndings in their research, they believed it important to take spatial variability of material parameters into account when designing and analysing unreinforced structural masonry. Stewart and Lawrence (2002) also showed the signicant eect that unit-to-unit spatial variability of certain masonry material properties can have on exural strength prediction and structural reliability. Lawrence (1991) suggested that assuming uncorrelated spatial variability of unit exural bond strength provides predictive strength results consistent with those found in experiments. Nonetheless, there are some who believe that, logically, variation in material properties between batches of one type of brick and mortar may transfer into spatial variation in strength properties in the masonry structure so that it is possible to witness regions of strength and weakness in the wall, correlated with a particular batch of bricks or mortar. Before examining how spatial variability of unit exural bond strength can aect the variability of failure load in a clay brick URM wall in vertical bending using 3D non-linear FEA analysis, one rst needs to establish whether or how unit exural bond strengths are spatially correlated. That is, due to dierent mortar and brick batches being used in a masonry structure, are there spatial regions in the structure with consistently low or high strength masonry, due to batch dierences. The answer to this question is important to establish because it can potentially provide insight into where one can expect a wall to fail and therefore, assist one to better model and predict wall behaviour and magnitude of failure load.

To achieve this objective, six full sized clay brick URM walls were constructed using a variety of mason workmanship. Bricks of one type were used but the mortar types were varied. Changes in mortar batch locations were duly noted during construction. Then unit exural bond strengths for each unit in all walls were tested. This provided a matrix of unit exural bond strengths for each wall and with this information, issues of spatial variability in bond strength and best t probability distributions for the bond strength variability could be examined. The experimental program, analysis and results are presented in Chapter 3. Once an understanding of how unit exural bond strength spatially varies in a wall due to mortar batch and workmanship issues, the next step was to try to implement this variability into a 3D non-linear FEA model, and utilise stochastic modelling in the form of Monte Carlo simulations to see how bond strength variability aects wall behaviour and strength. In order to do this, a chosen FEA package needed to be evaluated for its ability to model and capture wall behaviour and failure load. It was also important to establish whether spatial variability of material parameters, and in particular, unit exural bond strength could be implemented into the model. It was also important to see if the model remained stable and produced meaningful results with the inclusion of bond strength spatial variability. The model then needed to be validated against reasonable experimental results in order to be sure the model was accurately predicting failure load. This process and results are presented in Chapter 4. The next step in the investigative process was to create a full sized clay brick URM wall using 3D non-linear FEA, create realistic boundary conditions (in the context of a wall in vertical bending), material properties and variability in bond strength, and by varying the COV of bond strength, examine how this aected the variability of wall strength and behaviour. These results were then compared with previous proposals for methods pertaining to how and at what strength the wall is expected to fail. Observed crack pattern development from the 3D non-linear FEA wall model results were also presented and discussed. The results from this part of the thesis are presented in Chapters 5 and 6. Chapter 2 presents a literature review and Chapter 7 discusses conclusions, recommendations and suggestions for future research.

Chapter 2

Literature Review
The literature review presents previous research on topics related to the masters research objectives. It covers topics including general research into how URM wall behaviour is aected by material property variability, factors aecting the strength of URM walls in exure, random variation of material properties in brickwork, experimental data for URM walls in exure and appropriate numerical models available for modelling URM walls in exure.

2.1

Understanding URM Wall Behaviour in Terms of Material Parameter Variability

Baker (1981) states that up until 1980, no satisfactory general theory which adequately predicted the lateral strength of masonry panels in terms of that which reected the real behaviour of the material had been developed. Therefore Baker (1981) attempted to arrive at a theoretical understanding of the behaviour of walls under lateral load in terms of basic material properties. He examined masonry panels in vertical and two-way bending. In particular, he attempted to account for the variability of exural properties of units in a single leaf masonry wall panel and predict not only the mean strength that can be expected for a given panel conguration but also the variation in panel strengths. He showed that although the primary action of simply supported, laterally loaded masonry panels is assumed to be exural, other secondary eects associated with scale, loading, self weight, arching, rotational restraint at supports and translational yielding at supports may be

2.1 Understanding URM Wall Behaviour in Terms of Material Parameter Variability

present and can potentially aect the strength of the panel quite signicantly given that the exural strength of masonry is low. Baker (1981) states that the literature survey conducted as part of the Ph.D. showed that the most basic material property determining the resistance of masonry panels to lateral (out-of-plane) loads is the exural bond strength. In testing the variability of exural bond strength, he conducted a Kilmogorov-Smirnov test at the 10% level and found that the normal distribution adequately describes the statistical variability of joints in exure. For vertical exure investigations, Baker (1981) found that a partially plastic failure criterion which assumed that load sharing took place over 3 or 4 adjacent bricks closely predicted the mean and coecient of variation of panel strengths. It also explained the observed failure mechanism. He also states that fully plastic behaviour did not occur in these ideal circumstances where the applied moment was uniform along the potential yield-line and concludes that yield line theory is therefore inappropriate to describe the actual behaviour of panels, at least where a full contribution from yielding in the bed joints is assumed. Ellingwood (1981) summarises available strength data for unreinforced masonry load bearing walls in compression and bending and illustrates, through reliability calculations, how masonry compares with other engineering construction materials. The paper states that although unit strength and full-size elements are correlated, the strength of structural elements is substantially higher than would be predicted on the basis of the strength of the mortar. Ellingwood (1981) states that full-sized structural elements usually behave more uniformly than prisms and wallettes and are less aected by small variations in workmanship and materials. Ellingwood (1981) also mentions that laboratory tests, with their carefully controlled workmanship, curing conditions etc., tend to exhibit less variability in performance and comments that alignment of walls, thicknesses of mortar joints and completeness of joints, particularly hollow core units, are more dicult to control in the eld. Ellingwood (1981) focused on a reliability analysis for non-reinforced eccentrically loaded walls. Ellingwood (1981) found that the criteria for the design of unreinforced masonry walls are much more conservative when the vertical load eccentricity is small. However, he found that the reliabilities appear to be comparable in situations where there

2.2 Factors Aecting the Strength of URM Walls

is substantial moment applied to the wall either as a result of eccentric axial load or lateral forces. Lawrence (1983) considered various proposed methods of analysis for approximating the cracking and ultimate loads observed for full-scale two-way spanning wall panels. Lawrence (1983) states that an approach by Baker which involved the consideration of elastic principle moments and a Monte Carlo approach to the problem of variability made good agreement with ultimate loads but was in poor agreement for cracking loads, with no apparent explanation. He states that this method was the most promising of those he examined, which included elastic plate theory, yield line theory and Bakers empirical strip method.

2.2

Factors Aecting the Strength of URM Walls

There are several material parameters and site factors which have been identied as inuencing the strength of clay brick URM walls in exure. This section of the literature review documents some of these ndings and also aims to understand at the chemical level, through a review of the work by Sugo (2001a), Sugo (2001b) and Sugo (2007) how the brick/mortar bond develops. This was done to facilitate the identication of important material parameters which could signicantly aect the strength of URM walls in exure.

2.2.1

Brick/Mortar Interface

Sugo (2001a) over ve years conducted a detailed study of brick/mortar bond by examining the micro-structure of broken samples using both optical and scanning electron microscopy with polished sections across the joint. He found that both macro and micro scale factors control bond strength. He found that capillary suction by the brick causes the transport of cementitious material to the interface. He found that the formation of a coherent layer of paste at the interface ensures mortar-to-brick contact at the macro level. He found that the de-watering process also increases the packing density of the cement particles allowing the hydration products to form a dense structure at the micro level,

10

2.2 Factors Aecting the Strength of URM Walls

increasing the cohesive strength of the paste at the interface and within the bulk of the mortar. Sugo (2001a) states that suction of the mortar uids and associated transport of solids to the brick/mortar interface form an important role in the development of bond, but states that the interaction observed between brick and mortar during the experiments revealed that basic tests like IRA are of limited use when trying to predict bond strength. He suggests a more complex model incorporating the unit suction characteristics, rheology of the paste and its suction properties is required to assess mortar-to-brick compatibility.

2.2.1.1

Eect of Moisture and Ambient Temperature on Unit Flexural Bond Strength

Sugo (2001a) states that during the mixing of the mortar, the contact of Portland cement and water lead to the rapid hydration of the tricalcium aluminate compounds forming ettringite. The hydration of the tricalcium silicate and dicalcium silicate compounds does not occur till the end of the dormant period. For general purpose Portland cement the dormant period may last 4-6 hours under normal ambient conditions. He explains that the rapid increase in the rate of formation of these two compounds at the end of the dormant period is the beginning of actual bond formation and that hydration of these products will continue until they are consumed or until hydration ceases due to insucient moisture availability. Discussions with Sugo (2007) revealed a few noteworthy observations summarised in the following paragraph. Firstly, if the ambient temperature is high, then the dormant period may be less than the average time of 4-6 hours. If the dormant period is short then the tricalcium aluminate compounds may not have time to completely hydrate before the hydration of tricalcium silicate and dicalcium silicate compounds commence. Under these conditions, this situation may lead to a lowering of unit exural bond strength over the time taken to use up that mortar batch. If, due to high ambient temperature, mortar dries out rapidly and is not reasonably retempered, not only may the hydration period be less than the average time, but due to

11

2.2 Factors Aecting the Strength of URM Walls

drying out of mortar, insucient moisture will exist to completely hydrate all cementitious compounds in the mortar and a low unit exural bond strength may result. Under any climatic conditions, if too much water is added to the mortar, this can also reduce the unit exural bond strength for that mortar batch. The reason for this is that the mortar does not remain dense enough so that the cement particles are in suciently close contact to one another in order to bond together. If a mortar batch has an initially below optimum moisture content and small amounts of water are added to the batch during the laying of the batch, and the ambient conditions are mild, then this can actually act to increase the unit exural bond strength over the time it takes to use up that batch. This is because additions of small amounts of water to a mortar batch which has an initially below optimum moisture content under mild ambient temperatures will have the eect of allowing the more complete hydration of the tricalcium aluminate compounds to occur during a relatively long dormant period, which will facilitate a stronger exural bond strength.

2.2.1.2

Eect of Air Entraining Agents and Hydrated Lime on Unit Flexural Bond Strength

Air entraining agents are widely used in Australia to impart plasticity to fresh mortar. Sugo (2001b) studied the eect of air entraining agents on unit bond strength and found that the use of air entraining agent within the manufacturers recommendations was observed to produce a slightly reduced bond strength. However, overdosing of air entraining agent was found to essentially reduce the contact layer of cemetitious material along the brick/mortar interface which reduced bond strength signicantly. The experiments also showed the the addition of hydrated lime is benecial to bond. The hydrated lime material increases workability, moisture retentivity and contributes to the volume of paste.

12

2.3 Random Variation of Material Properties in Brickwork

2.2.2

Site Factors

Hendry (1976) identies some of the site and workmanship factors which aect masonry performance. Included in the discussion are factors like incorrect proportioning and mixing of mortar, incorrect adjustment of suction rate of masonry units, incorrect jointing procedures, disturbance of units after laying, failure to build walls plumb and true to line and level, and failure to protect new work from the weather. The paper concludes by saying that the most important eect of workmanship on masonry performance is the incorrect adjustment of suction rate of low strength bricks and the second most important eect was stated as being a failure to ll bed joints.

2.3

Random Variation of Material Properties in Brickwork

Random variation of material and mechanical properties of brickwork has been well documented to exist. The following section reviews literature regarding those studies which aimed to measure and quantify random variation in brickwork material and mechanical properties.

2.3.1

Proof of Random Variation in Brickwork Properties

Part of Lawrence (1983) thesis was to examine behaviour of wall panels and small sections of brickwork in exure. Test results were presented for over 600 beam specimens, 900 shear specimens, 32 full-scale wall panels, and a large number of tests were conducted on brick and mortar properties. These small specimen tests were carefully executed laboratory tests, with the materials used in the tests being as constant as possible throughout. This investigation provided valuable insight into the nature of statistical variation in brickwork properties. Based upon ndings from this research, Lawrence (1983) showed that consideration of random variation in brickwork properties is essential for a rational analysis of wall panels, but that approximate methods may be used in certain circumstances for simplied design. Lawrence (1985) took the small specimen test results which were conducted as part of Lawrence (1983) and conducted a statistical analysis in an attempt to characterise the

13

2.3 Random Variation of Material Properties in Brickwork

probability distributions which most accurately represent the properties tested using the small specimen tests. Lawrence (1985) points out that the histograms of the data dier signicantly in the tails, where they can have a marked eect on analysis results and probability of failure calculations. For exural and shear properties, Lawrence (1985) points out that the normal distribution is the best t for the data. For brick modulus of rupture, a signicant negative skew forced the best t distribution to be that of the Weibull distribution. Lawrence (1985) proposes that the negative skew was caused by the presence of ne cracks between perforations in the bricks. The histogram for mortar compressive strength indicated a signicant positive skew. Therefore the log-normal distribution was recommended as the most appropriate form to describe the probability distribution of data. Through these results, the paper stresses the importance of taking into account the random variation in brickwork properties when conducting any design and analysis of masonry.

2.3.2

Modelling Random Variation in Masonry for Walls Under Lateral Loading

Baker and Franken (1976) presented a theory for determining the mean strength and variability of brickwork beams or panels in vertical bending, taking into account the mean unit exural strength and variability of joints. The theory postulates that failure conditions for a course exist over a certain number of adjacent units when the average tensile stress produced by the total applied bending moment over those joints, less the self-weight compressive stress, is equal to or greater than the average strength of that number of joints. Based upon the experimental tests that accompanied the research, it was proposed that joint strengths can be averaged over three adjacent units. This theory, for one way vertical bending was in good agreement with experimental results. Baker and Franken (1976) state that the theory and experiments indicate that the exural strength of small brickwork beams depend upon the size of the specimen and the loading conguration applied in the test. They also report that brickwork panels are less variable than the individual joint strengths. Lawrence and Cao (1988) conducted a study which used a method of analysis based on elastic plate theory for the prediction of rst cracking load for non-load bearing masonry

14

2.3 Random Variation of Material Properties in Brickwork

walls under lateral loading. The approach used Monte Carlo simulations to take into account random variation of unit exural bond strengths with an isotropic plate analysis to determine the bending moments in the wall. Two support congurations were examined, namely one with all four sides of the wall simply supported and the other with sides and bottom simply supported and top edge free. The deterministic analysis utilized grid points over the rectangular plate (the wall). For each point, random horizontal and vertical exural tensile strengths were assigned by sampling from a log-normal distribution of given mean and coecient of variation (COV). Comparing the random strengths with calculated horizontal, vertical and twisting bending moments for each point and applying a failure criterion lead to the determination of where rst cracking on the plate would occur as the load was progressively increased. The process was continued 1000 times to produce a cracking load probability distribution. Various failure criteria were applied. The results showed the eect of considering variability in strength, with a 50% drop in rst cracking load when the COV went from 0.05 to 0.40. Lawrence and Cao (1988) state that the principal moment/elliptical failure criteron (see Baker (1982)) signicantly underestimated cracking load. They state that experimental results fell within the band of analytical results for longer panels but the analytical predictions for panels shorter than 4 meters were conservative. Lawrence and Cao (1988) concludes that considering random variation in bond strength and size of brick units can partially explain the higher observed strengths of blockwork panels compared with brickwork of the same basic material strength. The paper also rearms the suggestion that it is essential to consider inherent random variation in brickwork in any analytical approach. Stewart and Lawrence (2002) conducted a study to calculate the structural reliability of typical masonry walls subject to vertical bending. They considered rst cracking and collapse. Reliability indicies were obtained for lateral wind loading of non-load bearing masonry panels considering variations in discretized masonry unit thickness, exural bond strength variability, structural conguration and load intensity. Unit-to-unit spatial variability of these parameters were implemented into the model using three dierent hypothesis, namely the weakest link, averaging and load redistribution hypothesis. Stewart and Lawrence (2002) proposed a limit state function for the j th course from the base based upon the exural bond capacity of the course, bending moment caused by a 50 year wind load for that course, the height of the wall, the section modulus and the bulk

15

2.3 Random Variation of Material Properties in Brickwork

density of the masonry. A probability of failure that load eects will exceed resistance across a course was provided by a Monte Carlo simulation. Stewart and Lawrence (2002) found that as the number of joints in a course increased, the mean wall strength and wall strength variability decreased. The study also found that as wall width increases, structural resistance increases and so reliability also increases. They also found that as wall height increases, the reliability index reduces. Stewart and Lawrence (2002) propose the reason for this is because with increasing wall height, the number of failure modes increases therefore increasing the possibility of failure of an entire course of bricks. Also noted was the fact that load intensity has a negligible eect on the reliability indicies. Through the reliability indicies obtained using the dierent failure mechanism hypotheses, the study highlighted the importance of correctly modelling the failure mechanism if a realistic assessment of reliability indicies are to be obtained. Fyfe et al. (1999) conducted a study with the objective to examine and quantify the effects of random variation in material properties of masonry on the strength of laterally loaded masonry through the application of a nite element macro model coupled with Monte Carlo simulations. The variable properties examined were random variation in bond tensile strength and the random variation in bed joint thickness. Each constituent was assumed to behave as an isotropic linear elastic-brittle material. For the deterministic model, a two-stage homogenisation (macro-modelling) technique was used in which the constituent materials (bricks and mortar joints) were smeared to form a homogeneous orthotropic material which captures the average mechanical response of the masonry. The FEA model used 20 noded isoparametric quadratic 3D nite elements and the failure criterion was based on a tension cut-o dened by F = 1 ft = 0, where 1 is stress at the node. Each wall panel, of size 1000 mm1000 mm110 mm was analysed under a uniformly distributed lateral load. The loading was applied incrementally via a load factor which represented a percentage of the total load. The random bond tensile strengths and bed joint thicknesses were assigned to each gauss point in the FEA mesh by sampling a normal distribution of each property with a given mean and COV. The exural bond strength mean and COV was calculated as a fraction (10%) of compressive strength. This came from relationships dened in Grimm (1974). Thirty realisations with random variations in the material properties were run. Conclusions reported a mean failure load and

16

2.3 Random Variation of Material Properties in Brickwork

a corresponding specic partial factor of safety m .

2.3.3

Modelling Random Variation in Masonry for Other Structural and Loading Congurations

Lawrence (1991) conducted probabilistic analyses using various theories including elastic beam theory, simple nite element analysis and stochastic analysis to examine predictions when spatial variability of unit exural bond strength was implemented into the models for masonry beams spanning across the bed joints, one-way spanning walls, beams spanning across the perpends and wall panels supported on three or four sides. These analytical results were compared with the various small-specimen tests from Lawrence (1983), which provided exural strength and elastic modulus results. Spatial variability of bond strength was implemented into the model using several failure hypothesis, namely the weakest link and averaging hypothesis. Lawrence (1991) proposes that the weakest-link hypothesis will lead to an estimate of strength based on the weakest single joint unit in each cross-section, and the averaging hypothesis will lead to an estimate of strength based on averaging all joint units across each cross-section. For beams in exure across the bed joints, it was found that for beams one brick wide, there was a marked reduction in predicted beam strength for high COV and that for a typical COV of 0.35, the mean beam strength was about 60% of the mean joint strength. The analysis found good agreement with the experimental results. For the vertically spanning beams two bricks in width, it was found that the weakest link hypothesis was an adequate method for predicting the strengths, whereas the averaging hypothesis does not give acceptable estimates. For predicted strength of one way spanning walls, it was found that a reduction in wall strength with increasing COV was more severe as the number of courses in the wall increased. Lawrence (1991) proposed that this was because of the greater likelihood of a weak joint in the region of high bending moment. Lawrence (1991) reports that the weakest link hypothesis is the better means of predicting the strengths of these narrow walls. For masonry beams in exure across the perpend joints (i.e. a horizontal beam), the study found that the initial cracking stress was highly dependent on variability in the perpend strengths and the ultimate strength was highly dependent upon variability in brick modulus of rupture.

17

2.4 Experimental Data of URM Walls in Flexure

Fyfe et al. (2000) extended the Fyfe et al. (1999) study by examining the individual effects on the strength reduction of masonry from misalignment in wall panels, excessive thickness of mortar joints and excessive variations in mortar thickness. These eects are described in terms of a partial factor of safety which accounts for the variability in workmanship quality and material properties, as outlined in Prestandard (1996). The same base model as was used in Fyfe et al. (1999) is used in this study and the failure of each wall is judged from the percentage of cracking (failed gauss points) which was assumed to occur at 45% of total gauss points. Mislalignment of a wall causing eccentricities were examined by considering several workmanship defect scenarios including a panel with a bow horizontally across the middle and a deviation at the top running linearly to the base of the panel. By comparing the failure loads of the various misaligned panels with a control panel, the relative strength reductions and hence a partial factor of safety could be determined. The study found that the largest eect on the reduction of wall strength came from an increase in bed joint thickness. Variability in bed joint thickness and panels being out of plumb did not produce overly large strength reductions in the panels.

2.4

Experimental Data of URM Walls in Flexure

In order to check the accuracy of the predictive ability of various 3D non-linear FEA masonry models, experimental test results of full scale walls in exure needed to be collected. The wall data found are discussed briey in this section. Edgall (1993) generously provided a database of experimental wall and wallet test results where the walls were subject to lateral loading. The database is relatively old and many details are missing like the size of the units, thickness of mortar joints and average bond strength from for example, prism tests. However, relevant information about loading conditions, size of the walls, type of brick used (e.g. clay, perforations) have been provided. In particular, there are several test results for walls subject to vertical one-way bending with no axial force. In Dohertys research thesis (Doherty 2000), he conducted three uncracked static push tests with no precompression on 950 mm 1485 mm clay brick unreinforced masonry

18

2.4 Experimental Data of URM Walls in Flexure

panels. The walls were laterally line loaded at close to mid-height of the wall (course 10 of 18 courses) using a hand pump driven hydraulic actuator. The top of the walls were restrained by a cornice type support. The base of each wall was situated on an embossed polythene membrane DPC connection. The test boundary conditions amounted to simple supports top and bottom. The reader is referred to Doherty (2000) for a detailed description of the test rig set up in which the walls were tested. He also conducted several material tests which accompany the wall tests. Other statistics recorded in Dohertys thesis for the three walls tested include peak load, deection, and masonry Youngs modulus. One of the most extensive masonry wall testing programs was that conducted by Lawrence (1983) in which 32 walls were tested with support congurations which represented those most commonly found in buildings. The walls were either 2.5 or 3 meters in height and varied in length between 2.5 and 6 meters. A single brick layer and uniform materials were used for all tests, to minimise the variability eects from these factors. The walls were uniformly laterally loaded using inatable bags of plastic or rubber. The results from the experimental program include crack patterns and pressures for three signicant stages of behaviour. These are the formation of rst crack, the formation of the full crack pattern, and ultimate failure. Four failure modes were also identied. In 2001, Grith et al. (2001) reported conducting tests on two full-scale clay brick URM walls to investigate the relative contributions to the exural capacity of URM walls from each masonry mechanism. One wall was 4.5 m 2.7 m and the other was 3.5 m 2.7 m. Each wall had dierent boundary conditions. Amongst other things, the paper reports load-displacement data and failure loads for each wall. Grith and Vaculik (2005) conducted static, uniformly distributed lateral load tests on eight clay brick URM walls, of which, six contained window openings. Some walls were pre-compression loaded. The top and bottom of all walls were simply supported, and the vertical edges were xed. The paper reports the ultimate strength of each wall and compares them with current AS 3700 - 2001 and Willis et al. (2004a) analytical design exural strengths.

19

2.5 Appropriate Numerical Models for Unreinforced Masonry in Flexure

2.5

Appropriate Numerical Models for Unreinforced Masonry in Flexure

In order to accurately simulate failure of a clay brick URM wall in out-of-plane bending and implement variability of material properties into the model, two commercial FEA software packages were reviewed for their appropriateness. DIANA is a tested, well proven nite element analysis software package with a reputation for handling dicult technical problems relating to design of concrete, steel, masonry, soil, rock and soil-structure interaction. The program includes material, element and procedure libraries based on advanced database techniques, linear and non-linear capabilities and full 2D and 3D modelling features. In particular, it has a well developed capability to analyse masonry structures, at both the meso and macro levels. The theory used to develop DIANAs nite element handling of masonry structures is based on the theory developed in Lourenco (1996a). At the meso-level, the bricks are modelled with continuum elements and the joints with interface elements. The brick behaviour can be described with an elastic or visco-elastic model, and the interface behaviour can be described using either a discrete crack, coulomb friction or a combined coulomb friction/tension cut-o/compression cap model, or using a user supplied subroutine. The only current shortcoming of using the DIANA package to model masonry mesoscopically in 3D, is that no three-dimensional compression cap is implemented. At the macro-level, the masonry is modelled with continuum elements, with the orthotropic nature of the masonry being modelled with the anisotropic Rankine-Hill plasticity model. Other additions take into account rate-dependent crack behaviour, shrinkage and visco-elasticity. Han (2007) used the ABAQUAS commercial nite element package to model large scale masonry walls subject to out-of-plane bending for various boundary conditions. A simplied micro-model (meso-level) was used whereby the bricks units were expanded in dimension and modelled with continuum elements. The behaviour of the mortar, and the bond between brick and mortar were modelled using non-linear interface elements with tangential interaction being dened by a Mohr-Coulomb rule. Normal interaction was represented as elastic brittle under tensile forces and innite capacity under compression.

20

2.5 Appropriate Numerical Models for Unreinforced Masonry in Flexure

Shear strength was assumed to soften linearly.

21

Chapter 3

Experimental Program
As stated previously, the rst part of the research project was to conduct an experimental program to examine unit-to-unit spatial strength correlation within six full sized clay brick unreinforced masonry walls and to characterise a unit exural bond strength probability distribution. The experimental program and results are presented in detail in this section and accompanying information and results are presented in Appendix A.

3.1

Wall Materials and Construction

For the experimental program, two sets of experiments were conducted. The rst set, conducted in Newcastle, Australia (NA), used four single leaf clay brick URM walls, each with dimensions of 4 m long and 2.4 m high. The second set of experiments were conducted in Sao Carlos, Brazil (SCB), and used two single leaf clay brick URM walls both with dimensions of 4 m long and 2.5 m high.

3.1.1

Newcastle, Australia Experiments

The four walls were built with one type of brick; namely an extruded clay brick with no holes and the absence of any type of indentation. The brick dimensions were nominally 230 mm long 110 mm wide 76 mm high and the mortar joints nominally 10 mm in thickness (standard practice in Australia). Each wall was 16.5 bricks in length and 28

22

3.1 Wall Materials and Construction

courses high. Two types of mortar were used. Walls NA1 and NA3 (Newcastle, Australia, walls 1 and 3) were constructed using standard 1:1:6 (cement:lime:sand proportions by volume) mortar with no additives (Type M3 mortar in AS3700 (2001)). Walls NA2 and NA4 were constructed with a non-standard mortar mix (1.5:1:6 + Mortar Mate air entrainer (see http://www.geocel.co.uk)). This mortar is used extensively by masons in the Newcastle region in order to increase the workability of mortar; as beach sand sourced from the region is notorious for its narrow grading, which makes the mortar dicult to work. All four walls were constructed by professional masons. Walls NA1 and NA2 were constructed by a reputedly high quality mason and walls NA3 and NA4 were constructed by a fair quality mason. The walls were built sequentially in the laboratory (indoors) over a one month period. Each wall was constructed to a height of 1.6 m the rst day, and to completion the second day. As per Australian practice, two or three mud-boards were located equidistance along the proposed length of the wall. A mud-board is a at board on which shovels of mortar from the initial mixing tub of mortar are placed, to facilitate ready access to the mortar. In the Newcastle built walls, because dierent amounts of water were added to each mud-board in the process of retempering the mortar, this produced sub-batches from the main mortar batch, with a sub-batch being the mortar with a unique amount of water, located on a mud-board.

3.1.2

Sao Carlos, Brazil Experiments

Two walls were made with one type of brick; namely a pressed clay brick containing a frog on one face. The brick dimensions were nominally 210 mm long 100 mm wide 45 mm high and the mortar joints nominally 15 mm in thickness (standard practice in Brazil). Each wall was 17.5 bricks in length and 40 courses high. One type of mortar was used to build the two walls, namely a 1:2:9 mortar with no additives. The 1:2:9 mortar was chosen for its low average exural bond strength to ensure joint rather than brick failures during bond wrench testing. Clay bricks manufactured in Brazil are very low in strength, hence the need for low strength mortar. Wall SCB1

23

3.2 Method of Testing

(Sao Carlos, Brazil, wall 1) was constructed by a reputedly high quality mason and wall SCB2 was constructed by a fair quality mason. The two walls were constructed outdoors, simultaneously over a ve day period. Approximately ten courses of each wall were constructed per day. Mud-boards were not used during construction, so the mason trowelled mortar from the same tub of mortar. Figure 3.1 shows walls SCB1 and SCB2.

Figure 3.1: Walls SCB1 and SCB2 constructed in Sao Carlos, Brazil

Table 3.1 gives a summary of mortar types and mason quality used to build each of the experimental walls.

3.2

Method of Testing

The Newcastle, Australia walls were cured for a minimum period of 28 days, but no more than 40 days before they were tested. The Sao Carlos, Brazil walls were cured for 21

24

3.2 Method of Testing

Location Newcastle, Australia

Wall Reference Number NA1 NA2 NA3 NA4 SCB1 SCB2

Mortar Type 1:1:6 1:1.5:6 plus a.e. 1:1:6 1:1.5:6 plus a.e. 1:2:9 1:2:9

Mason Quality High High Fair Fair High Fair

Sao Carlos, Brazil

Table 3.1: Summary of mortar type and mason quality for the experimental walls days before testing. The exural bond strength for each unit in each wall was then tested using the bond wrench test in accordance with AS3700 (2001) (see Figure 3.2). All units in one wall were tested before moving onto the next wall. For each wall, starting at the top course of the wall, the perpend (vertical) joints were cut through the depth of the course, down to the top face of the brick on the underlying course using a hack saw. The bond wrench was then used to remove each brick, one by one, recording its exural bond strength and location in the wall. Once the top course was removed, the process was repeated for the second course from the top, and so on, until the entire wall was dismantled. The unit exural bond strengths for the bottom course of units attached to the underlying concrete base were not recorded as the exural bond strength for those units would not have been representative of a common unit exural bond strength due to the dierence in extruded clay brick and concrete suction properties. During the bond wrench testing of a course, the course beneath it was not restrained. However, in the SCB tests, it was found that in a small number of cases, the joint of the unit being tested did not fail but rather a small section of wall which included several bricks failed. Still pertaining to the SCB tests, in other units, it was found that the exural bond strength of the joint was so low that the self-weight of the bond wrench could not be supported by the joint and therefore the joint failed before a reading could be taken. Tables 3.2 and 3.3 show an example of the results obtained from the wall tests. Table 3.2 shows the location of the mortar batches for wall NA2. Batch 111 refers to batch one, mud-board 1, start of batch 1. Batch 112 refers to batch 1, mud-board 1, rst retempering of mortar on mud-board 1. Table 3.3 shows the unit exural bond strength

25

3.3 Spatial Correlation Method

Figure 3.2: Bond wrench testing units in wall SCB2

values (in N/mm2 ) obtained from the bond wrench tests on wall NA2. Similar results for the other test walls can be found in Appendix A.

3.3

Spatial Correlation Method

Generally speaking, correlation indicates the strength and direction of a linear relationship between two random variables. In order to conduct a spatial correlation analysis of units within and between courses in all of the six walls tested, each course of a wall was considered as a series of unit exural bond strengths. The mathematical tool used to measure spatial correlation for a series of equispaced data points in the space domain is the autocorrelation function k , which is calculated for lag k using all pairs of values in a series separated by k. The number of units separating two points in a series is dened by k and can take on a value in the integer set 0, 1, 2 . . . , (N 2), (N 1), where N equals

26

3.3 Spatial Correlation Method

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111

Unit 2 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111

Unit 3 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111

Unit 4 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111

Unit 5 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 221 112 112 112 111 111 111 121 111 111

Unit 6 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 221 112 112 112 111 111 111 121 121 111

Unit 7 411 411 322 312 312 311 311 311 311 311 311 222 212 212 211 211 211 221 221 112 112 112 121 121 121 121 121 111

Unit 8 411 411 322 312 312 311 311 311 311 311 311 222 212 212 211 211 221 221 133 112 122 112 121 121 121 121 121 111

Unit 9 411 411 322 312 312 311 321 311 311 311 311 222 222 212 211 221 221 221 133 112 122 122 121 121 121 121 121 121

Unit 10 421 411 322 312 322 321 321 321 311 321 311 222 222 212 211 221 221 221 133 112 122 122 121 121 121 121 121 121

Unit 11 421 411 322 312 322 321 321 321 311 321 321 222 222 212 211 221 221 221 133 112 122 122 121 121 121 131 121 121

Unit 12 421 411 322 322 322 321 321 321 311 321 321 222 222 222 221 221 221 221 133 112 122 122 121 121 131 131 121 121

Unit 13 421 421 322 322 322 321 321 321 321 321 321 222 232 222 221 221 221 221 133 112 122 122 121 121 131 131 121 121

Unit 14 421 421 322 322 322 321 321 321 321 321 321 222 232 222 231 221 221 221 133 112 122 122 121 121 131 131 121 121

Unit 15 421 421 322 322 322 321 321 321 321 321 321 222 232 232 231 231 231 221 133 112 122 122 121 131 131 131 121 131

Unit 16 421 421 322 322 322 321 321 321 321 321 321 232 232 232 231 231 231 231 133 112 132 132 131 131 131 131 131 131

Unit 17 421 421 322 322 322 321 321 321 321 321 321 232 232 232 231 231 231 231 133 112 132 132 131 131 131 131 131 131

Table 3.2: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Location of Mortar Batches in the Wall the number of equispaced data points in the series. Intuitively, the autocorrelation function k is interpreted as a measure of the similarity or degree of correlation between each value in the series and a value located k units away from it. The value for the autocorrelation function ranges between 1 and -1. A correlation function value of k = 1 indicates that each value in the series is perfectly correlated with a value in the series k units away from it. A correlation function value of k = 0 indicates that each value in the series is not correlated with a value in the series k units away from it (i.e. statistically independent). A correlation function value of k = 1 indicates that each value in the series is perfectly negatively correlated with a value in the series k units away from it. For a series of N observations z1 , . . . , zN measured at locations i1 iN , Priestley (1981) shows that there are a number of estimates for the autocorrelation function but states that the bias estimate is the most popular variance estimator in time series analysis. Moreover, Fenton (1999) states that the bias estimate produces a slightly smaller expected

27

3.3 Spatial Correlation Method

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 0.40 0.00 0.25 0.00 0.18 0.00 0.18 0.00 0.27 0.00 0.23 0.00 0.52 0.00 0.03 0.00 0.31 0.00 0.31 0.00 0.49 0.00 0.17 0.00 0.19 0.00 0.29 0.00

Unit 2 0.62 0.42 0.51 0.63 0.56 0.39 0.47 0.41 0.54 0.49 0.48 0.35 0.23 0.55 0.04 0.58 0.33 0.65 0.12 0.44 0.14 0.19 0.09 0.07 0.18 0.17 0.24 0.00

Unit 3 0.59 0.24 0.45 0.50 0.59 0.33 0.40 0.32 0.45 0.38 0.36 0.19 0.42 0.59 0.11 0.60 0.78 0.34 0.27 0.32 0.49 0.07 0.07 0.26 0.20 0.18 0.15 0.00

Unit 4 0.53 0.40 0.44 0.58 0.45 0.24 0.54 0.40 0.15 0.36 0.48 0.11 0.45 0.54 0.17 0.44 0.54 0.68 0.14 0.10 0.22 0.22 0.09 0.08 0.18 0.26 0.25 0.00

Unit 5 0.44 0.38 0.44 0.59 0.31 0.16 0.48 0.43 0.56 0.55 0.48 0.28 0.03 0.37 0.37 0.39 0.42 0.26 0.18 0.33 0.26 0.10 0.34 0.14 0.24 0.20 0.19 0.00

Unit 6 0.57 0.44 0.39 0.39 0.43 0.49 0.28 0.47 0.40 0.52 0.61 0.38 0.03 0.15 0.31 0.44 0.41 0.52 0.40 0.16 0.43 0.20 0.05 0.20 0.05 0.44 0.20 0.00

Unit 7 0.41 0.44 0.28 0.65 0.54 0.53 0.47 0.40 0.60 0.29 0.47 0.46 0.06 0.46 0.10 0.33 0.47 0.47 0.19 0.49 0.34 0.34 0.07 0.14 0.15 0.30 0.30 0.00

Unit 8 0.57 0.58 0.51 0.56 0.50 0.26 0.47 0.42 0.46 0.24 0.61 0.09 0.05 0.50 0.07 0.34 0.50 0.48 0.23 0.27 0.09 0.30 0.20 0.35 0.22 0.10 0.23 0.00

Unit 9 0.50 0.60 0.28 0.40 0.42 0.53 0.33 0.49 0.52 0.38 0.31 0.11 0.39 0.49 0.07 0.35 0.72 0.58 0.11 0.18 0.37 0.11 0.24 0.33 0.22 0.13 0.28 0.00

Unit 10 0.54 0.50 0.33 0.58 0.65 0.59 0.25 0.34 0.52 0.45 0.62 0.11 0.43 0.60 0.45 0.50 0.51 0.54 0.32 0.05 0.26 0.08 0.29 0.25 0.34 0.16 0.25 0.00

Unit 11 0.41 0.38 0.38 0.71 0.48 0.55 0.30 0.57 0.46 0.49 0.39 0.31 0.06 0.82 0.23 0.32 0.42 0.40 0.37 0.29 0.17 0.10 0.13 0.21 0.20 0.20 0.29 0.00

Unit 12 0.68 0.47 0.57 0.67 0.54 0.37 0.12 0.49 0.38 0.46 0.42 0.47 0.40 0.58 0.44 0.42 0.49 0.41 0.17 0.42 0.12 0.32 0.25 0.16 0.24 0.11 0.32 0.00

Unit 13 0.22 0.38 0.50 0.68 0.35 0.42 0.55 0.49 0.33 0.53 0.76 0.30 0.50 0.41 0.63 0.37 0.47 0.17 0.44 0.12 0.07 0.12 0.40 0.16 0.23 0.14 0.36 0.00

Unit 14 0.52 0.37 0.37 0.48 0.77 0.49 0.28 0.51 0.48 0.40 0.54 0.37 0.54 0.40 0.27 0.46 0.30 0.25 0.44 0.28 0.24 0.38 0.21 0.27 0.16 0.29 0.31 0.00

Unit 15 0.83 0.39 0.40 0.42 0.52 0.50 0.27 0.66 0.42 0.54 0.55 0.62 0.59 0.09 0.76 0.09 0.42 0.50 0.51 0.06 0.27 0.09 0.24 0.28 0.17 0.19 0.44 0.00

Unit 16 0.77 0.55 0.56 0.43 0.45 0.45 0.23 0.26 0.29 0.50 0.82 0.10 0.45 0.19 0.43 0.42 0.40 0.46 0.51 0.03 0.11 0.06 0.26 0.23 0.38 0.28 0.34 0.00

Unit 17 0.00 0.42 0.00 0.38 0.00 0.29 0.00 0.36 0.00 0.53 0.00 0.10 0.00 0.64 0.00 0.44 0.00 0.53 0.00 0.03 0.00 0.36 0.00 0.16 0.00 0.23 0.00 0.00

Table 3.3: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Bond Strength Values (MPa) error variance than that of the unbias case. For these reasons, the bias estimate of the autocorrelation function was used for the spatial correlation analysis. The bias estimate of the kth lag autocorrelation function value is dened by:

N k 1 N

(zi (z))(zi+k (z))

i=1

k =
1 N

, 1 N
N k

N k

k = 0, 1, 2, ..., N 1

(3.1)

(zi (z))2
i=1

(zi+k (z))2
i=1

where (z) is the mean of the series. When calculating autocorrelation function values for a course of unit exural bond strengths in a wall, N is the number of whole units in a course (e.g. for wall NA2, N = 16 for each course), z1 is equal to the rst unit exural bond strength value in the course and zN is equal to the last unit exural bond strength value in the course. When examining correlation between courses in a wall, the value N

28

3.3 Spatial Correlation Method

used in Equation 3.1 is equal to the total number of courses that were bond wrench tested in the wall (e.g. for wall NA2, N = 27), z1 is equal to the unit exural bond strength value at a certain distance along the length of the wall in the second course of the wall (note that course 1 bond strengths were not tested), zN is equal to the unit exural bond strength value at the same location but in the 28th course of the wall. Taking wall NA2 as an example, the correlation between unit bond strengths for each course in wall NA2 are represented by 27 corresponding correlograms, one for each course, and the correlation between bond strengths between courses in wall NA2 are represented by 16 correlograms. A correlogram is a plot of the autocorrelation function value k versus the lag k for a series. When examining correlation within a course of unit exural bond strengths in a wall, if (z) is taken to be the mean of the unit exural bond strengths (z1 , . . . , zN ) in the course, this mean will reect the variation amongst values in the course. Using this course mean to calculate the autocorrelation function value for lag k, will result in seeing a less than expected level of correlation for a course. This can also be seen by examining Equation 3.1 more closely. Observe that approximately half the bond strength values in a course are above the course mean and approximately half of the values are below the course mean. This could, in many instances, result in positive and negative terms in Equation 3.1 cancelling each other out to a large extent, providing a correlation function value close to zero. Therefore, to avoid this local mean eect when calculating the autocorrelation function values for within and between courses in a wall, the mean (z) was taken as the mean unit exural bond strength (unit ) for all the joints in the wall. Priestley (1981), Fenton (1999) and others suggest that autocorrelation function value estimates lying within a band of 2 1/N are deemed to be not signicantly statistically dierent from = 0. For the purpose of examining spatial correlation, the upper band (hereafter referred to as the limiting value for correlation) location of 2 1/N is plotted on each correlogram. This was used to establish if there exists unit exural bond strength correlation within a course after this band was taken into account. For example, given a course of unit exural bond strengths, if the autocorrelation function value falls below 2 1/N at k = 4, then units may be taken as being statistically correlated for up to 3 adjacent units.

29

3.4 Results

3.4

Results

The experimental testing program resulted in a set of data for each wall consisting of the location of the unit in the wall, its corresponding exural bond strength, and the batch from which the mortar for that unit came (see Appendix A, Section A.1). Walls NA1, NA2, NA3 and NA4 produced full wall data sets. However, as discussed earlier, the SCB1 and SCB2 wall data sets lack some unit exural bond strength values.

3.4.1

Probability Distribution for Unit Flexural Bond Strengths

A range of probability distributions were tted to wall NA1, NA2, NA3 and NA4 data sets. Note that a truncated Normal distribution instead of a Normal distribution was considered because unit exural bond strengths are relatively low (when measured in N/mm2 ) and cannot assume negative values. Due to the fact that SCB1 and SCB2 did not produce full data sets, it is not meaningful to calculate the statistical parameters for these walls. The maximum likelihood method was used to t the distributions to wall NA1, NA2, NA3 and NA4 unit exural bond strength data. The histogram and tted probability distributions for NA2 are shown in Figure 3.3. A Kolmogorov-Smirnov test at the 5% signicance level was also performed in order to check whether the exural bond strength data set for a wall could be obtained from any of the hypothesized distributions. An Inverse Cumulative Distribution Function (CDF1 ) plot was also used to infer a goodness-of-t for the probabilistic models. When the CDF1 of a particular probabilistic model sits on the 1:1 line, this indicates that the probabilistic model ts well to the data (see Figure 3.4 for wall NA2 CDF1 ). On the CDF1 plot, when a value is below the 1:1 line, the probability density is higher than that exhibited by the data. By overestimating the lower tail of the histogram (when the CDF1 values are plotted below the 1:1 line) for a material strength, the resulting probability of failure is over-estimated and therefore provides a conservative result (e.g. see Stewart and Lawrence (2007)). The statistical parameters for Newcastle walls NA1 to NA4 are summarised in Table 3.4. In the case of wall NA2, the Kolmogorov-Smirnov test at the 5% signicance level indicates that only the truncated Normal distribution could not be rejected as the underlying

30

3.4 Results

3.5 3 Truncated Normal Lognormal Weibull Gamma Gumbel

Probability Density

2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6

0.7

0.8

0.9

Unit Flexural Bond Strength (MPa)

Figure 3.3: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Probability Distribution Fits

distribution for the wall unit exural bond strengths. By examining the CDF1 plot for wall NA2 (Figure 3.4), the lower tail of the truncated Normally distributed histogram is below the 1:1 line, which provides a slightly conservative estimate for the probability of failure when this distribution is used to generate random variables of unit exural bond strength for a reliability analysis. For all four NA wall data sets, a truncated Normal distribution appeared to be the best t. The condence with which a truncated Normal distribution can be adopted to represent the variability in unit exural bond strength data is high because the number of data points in the sets used to obtain this distribution is large. The wall unit exural bond strength means and coecients of variation in Table 3.4 are within the limits reported by McNeilly et al. (1996) recorded for 19 building sites in Melbourne, Australia; where means varied from 0.21 N/mm2 to 0.85 N/mm2 and coecients of variation from 0.16 to 0.49. Using the truncated Normal distribution to represent variability in unit exural bond strength is consistent with ndings by Baker (1981). Lawrence (1983) also found that although the Lognormal, Normal and Weibull distributions were

31

3.4 Results

0.8

CDF(1) (Predicted Unit Flex. Bond Strength (MPa)

0.7

0.6

Perfect Fit Truncated Normal Lognormal Weibull Gamma Gumbel

0.5

0.4

0.3

0.2

0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Unit Flexural Bond Strength (MPa)

Figure 3.4: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, CDF1

not rejected for the Kolmogorov-Smirnov test at the 20% signicance level, the Normal distribution gave the lowest statistic and was used for subsequent analysis in that research. It is observed that walls NA1 and NA3 have higher wall mean unit exural bond strengths than do walls NA2 and NA4. Both NA1 and NA3 walls used 1:1:6 mortar with no additives, which is considered a high strength mortar (Sugo 2001a). An air entraining agent used in the mortar (1.5:1:6 + air entrainer) to build walls NA2 and NA4 has been shown to provide slightly lower mean unit exural bond strengths than in mortars which do not use an air entraining agent (Sugo 2001a). It is also observed that the mean unit exural bond strength for wall NA1, built by the high quality mason, is slightly less than for wall NA3, built by the fair quality mason. A similar result can be seen between walls NA2 and NA4. This result seems counter-intuitive. One would expect the high quality mason to produce a higher mean unit exural bond strength than that produced by the lesser quality mason. The reason for this result might be attributed to the fact that walls NA1

32

3.4 Results

Wall

Number of Data Points

Kolmogorov-Smirnov Test 5% Signicance Truncated Normal, Weibull Truncated Normal Truncated Normal Truncated Normal

Best Fit Distribution

Mean Unit Flexural Bond Strength (uunit ) (N/mm2 ) 0.51 0.36 0.56 0.40

Coecient of Variation (COV) 0.41 0.47 0.45 0.42

NA1 NA2 NA3 NA4

432 432 432 427

Truncated Normal Truncated Normal Truncated Normal Truncated Normal

Table 3.4: Statistical parameters for NA wall data sets and NA2 were built shortly after very heavy rainfall. The bricks used to build these two walls were exposed to the torrential downpour before being brought into the laboratory (indoors). The bricks used in walls NA3 and NA4 were left outside to dry in the ne weather conditions that followed the storm for several weeks afterwards. Due to the mild conditions within the laboratory, the bricks used in walls NA1 and NA2 may not have dried out to the extent that bricks used in walls NA3 and NA4 were able to. Using relatively moist bricks for walls NA1 and NA2 would have had the eect of reducing the suction potential between bricks and mortar, which if too low, can in some circumstances reduce the unit exural bond strength produced (Sugo 2007).

3.4.2

Wall Course Unit Flexural Bond Strength Fluctuations

Unit exural bond strength values for each course were examined for walls NA1, NA2, NA3, NA4, SCB1 and SCB2. Figure 3.5 shows, for each course of wall NA2, a plot of unit exural bond strength (in N/mm2 ) versus its location within a course. This was done in order to examine whether trends or periodicities in unit exural bond strength within a course could be seen. A trend might be seen in a series of unit exural bond strengths if, for example, the unit bond strengths become progressively weaker or stronger over the series. A periodicity might be observed in a series of unit exural bond strengths if, for example, unit bond strengths for a certain section of the series are consistently higher or lower than for another section of the series. It is evident from Figure 3.5 that no trends in unit exural bond strength existed within each course for this (or other) NA walls. It was expected that unit exural bond strength

33

3.4 Results

Figure 3.5: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa)

trends for a course within these walls was unlikely to be observed because of the several mortar sub-batches used to construct the units in each course of a wall. However, a trend in unit exural bond strength for each course is also not observed in the SCB walls, for which the same mortar batch was used to build a course, ambient conditions were sunny, hot and dry, and no retempering of the mortar was performed. Under such conditions, one might expect to see a decreasing trend in unit exural bond strength values over a course due the eects that aggressive ambient conditions can have on mortar and hence unit exural bond strengths (Bye 1983). However, as the approximate laying time for a course of bricks was 10 minutes, it is supposed that this time frame was too short to see any strength reductions. Periodicities along a course are also not consistently evident, as the unit exural bond strengths oscillated in a random fashion along each course. Course mean unit exural bond strengths were examined for walls NA1, NA2, NA3 and NA4. Figure 3.6 shows a plot of course number versus course mean unit exural bond strength (in N/mm2 ) for wall NA2. Similar plots for the other test walls can be found in Section A.1, Appendix A. Note that the horizontal lines located on each plot indicate the commencement of a new mortar batch. It is noted that walls NA1 and NA3 show a larger mean bond strength uctuation from course to course than do walls NA2 and NA4.

34

3.4 Results

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Average Unit Flexural Bond Strength (MPa)

Batch 4

Batch 3

Course Number

Batch 2

Batch 1

0.8

0.9

Figure 3.6: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strength for Each Course (MPa)

Mortar batches used in walls NA1 and NA3, built with 1:1:6 mortar (no additives), were retempered more frequently in an attempt to improve workability, than that which was necessary for the 1.5:1:6 + air entrainer mortar batches used to build walls NA2 and NA4. As masons found the 1:1:6 mortar more unworkable, it may have aected the rhythm with which they laid the units. The combination of these reasons may have contributed to the larger course mean uctuation seen in walls NA1 and NA3.

3.4.3

Spatial Correlation of Unit Flexural Bond Strengths

An autocorrelation analysis using Equation 3.1 was performed on units within each course and between courses for all walls. For walls SCB1 and SCB2, there were exural bond strengths for units located at various locations in the wall that were not recorded. When conducting a spatial correlation analysis on a course which contains non-recorded unit exural bond strength values, the following approach was taken. Within a course (series) of unit exural bond strength values, if there were non-recorded locations which isolated

35

3.4 Results

two or less values, these values were discarded from being included in the spatial correlation analysis for that course. If spatial correlation between unit exural bond strengths are to exist in a course, it would be most evident within 3 adjacent units, and a minimum of three adjacent values in succession are required to calculate the autocorrelation function value at k = 3.

Autocorrelation Coefficient (k)

0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4

Lag k

10

Figure 3.7: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall

The 27 course correlograms were plotted together for each wall and are shown in Section A.1, Appendix A, Figures A.21 to A.26. Between course correlograms are shown in Section A.1, Appendix A, Figures A.27 to A.32. The horizontal dotted line in each plot is the limiting value for correlation dened by 2 1/N for which correlations below this line are not signicant. The course and between course correlograms for Wall NA2 are conveniently shown in Figures 3.7 and 3.8 respectively. Initially, the percentage of courses and between courses in a wall which contained some adjacent unit (i.e. when k = 1 in Equation 3.1) correlation, and the wall average of k=1 for courses and between courses were examined to see if there was consistently observed correlation between adjacent unit exural bond strengths within a wall. If correlation between adjacent units was consistently seen throughout the walls, and autocorrelation

36

3.4 Results

Autocorrelation Coefficient (k)

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure 3.8: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall

function values at k = 1 were relatively high compared with other construction materials, then further autocorrelation lags were examined in more detail. If the autocorrelation function values at k = 1 were relatively low compared with other construction materials, and if correlated units within and between courses were not consistently seen and predictably locatable within the walls, then further autocorrelation lags were not examined in more detail, because for the sake of a probability or reliability analysis, unit exural bond strengths can be considered statistically independent for clay brick masonry walls. Weak adjacent unit correlation (i.e. k=1 > 2 1/N ) existed for between 16% and 48% of courses in the walls tested, and for between 0% and 44% of between courses in the walls tested. The average autocorrelation function value at a lag of k = 1 (i.e.k=1 ) for each wall ranged between 0.22 and 0.50 within courses and 0.09 and 0.36 between courses. Table 3.5 summarises the correlation statistics for adjacent unit correlation within and between courses. Note that for all but one wall, the average course and between course adjacent unit autocorrelation function values were less than the limiting value for correlation (2 1/N ). The one exception is marginal, as the average course adjacent unit autocorrelation function value in this case was only 0.01 greater than the limiting value for correlation. This result means that adjacent unit correlation was not signicant. Note

37

3.4 Results

that even an autocorrelation function value of 0.5 to 0.7 is not considered a strong correlation but rather a weak correlation.

Wall

Percentage of courses where k=1 > 2 1/N (%) 41% 37% 48% 41% 28% 16%

Average k=1 for courses in wall 0.44% 0.46% 0.50% 0.44% 0.26% 0.22%

Limiting value of correlation for courses in wall (%) 0.49% 0.49% 0.49% 0.49% 0.48% 0.48%

Percentage of between courses where k=1 > 2 1/N in wall 13% 0% 44% 37% 12% 6%

Average of k=1 between courses 0.24% 0.23% 0.36% 0.35% 0.12% 0.09%

Limiting value of correlation between courses in wall 0.38% 0.38% 0.38% 0.38% 0.34% 0.34%

NA1 NA2 NA3 NA4 SCB1 SCB2

Table 3.5: Within and between course statistics for adjacent unit correlation (i.e when k=1 ) The majority of courses which did show weak adjacent unit correlation were situated one course after another in the wall. Other weakly correlated courses within the walls were randomly situated. Between course weak adjacent unit correlation was not consistently situated at any particular location in the walls. During the construction of the walls, there was nothing in particular that was done dierently with the preparation of the mortar batches where sequential weak course correlation was found, compared with all the other batches. The weakly correlated courses did not contain unit exural bond strengths which were consistently higher or lower than the mean for the wall (e.g. weak course correlation was not only found in courses which had consistently high unit exural bond strengths within the course). Reasons for seeing a higher number of courses which contained weak correlation in the NA walls compared with the SCB walls might be due to the eect that diering ambient environmental conditions had on the mortar during construction, dierences in brick suction properties between the bricks used in Australia and Brazil, and the general quality of workmanship produced by the Australian masons compared with the Brazilian masons. It can also be observed that the conditions present for the building of the SBC walls (constructed outdoors) were closer to what one would expect to nd on most construction

38

3.5 Discussion of Experimental Results

sites, compared to those that existed for the construction of the NA walls (constructed in the laboratory). Given that on construction sites, there is a higher possibility for incorrect batch mixing and several masons with dierent skill levels all working on the one wall, and noting that there was a very low incidence of adjacent unit correlation within the SCB walls, one would expect to see reduced presence of unit exural bond strength correlation in walls built under normal construction conditions. A rst glimpse of the correlograms in Section A.1, Appendix A may lead one to conclude that unit exural bond strength correlations are high. However, upon closer examination, it has been shown that average adjacent unit correlation for the experimental masonry walls was low (between k=1 = 0.22 and k=1 = 0.50 for courses in a wall), with a low incidence rate (between 16% and 48% of courses in a wall) and unpredictably situated within a wall. For all but one wall which gave only a slightly higher average, the wall average adjacent unit correlations were less than the limiting value for correlation (2 1/N ), thus indicating statistically insignicant correlations. Given that controlled laboratory experiments showed relatively little adjacent unit correlation, for the purpose of conducting a probability or reliability analysis, it is recommended that exural bond strengths for all units in a masonry wall are statistically independent, with each unit exural bond strength being modelled as a truncated Normal distribution.

3.5

Discussion of Experimental Results

The most obvious and well known factors for high unit-to-unit exural bond strength variability (i.e. lack of spatial correlation) are changes in mortar and brick batch, frequency of water addition and the use of additives in the mortar, variations in weather conditions during construction, and variations in brick suction rates due to factors like location of each brick in the kiln during production and its location and exposure to weather conditions on a pallet. Other factors which are thought to have contributed to the high unit-to-unit exural bond strength variability witnessed in the experimental results were: The thickness of the mortar bed laid along a course and the number of taps applied to the brick in order to align the unit to the nished level may have aected its exural

39

3.5 Discussion of Experimental Results

bond strength. In practice, for each unit in a course, it is possible that the mason laid a marginally dierent thickness of mortar bed before adding each brick and tapped each brick a dierent number of times. In order to realign a brick already laid, the mason detached the brick from the mortar that it was initially bonded to. The bricks ability to eectively rebond with the mortar at the new location is thought to have diminished (Sugo 2007). Each brick on a pallet potentially possessed dierent amounts of surface dust due to reasons like varying exposure to weather conditions because of its location on a pallet, which in turn may have aected its suction rate. Elapsed times between mixing the mortar, spreading the mortar and laying the bricks onto the mortar bed may have aected each units exural bond strength. It would appear that brick suction rate, which, as stated previously can have a signicant inuence on the suction potential between bricks and mortar, can signicantly aect the overall strength of the wall. This can be seen by observing that although the fair quality mason was generally more reckless, both of his walls provided higher mean bond strengths than those built by the high quality mason. This is because the high quality mason used relatively moist bricks in the construction of his walls, which can lead to an overall decrease in bond strength due to the reduced suction potential between brick and mortar. It was also observed that although weak correlation in unit exural bond strength exists in some courses and between courses, these locations were dicult to predict and didnt follow any particular pattern relating to for example, mortar batch. Therefore, although somewhat counter-intuitive, the results indicate that statistically signicant correlation between adjacent unit exural bond strengths is not likely to be observed. It was also observed that clay brick wall unit exural bond strengths obtained for all of the walls tested best t a truncated Normal probability distribution.

40

Chapter 4

Developing the 3D Non-Linear FEA Model

This chapter discusses the development of the 3D non-linear FEA model and examines the ability of the chosen FEA package to accurately estimate wall failure for clay brick URM walls in vertical bending.

4.1

The 3D FEA Modelling Package

All 2D and 3D walls were modelled using TNO iDIANA 9.2 Finite Element Package (see www.tnodiana.com/). This FEA package was chosen because of its ability to model masonry using a simplied micro-modelling approach. It is considered the most advanced nite element package for the modelling of masonry available and is founded on sound theory and extensive experimental research (see Lourenco (1996a)). In the simplied micro-modelling approach, expanded bricks are represented by continuum elements and the behaviour of the mortar joints and the brick-mortar interface are captured by the interface elements (which have zero thickness in the FEA model). Using this approach, masonry is considered as a set of elastic blocks bonded by potential fracture/slip lines at the joints (Lourenco 1996a). Lourenco (1996a) states that accuracy is lost since Poissons eect of the mortar is not included. However, when modelling a relatively large size clay brick URM wall, this modelling approach makes for an acceptable compromise in terms of run time and accuracy, when compared to the detailed micro-modelling and

41

4.1 The 3D FEA Modelling Package

macro-modelling approaches. For a large size clay brick URM wall, the micro-modelling approach is too computationally expensive and in the macro-modelling approach, local stress eects are lost. The simplied micro-model also allows for the inclusion of variation in certain brick/mortar interface properties like unit exural bond strength and cohesive shear strength.

Wall Dimensions Number of full bricks per course Number of courses high Mortar joint thickness Brick dimensions Brick type

960 mm long x 1548 mm high 4 18 10 mm 230 mm x 110 mm x 76 mm clay brick extruded with no holes or indentations

Table 4.1: Summary of dimensions used in the 2D and 3D FEA model test wall

Figure 4.1: 2D FEA model test wall set up to compare with linear/elastic theory (no self-weight)

For the simplied micro-modelling approach, there are several interface material models which can be used to describe the behaviour usually exhibited in experiments at the

42

4.2 Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

brick/mortar interface. These constitutive relations include denitions for discrete cracking (including brittle cracking, linear tension softening and non-linear tension softening), crack dilatancy, bond-slip, coulomb friction and a combined crack-shearing-crushing model (also known as the composite interface model). Technical details regarding these dierent models can be found in Hendriks and Wolters (2007) and will not be reproduced here.

Figure 4.2: 3D FEA model test wall set up to compare with linear/elastic theory

4.2

Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

Initially a plane strain 2D non-linear model was made to check the numerical accuracy of the FEA package with linear/elastic theory. A plane strain 2D modelling approach was used with expanded brick elements and zero thickness interface elements to capture the behaviour at the brick/mortar interface. The 2D model consisted of 18 courses high, with a wall thickness of 110 mm, expanded brick height of 86 mm, simply supported top and bottom, with a lateral load on the 10th course of bricks from the bottom of the wall (see Table 4.1 and Figure 4.1). For the plane strain model, the length of the wall was given

43

4.2 Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

FEA Model Element Types 2D Model see also Figure 4.3(a)

TNO Diana 9.2 Bricks

Simplied micro model using expanded brick elements and interface elements QU8 CQ16E: An eight-node, 2D quadrilateral isoparametric plane strain element IL33 CL12I: A six-node, 2D interface element based on quadratic interpolation 4x4 4x1 (interface elements are only allowed to be one element thick) HE20 CHX60: A twenty-node, 3D isoparametric solid brick element based on quadratic interpolation and Gauss integration IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation 2x4x1 2x4x1 (interface elements are only allowed to be one element thick) 1x4x1 (interface elements are only allowed to be one element thick)

Mortar joints For brick thickness/height For mortar thickness/height

Mesh Density

Element Types

3D Model see also Figure 4.3(b)

Bricks

Mortar joints Mid-length brick interface elements

Mesh Density

For half brick For half mortar joint Mid-length brick interface elements

Table 4.2: Summary of 2D and 3D FEA element type and mesh selection for the test wall as 1 mm. Element types used for the 2D model were the QU8 - CQ16E plane quadratic elements for the bricks and IL33 - CL12I interface elements for the mortar interface elements (see Table 4.2 and Figure 4.3(a) for summary of element types and mesh densities). Failure of slender URM walls subject to vertical one-way bending is expected to initiate by exural tensile failure at the brick/mortar bond. Therefore, the interface element material behaviour initially chosen was that dened by the non-linear tension softening interface model, taken from the list of options available in Hendriks and Wolters (2007) and shown in Figure 4.4. This model was developed by Cornelissen et al. (1986) and Hordijk (1991). It is generally used to dene interface behaviour in concrete materials, but also may be suitable to describe behaviour at the brick/mortar interface. It was specically chosen because of its simplicity (only requiring unit exural bond strength (ft ) and tensile fracture energy (Gf I ), as well as linear normal and tangential stiness moduli as its list of material parameters) (note: n.ult in Figure 4.4 refers to the ultimate crack opening thickness). In order to compare the model linear rst cracking approximation with the hand calculated linear approximation, the post peak tensile softening behaviour is minimised (by letting Gf I = 1e05 N/mm). Any interface model could have

44

4.2 Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

been used for this step, as long as tensile fracture energy was minimised (assuming that the dominating mechanics of a slender wall subject to a lateral load is tensile failure). Material model parameters for the 2D model are listed in Table B.1 in Appendix B.

half brick - 2x4x1 mesh density

brick mesh = 4x4

expanded brick to account for zero thickness interface elements zero thickness interface element interface mesh = 4x1
half mortar joint (using zero thickness interface elements) - 2x4x1 mesh density interface element mid-length of brick

(a) 2D mesh density

(b) 3D mesh density

Figure 4.3: Mesh Densities for the 2D and 3D brick/mortar bodies in the FEA model

Figure 4.4: Non linear tension softening model for interface elements (Cornelissen et al. 1986) and (Hordijk 1991)

Four unit exural bond strengths were examined for the simulations. These bond strengths were 0.2 MPa, 0.4 MPa, 0.6 MPa and 0.8 MPa. The rst cracking loads obtained using the 2D model for these bond strengths are listed in Table 4.3. Next a 3D FEA model was built with the same structural conguration and boundary conditions (See Figure 4.2). For this model, Diana quadratic HE20 - CHX60 elements

45

4.2 Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

were used for the brick elements and IS88 - CQ48I elements were used for the interface elements (see Table 4.2 and Figure 4.3(b) for summary of element types and mesh densities). Again the interface element material behaviour was dened by the non-linear tension softening model used for the 2D model, with tensile fracture energy set to virtually zero (1e05 N/mm). Other material properties for the 3D equivalent of the 2D model with non-linear tension softening interface elements are listed in Table B.1 in Appendix B. The model was simulated with bond strengths of 0.2 MPa, 0.4 MPa, 0.6 MPa and 0.8 MPa, and the predicted rst cracking loads are listed in Table 4.3.

Figure 4.5: Diagram used for linear/elastic theory calculation (no self-weight)

Next, using linear/elastic theory, the load at which the peak stress at the extreme ber on the tension side of the wall at the location of maximum bending moment equaled the bond strength of the interface element at that location was calculated. For the purpose of simplicity, self-weight was not included in the model and hand calculation. The structural conguration for the equivalent hand calculation of 2D FEA model simulation is shown in Figure 4.5. Referring to Figure 4.5, for a wall with a point load at a distance a from the base of the wall, and for a given unit exural bond strength , the load at which rst cracking

46

4.2 Checking the 2D and 3D FEA Model Test Wall With Linear/Elastic Theory

will occur, i.e. the load (P ) at which stress at maximum bending moment rst exceeds strength at the extreme ber on the tension side of the wall, considering all unit exural bond strengths in the wall are equal, is given by Equation 4.1. Note that Equation 4.1 does not include self-weight.

P =

I ya(1 a ) l

(4.1)

I is equal to the second moment of area of the wall cross-section about a horizontal axis in the plane of the wall, y is the distance from half of the thickness of the wall to the tension edge and l is the height of the wall. The formula comes from resolving forces in the system (shown in Figure 4.5, taking moments and using the formula = relates moments and hence forces to stress in the wall).
My I

which

Bond Strength (MPa) 0.2 0.4 0.6 0.8

Load to Cause First Cracking (kN) Linear/Elastic Theory 2D FEA Model 3D FEA Model Prediction Result Result 1.0 2.1 3.1 4.2 1.0 2.1 3.1 4.2 1.0 2.1 3.1 4.2

Table 4.3: Summary of linear/elastic approximation (rst cracking load) compared with 2D and 3D FEA predictions using the structural conguration shown in 4.1 and 4.2, (taking a wall length of 960 mm for the 3D model)(no self-weight) Upon using the relationship in Equation 4.1 and substituting for unit exural bond strengths of 0.2 MPa, 0.4 MPa, 0.6 MPa and 0.8 MPa accordingly, with the other values being I =
bd3 12 = l 1.1092e+05 mm4 , y=55 mm, l=1538 mm, a = 2 , the linear/elastic calcu-

lations for the structural conguration in Figure 4.1 are listed in Table 4.3. Comparing the linear/elastic approximation of applied load for a given uniform unit exural bond strength for the wall conguration shown in Figure 4.1, with the corresponding 2D and 3D equivalent FEA model results summarised in Table 4.3, all three show perfect

47

4.3 The Consideration of Two Non-Linear Interface Behaviour Material Models

agreement. Therefore, it is fair to say that the FEA model is working correctly at the linear level.

4.3

The Consideration of Two Non-Linear Interface Behaviour Material Models

The next step in the 3D FEA model development process was to examine how sensitive the wall model is to the interface element model behaviour chosen. The 2D wall model was used to examine the eect of interface material models on the failure load of the conguration shown in Figure 4.1. Two interface material models were chosen to be examined more closely. The rst material model was the non-linear tension softening model shown in Figure 4.4 and presented by Cornelissen et al. (1986) and Hordijk (1991). The material model parameters for the non-linear tension softening model are listed in Table B.1 in Appendix B. The second interface material model is the crack-shear-crush model, also known as the composite interface model. It is appropriate to simulate fracture, frictional slip and crushing along material interfaces (Hendriks and Wolters 2007). The two-dimensional composite interface model enables the description of delamination (tension cut-o) and relative shear-slippage of two planes (coulomb friction). The mathematical details for the the composite interface model will not be reproduced here but can be found in the Material Library section of Hendriks and Wolters (2007).

Figure 4.6: The composite interface model two-dimensional interface yield function (Hendriks and Wolters 2007)

48

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

Figure 4.6 shows the two-dimensional interface material law for the composite interface model, where is the shear stress, is the stress normal to the plane of the interface element, is the friction angle, c0 is the initial adhesion, fc is the compressive strength and ft is the tensile strength. For the composite interface model, material parameters are listed in Table B.2, Appendix B. How these values were determined is described in Section 4.7.2.

4.4

Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

In order to examine the eect of the interface material models on peak wall load, a suitable tensile fracture energy Gf I for the interface element models needed to be determined. Tensile fracture energy is an important consideration when modelling out of plane bending because it describes how the material will behave at a location where the stress in the element has exceeded the tensile strength for the element (where a crack has occurred). In other words, it describes the tensile softening behaviour of the element and can have a signicant impact on how the wall responds to load. Conceptually, the tensile fracture energy can be visualised as the area under the normal stress versus normal displacement curve at the mortar joint between two bricks which are uniaxially loaded (see Figure 4.7).

Figure 4.7: Typical behaviour of the mortar joint between two bricks which are uniaxially loaded (taken from (Lourenco 1996a))

It can be seen in Table 1 in the Lourenco (1996b) report (sourced from (CUR 1994)) (reproduced for convenience in Figure 4.8), and the plots of tensile fracture energy ver-

49

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

sus tensile bond strength in Figures 4.9, 4.10 and 4.11 which come from the papers Van Der Pluijm (1992) and Van Der Pluijm (1997), that there is no clear correlation between tensile bond strength and tensile fracture energy except that it can be seen that with increasing bond strength, tensile fracture energy tends to increase. What can also be observed from Figures 4.9 and 4.10 is that tensile fracture energy would appear to be dependent upon mortar type and brick type. For example, examining Figures 4.9 and 4.10, it can be seen that the 1:2:9 mortar and the Vijf Eiken soft mud brick (which has a high suction rate), both have steeper tensile fracture energy versus tensile bond strength
1 1 slopes than that which can be observed for the 1: 2 :4 2 mortar and the Joosten wire cut

clay brick (which has a low suction rate).

Figure 4.8: Tensile fracture energy material data from Lourenco (1996b) report

50

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

Figure 4.9: Tensile bond strength versus tensile fracture energy for clay brick masonry arranged by mortar type (Van Der Pluijm 1992)

Figure 4.10: Tensile bond strength versus tensile fracture energy for clay brick masonry arranged by brick type [Joosten: wire cut clay bricks, Vijf Eiken: soft mud clay bricks] Van Der Pluijm (1992)

51

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

Figure 4.11: Tensile bond strength versus tensile fracture energy for clay brick masonry with general purpose mortar (1:1:6 + air-entrainer) from (Van Der Pluijm 1997)

4.4.1

Examining the 2D Non-Linear Model For the Two Interface Material Models with a Gf I = 0.012 N/mm

Lourenco (1996b) suggests that a tensile fracture energy of 0.012 N/mm can be, in principal, adopted for URM walls subject to shear loading (he did not make any recommendations for walls subject to lateral loading). Using a Gf I = 0.012 N/mm and other specied material parameters for the two interface models (see Appendix B), the 2D model peak loads obtained for the wall conguration in Figure 4.1 (without self-weight) are listed in Table 4.4. Upon examination of the peak loads in Table 4.4, it can be seen that using the two different interface models gives slightly dierent results. The non-linear tension softening model consistently gives a slightly lower peak load than the composite interface model for a variety of unit exural bond strengths ft . To examine what could be causing this peak load dierence, it was decided that a closer examination of parameters making up the two interface material models was warranted.

52

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

Bond Strength (MPa) 0.2 0.4 0.6 0.8

Predicted Peak Load on Wall (kN) No Tensile Fracture Non-linear Tension Composite Interface Energy (Linear Approx.) Softening Model Model 1.0 2.1 3.1 4.2 1.8 2.8 3.5 4.7 1.9 3.0 3.9 4.8

Table 4.4: Summary of results for the two interface material models examined, with a tensile fracture energy Gf I of 0.012 N/mm (suggested from Lourenco (1996b)), using the 2D non-linear FEA model shown in Figure 4.1 (no self-weight)

Figure 4.12: Stress displacement curves for tests which used clay bricks and 1:1:6 mortar. Average theoretical descending branches according to the non-linear tension softening model (Hordijk 1991) and the composite interface model (Lourenco 1996a)

53

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

4.4.2

Material Parameters Aecting Peak Load in the Non-Linear FEA Model

Upon varying the shear and compressive parameters in the composite interface model, it was discovered that the only interface material properties that aected the peak load in the 2D model were the unit exural bond strength ft and the tensile fracture energy Gf I . In reality, if joint compressive strength was relatively small, it would be expected to have an eect on peak load in the 2D and 3D wall model, but realistic values of compressive strength for a brick/mortar joint far exceed the tensile strength of the joint. Therefore at realistic values of compressive joint strength, wall peak load is not aected by variations in this parameter. What was causing a slight dierence in peak loads achieved by using the two dierent interface models might be attributed to the slightly dierent formulas used to describe the descending branch of the stress versus crack opening width curve for masonry in tension (see Van Der Pluijm (1997), Equations 3 and 4 in that paper) and Figure 4.12. Although both equations give similar results, the descending branch used in the non-linear tension softening inteface model is steeper in the rst part of the descending branch as shown in Figure 4.12. This mean that once peak stress on the tension side of the wall has been reached at the interface node at the location of peak bending moment, the stress is redistributed to the surrounding nodes and therefore elements. If the descending branch of the stress versus crack opening displacement is steeper in the non-linear tension softening model, this means that the stress on the surrounding nodes will be redistributed more quickly and the crack will propagate more quickly through the thickness of the wall. It would therefore be expected that the load for a wall using the non-linear tension softening model may produce a slightly lower peak load than for a wall using the composite interface model, as was observed.

4.4.3

Sensitivity of Wall Peak Load to Gf I

The next step was to examine how sensitive the model was to varying tensile fracture energy for a particular bond strength. In order to do this, unit exural bond strengths of 0.2 MPa and 0.8 MPa were used and random tensile fracture energies were simulated.

54

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

Bond Strength (MPa) 0.2 0.2 0.2 0.2 0.2 0.8 0.8 0.8 0.8

Tensile Fracture Energy (N/mm) 0.005 0.012 0.020 0.028 0.036 0.012 0.020 0.028 0.050

Peak Wall Load (kN) Non-linear Tension Composite Interface Softening Model Model 1.6 1.8 2.0 2.1 2.2 4.7 4.8 5.0 5.7 1.7 1.9 2.1 2.2 2.2 4.8 5.1 5.4 6.1

Table 4.5: Varying tensile fracture energy for specic unit exural bond strengths to examine if there is an upper limit on peak load with increased tensile fracture energy (no self-weight) The results are shown in Table 4.5. It would appear that in the model, as fracture energy for a particular bond strength increases, so does the peak load. There didnt appear to be a maximum tensile fracture energy for either bond strength which created a convergence of peak failure load for that particular bond strength. As mentioned previously, extensive tests conducted by Van Der Pluijm (1997) observed that there is no clear correlation between unit exural bond strength and tensile fracture energy (see for example Figure 4.11), but with increasing unit exural bond strength, tensile fracture energy also tends to increase. It can be seen in Figure 4.11 that for unit exural bond strengths ranging from approximately 0.1 MPa to 0.8 MPa, tensile fracture energies range from approximately 0.001 N/mm to 0.020 N/mm. However, as shown in Figure 4.11, a tensile fracture energy of 0.001 N/mm could be associated with bond strengths over a large range (Gf I < 0.004 N/mm for bond strengths up to 0.5 N/mm2 ). Conversely, for a unit exural bond strength of approximately 0.39 MPa, a tensile fracture energy of approximately 0.019 N/mm was recorded. Such variability in results may be due to the various errors in measurement due to the testing procedures employed in the study. It can be seen from Table 4.5, that for a certain unit exural bond strength, the peak load obtained in the 2D model varies signicantly based on the tensile fracture energy

55

4.4 Determining a Suitable Tensile Fracture Energy Gf I for the Non-Linear FEA Model

0.02 Tensile Fracture Energy (N/mm) 0.01571x+0.0004882 0.015

0.01

0.005

0 0 0.1 0.2 0.3 0.4 0.5 0.6 Tensile Bond Strength (MPa) 0.7 0.8

Figure 4.13: Fitting a linear relationship between tensile bond strength versus tensile fracture energy for clay brick masonry with general purpose mortar (1:1:6 + air-entrainer) using Figure 4.11 from Van Der Pluijm (1997)

specied. In order to correctly deal with tensile fracture energy as it relates to unit exural bond strength in the non-linear FEA model, it was decided to create a tensile fracture energy versus tensile bond strength relationship from an appropriate set of experimental results for a specic type of mortar. The general purpose (1:1:6 + air-entrainer) plot of tensile fracture energy versus bond strength (from (Van Der Pluijm 1997) - see Figure 4.11) was chosen for tting a mathematical relationship. It was decided that the most appropriate relationship (over a range of bond strengths) that could be determined would employ tting a linear relationship to the data in Figure 4.11. Using a quadratic function poses the danger of overestimating tensile fracture energies for higher bond strengths for which there is no data available in Figure 4.11. The best t linear relationship for the data points in Figure 4.11 is shown in Figure 4.13. This relationship is presented in Equation 4.2.

Gf I = 0.01571 ft + 0.0004882

(4.2)

56

4.5 An Aside: The Issue of Load Step Size for the 2D Non-Linear FEA Model

4.5

An Aside: The Issue of Load Step Size for the 2D NonLinear FEA Model

The 2D wall models used load stepping control instead of displacement control to apply force to the walls. For the 2D models, it was found that for the type of analyses conducted, with the given boundary and loading conditions, the most appropriate load stepping type was that of a user specied load step size. However, it was found that the size of the load step specied had a large impact on the accuracy of the results. Particularly for the case where there existed no self-weight and the absence of a tensile fracture energy when using the non-linear tension softening interface material model. If the load step size was too large, it appeared to over-estimate the cracking load before the simulation started unloading. Therefore, for all the 2D simulations, a load step of 0.0001 predetermined applied force was used. The predetermined applied force to the wall was taken as 10 kN. It should be noted here that if the simulation arrives at the maximum number of load steps specied (which in this case was 10000) and it hasnt reached peak load, the analysis package automatically continues to increase the load at the specied load step size until a maximum load is reached and then unloading will occur. Spherical path arc-length control method was applied to one interface node located at mid-height of the wall. Using an arc-length control method is considered something of an indirect displacement control and may be useful in the case of local snap-through or snap-back behavior (see Analysis Procedures, Section 29.1.5 in Hendriks and Wolters (2007) for more information). There is no unloading determination available for the user specied load step size.

4.6

Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

The non-linear tension softening model and the composite interface model were examined for their computational stability when implemented into the 3D equivalent of the 2D wall model. Self-weight and an appropriate tensile fracture energy (based upon Equation 4.2) were also added to the model. Once model stability was established, unit exural bond

57

4.6 Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

strengths were spatially varied and again the model results were examined for quality and computational stability.

Applied Load at Tenth Course (kN)

2.5

1.5

0.5

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.14: Load step versus deection curve for the 3D non-linear model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.2, peak load = 1.8kN

Just to reiterate, for the 3D test model, the boundary conditions are that of a simply supported wall top and bottom, and restricted from vertical translation at the base, on the unloaded side of the wall (see Figure 4.2). A unit exural bond strength of 0.2 MPa was used, with an associated tensile fracture energy of 0.0036 N/mm based upon Equation 4.2. Brick material properties were the same as used in the 2D model (see Section 4.2). For the shear and compression properties required for the interface elements in the model which utilizes the composite interface model, see Table B.2 in Appendix B. The plot of load versus deection for the non-linear tension softening model and the composite interface model are shown in Figures 4.14 and 4.15 respectively. The 3D model with non-linear tension softening interface elements predicts a peak load of 1.8 kN, whereas the 3D model with composite interface elements predicts a peak load of 1.9 kN. These results are higher (because of the inclusion of Gf I ) than the linear approximation (1.3 kN for ft = 0.2 MPa and self-weight included) as expected and the model appeared to stay

58

4.6 Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

Applied Load at Tenth Course (kN)

2.5

1.5

0.5

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.15: Load step versus deection curve at point of maximum deection for the 3D non-linear model shown in Figure 4.2 using the composite interface elements, ft = 0.2 MPa, peak load = 1.92 kN

computationally stable.

4.6.1

Inclusion of Spatial Variability in Bond Strength in the 3D Interface Element Models

Uncorrelated spatial variability of unit exural bond strength (recall: it was found in Chapter 3 that unit exural bond strengths are statistically independent) for the horizontal and vertical mortar joints in the 3D URM wall was implemented and the stability of the model was examined for using both the non-linear tension softening model and the composite interface model. A script was written which took the data input le for the model and then assigned randomly generated unit exural bond strengths (according to a truncated normal distribution and associated fracture energy (according to the relationship in Equation 4.2)) to each horizontal and vertical mortar joint in the wall. A mean unit exural bond strength of 0.8 MPa with a COV of 0.5 (i.e. standard deviation of 0.4) were used in both models, and both models had the same randomly generated unit exural bond strength assigned to each joint so that a direct comparison between the two mod-

59

4.6 Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

els could be made. All other material parameters remained the same as previously stated.

Applied Load at Tenth Course (kN)

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.16: Load step versus deection curve at point of maximum deection for the 3D non-linear test wall model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.8, COV = 0.5, peak load = 3.85 kN

The results for the load/deection curve for a set of randomly generated spatially varying bond strengths for the non-linear tension softening material elements and the composite interface material elements can be seen in Figures 4.16 and 4.17 respectively. As can be seen from the load/deection plots in Figures 4.16 and 4.17, the 3D non-linear FEA model with implemented spatial variability in bond strength and corresponding tensile fracture energy produce stable results using either the non-linear tension softening or composite interface elements. In Figure 4.17, it appears that the model crashed before a peak load was obtained but upon closer investigation, this was not found to be the case.

60

4.6 Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

Applied Load at Tenth Course (kN)

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.17: Load step versus deection curve for Dohertys 3D non-linear model shown in Figure 4.2 using the composite interface elements, ft = 0.8, COV = 0.5, peak load = 4.35 kN

Figure 4.18: Final deected wall shape for a wall using non-linear tension softening interface elements, with variable bond strength according to ft = 0.8, COV = 0.5, peak load = 3.85 kN

61

4.6 Examining Failure Load Using Interface Models in the Equivalent 3D Non-Linear Wall Model With Self-Weight

Figure 4.19: Final deected wall shape for a wall using composite interface elements, with variable bond strength according to ft = 0.8, COV = 0.5, peak load = 4.35 kN

Figures 4.18 and 4.19 show the nal deected shape of the 3D wall simulations using spatial variability in bond strength with a ft = 0.8 and a COV = 0.5, using non-linear tension softening interface elements and composite interface elements respectively. It was observed that the weak mortar joints which crack independently of the ultimate failure course were captured with both interface models.

62

4.7 Modelling Doherty (2000) Experimental Walls

4.7

Modelling Doherty (2000) Experimental Walls

An appropriate set of experimental results were then used to validate the 3D non-linear FEA model. From a search of the literature, within the limited experimental results of laterally loaded URM walls in vertical bending available, only one set of experimental results contained more than one of exactly the same wall size, boundary conditions and loading specications, with sucient small specimen test data which could be used to provide a reasonable set of model material parameters. These walls were part of Dohertys research thesis (Doherty 2000), where he conducted three uncracked static push tests with no precompression on 960 mm 1548 mm vertically spanning clay brick unreinforced masonry panels. The walls were laterally line loaded at close to mid-height of the wall (course 10 of 18 courses) using a hand pump driven hydraulic actuator. The top of the walls were restrained by a cornice type support. The base of each wall was situated on an embossed polythene membrane DPC connection. Doherty (2000) states that the test boundary conditions were therefore such that the vertical reactions are constant but pushed to the compressive zone with mid-height displacement. This amounts to boundary conditions modelled in FEA by simple supports top and bottom. The reader is referred to Doherty (2000) for a detailed description of the test rig set up in which the walls were tested. The bricks used in the experimental tests were standard extruded clay bricks having the dimensions 76 mm 110 mm 230 mm and contained 3 45 mm diameter holes. The mortar used was 1:1:6 and the horizontal and vertical mortar joints were 10 mm. See Figure 4.20 for a photograph of one Dohertys walls in the test rig. For simplicity, the brick bodies in the FEA model were modelled as solid, whereas the experimental bricks had 3 45 mm holes. Normally holes in bricks are partially lled with mortar, which is an unnecessary complication to model in FEA. Therefore the solid brick is considered a reasonable approximation to the experimental brick. The size specications used in the 3D non-linear FEA to model the three walls tested from Doherty (2000) are listed in Table 4.6.

63

4.7 Modelling Doherty (2000) Experimental Walls

Figure 4.20: Static push test rig setup from (Doherty 2000)

Wall Dimensions Number of full bricks per course Number of courses high Mortar joint thickness Brick dimensions Brick type

960 mm long x 1548 mm high 4 18 10 mm 230 mm x 110 mm x 76 mm clay brick extruded with no holes or indentations

Table 4.6: Summary of dimensions used in the 3D non-linear FEA to model the three Doherty (2000) walls

64

4.7 Modelling Doherty (2000) Experimental Walls

4.7.1

Material Properties for the Doherty (2000) Walls Recorded in His Thesis
kg . m3

In Doherty (2000), several material tests were conducted to accompany the wall tests. Doherty (2000) determined the average density of the brickwork to be 1800 It was dicult to determine the actual unit exural bond strength for each of the walls tested, based on Dohertys thesis. He reports a unit exural bond strength of 0.46 MPa for each of the walls tested, whether they were built and tested in March 1998 or September 1998. It does not make sense that walls built at dierent times throughout the year could have exactly the same average unit exural bond strength and is not consistent with each set of bond wrench test data tested at the corresponding time that each wall was built (Recorded in Appendix E of (Doherty 2000)). Therefore, for the purpose of comparing the experimental wall results with the 3D non-linear FEA model, an average unit exural bond strength was calculated using the bond wrench two-brick prism test results conducted at the same time as the three walls were built. For each wall, three two-brick bond wrench tests were conducted. That provided three bond wrench exural tensile strengths for each wall. That provided nine bond wrench exural tensile strength results for the three walls. The mean of these strengths is ft = 0.328. The standard deviation is ft = 0.087. This corresponds to a COV of 0.27. Other statistics recorded in Dohertys thesis for the three walls tested, including peak load, deection, and masonry Youngs modulus are presented in Table 4.7.

Wall Number

Date Built

Overburden Pressure (MPa)

Flexural Tensile Strength ft (MPa) Mean COV 0.4 (P) 0.3 (P) 0.29(P) 0.25 0.10 0.34

Masonry (Brickwork) Youngs Modulus (MPa) 5000 11000 ?

Peak Force (kN)

Displ. at Peak Force (mm) 0.5 0.3 0.59

3 12 13

March98 Sept98 Sept98

0 0 0

3.18 3.0 2.2

Table 4.7: Summary of results from (Doherty 2000) for the uncracked static push tested walls (P indicates results are from prism bond wrench tests) The three Doherty (2000) static push test wall load verses displacement diagrams are

65

4.7 Modelling Doherty (2000) Experimental Walls

shown in Figures B.1, B.2 and B.3 in Appendix B. These plots are of limited use because the load/deection plot for each wall from 0 kN to peak load cannot be zoomed in upon to examine any slight changes in slope of the load deection curve in more detail. The average peak load for the three static push test walls was 2.8 kN. However, examining the third static push test result (Wall Number 13), it appears that this wall fails at a lower load and a higher deection than Wall Numbers 3 and 12. That wall also appeared to have a higher COV of bond strengths based on the corresponding prism tests, which, based on the study conducted by Lawrence and Cao (1988), infers that the wall should fail at a lower load than for similar walls which have a lower COV in bond strength (see Chapter 5 for the theoretical explanation of why this should occur). Doherty (2000) states that the modulus of elasticity for the masonry in each wall was determined by conducting compressive tests on ve-brick tall prisms. The modulus test results ranged from 5000 MPa to 11000 MPa (see Table 4.7). Doherty suspected the large variation in brickwork modulus of elasticity was due to inconsistencies in the preparation of the ve brick prisms. Willis (2004) compared Doherty (2000) experimental results with his non-linear stressstrain model and a rigid body analysis and discovered that the non-linear model overestimated the cracking load and underestimates deection at cracking. Williss non-linear elastic model was thoroughly examined and agrees reasonable well with elastic theory, using an eective exural rigidity for the brickwork. Therefore Willis (2004) suspected that the elastic material properties of Doherty (2000) were much softer and more non-linear than indicated by Dohertys material tests. Willis (2004) states that if a probabilistic method for determining cracking load were used, then his non-linear model could be expected on average to give better predictions for the cracking load.

4.7.2

Set of Material Parameters Used for the 3D Simulations

Calculating reasonable normal and tangential stiness moduli to model Dohertys walls using a 3D non-linear FEA model is outlined in the proceeding section. The other material property parameter values were determined by examining several sources, namely

66

4.7 Modelling Doherty (2000) Experimental Walls

(Lourenco 1996a), (Lourenco 1996b), (Petersen et al. 2007) and (Van Zijl 2004) and then determining appropriate values based upon those sources. Cohesion is dependent upon the type of bricks and mortar used. From a sensitivity analysis conducted using the 3D non-linear model and from previous theory developed for masonry walls subject to one-way vertical bending (Willis 2004), the shear properties including cohesion, friction angle, shear fracture energy, conning normal stress and dilatency dont appear to have an eect on the failure load of clay brick URM walls subject to lateral loading. So for the values of these parameters, a set was chosen based on observations of what was used in the literature sources listed above. Similarly, due to conducting a sensitivity analysis on the following parameters, it appears that exponential degradation coecient, capped critical compressive strength, shear traction control factor, compressive fracture energy, equivalent plastic relative displacement and the shear fracture energy factor also dont have an eect on peak failure load of a clay brick URM wall subject to one-way vertical bending and are therefore assigned standard parameter values as discussed in Lourenco (1996b). For the brick material properties, a low Youngs modulus of 5884 N/mm2 was prescribed, the reasoning for this is outlined in the next section. The brick poissions ratio and density were those determined in Doherty (2000). The potential failure interface elements in the brick located at mid-length of each brick are not expected to have any impact on the failure load of a clay brick URM wall subject to one way vertical bending and were therefore prescribed articially high linear normal and tangential stinesses as well as a high tensile strength and tensile fracture energy. The crack model describing this potential brick crack interface behaviour was described by the non-linear tension softening interface model. Material parameters used for the 3D model are summarised in Table B.3 in Appendix B.

4.7.3

Determining Appropriate Elastic Stiness Moduli for the Doherty 3D Non-Linear FEA Model

For the experimental tests being simulated, the ultimate strength of a 3D modelled masonry wall subject to lateral loading is not aected by the stiness moduli. However for

67

4.7 Modelling Doherty (2000) Experimental Walls

completeness, determination of a reasonable stiness moduli will be addressed. To this end, given the experience of Willis (2004) where he found that elastic material properties of Doherty (2000) were much softer and more non-linear than indicated by Dohertys material tests, it was decided more suitable to assign a lower than recorded average Youngs moduli for the brickwork, choosing it in such a way as to provide realistic values for the Youngs modulus of the bricks and the mortar which result in a model deection reective of what was found for the experimental wall test deections. The amount of linear deection (i.e. before the model becomes non-linear) obtained in the 3D FEA simplied micro model of a masonry wall subject to a lateral load is controlled by the elastic modulus of brick material as well as the values of the linear normal and tangential stiness moduli (kn and kt respectively) associated with the joint interface elements. The latter are two of several material property input parameters for the interface elements, and help dene the behaviour at the brick/mortar interface during wall loading, up until the wall starts being dened by non-linear material properties (until cracking occurs in the mortar joints). The linear normal and tangential stiness moduli for the interfaces are calculated using Equations 4.3 and 4.4 (Lourenco 1996b).

kn =

Eb Em hm (Eb Em ) Gb Gm hm (Gb Gm )

(4.3)

kt =

(4.4)

where Gb and Gm are the average shear moduli for the bricks and mortar respectively, hm is the thickness of the mortar joint, and Eb , Ebw (see Equation 4.7) and Em are the average Youngs moduli for the bricks, the brickwork and the mortar respectively. Gb and Gm can be calculated using the relationships in Equations 4.5 and 4.6 respectively.

Gb =

Eb 2(1 + ) Em 2(1 + )

(4.5)

Gm =

(4.6)

68

4.7 Modelling Doherty (2000) Experimental Walls

Assuming ve brick high and ve mortar joint high prisms (which were used to determine the Youngs modulus of the brickwork for each wall in Dohertys thesis), various mortar Youngs moduli were assumed and the corresponding brick Youngs moduli were calculated using compatibility theory shown in Equation 4.7.

Eb =

Lb
Lbw Ebw

Lm Em

(4.7)

where Lb , Lbw and Lm are the total height made up by the bricks, the brickwork and the mortar in a prism respectively, and the other terms were dened previously. Several combinations of brick and mortar Youngs moduli were then used to obtain a value for the linear stiness moduli of the interface elements in the 3D FEA model and the brick Youngs modulus which is a direct input model parameter. The most representative Eb and Em , and hence kn and kt are provided in Table B.3, Appendix B.

4.7.4

Element and Mesh Specications for the 3D Non-Linear FEA Model

3D Model see also Figure 4.3(b) Bricks HE20 CHX60: A twenty-node, 3D isoparametric solid brick element based on quadratic interpolation and Gauss integration IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation 2x4x1 2x4x1 (interface elements are only allowed to be one element thick) 1x4x1 (interface elements are only allowed to be one element thick)

Element Types

Mortar joints Mid-length brick interface elements

Mesh Density

For half brick For half mortar joint Mid-length brick interface elements

Table 4.8: Summary of FEA element selection for the walls in Doherty (2000) The element specications used to model the uncracked static push tests using 3D nonlinear FEA, are listed in Table 4.8. Each unit (one brick and the underlying mortar that

69

4.8 Doherty (2000) Wall 3D Non-Linear FEA Results

brick is attached to) is modelled with two half brick units and an interface element between them (to allow potential cracking at mid-length of the brick). The reason for this is two-fold. The rst reason is so that points which describe the geometry of the bodies of the brick and underlying mortar are common to each body. This alleviates any problems associated with generating a mesh amongst bodies which dont contain common nodes at the common surfaces which make up two adjacent bodies. The second reason is so that the brick has the potential to crack at its mid-length, which can occur in the case of two-way bending of laterally loaded walls. Several mesh densities were trialed, however the mesh density decided upon and listed in Table 4.8 was chosen for its compromise between accuracy and computational speed. The boundary conditions for the 3D non-linear FEA model are that of a simply supported wall top and bottom, and restricted from vertical translation at the base, on the unloaded side of the wall (see Figure 4.2).

4.7.5

Load step size determination

For the simulation of both the Doherty (2000) and the full sized wall which is discussed in Chapter 5, various load step size determinations were tested. These included explicitly specied load step sizes, iteration based automatic step size control, energy based automatic step size control and automatic step size control. From the tests, it was determined that the energy based automatic step size control consistently produced the best results for this model. It allowed loading and unloading of the wall to be captured and appeared to be robust with respect to maximum load step size chosen. See Table B.4 (Appendix B) for a summary of wall loading specications used in the Doherty 3D model.

4.8

Doherty (2000) Wall 3D Non-Linear FEA Results

Simulations with the uniform unit exural bond strength of ft = 0.328 (obtained from Dohertys prism bond wrench tests as discussed previously) and an associated tensile fracture energy according to Equation 4.2 which equates to 0.0056 N/mm were conducted using

70

4.8 Doherty (2000) Wall 3D Non-Linear FEA Results

both the non-linear tension softening and the composite interface element models. The results are shown in Figures 4.21 and 4.22 respectively. Amongst other output, DIANA provides x, y and z axis wall translation output for each node in the wall. In this instance, the load/defection output is examined for a node at which the wall has maximum y deection (out-of-plane displacement). The node is one at an interface element located mid-length along the wall in course 10 on the side of the wall which is in tension.

Analysis Method Dohertys three uncracked static push tests 3D model utilizing non-linear tension softening interface elements 3D model utilizing composite interface elements

Failure Load (kN) 2.8 (mean of three walls) 2.5

2.7

Table 4.9: Summary of failure loads for Dohertys uncracked static push tests and the 3D FEA model with two dierent interface material models As can be seen from the plots and in keeping with previous ndings, the 3D FEA model which utilises the composite interface elements produces a slightly higher peak load than the model which utilises the the non-linear tension softening interface elements. The peak load produced by the model utilising the composite interface elements was 2.7 kN, which was closer to the average experimental result of 2.8 kN than the model utilising the non-linear tension softening interface elements which produced a peak load of 2.5 kN (see Table 4.9 for a summary).

71

4.8 Doherty (2000) Wall 3D Non-Linear FEA Results

Applied Load at Tenth Course (kN)

2.5

1.5

0.5

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.21: Load versus deection curve for the 3D non-linear FEA model shown in Figure 4.2 using the non-linear tension softening interface elements, ft = 0.328, peak load = 2.5 kN

Applied Load at Tenth Course (kN)

2.5

1.5

0.5

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

OutofPlane Wall Deflection (mm)

Figure 4.22: Load versus deection curve for the 3D non-linear FEA model shown in Figure 4.2 using the composite interface elements, ft = 0.328, peak load = 2.7 kN

72

4.8 Doherty (2000) Wall 3D Non-Linear FEA Results

4.8.1

Linear/Elastic (Lower Bound) and Plastic (Upper Bound) Peak Load Approximations

The values in Table 4.9 were compared with the expected rst cracking load using linear/elastic theory and with elastic plastic theory. Linear beam and rigid plastic analysis are considered to give the bounding wall failure loads for such a structural conguration. Both theories used the boundary conditions of a simply supported wall top and bottom. For linear/elastic theory, rst cracking (with self-weight) was calculated to be 2 kN.

0.5 0.4 0.3 0.2 2.0 kN (LS 82) 2.7 kN (LS 114)

Stress at node (MPa)

0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 13.75 27.5 41.25 55 68.75 82.5 96.25 110

Distance from compressive edge (mm)

Figure 4.23: Stress distribution across the thickness of the wall at the location of peak deection (course 10) for the 3D non-linear wall model shown in Figure 4.2 using the composite interface elements, ft = 0.328

For the plastic analysis, we rst examine the stress distribution across the joint where peak deection occurs with increasing load in the model utilizing the composite interface elements. The stress distribution across the thickness of the mortar joint located at maximum deection (course 10) is shown for rst cracking (2 kN) and peak load (2.7 kN) in Figure 4.23. Compressive strength of masonry is signicantly higher than the tensile strength. Therefore, observe in Figure 4.23 that as load increases and the joint softens on the tension side of the wall, stress on the compressive bers continue increasing linearly,

73

4.8 Doherty (2000) Wall 3D Non-Linear FEA Results

well after the tensile side of the joint has softened. Because the compressive strength of the joint is signicantly higher than the tensile strength, the moment capacity of the joint is more than what could be expected from a normal plastic stress distribution where tensile strength equals compressive strength (as is found in steel for example).

0.5 0.4 0.3 0.2

Stress at node (MPa)

0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 13.75 27.5 41.25 55 68.75 82.5 96.25 110

Distance from compressive edge (mm)

Figure 4.24: Limiting stress distribution across the thickness of the wall at the location of peak deection (course 10) for the wall shown in Figure 4.2

Therefore, the limiting stress distribution for the calculation of a plastic moment utilized in a plastic analysis for the case of a brick/mortar interface (and the given structural conguration) is suggested in Figure 4.24. That is, at the plastic limit, the stresses on the tensile side of the joint exhibit the usual limiting plastic behaviour, but on the compression side of the joint, the stresses remain linear/elastic.

74

4.9 Discussion of 3D FEA Model Development Results

Wall Linear/elastic analysis Plastic analysis 3D Model utilizing composite interface elements

Predicted Failure Load (kN) 2.0 3.0

2.7

Table 4.10: Predicted failure loads using various approximations Using this stress distribution, the value of Mp for a plastic analysis with the dimensions, specications, boundary conditions and self-weight of the wall in question, the failure load is calculated to be 3 kN. The results of the three forms of wall analysis (linear/elastic, plastic, and 3D non-linear FEA) are summarised in Table 4.10. As would be expected, the results of the 3D nonlinear FEA analysis lie between the linear/elastic and plastic analysis predicted wall capacities loads.

4.9

Discussion of 3D FEA Model Development Results

Based upon the investigation provided in this chapter, TNO DIANA 9.2 is deemed to provide reasonable results. It was decided that the composite interface model should be used to model the brick/mortar interface elements for the full sized wall with spatial variability in bond strength (presented in Chapter 5), as it gave a closer approximation to the mean failure load of Dohertys three walls.

75

Chapter 5

Stochastic FEA and Wall Strength Variability

5.1 Introduction

Once the 3D non-linear FEA model was validated, the next step in the thesis was to build a full sized 3D non-linear FEA wall model and conduct various stochastic analysis by incorporating spatial (non-correlated) and non-spatial (fully correlated) variability of unit exural bond strength (ft ) and corresponding tensile fracture energy (Gf I ) into the model. This was done with the objective of examining how variability in ft aects the mean and variability of wall failure load. Wall failure load is dened as the peak load obtained just before unloading of the wall occurs.

Stochastic Analysis Spatial (uncorrelated) ft Non-spatial (fully correlated) ft

Distribution Type Trunc. Norm. Trunc. Norm. Trunc. Norm. Trunc. Norm. Trunc. Norm. Trunc. Norm.

Bond Strength Mean (MPa) 0.4 0.4 0.4 0.4 0.4 0.4

COV 0.1 0.3 0.5 0.1 0.3 0.5

Number of Simulations 50 50 50 50 50 50

Table 5.1: Statistical parameters to be used to obtain failure load distributions for the 3D FEA model of the full wall

76

5.1 Introduction

In order to investigate this issue, unit variability models were examined using stochastic analysis of the 3D non-linear FEA wall model. Spatial Analysis: The rst scenario considers a stochastic analysis with the full sized wall having spatially varying unit exural bond strengths, where no spatial correlation in bond strength exists between each unit in the wall, as found in the experimental results in Chapter 3. That is, each unit exural bond strength in the wall is statistically independent of adjacent unit exural bond strengths. A stochastic analysis was conducted using 50 non-linear FEA wall simulations for each COV(ft ) of 0.1, 0.3 and 0.5, all with a mean bond strength (ft ) = 0.4 MPa. Non-Spatial Analysis: The second scenario considers a stochastic analysis with the full sized wall having a non-spatially varying unit exural bond strength for each realisation, i.e. where wall bond strength is fully correlated (i.e. uniform bond strength in the wall). A stochastic analysis was conducted using this scenario with 50 simulations for each COV(ft ) of 0.1, 0.3 and 0.5, all with a mean bond strength (ft ) = 0.4 MPa. The non-spatial stochastic analysis was conducted because many existing stochastic analysis of structures assume this scenario and so it is important to examine how dierent the results are between this commonly examined scenario and that of the spatial scenario, which is considered a more realistic approach to examining material variability in a clay brick URM wall. See Table 5.1 for a summary of simulations conducted for the full sized wall. Fifty simulations for each COV in each scenario was deemed a reasonable number of simulations from a convergence check of failure load and COV (see Figure 5.1). Due to time limitations, only one mean unit exural bond strength was examined. This ft = 0.4 MPa was chosen because based upon an extensive literature review and experience, a mean bond strength of ft = 0.4 MPa is considered a bond strength reective of many realistic mean bond strengths found in practice on construction sites. Chosen COV(ft ) of 0.1, 0.3 and 0.5 encompass those reported in the paper by McNeilly et al. (1996) for 19 building sites in Melbourne. They are considered generally observed

77

5.1 Introduction

Wall Mean Failure Load (kPa)

Wall Failure Load COV

2.2

0.06

COVf = 0.1 2.1

t t t

0.04

COVf = 0.1
t t t

COVf = 0.3 COVf = 0.5

COVf = 0.3 COVf = 0.5 0.02

2 10

Number of Spatial NonLinear FEA Simulations

20

30

40

50

10

Number of Spatial NonLinear FEA Simulations

20

30

40

50

(a) Convergence of mean failure load with increased (b) Convergence of failure load COV with increased number of simulations number of simulations

Figure 5.1: Convergence of mean failure load and COV with increased number of simulations small, medium and large COVs respectively which can be expected in a clay brick URM wall in practice. Correa et al. (2008) found that unit exural bond strengths measured from the walls studied in Chapter 3 and unit exural bond strengths measured from the corresponding prisms were not statistically similar in most cases. Therefore, accurately trying to see how mean and COV of bond strength (i.e. ft and COV(ft ) respectively) aects the mean and COV failure load for a wall using physical experimental testing techniques of measuring bond strength ft and COV(ft ) using the prisms and then measuring the peak failure load for the wall by physically testing the wall to failure could be considered as an inaccurate method to achieve such an objective. However, using a 3D non-linear FEA model where randomly generated bond strengths can be assigned to each unit in a wall, could, one would hope, produce a more accurate and reliable indication of how spatial variability in bond strength aects the mean and variability of failure load of a clay brick URM wall in vertical bending. Also, the non-linear 3D FEA analysis of a clay brick URM wall can capture non-linear wall behaviour and hence ultimate failure load as opposed to just rst cracking load (which is obtained using a linear model), and as shown in Chapter 4, predicts experimental wall failure load reasonably well. The research presented in Chapter 4 indicated that the traditional denition of rst cracking load (i.e. the load at which stress exceeds strength at the outer most tensile ber of the wall at the location of maximum bending moment) doesnt have any signicant implications for wall

78

5.1 Introduction

behaviour at a practical level. The results of the stochastic analysis were compared with one current method outlined in Stewart and Lawrence (2002) which previously attempted to include spatial variability of material parameters using several unit failure hypotheses in combination with linear/elastic theory. This was done to see if that method is an appropriate analytical approximation to the more accurate method of conducting a stochastic non-linear FEA analysis. Structural mechanics for the structural conguration used in the stochastic analysis in this Chapter are also examined in more detail in order to justify the relatively signicant increase in magnitude observed between rst cracking load (approximated with linear/elastic theory) and failure (peak) load approximated by the non-linear FEA model. This was done to accompany aspects of vertical bending wall behaviour previously observed in Chapter 4 but for the boundary conditions present for the full size wall.

5.1.1

Vertical Bending Wall Structural Conguration to be Examined

An approximately 2.5 m 2.5 m (exact dimensions are shown in Table 5.2) wall is chosen to examine how spatial variability of ft aects failure load. It is thought that it is important to consider realistic dimensions for both the height and length of the wall. When spatial variability in unit exural bond strength is introduced into an FEA model, a 2.5 m long wall may potentially react dierently to a 1 m long wall. If there are more joints across the length of the wall, there is potentially a greater chance of a weak joint at which cracking could initiate. This assumption is based in the context of assuming the load redistribution system hypothesis of failure (see Stewart and Lawrence (2002)). This hypothesis presumes that if a unit in a course fails, then the moment carried by that unit will redistribute to the other unfailed units in the course. Once sucient units in that course have failed, the resistance of that course will drop below that required to resist the applied load and the wall will fail. Stewart and Lawrence (2002) state that this failure hypothesis is more realistic than the averaging or weakest link hypothesis for a wall with a large number of units along a course of a wall, where there is ample scope for

79

5.1 Introduction

Simply supported at the top

Bonded to the ground by a bed of mortar

Figure 5.2: Boundary conditions for the 3D full sized wall

redistribution of load. Because of the computational requirements to simulate a 3D non-linear FEA model of a full size wall; a single leaf, clay brick URM wall of dimensions 2.5 m 2.5 m was the longest and the highest wall that could be modelled and simulated within the time frame for the masters research (as each simulation took approximately 20 hours to run). Also for this reason, a parametric study where size and eccentric loading aects are investigated will not be conducted as part of this research. The objective of this research is to take the largest sized wall possible, based on time and computational limitations, and examine

80

5.1 Introduction

Wall Dimensions Number of full bricks per course Number of courses high Mortar joint thickness Brick dimensions Brick type Expanded unit density

2430 mm long x 2464 mm high 10 28 10 mm 234 mm x 112 mm x 78 mm clay brick extruded with no holes or indentations 1800 kg/m3

Table 5.2: Summary of wall dimensions and details used for the 3D FEA model for the full sized wall how spatial variability of unit exural bond strength aects the mean and COV of failure load of a URM wall subject to vertical bending. Realistic boundary conditions are also employed in the model (simply supported at top and bonded to the ground at the base by a course of mortar which also has the potential to fail (see Figure 5.2)). The bricks to be used for all of the full wall simulations are solid clay bricks of dimensions 234 mm 112 mm 78 mm. The mortar joints are 10 mm thick. For a 2.5 m 2.5 m wall, this corresponds to a wall consisting of 10 full units per course and 28 courses in height (see Table 5.2). The non-standard sized brick of 234 mm 112 mm 78 mm was chosen so that when, in future research, the model is used to examine variability in bond strength for walls in two-way bending, the experimental results of Lawrence (1983) can be used to examine the accuracy of FEA model predictions and the wall dimensions between the various modes of bending will have remained constant.

5.1.2

Material Parameters

Based upon various values reported in the literature, average values of Youngs Moduli for brick and mortar were used. To this end, a Youngs modulus of 20000 MPa for bricks and 3000 MPa for mortar were used. These values are important in determining the values for the linear normal and tangential stiness moduli (i.e. kn and kt respectively), as described in Chapter 4. For the linear normal and tangential stiness moduli of the potential brick cracks located mid-length along each brick, these were given articially high values so that the bricks behave as would be expected in reality. Other material properties for the brick

81

5.2 Spatial Analysis

and composite interface element types are listed in Table C.2 which can be found in Appendix C. Mesh densities for the brick and interface elements are the same as those listed and shown in Table 4.2 (3D) and Figure 4.3(b) respectively (also reproduced in Table C.1).

5.1.3

Boundary Conditions

For the full sized wall, the boundary conditions were chosen as those one would expect to see in many practical situations (see Figure 5.2). That is, at the base of the wall, the bottom course of bricks are bonded by a layer of mortar 10 mm thick. At the top of the wall, the wall is restricted from lateral (out-of-plane) displacement but can move in the vertical (z) direction. A detailed summary of the boundary conditions used for the full sized wall are given in Appendix C, Table C.3. In visualizing the boundary conditions, the orientation of the axis should be noted. The global axis of orientation associated with the wall is such that global x is in the direction of the length of the wall, global y is in the direction of the thickness of the wall and global z is in the direction of the height of the wall.

5.1.4

Loading Conditions

The loading conditions for the wall are that of a uniform lateral pressure load over the full face of the wall. The pressure load is slowly applied to the wall, with each load step being determined by an energy based gradient load step size, with a minimum allowable step size of 0.001 kPa and a maximum step size of 0.05 kPa. See Table C.4 in Appendix C for full wall loading specications.

5.2

Spatial Analysis

The mean failure load and COV of the 3D non-linear FEA stochastic analysis with spatially varying wall unit exural bond strengths are summarised in Table 5.3. A plot of mean wall failure load versus wall bond strength COV(ft ) is shown in Figure 5.3(a). The histogram for each bond strength COV with best ts for various distributions are shown in Figures 5.4 to 5.6, and the inverse cumulative distribution function (CDF1 ) with best

82

5.2 Spatial Analysis

2.3

Wall Mean Failure Load (kPa)

2.2

2.1

1.9 0.1

Wall Unit Flexural Bond Strength COV

0.2

0.3

0.4

0.5

(a) Unit exural bond strength COV versus mean failure load (kPa) for the full wall with spatially varying bond strengths
0.08

Wall Mean Failure Load COV

0.06

0.04

0.02

0 0.1

Wall Unit Flexural Bond Strength COV

0.2

0.3

0.4

0.5

(b) Unit exural bond strength COV versus mean failure load COV for the full wall with spatially varying bond strengths

Figure 5.3: Change in mean failure load (kPa) and COV with unit exural bond strength COV

83

5.2 Spatial Analysis

Bond Strength Mean (MPa) ft

Bond Strength COV COV(ft )

Failure Load Mean (kPa)

Standard Deviation of Failure Load (kPa) 0.05

COV of Failure Load

Best Fit Distribution

95% condence interval for population mean using t distribution (2.23,2.25)

0.4

0.1

2.24

0.02

Weibull Gumbel

0.4

0.3

2.16

0.09

0.04

Weibull Gumbel

(2.13,2.19)

0.4

0.5

2.00

0.12

0.06

Weibull Gumbel

(1.96,2.04)

Table 5.3: Summary of wall failure load means and COVs for the full wall with spatially varying unit exural bond strength ts for various distributions are shown in Figures C.1 to C.3 in Appendix C.

18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 2.1 2.2 2.3 2.4 Normal Lognormal Weibull Gamma Gumbel

Probability Density

Spatially Variable Bond Wall Failure Load (kPa)

Figure 5.4: Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.1

84

5.2 Spatial Analysis

10 9 8 7 Normal Lognormal Weibull Gamma Gumbel

Probability Density

6 5 4 3 2 1 0 1.9 2 2.1 2.2 2.3 2.4

Spatially Variable Bond Wall Failure Load (kPa)

Figure 5.5: Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.3

6 Normal Lognormal Weibull Gamma Gumbel

Probability Density

0 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4

Spatially Variable Bond Wall Failure Load (kPa)

Figure 5.6: Peak load probability distribution ts for a ft = 0.4MPa and COV = 0.5

For a mean unit exural bond strength of ft = 0.4 MPa, as COV(ft ) is increased from 0.1 to 0.5, mean wall failure load decreases from 2.24 kPa to 2 kPa. This corresponds to an 11% drop in failure load when bond strength COV is increased from 0.1 to 0.5. To

85

5.2 Spatial Analysis

check to see if the wall failure load means for each COV(ft ) are signicantly dierent to one another, a condence interval of mean wall failure load for each COV was determined. As can be seen in the last column of Table 5.3, the mean wall failure load for each COV(ft ) does not lie in the condence interval of any of the mean wall failure loads for the other COV(f t ). This would indicate that the dierences in the mean wall failure loads are statistically signicant. This was conrmed by performing a two tailed t-test to test the dierence between means. Although the Normal, Lognormal, Weibull, Gamma and Gumbel distributions all described the wall failure data at the KS 5% signicance level for all three failure load COVs, examining the CDF1 plots show that the functions that sit closest to or below the 1:1 line at the lower tail of the distributions (thus overestimating the lower tail of the histogram and providing a conservative approximation) were the Weibull and Gumbel distributions in all three cases. The statistically signicant dierence between wall failure load standard deviations was also tested using the f-test. It was found that the only standard deviations which could not be considered statistically dierent from one another were those of 0.09 and 0.12 for wall failure load COVs of 0.3 and 0.5 respectively. This might indicate that as COV in the spatial variability of unit exural bond strength increases, the corresponding variability in the wall failure loads produced doesnt increase signicantly. A plot of unit exural bond strength COV versus mean peak failure load COV is shown in Figure 5.3(b). It can be seen from Figure 5.3 that the smallest bond strength COV produced the smallest wall failure load COV and the largest mean failure load. The reason for this situation producing the largest mean wall failure load can be attributed to the fact that generally there will be a smaller number of extreme value smaller unit exural bond strengths in the wall and so that generally the bond strengths will be closer to the mean than what could be expected with a larger bond strength COV. Therefore in theory, there wont be as many or as extreme smaller unit exural bond strengths which could initiate failure or instability in the wall. The reason why the COV of wall failure loads is smallest for the smallest bond strength COV, is because as unit exural bond strengths will be situated closer to the mean (compared with the magnitude of bond strengths generated for a larger COV) there will be less percentage of relatively lower or higher value bond strengths which could make the wall overall a lot weaker or stronger than what would be

86

5.2 Spatial Analysis

expected if the bond strength for the whole wall was equal to the mean. The same reasoning can be used to explain why the largest unit exural bond strength COV produces the largest wall failure load COV and the smallest mean wall failure load.

5.2.1

Explaining Small Wall Failure Load COVs Found in Spatial Stochastic Analysis

The next issue to address is how the crack patterns and unit failure progression observed in Chapter 6 could lead to relatively low wall failure load COVs of 0.02, 0.04 and 0.06 being observed in walls with spatially variable bond strengths with COVs of 0.1, 0.3 and 0.5. The reason for this may be explained by referring back to the rst example of the 3D non-linear FEA wall model which had a uniform unit exural bond strength of ft = 0.4 MPa (see Section 6.3 in Chapter 6). Crack development occurred over four courses in the middle of the wall and one course at the base of the wall. Also, although a total of ve courses were observed to possess signicant crack development, it was also observed that several other courses in the vicinity of the cracked courses had also cracked at the extreme ber on their tensile edge and were tension softening to varying degrees, all be it, by relatively small amounts in some cases. Unloading in the non-linear FEA model occurs when negative pivots appear in the global system of equations. Often a negative pivot indicates unstable structural behaviour. If this occurs the sign of the load increment changes from loading to unloading. Taken as a whole, it would appear that the 3D non-linear FEA model wall failure load is more determined, not by the few courses which exhibit cracking, but perhaps also by the courses which have started to crack and are tension softening. This means that in total, many courses are involved in the determination of failure load. If several more courses than expected were involved in determining when peak load and therefore failure load in the wall occurs, then overall, one would expect to see less variability in the failure load results for walls with spatial varying bond strengths which all have the same ft . This is because the average bond strength causing failure due to several courses could be less variable than if failure in the wall was dependent upon one or two courses only.

87

5.3 Non-Spatial Analysis

5.3

Non-Spatial Analysis

In order to examine how the spatially varying bond strength failure load distribution compares with the non-spatially varying bond strength failure load distribution for the full size wall, the following non-spatial analysis was conducted.

Wall Peak Failure Load (kPa)

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4

Wall Unit Flexural Bond Strength (MPa)

Figure 5.7: Cubic spline interpolation t to 3D non-linear FEA wall peak loads for uniform bond strengths ranging from 0.01 MPa to 1.4 MPa

Using a very low probability (i.e. p = 1e6 ) of obtaining bond strengths at the extreme ends of a truncated normal distribution dened by ft = 0.4 MPa, COV(ft ) = 0.5 (the largest variability in bond strength examined), the bond strength values at the extreme ends of the distribution were calculated as being 0 MPa and 1.4 MPa. With this information, a set of 3D non-linear FEA full wall simulations were conducted. Each simulation was the same except that the non-spatial (uniform) unit exural bond strength for the wall (ft ) ranged from 0.01 MPa to 1.4 MPa. The wall failure loads for these simulations are shown in Figure 5.7. With these results, a response surface could be constructed which

88

5.3 Non-Spatial Analysis

describes the relationship between wall failure load and wall unit exural bond strength. Then for a randomly generated bond strength, a failure load could be obtained for the wall by using the best t cubic spline interpolation of the non-spatial runs, shown as a curve t through the points in Figure 5.7. One hundred thousand wall runs were simulated for each bond strength COV to produce a wall failure distribution. The mean wall failure load and corresponding COV for each COV in bond strength is summarised in Table 5.4. The spatial wall failure load means and COVs are also included in Table 5.4 to provide easy comparison with the non-spatial results. The histogram of non-spatial failure loads for each COV, with superimposed spatial histograms are shown in Figures 5.8 to 5.10.

Bond Strength Mean (MPa) 0.4 0.4 0.4

Bond Strength COV 0.1 0.3 0.5

Spatial Wall Failure Mean (kPa) 2.24 2.16 2.00

Spatial Wall Failure COV 0.02 0.04 0.06

Non-Spatial Wall Failure Mean (kPa) 2.25 2.24 2.24

Non-Spatial Wall Failure COV 0.07 0.20 0.32

Table 5.4: Summary of peak wall failure load means and COVs for the full wall with non-spatially generated unit exural bond strengths Comparing the results of the non-spatial and spatial runs, the following is observed. It can rst be observed that the mean wall failure load for all non-spatial runs with a ft = 0.4 MPa and a COV of either 0.1, 0.3 and 0.5 all have essentially the same mean of 2.24 kPa (given that 2.25 - 2.24 kPa). This could be expected in the non-spatial runs because for each wall simulation, one randomly generated bond strength is assigned to all units in the wall. As the non-spatial randomly generated bond strengths are randomly selected from a truncated normal distribution, then it is expected that for all bond strength COVs, the selection of bond strengths will be relatively evenly distributed about the mean. As there is no spatial variation in unit exural bond strength in these walls, the failure load cannot be inuenced by weaker bonds initiating failure, since all bond strengths in the wall are the same. It can be seen in Table 5.4 that as the COV in randomly generated non-spatial bond strengths increase, the COV in wall failure load also increases. This is expected because as the range of values in bond strength increases with an increase in

89

5.3 Non-Spatial Analysis

COV, so will the range in values of wall failure load.

16 14 12

Spatial COV=0.1 NonSpatial COV=0.1

Frequency

10 8 6 4 2 0 1.8 2 2.2 2.4 2.6 2.8

Wall Failure Load (kPa)

Figure 5.8: Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.1

Spatial COV=0.3 NonSpatial COV=0.3

Frequency

0 0.5 1

Wall Failure Load (kPa)

1.5

2.5

3.5

Figure 5.9: Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.3

90

5.3 Non-Spatial Analysis

3 2.5

Spatial COV=0.5 NonSpatial COV=0.5

Frequency

2 1.5 1 0.5 0 0.5 1 1.5

Wall Failure Load (kPa)

2.5

3.5

4.5

Figure 5.10: Peak load distribution for the spatial and non-spatial wall simulations, ft = 0.4MPa and COV = 0.5

Comparing the non-spatial and spatial stochastic analysis, it can be observed and also shown using the t-test, that the mean failure load for the spatial and non-spatial results with a COV of 0.1 are considered not signicantly dierent from one another. This is because when the COV of the spatial bond strengths is small, there will be less smaller valued unit bond strengths which could contribute to weakness in the wall and the wall should behave more like one with a uniform bond strength. Because the spatially variable bond strengths are distributed close to the mean, then the failure load will also be close to the mean failure load of the non-spatial mean failure load. The spatial and non-spatial failure wall load means for the COV in bond strengths of 0.3 and 0.5 are signicantly dierent from one another. Also, as bond strength COV increases, the mean wall failure load for the spatial runs decreases signicantly compared with the means of the non-spatial runs with the same COV. This trend is observed because in the non-spatial runs, if a wall is assigned a larger than average bond strength, then all units in the wall will have this high bond strength and the wall failure load will be larger than average. With the spatial runs, if a percentage of larger bond strengths are assigned to the wall, there will still be a percentage of smaller bond strengths assigned

91

5.4 Comparing Wall Conguration Failure Load with Lower Bound (Linear/Elastic) and Upper Bound (Plastic) Analysis

to the wall. These smaller bond strengths create weaknesses in the wall and the wall is expected to fail at a lower load than if all bond strengths in the wall were high. For the spatial and non-spatial runs with a COV = 0.1, the mean wall failure load is approximately the same as that found for one simulation of the wall with a uniform bond strength of 0.4 MPa. The failure load for this wall is 2.25 kPa. This result is encouraging in that it is a fair indication that the model is providing accurate results over many wall simulations where there is spatial variability in unit exural bond strength.

5.4

Comparing Wall Conguration Failure Load with Lower Bound (Linear/Elastic) and Upper Bound (Plastic) Analysis
Lower Bound Failure Load (Linear/Elastic) Approximation

5.4.1

Examining a lower bound for expected failure load for the full size wall, a linear/elastic analysis was conducted. In order to calculate failure load w using linear beam theory (including self-weight) for the wall, the full sized 3D FEA wall is considered as simply supported at the top and bonded to the ground by mortar, thereby producing another row of units with bond strengths at the base of the wall. Therefore, when a lateral pressure load is initially applied to the full sized wall, it will act as a propped cantilever. Once a course close to the base of the wall fails, the wall then acts as a simply supported beam top and bottom. Examining a propped cantilever in linear beam theory, with the conguration shown in Figure 5.11, and neglecting the eect of self-weight (temporarily - for the sake of explanatory purposes), the bending moment at the xed end (location C) is mC = bending moment at location B is mB = and L is the height of the wall. Based upon the magnitude of these moments, when a pressure load is rst applied to the wall, the maximum moment in the wall will develop at the base of the wall, because the
9wL2 128 , wL2 8

and the

where w is the pressure load on the wall

92

5.4 Comparing Wall Conguration Failure Load with Lower Bound (Linear/Elastic) and Upper Bound (Plastic) Analysis

Figure 5.11: Initially the full sized wall acts as a propped cantilever until stresses due to the moment at the base exceeds bond strength at the extreme tensile bers of the brick/mortar bond

bending moment at the xed end is greater than the bending moment which develops close to mid-height. Therefore, it is expected that the course at the base of the wall will crack rst. In linear/elastic theory, there is no tensile fracture energy and therefore no tensile softening. This means that once the base of the wall cracks, the load will, in the next instant, be transferred to close to mid-height of the wall and the wall will act like it is simply supported. As it happens, for a simply supported wall (neglecting self-weight for explanatory purposes), the peak bending moment located near the mid-height of the wall will also be mmax =
wL2 8 .

Therefore, the failure load of the propped cantilever wall

wL2 8 .

and a simply supported wall both have approximately the same magnitude in maximum moment (if we neglect self-weight) of mmax =

93

5.4 Comparing Wall Conguration Failure Load with Lower Bound (Linear/Elastic) and Upper Bound (Plastic) Analysis

Using this linear/elastic theory and the reasoning in the preceding paragraph, and accounting for the eects of self-weight, the lower bound approximation for failure load, which will occur rst at the base of the wall and in the next instant at close to mid-height of the wall (because of almost equal bending moments as described in the previous paragraph), the rst cracking load is calculated as 1.3 kPa.

5.4.2

Upper Bound Failure Load (Plastic) Approximation

To record the upper bound for expected failure load, a plastic analysis is conducted, based upon the limiting stress distribution for the brick/mortar interface as shown in Figure 5.12. This limiting plastic stress distribution was determined based upon examining the non-spatial wall simulation for a uniform ft = 0.4 MPa and recording the stresses across the joint in the bottom course of the wall until failure, as can observed in Chapter 6, Section 6.3. The limiting plastic stress distribution encapsulate the stress development witnessed.

0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 0 14 28 42 56 70 84 98 112

Stress at node (MPa)

Distance from tensile edge (mm)

Figure 5.12: Plastic limit denition of behaviour obtained from observing the behaviour of the brick/mortar interface subject to loading as shown in Figure 6.5, Chapter 6

94

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

Conducting a plastic analysis with Mp determined by using the stress distribution in Figure 5.12, the plastic failure load is calculated to be 3.9 kPa. Therefore, the failure load of 2.25 kPa, lies between the bounding limits (i.e. linear/elastic = 1.3 kPa, plastic = 3.9 kPa) of what one would expect for failure load of a wall composed of a quasi-brittle material.

5.5

Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory
Introduction

5.5.1

Prior to this masters thesis, many attempts to model spatial variability of material properties for URM walls subject to lateral loads had employed linear/elastic analytical methods which predict rst cracking load. As observed in Chapter 4, rst cracking load as it is dened in the traditional sense (i.e. when stress exceeds strength at the location of peak bending moment at the extreme ber on the tensile edge) doesnt have any signicant physical meaning regarding a stage of key change in wall behaviour or wall failure load. In the context of conducting a reliability analysis, an analytical approach is generally sought because stochastic modelling using non-linear FEA analysis is computationally expensive (and therefore time expensive) and there are size limitations on the structural congurations that can be examined (because of the number of nodes in combination with the size of mesh required to produce reasonably accurate results), especially when there is a requirement to take spatial variability of material parameters into account (which for the case of masonry, currently requires a meso-scale approach). Based upon the non-linear FEA model development presented in Chapter 4 and the stochastic results from this Chapter, it can now be stated that previous stochastic analysis methods utilizing models which predict rst cracking load can be considered as providing something of a lower bound approximation to the problem of variability in failure load due to material variability. The objective of this section is to compare the results of the stochastic analysis with the previous linear/elastic approach outlined in Stewart

95

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

and Lawrence (2002) in order to identify shortcomings of the linear/elastic method and associated failure hypotheses.

5.5.2

Unit Failure Hypothesis Described and Dened

The full wall stochastic FEA results were compared with linear/elastic theory in combination with the three failure hypotheses presented in Stewart and Lawrence (2002). These hypotheses are namely the weakest link, the averaging and the load redistribution hypothesis. These hypotheses are all based on assumptions about how units fail and how these failures collectively progress to wall failure. The weakest link hypothesis doesnt have any real rational for walls which are more than one or two units long, but is included because it gives the lowest of the lower bound approximations. The averaging and load redistribution hypothesis are expected to produce a load closer to the real expected failure load of the wall. Stewart and Lawrence (2002) describe the dierent hypotheses very well. The weakest-link hypothesis assumes that if the weakest unit in a course fails, then the entire course fails. The averaging hypothesis assumes that the unit exural bond strengths along a course will provide a strength corresponding to the average unit exural bond strength of the course. The load redistribution hypothesis assumes that if a unit in a course fails, then the moment carried by that unit will redistribute to the other unfailed units in the course. Once sucient units in that course have failed, the resistance of that course will drop below that required to resist the applied load and the wall will fail. There are several other hypotheses which could have been used (see Baker (1981) for example), but these three hypotheses where considered for the time being. Mathematically, these three hypotheses are dened as follows: Weakest link hypothesis: The weakest joint across the course which will initiate failure is: ft weak = min(r1 , r2 , , rn ) (5.1)

where r1 , r2 , , rn are the bond strengths for the n joints which make up the course.

96

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

Averaging hypothesis: The bond strengths of the joints across the course will average due to plastic behaviour, with the eective bond strength becoming: ft averaging = Load redistribution hypothesis: ft loadredist. = nr1 , (n 1)r2 , 2rn1 , rn , r 1 < r 2 < < rn n (5.3) r1 + r2 + + rn n (5.2)

These failure mechanisms have remained nothing more than hypotheses due to the dicult nature of testing and measuring experimentally the failure status of each unit in a wall, up until the ultimate failure load of a single leaf clay brick URM wall is reached. However, due to the micro-modelling aspect of the simplied micro 3D model developed herein, it is possible to examine the wall mechanics to failure at the unit level and therefore, recommend which hypothesis best represents a lower bound for failure load.

5.5.3

Structural Conguration for the Linear/Elastic Analysis

The boundary conditions for this analysis departs slightly from the FEA simulated wall in that, the bottom is assumed to be simply supported (not bonded by masonry to the oor), whereas the 3D FEA wall was modelled with the base being bonded to the oor by mortar so that failure could also occur near the base of the wall. The rational behind the simple support conditions for the linear/elastic theory aspect of the calculation was described previously in Section 5.4.

5.5.4

Procedure for the Linear/Elastic Analysis

The wall was set up with boundary conditions and loading as shown in Figure 5.13. For a resultant unit exural bond strength of ft result (ft result obtained using one of the three hypotheses) for a course being a vertical distance (j) from the top of the wall, the lateral pressure load w which is required to exceed ft result for the course will be:

w=

I(ft result + P ) A + y 2 lhjlj 2

slk 2 j

(5.4)

where I is the moment of inertia, P is the axial force on the wall due to self-weight, s is the force density of the brickwork, k is the thickness of the wall, j is the distance from the top

97

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

of the wall to the course of interest, l is the length of the wall and h is the height of the wall. Therefore, for each course in the wall, with a resultant ft result obtained according to one of the unit failure hypotheses in Equations 5.1, 5.2 or 5.3, a pressure load w for which that course will fail can be calculated. Then the course which requires the smallest moment at which the resultant bond strength will be exceeded at the extreme ber, is taken as being the failure load for the wall.

Figure 5.13: Structural conguration used for analysis of the wall using the three failure hypotheses and linear beam theory

The process of examining these three failure hypotheses with linear/elastic theory is therefore as follows: 1. Using a truncated normal distribution with a ft = 0.4 MPa and a COV of 0.1, 0.3 and 0.5, generate a set of unit exural bond strengths for each course in the full sized

98

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

wall. 2. For each hypotheses, calculate the resultant ft result for each course in the wall. 3. For each course in the wall, and each hypotheses, calculate the minimum load required to cause each course to exceed the available bond strength of that course. 4. For each hypotheses, determine which course requires the smallest load in order to exceed its resultant bond strength and choose that as the failure load for the wall. 5. Repeat the process for 10000 simulation runs of the wall, and generate a failure load distribution for each hypotheses.

5.5.5

Results of the Hypotheses in Combination with the Linear/Elastic Analysis

The results of this analysis are shown in Table 5.5. The wall capacity histogram for the three bond strength COVs and hypotheses are shown in Figures 5.14 to 5.16. To facilitate evaluation of the analysis, it should rst be noted that the rst cracking load obtained using the 3D FEA model for the given structural conguration with a uniform ft = 0.4 MPa, is 1.3 kPa.

Bond Strength Mean=0.4 (MPa) Weakest link Mean (kPa) Weakest link COV Averaging Mean (kPa) Averaging COV Load Redist. Mean (kPa) Load Redist. COV

Bond Strength COV=0.1 1.08 0.05 1.3 0.02 1.11 0.03

Bond Strength COV=0.3 0.49 0.27 1.21 0.05 0.85 0.07

Bond Strength COV=0.5 0.26 0.27 1.15 0.08 0.75 0.09

Table 5.5: Summary of wall failure load using the weakest link, averaging and load redistribution hypothesis

99

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 0.5

Weakest Link Averaging Load Redistribution

Frequency

Wall Peak Load (kPa)

1.5

Figure 5.14: Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.1

7 6 Weakest Link Averaging Load Redistribution

Frequency

5 4 3 2 1 0 0

Wall Peak Load (kPa)

0.5

1.5

Figure 5.15: Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.3

100

5.5 Comparing the Stochastic FEA Results with Three Failure Hypotheses in Combination with Linear Beam Theory

7 6

Weakest Link Averaging Load Redistribution

Frequency

5 4 3 2 1 0 0 0.5 1 1.5

Wall Peak Load (kPa)

Figure 5.16: Peak load distribution for the three failure hypothesis, ft = 0.4MPa and COV = 0.5

Comparing the stochastic FEA simulation results in Table 5.4 with Table 5.5, there is a signicant dierence in the magnitude between the two sets of wall strength capacities. As can be seen for a ft = 0.4 MPa and COV of 0.1, the spatially variable non-linear FEA result provides a peak wall load of w = 2.24 kPa, the non-spatial non-linear FEA result provides a peak wall failure load of w = 2.25 kPa and the linear/elastic theory with various hypothese provided wall capacity loads of w = 1.08, 1.3 and 1.11 kPa. Comparing the linear/elastic hypotheses results with the stochastic FEA spatial results when the COV = 0.1, the averaging hypothesis provides the closest approximation to rst cracking load and variability in failure load found for the spatial stochastic analysis. The load redistribution hypothesis underestimated mean rst cracking load but relatively accurately approximated variability in failure load found in the spatial stochastic FEA. The weakest link hypothesis inaccurately predicted mean rst cracking load and variability in cracking load compared with the stochastic FEA spatial results. This investigation clearly shows that the use of linear/elastic beam theory in combination with the various hypotheses, at best produce a lower bound solution to the problem of es-

101

5.6 Accounting for the Large Dierence in Magnitude Between First Cracking and Failure Load for the the Full Size Wall Structural Conguration

timating mean and COV of wall failure load considering spatial variability in unit exural bond strength. However, it is recommended that linear/elastic theory with any of these hypothesis should not be used for future eorts which attempt to predict mean failure load and failure load variability as they all signicantly underestimate wall capacity.

5.6

Accounting for the Large Dierence in Magnitude Between First Cracking and Failure Load for the the Full Size Wall Structural Conguration

A point of contention in this research was addressing the issue of how the inclusion of tensile fracture energy can have a signicant eect on the failure load of a wall, especially given that linear/elastic theory which predicts rst cracking load is considered an acceptable approximation for wall capacity for clay brick URM walls subject to vertical bending. For example, in the case of the 3D modelled full sized wall, bonding the bottom course of bricks to the ground with a mortar joint and including tensile fracture energy in the model increased the load capacity of the wall from a rst cracking load of 1.3 kPa to a failure (peak) load of 2.25 kPa. In order to show how this signicant increase in wall capacity occurs, the 2D non-linear full sized FEA wall model was examined in more detail. When the 2D model has a uniform ft = 0.4 MPa, with the base bonded to the ground by a mortar joint, the peak load on the wall is 2.25 kPa, with maximum deection occurring at course 17 of 28 from the bottom of the wall. At peak load and at the location of maximum wall deection, the stress at the extreme ber on the tensile side of the wall is 0.364 MPa. That is, the exural bond strength of 0.4 MPa has been exceeded at the extreme ber prior to peak load and the course is now tension softening. At peak load, tensile softening in the course at the base of the wall on the loaded side is in an advanced development stage and is shown in Figure 5.17. This residual strength capacity at the base of the wall creates something of a resistance moment to the load on the wall. Using this information, it is possible, using linear/elastic theory with the case of a simply supported wall top and bottom with an applied bending moment at the base of the wall, to see how much extra load capacity it provides to the wall. The moment is, as shown

102

5.6 Accounting for the Large Dierence in Magnitude Between First Cracking and Failure Load for the the Full Size Wall Structural Conguration

0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 0 14 28 42 56 70 84 98 112

Stress at a node (MPa)

Distance from tensile edge at base (mm)

Figure 5.17: Stress distribution across the thickness of the brick/mortar joint at the base of the wall at peak load for the wall with uniform ft = 0.4 MPa

above, the result of the stress distribution which produces a restoring moment in the wall (see Figure 5.18). This is a crude approximation which doesnt take into account other small moments which have been created within the non-linear FEA analysis due to the fact that the courses above and below the peak deection course have also cracked and are now tension softening. It also assumes purely brittle failure (i.e. no consideration for the eects of strain energy). However, it should be sucient to show that the magnitude of the increase in peak load with the resistance of stresses (due to tensile softening in combination with the compressive strength capacity at the compressive edge of the wall) at the base of the wall is signicant to see a jump in peak pressure load on the wall of 1.3 kPa to 2.25 kPa as shown in the 2D non-linear FEA model. The the new linear beam situation which accounts for a restoring moment at the base of the wall due to residual tensile/compressive resistance is shown in Figure 5.18. For the structural conguration shown in Figure 5.18, the pressure load on the wall w, is then calculated by Equation 5.5.

103

5.6 Accounting for the Large Dierence in Magnitude Between First Cracking and Failure Load for the the Full Size Wall Structural Conguration

Figure 5.18: Modied linear beam conguration to account for restoring moment at the base of the wall due to tensile softening of the bottom brick/mortar joint

w=

I(+ P ) A y

+ slk 2 j +
lhjlj 2 2

M1 j h

M1

(5.5)

where M1 is the moment at the base of the wall due to the stress distribution along the elements at the base of the wall, is the residual tensile strength at the outermost ber of the course at the location of maximum deection, j is the distance from the top of the wall to the course of maximum deection, and the other terms were described previously. The restoring moment which acts at the base of the wall due to tension softening in combination with the compressive capacity of the base brick/unit bond can be calculated

104

5.6 Accounting for the Large Dierence in Magnitude Between First Cracking and Failure Load for the the Full Size Wall Structural Conguration

using Equation 5.6:

M1 =
A

ydA

(5.6)

where yda is the moment of force on an area dA about the neutral axis at which the moment acts. Taking the stress distribution shown in Figure 5.17 and using Equation 5.6, a base restoring moment of M1 = 1.36 Nm is obtained. Substituting M1 into Equation 5.5, with = 0.364 MPa, j = (2817) 0.088 m = 1.056 m, yields a wall pressure load of w = 2.29 kPa. This value is expected to be slightly higher than the 2D and 3D non-linear model peak pressure load of 2.25 kPa because of the fact that the simple linear/elastic theory calculation (Equation 5.5), as stated previously, doesnt take into account the softening of the courses above and below the course for which peak deection occurs and doesnt consider the eects of strain energy. Therefore, due to the existence of tensile fracture energy combined with the compressive capacity of the joint, the base of the wall still has strength capacity at peak load and doesnt lose all strength when the outer most ber ft is exceeded by the stress, as is assumed in linear/elastic rst cracking theory. Van Der Pluijm (1992) states that from fracture mechanics, it is known that the post-peak behaviour (softening) plays a very important role in the failure of a URM wall subject to lateral loading. They state that due to the tensile softening, it is still possible to reach an equilibrium and a higher internal moment after the tensile strength is reached at the extreme ber. It can also be observed in Figure 5.12 that compressive strength of the material exceeds tensile strength and therefore, in combination with tensile fracture energy, enables a higher joint capacity.

105

5.7 Discussion of Stochastic Analysis Results

5.7

Discussion of Stochastic Analysis Results

Upon reection of the stochastic FEA results a few noteworthy observations were made and a better, more realistic approach to the spatial bond strength stochastic analysis realised. In the case of the spatial stochastic analysis, each wall simulated with spatial variability in bond strength had a ft = 0.4 MPa and a specied COV. This assumes that many constructed walls with the same structural conguration and material properties will have the same wall ft . However, this will not be the case. For example, in reality many walls built with one structural conguration might be built using the specications of, for example, extruded clay bricks with a 1:1:6 mortar. As was seen in the experimental walls built and presented in Chapter 3, several workmanship and environmental factors can play a role to signicantly change the ft of walls which have the same structural conguration and material properties (in that case it was brick suction rate due to environmental factors). Therefore, each walls ft can be considered instead as a sample mean ft sample . The population mean ft pop should be considered as the mean of all ft sample (where each wall has the same structural conguration and material properties). Taking this into consideration, a more realistic approach to the analysis would be rst to generate a sample mean bond strength ft sample from a predetermined distribution with a population mean bond strength ft pop and a predetermined COV which reects workmanship and environmental factors aecting wall mean bond strength variability. ft sample could then be used to generate spatially varying bond strengths in the wall using a predetermined distribution with specied COV reecting inherent material variability (e.g. variability in graded aggregate used in the mortar). This process could be repeated many times to generate a failure probability distribution more reective of what could be expected in reality. This would address the issue of external factors such as workmanship and weather on the overall strength of the wall, and inherent bond strength variability due to material variability. In this case, the COVs of wall failure load will be signicantly higher and more realistic than what was found for the spatial stochastic analysis results in Section 5.2 and will reect similar, but very slightly higher wall failure COVs than those found for the non-spatial stochastic analysis (however the result will depend upon the distribution and

106

5.7 Discussion of Stochastic Analysis Results

COV used to generate ft sample from ft pop ). This will occur due to the relationship COVresultant = COVsample 2 + COVpop 2 . For example, for a COV(ft ) = 0.5, assuming COVpop = 0.32 and COVsample = 0.06, then COVresultant = 0.33. To the authors knowledge, tensile fracture energy or residual strength in the mortar joints due to tension softening in combination with the compressive strength capacity of the joint is not a parameter which has been considered in previous analytical methods. It would seem evident, based upon the detailed investigation conducted herein with stochastic non-linear FEA, that none of the three failure hypotheses presented would reasonably approximate wall strength capacity and can signicantly underestimate it. It is suggested, based upon research ndings, that a failure hypothesis for how the units fail in the wall should also include residual strength in the wall due to the eects of tensile softening in combination with the compressive edge strength capacity of the joint, given a certain set of wall boundary conditions. A failure hypothesis should also accommodate global eects which can signicantly eect overall strength of the wall, such as weather conditions and workmanship factors.

107

Chapter 6

Unit Failure Progression in a Wall

6.1 Introduction

With 3D non-linear FEA modelling of the full sized wall presented in Chapter 5, it is possible to examine unit crack opening development as the wall progresses to failure. This is best done visually by examining the crack openings in the wall for various load steps in a simulation. This unit crack opening development pattern was examined for both the full sized wall consisting of a non-spatial (fully correlated/uniform) bond strength, and for the wall with spatially varying bond strengths.

6.2

Expected Behaviour Based Upon Statics

When the pressure load is rst applied to the full sized wall, the maximum moment in the wall will develop at the base of the wall, because as was shown previously (see Figure 5.11) the bending moment at the base (xed end) is greater than the bending moment which develops close to mid-height. Therefore, in the FEA model, it is expected that the course at the base of the wall will crack before the a course close to mid-height will crack. This is under the assumption that a uniform bond strength exists for the entire wall. For the wall with spatial variability in bond strength, the course near the base of the wall which cracks rst will depend upon which course has the overall weaker bond strengths in combination with the most signicant bending moment.

108

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

Once a course towards the base of the wall cracks, the wall will then act like a simply supported wall top and bottom. Then the wall is expected to fail in a course close to mid-height of the wall at the location of maximum bending moment (which if self-weight is considered should be just above mid-height).

6.3

Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

The expected behaviour is rst compared with the 3D non-linear FEA model with a non-spatial (fully correlated) bond strength. The bond strength ft is 0.4 MPa with an associated tensile fracture energy Gf I of 0.0068 N/mm as in accordance with Equation 4.2. For the simulated results, the following is observed. Stress exceeds strength at the extreme ber at the bottom course of the wall on the tensile edge (loaded side of wall) at a load of 1.3 kPa (rst cracking), as was also seen in the 2D non-linear full wall model. From a pressure load of 1.3 kPa to 2 kPa, the bottom two courses show tensile softening, with stress exceeding strength on the tensile edge of the third course horizontal mortar joints at a load of 2 kPa. Towards mid-height of the wall, stress rst exceeds strength of the extreme bers of the horizontal mortar joints in courses 18 and 19 (counting from the bottom course) at a load of 2 kPa. From a pressure load of 2 kPa to 2.25 kPa, stress exceeds strength in the extreme tensile bers of courses 14 to 21, with various degrees of signicant tensile softening occurring, in particular for courses 15 to 18. Immediately after the model begins to unload (after the peak load of 2.25 kPa is reached), signicant crack opening occurs at the bottom course of the wall (but not in courses 2 and 3). In the next instant, after crack opening begins to be visually signicant (> 0.01 mm) in the bottom course of the wall, crack opening also develops in courses 15 to 18. Courses 15 to 18 appear to open together with approximately the same size crack widths for each of these courses. After crack opening has developed at the base and near mid-height of the wall, the simulation stops unloading at a load of 2.07 kPa.

109

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

Course Number 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Bottom course appears to remain relatively bonded to the ground despite first cracking having occurred

Figure 6.1: For wall with uniform ft = 0.4 MPa, despite rst crack (rst crack load = 1.3 kPa) being reached at the base of the wall, it can be seen from the reverse curvature at the base, that the base remains relatively bonded to ground by the mortar joint at a load of 1.4 kPa (load step 8)

Summarising, Figure 6.1 shows the deection of the wall at a load of 1.4 kPa. Figure 6.1 shows that successive cracking at the base of the wall did not occur automatically after the stress at the extreme tensile ber at the bottom course occurred (which occurs at a load of 1.3 kPa). Figure 6.2 shows that at a pressure load of 2.25 kPa (peak load) how cracking at the base of the wall has occurred and the wall now acts like it has simple supports top and bottom. Figure 6.3 shows crack opening in courses 15 to 18 on the unloaded face (predominately tensile side) of the wall which occurs after wall unloading commenced.

110

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

Cracking of base has occurred and now wall acts simply supported top and bottom

Figure 6.2: For wall with uniform ft = 0.4 MPa, with continued loading, bottom mortar joint cracks and tension softening occurs, Failure (Peak) Load = 2.25 kPa (load step 16)

Maximum out-of-plane deection occurs at course 16. Examining a node in course 16, the load/deection curve is shown in Figure 6.4. For three load stages, i.e. just after rst cracking load (1.4 kPa), at peak load (2.25 kPa) and after peak load (2.08 kPa), the stress and crack opening distribution across the thickness of the mortar joint at both the bottom course of the wall and at the course of maximum lateral deection (course 16) are shown in Figures 6.5 to 6.8.

111

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

Significant cracking has occurred over courses 15 to 18

Figure 6.3: For wall with uniform ft = 0.4 MPa, after failure (peak) load, cracks developed over courses 15 to 18 of the pressure loaded wall, Load = 2.08 kPa (load step 18)

Figure 6.5 shows that at the base of the wall, at approximately rst cracking load, the extreme ber stress has reached the strength of the material. Then, as load continues, the extreme bers tensile soften and as the stress in the joint increases, the internal ber strength is exceeded and also tension softens. As discussed in Chapter 4, compressive side stresses increase linearly during wall loading. Figure 6.7 shows a similar plot for the stress development across the joint with increasing load for course 16 of the wall. Figures 6.6 and 6.8 show that as load on the wall increases, crack opening in the bottom

112

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

Figure 6.4: For wall with uniform ft = 0.4 MPa, load versus deection curve for a node located at course 16, where peak out-of-plane deection occurs

0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 0

Stress at node (MPa)

1.4 kPa (LS 8) 2.25 kPa (LS 16) 2.08 kPa (LS 18) 14 28 42 56 70 84 98 112

Distance from tensile edge (mm)

Figure 6.5: The stress distribution in the bottom course across the thickness of the joint at various load steps (uniform ft = 0.4 MPa)

course and at the course of maximum deection (course 16) increases. After peak load is reached at a load of 2.25 kPa and unloading of the wall occurs, crack opening increases rapidly which leads to wall instability.

113

6.3 Wall Behaviour for a Wall with Non-Spatial (Fully Correlated) Bond Strength

0.04 0.035 0.03 1.4 kPa (LS 8) 2.25 kPa (LS 16) 2.08 kPa (LS 18)

Crack Opening at node (mm)

0.025 0.02 0.015 0.01 0.005 0 0.005 0.01 0 14 28 42 56 70 84 98 112

Distance from tensile edge (mm)

Figure 6.6: (The crack opening thickness in the bottom course across the thickness of the joint at various load steps (uniform ft = 0.4 MPa)

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 0 14 28 42 56 70 84

Stress at node (MPa)

1.4 kPa (LS 8) 2.25 kPa (LS 16) 2.08 kPa (LS 18) 2.07 kPa (LS 20) 98 112

Distance from compressive edge (mm)

Figure 6.7: The stress distribution at the course of maximum deection (course 16) across the thickness of the joint at various load steps (uniform ft = 0.4 MPa)

114

6.4 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 1

11 10 9 8

x 10

1.4 kPa (LS 8) 2.25 kPa (LS 16) 2.08 kPa (LS 18) 2.07 kPa (LS 20)

Crack opening at node (mm)

7 6 5 4 3 2 1 0 1 2 3 0 14 28 42 56 70 84 98 112

Distance from compressive edge (mm)

Figure 6.8: The crack opening thickness at the course of maximum deection (course 16) across the thickness of the joint at various load steps (uniform ft = 0.4 MPa)

6.4

Wall Behaviour for a Wall with Spatially Variable (NonCorrelated) Bond Strength - Example 1

Figures 6.9 and 6.10 show an example of wall failure progression for a wall with spatial variability in bond strength (ft = 0.4 MPa, COV = 0.5). In Figure 6.9, at the base of the wall, not only a section of the bottom course cracks, but also a section of the second course from the bottom shows signicant cracking. Various weak joints located around maximum bending moment near the mid-height of the wall crack due to their weak bond strengths, but nally, full course cracking occurs along one course near mid-height (see Figure 6.10).

115

6.4 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 1

Crack opening occurring in the weaker joints in courses close to the base of the wall

Figure 6.9: Initially, crack opening occurs in the weaker joints in courses close to the base of the wall, (ft = 0.4 MPa, COV = 0.5)

116

6.4 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 1

Full crack development occurring along one course close to mid-height of the wall

Figure 6.10: At failure, full crack development occurs at a course near mid-height of the wall, (ft = 0.4 MPa, COV = 0.5)

117

6.5 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 2

6.5

Wall Behaviour for a Wall with Spatially Variable (NonCorrelated) Bond Strength - Example 2

Cracking has occurred over several courses at the base of the wall at locations of weaker bond strength

Figure 6.11: Initially, crack opening occurs in the weaker joints in courses close to the base of the wall (ft = 0.4 MPa, COV = 0.5)

More interesting, and more often witnessed was the following crack development pattern. The fully developed crack pattern shows crack development occurring over more than one course near mid-height of the wall. This type of crack pattern occurred for wall failures where the bond strength COVs were 0.1, 0.3 and 0.5. An example of this crack pattern is shown in Figure 6.13. For the specic wall simulation used in this example, initially, as shown in Figure 6.11, cracking occurs through sections of the bottom two courses.

118

6.5 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 2

Cracking near mid-height first develops at several relatively weak bond strengths located in succession (adjacent to one another) in course 16 of the wall

Figure 6.12: As more load is applied, cracking develops rst at a location of extremely weak bond strengths in course 16 (ft = 0.4 MPa, COV = 0.5)

Course 17 16

half unit 0.065

Unit 1 0.26 0.008

Unit 2 0.46 0.11

Unit 3 0.32 0.51

Unit 4 0.17 0.22

Unit 5 0.4 0.29

Unit 6 0.27 0.47

Unit 7 0.52 0.59

Unit 8 0.38 0.27

Unit 9 0.38 0.25

Unit 10 0.32

half unit 0.55

Table 6.1: Unit exural bond strengths (in MPa) for course 16 and 17 in Figure 6.13 At the end of the simulation, near mid-height of the wall, two courses show signicant cracking (see Figure 6.13). These courses are course 16 and 17. To examine why this crack pattern has occurred in the model, the values for unit exural bond strength for

119

6.5 Wall Behaviour for a Wall with Spatially Variable (Non-Correlated) Bond Strength - Example 2

each unit in courses 16 and 17 are examined (see Table 6.1). These bond strengths, read left to right, correspond to the unit exural bond strengths when looking left to right in courses 16 and 17 in Figure 6.13.

Crack propagation continues in course 17 Initiated by weaker bond strengths in course 17 which were located adjacent to weaker bond strengths in course 16

Figure 6.13: At failure, crack has continued propogation in course 17 (ft = 0.4 MPa, COV = 0.5)

In course 16, the rst half unit and the next two units show signicant cracking. These units have relatively small bond strengths of 0.065, 0.008 and 0.11 MPa. Cracking rst occurs at this location near mid-height. The next unit along course 16 has a larger than average bond strength of 0.51 MPa (since ft = 0.4 MPa). Once cracking has occurred in course 16, stresses may redistribute further to the right hand end of the wall. In course

120

6.6 Summary of Crack Pattern Observations

17, the third and fourth bond strengths from the left hand side of the wall are relatively low in value (0.32 and 0.17 MPa). This is where cracking next begins to occur in the wall (see Figure 6.12). As wall displacement increases, the remaining units in course 17 unzip from the third to the tenth unit in the course. One can presume that course 17 is the ultimate course where maximum bending moment exceeds moment resistance capacity of the units in the course.

6.6

Summary of Crack Pattern Observations

In summary, what has been observed is that for the given boundary conditions, in walls with spatial variability in unit exural bond strength, generally sections of cracking initially occur in a course or courses close to the bottom of the wall, the course numbers depend upon the overall strength of the course and the associated bending moment which it has to resist. Simultaneously, cracks may also develop in weak units situated in a few courses either side of the location of maximum bending moment near mid-height of the wall. At some point, a course or several courses near mid-height of the wall, which has the highest bending moment in combination with the largest number of relatively weak neighbouring units develops crack opening. Therefore, it appears that the combination of crack opening at the base of the wall with crack opening near mid-height ultimately leads to failure of the wall. Sometimes some minor unit cracking occurred near mid-height of the wall before cracking occurred at the base of the wall, but cracking at the base of the wall did fully develop soon after as more pressure load was applied to the wall, and denitely occurred before full crack development occurred at mid-height of the wall. Statistics could not be obtained regarding the percentage of courses which cracked overall in the approximately 160 full wall, non-linear FEA Monte Carlo simulations which were conducted as part of the research. The best way one considered doing this was to extract the element strain information which would have provided the normal displacement of the interface elements. A maximum limiting normal displacement at which crack opening

121

6.7 Discussion of Unit Failure Progression Results

could be considered to have occurred could then be determined. Then if any element exceeded this limiting value, that unit could be considered to have cracked. However, based upon the crack patterns which in some cases occurred over several courses, this type of algorithm may have become thrawt with complications regarding determining exactly what was happening. Therefore it was not conducted. But it can be said, based upon visual cracking and deection observations, for walls with spatial variability in bond strength, that wall cracking generally occurred over the bottom three courses of the wall and, as was observed in the failure of the wall without variability in bond strength mentioned above, over courses in the range of courses 14 to 23. These courses are located in the region of maximum bending moment near mid-height of the wall.

6.7

Discussion of Unit Failure Progression Results

As stated in Chapter 5, it would appear that the 3D non-linear FEA model wall failure load is more determined, not by the few courses which exhibit signicant cracking, but perhaps also by the courses which have started to crack and are tension softening. This means that in total, many courses are involved in the determination of failure load. If several more courses than expected were involved in determining when peak load and therefore failure load in the wall occurs, then overall, one would expect to see less variability in the failure load results for walls with spatial varying bond strengths which all have the same ft . This is because the average bond strength causing failure due to several courses would be less variable than if failure in the wall was dependent upon one or two courses only. It would appear based upon a literature search, that it is always assumed that walls subject to vertical bending will crack just slightly higher than mid-height because of selfweight. This was certainly found to be the case in the non-linear FEA model. However, from a literature search, it could not be found if lateral pressure loaded walls in vertical bending, do in fact fail by such mechanisms. Based upon a literature search, information regarding crack patterns from experimental studies for walls subject to vertical bending were never provided. Perhaps one assumes that the case of vertical bending is too simple to waste experimental eort on. However, based upon the results herein, it would seem

122

6.7 Discussion of Unit Failure Progression Results

an appropriate course of action to build many full sized walls subject to vertical bending, with perhaps simple and propped cantilever supports and test the pressure loaded walls to failure, recording how they behave, what failure load they achieve and the ultimate damage pattern. This is suggested as an important aspect of further investigation which could support or disprove the validity of the 3D non-linear FEA results presented herein. As presented in Chapter 4, based upon the details provided in the experimental results of Doherty (2000), the 3D non-linear FEA model did predict mean failure load of the experimental walls particularly well. The assumption in research has been made that one would expect more brittle than plastic behaviour from clay brick URM walls in vertical bending. But the fact is that for reasonably sized walls, it would appear based upon the literature search that this assumption has not been conrmed.

123

Chapter 7

Conclusions and Recommendations

Although constructing structures with masonry has many advantages, unreinforced masonry is well known to be inherently weak when subject to out-of-plane lateral loads. As a structural material, it has proven dicult to model due to its composite structure and the complicated mechanical interactions which occur within the array of bricks and mortar. Therefore, the actual strength of masonry structures, particularly when subject to lateral loading, still cannot be accurately and consistently predicted analytically for a variety of loading congurations. Current design rules used in Australia, US, Canada and Europe are not based on reliability based calibration methods but rather calibrated to past experience. Although it is commonly believed that current design standards produce conservative results which translate into masonry being, in some cases, a relatively expensive and inecient use of construction material, this is the safest option for designers because of the unknown level of safety of masonry structures. It has been shown that masonry material properties, in particular, unit exural bond strength, vary signicantly throughout a structure, despite the fact that often only one type of brick and mortar is used. Unit exural bond strength was previously identied as one of the most important material parameters contributing to the strength of clay brick URM walls in exure. Therefore, in this thesis, this was the material parameter which

124

was focused on in the experimental and numerical study. It was the objective of this research, in the context of clay brick URM walls subject to vertical bending, to examine how unit exural bond strength varied spatially in a clay brick URM wall, determine a best t probability distribution function which can describe expected spatial variability in unit exural bond strength, and determine how this variability aects wall behaviour and variability in failure load using 3D non-linear nite element analysis. It was hoped that modelling a full sized clay brick URM wall in vertical bending using a 3D non-linear FEA model, would more accurately predict variability in wall failure load and allow the examination of crack pattern development as the wall progresses to failure upon being subject to vertical bending. The main ndings of the research are outlined in the subsequent subsections.

7.0.1

Experimental Findings for the Investigation of Spatial Variability of Unit Flexural Bond Strength in Clay Brick URM Walls

The results of the experimental program indicate that although weak correlation in unit exural bond strength existed in some courses and between courses in the experimental walls, these locations were dicult to predict and didnt follow any particular pattern relating to for example, mortar batch. Therefore, although somewhat counter-intuitive, the results indicate that statistically signicant correlation between adjacent unit exural bond strengths is not likely to be observed. Therefore unit exural bond strength can be considered statistically independent. It was also observed that clay brick wall unit exural bond strengths obtained for all of the walls tested best t a truncated Normal probability distribution. It was observed during testing, that the strength of the brick/mortar interface is governed by factors relating to workmanship, the environment and inherent material variability. It would appear that brick suction rate, which, as stated previously can have a signicant inuence on the suction potential between bricks and mortar, can signicantly aect the overall strength of the wall. This was observed in the experimental program because although the fair quality mason was generally more reckless, both of his walls provided

125

higher average bond strengths than those built by the high quality mason. This is because the high quality mason used relatively moist bricks in the construction of his walls, which can lead to an overall decrease in bond strength due to a reduced suction potential between brick and mortar.

7.0.2

Developing the 3D Non-Linear FEA Model

It was found that the TNO DIANA 9.2 FEA package could be used to implement spatial variability in various material parameters and reasonably accurately modelled failure of clay brick URM walls in vertical bending. Uncracked static push test wall results provided by Doherty (2000) were used to validate the models ability to predict load and deection for a clay brick URM wall subject to vertical bending. The 3D non-linear models predicted failure load was in good agreement with the mean failure load from the Doherty experimental wall results. TNO DIANA 9.2 FEA package remained computationally stable when variability of unit exural bond strength and associated tensile fracture energy were implemented into the model. During the model validation process, the importance of considering a realistic value for tensile fracture energy was emphasized. Apart from unit exural bond strength, tensile fracture energy was found to have a signicant eect on the failure load of the clay brick URM walls in vertical bending. A linear relationship between unit exural bond strength and tensile fracture energy was created based upon research results presented by Van Der Pluijm (1997) for a 1:1:6 mortar + air entrainer. This relationship was used to provide an associated tensile fracture energy for each randomly generated bond strength. Two material interface element types, which describe the interaction at the brick/mortar bond, were trialed. These two interface element types were the non-linear tension softening model and the composite interface model (also called the crack-shear-crush model). It was found that using composite interface elements to describe brick/mortar bond behaviour more closely approximated failure load when the model was compared with the Doherty (2000) experimental results.

126

It was also noted that the peak failure load provided by the model was within the bounding limits dened by rst cracking (using linear/elastic theory) and plastic analysis. For the plastic analysis, it was found that for stresses across the thickness of the wall (the brick/mortar bond), the limiting tensile behaviour employed the usual denition for plasticity, but the compressive behaviour of the joint remained dened by an elastic approximation (see Figure 4.24 and Figure 5.12 in Chapters 4 and 5 respectively). Because the brick/mortar bond has signicantly more strength capacity in compression, it is suggested that the lateral load resistance of the brick/mortar bond comes from a combination of the ability of the brick/mortar bond to tensile soften while providing signicant compressive resistance at the compressive edge of the unit.

7.0.3

Stochastic Non-Linear FEA Analysis for Walls With and Without Spatial Variability in Unit Flexural Bond Strength

A single leaf, clay brick URM wall structural conguration with dimensions of 2.5 m 2.5 m, bonded to the ground with mortar, simply supported at the top of the wall and subject to a lateral pressure load (a wall in vertical bending) was modelled with 3D nonlinear FEA to examine the eect of unit exural bond strength variability on wall failure load and behaviour. Spatial (non-correlated) and non-spatial (fully correlated) variability of unit exural bond strength (ft ) and corresponding tensile fracture energy (Gf I ) was incorporated into the model. A stochastic analysis for both scenarios was conducted. The spatially and non-spatially generated bond strengths were dened by a truncated normal distribution with a ft = 0.4 MPa, and COV(ft )s in bond strength of 0.1, 0.3 and 0.5 were examined. It was found that for the spatial stochastic analysis, for bond strength COVs of 0.1, 0.3 and 0.5, COV of wall failure loads were relatively small, being 0.02, 0.04 and 0.06 respectively. For the non-spatial stochastic analysis, for bond strength COVs of 0.1, 0.3 and 0.5, COV of wall failure loads were 0.07, 0.20 and 0.32 respectively. For the spatial stochastic analysis, it was found that with a COV(ft ) increase from 0.1 to 0.5, the mean wall failure load dropped from 2.25 kPa to 2.0 kPa (an 11% reduction). Despite the small magnitude of the mean wall failure load reduction with increased COV, the mean wall failure loads were statistically dierent to one another. For the non-spatial wall simulations, mean

127

failure load stayed relatively constant at 2.24-2.25 kPa. These results could be explained by examining the 3D wall progression to failure. For walls with spatial variability in bond strength, it is expected that wall failure load COVs would be smaller because those walls would consistently be composed of smaller valued bond strengths which would consistently contribute to weakness in the wall. For the non-spatial wall simulations, this eect would not occur as failure load is determined by one uniform weak or strong bond strength. It was proposed that failure of a clay brick URM wall is not governed by one course only cracking, but rather, instability in the wall is governed by several courses in the vicinity of locations of large bending moment. It was shown that various current stochastic approximations which employ a unit failure hypotheses in combination with the linear/elastic approximation for rst cracking load all underestimated wall capacity signicantly. The reason for this is suggested as being because all hypotheses only assume failure is governed by one course and linear beam theory only considers the tensile capacity of a joint and neglects strength capacity available as a result of joint tension softening and the resistance to failure provided by the joint compressive strength on the compression side of the wall. The hypotheses also dont take into consideration factors which aect overall wall bond strength mean which result from inuences such as workmanship and material variability factors, such as (for example), variation in brick suction rate due to weather conditions which can make the overall strength of the wall stronger or weaker. The mean wall failure loads for the spatial and non-spatial stochastic analysis were within the lower and upper bounds dened by linear/elastic and plastic theory respectively.

7.0.4

Unit Failure Progression

The 2.5 m 2.5 m full sized wall laterally loaded, with boundary conditions being that of the base of the wall bonded to the ground with mortar and simply supported at the top of the wall (thereby creating a wall in vertical bending) was examined for unit failure progression as pressure load increased. For the wall with uniform ft = 0.4 MPa, it was found that the bottom two courses exhibited tensile softening up until peak load, with crack opening occurring uniformly along the bottom course. Crack opening at the base

128

7.1 Future Research

was followed by crack opening of four courses in the vicinity of maximum bending moment close to mid-height. Uncontrolled crack growth occurred at the base and in the courses near mid-height after peak load (i.e. failure load) was reached. The stress distributions and crack opening for the base course and for a course close to mid-height were presented for load steps before, at and after peak load. For walls with spatial variability in bond strength, upon examination of numerous wall simulation results, several crack patterns were witnessed. The bottom two courses and one to three courses close to mid-height normally exhibited signicant cracking. If full cracking wasnt observed along only one joint at the base and close to mid-height, wall cracking at locations of maximum bending moment tended to occur in clumps over several courses. It appeared that at least one crack opening clump within the cracked courses tended to be instigated because of signicant joint weakness over several units adjacent to one another in one particular course.

7.1

Future Research

Based upon a comparison of wall failure load COV for the spatial and non-spatial stochastic analysis, a more realistic approach for future modelling attempts of spatial variability in masonry material properties is suggested. It involves rst generating a sample mean bond strength ft sample from a predetermined probability distribution, population mean bond strength ft pop , and COV. This ft sample could then be used to generate spatially varying bond strengths in the wall using a predetermined distribution with specied COV. The chosen probability distribution, ft pop and COV used to generate ft sample , might reect what could be expected to eect overall mean wall bond strength variability amongst like walls due to environmental and workmanship factors. The chosen probability distribution and COV used to generate the spatially varying bond strengths might reect what could be expected to be seen in bond strength variability within a given wall due to workmanship and material variability. This would address the issue of external factors such as workmanship and weather on the overall strength of the wall (which would be addressed through the selection of an appropriate probability distribution, ft pop and COV), as well

129

7.1 Future Research

as the inherent bond strength variability due to material variability (which would be addressed through the selection of an appropriate probability distribution and COV). This approach is expected to generate a wall failure probability distribution more reective of what could be expected in reality. Based upon the research presented herein, mean failure load and potential crack patterns exhibited in the model could still be considered a point of contention because of a lack of appropriate experimental data which could be used to conrm FEA model observations. Perhaps one assumes that the case of vertical bending is too simple to waste experimental eort on. However, based upon the results herein, it would seem an appropriate course of action to build many full sized walls subject to vertical bending, with perhaps simple and propped cantilever supports, and test the pressure loaded walls to failure, recording how they behave, what failure load they achieve and the ultimate damage pattern. This would enable one to improve the non-linear FEA model validation process and taken together (non-linear FEA model and experimental observations) would provide an improved understanding of the simplest type of wall behaviour (vertical bending) before attempting to gain a better understanding of masonry walls subject to horizontal and two-way bending. To be able to create a 3D non-linear FEA model of the equivalent experimental vertical bending wall test results, one would need to conduct the appropriate small specimen tests to measure material parameters like mean bond strength, Youngs moduli of brick and mortar and perhaps even tensile fracture energy (these parameters would be used in the non-linear FEA model). However, this in itself poses a huge obstacle. As was observed in the material test results of Doherty (2000) which accompanied his experimental walls, it was dicult to obtain consistent and therefore reliable material parameter results. These errors in measurement generally come from the inadequacies of the test method and equipment. Therefore, it is dicult to measure parameters in the real world as they are found in the ideal world of science. The next challenge is the fact that in some circumstances the small test specimens dont represent what is found in situ. A good example of this is found in the work of Correa et al. (2008) where, as stated previously, he found that unit exural bond strengths measured from the walls studied in Chapter 3 and unit exural bond strengths measured from the corresponding prisms were not statistically similar in most cases. The next issue is the error in measuring failure load of full sized walls because

130

7.1 Future Research

of slight variability in loading method and measuring devices used on the experimental walls. Then there is the issue of the FEA model errors due to the fact that theory often idealizes observed physical behaviour. Then there are also issues related to how closely the numerical model boundary conditions and loading conditions reect those of the experiments. There are two ways to deal with these issues. The rst long-term way would be to improved numerical and experimental techniques used to simulate and measure wall failure load and behaviour. The short term way to deal with these issues would be to account for them using various error measurement techniques.

131

Appendices

Appendix A

Appendix to Chapter 3
A.1
Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 411 411 411 312 312 312 312 312 311 311 311 213 213 213 213 212 212 212 211 211 114 113 113 111 111 111 111 111

Newcastle Wall Results

Unit 1 Unit 2 411 411 411 312 312 312 312 312 311 311 311 213 213 213 213 212 212 212 211 211 114 113 113 113 111 111 111 111 Unit 3 411 411 411 312 312 312 312 312 311 311 311 213 213 213 213 212 212 212 211 211 114 113 113 113 113 112 111 111 Unit 4 411 411 411 312 312 312 312 312 311 311 311 213 213 213 213 212 212 212 211 211 114 113 113 113 113 112 112 111 Unit 5 411 411 411 312 312 312 312 312 311 311 311 213 223 213 213 213 212 212 211 114 114 113 113 113 113 112 112 111 Unit 6 411 411 411 312 312 312 322 312 311 311 311 213 223 213 223 213 222 212 211 114 114 113 113 113 113 112 112 121 Unit 7 411 411 411 322 312 322 322 312 311 311 311 213 223 223 223 223 222 222 211 114 114 113 113 113 123 112 112 121 Unit 8 411 411 411 322 312 322 322 312 311 321 321 213 223 223 223 223 222 222 221 124 114 124 123 113 123 122 122 121 Unit 9 411 411 411 322 312 322 322 312 311 321 321 213 223 223 223 223 222 222 221 124 114 124 123 123 123 122 122 121 Unit 10 421 421 411 322 322 322 322 312 321 321 321 223 223 223 223 223 222 222 221 134 114 124 123 123 123 122 122 121 Unit 11 421 421 421 322 322 322 322 312 321 321 321 223 223 223 223 223 222 222 221 134 114 124 123 123 123 122 122 121 Unit 12 421 421 421 322 322 322 322 312 321 321 321 223 233 223 223 223 222 222 221 134 114 124 123 123 123 122 122 121 Unit 13 421 421 421 322 322 322 322 322 321 321 321 233 233 233 223 223 222 222 221 134 114 124 123 123 123 122 122 121 Unit 14 421 421 421 322 322 322 322 322 321 321 321 233 233 233 223 223 232 222 221 134 114 134 123 133 133 132 122 131 Unit 15 421 421 421 322 322 322 322 322 321 321 321 233 233 233 223 233 232 222 231 134 114 134 123 133 133 132 131 131 Unit 16 421 421 421 322 322 322 322 322 321 321 321 233 233 233 233 233 232 222 231 134 114 134 133 133 133 131 131 131 Unit 17 421 421 421 322 322 322 322 322 321 321 321 233 233 233 233 233 232 222 232 134 114 134 133 131 131 131 131 131

Table A.1: Newcastle Wall 1, 1:1:6, High Quality Mason, Location of Mortar Batches in the Wall

133

A.1 Newcastle Wall Results

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 0.00 0.76 0.00 0.43 0.00 0.63 0.00 0.83 0.00 1.06 0.00 0.03 0.00 0.50 0.00 0.42 0.00 0.48 0.00 0.62 0.00 0.03 0.00 0.91 0.00 0.06 0.00 0.00

Unit 2 0.55 0.39 0.49 0.33 0.34 0.20 0.49 0.66 0.82 0.60 0.69 0.03 0.34 0.39 0.34 0.36 0.55 0.48 0.58 0.65 0.55 0.12 0.73 0.32 0.93 0.56 0.86 0.00

Unit 3 0.80 0.82 0.65 0.38 0.46 0.35 0.55 0.66 0.63 0.81 0.79 0.06 0.44 0.55 0.27 0.36 0.40 0.44 0.69 0.56 0.73 0.56 0.64 0.53 1.23 0.73 0.36 0.00

Unit 4 0.55 0.76 0.54 0.33 0.36 0.23 0.49 0.66 0.77 0.91 0.34 0.06 0.43 0.36 0.34 0.21 0.35 0.49 0.50 0.40 0.67 0.42 0.82 0.39 0.66 0.65 0.69 0.00

Unit 5 0.70 0.44 0.80 0.18 0.33 0.40 0.52 0.58 0.73 1.07 0.64 0.03 0.54 0.50 0.33 0.35 0.47 0.38 0.47 0.67 0.60 0.48 0.67 0.50 0.53 0.28 0.29 0.00

Unit 6 0.58 0.72 0.49 0.27 0.40 0.27 0.28 0.70 0.53 0.76 0.42 0.05 0.53 0.39 0.29 0.28 0.47 0.38 0.45 0.66 0.73 0.33 0.72 0.41 0.48 0.44 0.30 0.00

Unit 7 0.60 0.79 0.50 0.35 0.37 0.37 0.34 0.52 0.51 0.81 0.69 0.03 0.46 0.30 0.36 0.38 0.40 0.43 0.62 0.62 0.56 0.42 0.55 0.71 0.26 0.46 0.30 0.00

Unit 8 0.45 0.32 0.50 0.26 0.25 0.46 0.37 0.81 0.74 0.93 0.63 0.04 0.49 0.27 0.27 0.46 0.35 0.39 0.47 0.74 0.61 0.33 0.38 0.31 0.38 0.33 0.46 0.00

Unit 9 0.82 0.60 0.42 0.23 0.29 0.35 0.42 0.62 0.33 0.78 0.60 0.05 0.38 0.45 0.22 0.34 0.60 0.42 0.60 0.80 0.66 0.54 0.51 0.52 0.60 0.25 0.58 0.00

Unit 10 0.64 0.56 0.46 0.39 0.49 0.36 0.67 0.78 0.84 0.79 0.61 0.12 0.53 0.40 0.53 0.18 0.47 0.58 0.46 0.53 0.70 0.42 0.47 0.74 0.64 0.73 0.68 0.00

Unit 11 0.61 0.64 0.50 0.17 0.47 0.67 0.68 0.68 0.81 0.60 0.57 0.14 0.44 0.23 0.28 0.28 0.55 0.35 0.44 0.64 0.58 0.50 0.57 0.74 0.30 0.59 0.43 0.00

Unit 12 0.48 0.67 0.71 0.38 0.40 0.80 0.82 0.62 0.55 0.67 0.79 0.13 0.61 0.34 0.44 0.36 0.47 0.46 0.22 0.87 0.71 0.51 0.31 0.60 0.73 0.65 0.31 0.00

Unit 13 0.53 0.34 0.50 0.36 0.36 0.72 0.74 0.52 0.47 0.76 0.70 0.16 0.13 0.37 0.19 0.36 0.41 0.36 0.39 0.63 0.76 0.61 0.39 0.37 0.27 0.49 0.29 0.00

Unit 14 0.44 0.39 0.57 0.26 0.35 0.75 0.75 0.88 0.61 0.81 0.88 0.13 0.43 0.28 0.23 0.45 0.33 0.45 0.61 0.40 0.73 0.64 0.33 0.58 0.47 0.11 0.42 0.00

Unit 15 0.49 0.38 0.62 0.53 0.48 0.91 0.77 1.03 0.52 1.05 0.74 0.34 0.48 0.40 0.27 0.51 0.39 0.63 0.57 0.52 0.71 0.90 0.37 0.40 0.32 0.30 0.68 0.00

Unit 16 0.67 0.27 0.37 0.35 0.48 0.90 0.63 1.18 0.82 0.98 0.55 0.41 0.33 0.42 0.15 0.53 0.73 0.66 0.53 0.59 0.74 0.52 0.52 0.56 0.70 0.94 0.59 0.00

Unit 17 0.60 0.00 0.54 0.00 0.09 0.00 0.55 0.00 0.82 0.00 0.57 0.00 0.32 0.00 0.32 0.00 0.58 0.00 0.49 0.00 0.82 0.00 0.45 0.00 0.70 0.00 0.30 0.00

Table A.2: Newcastle Wall 1, 1:1:6, High Quality Mason, Bond Strength Values (MPa)
Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Unit 1 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111 Unit 2 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111 Unit 3 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111 Unit 4 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 211 112 112 112 111 111 111 111 111 111 Unit 5 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 221 112 112 112 111 111 111 121 111 111 Unit 6 411 411 411 312 312 311 311 311 311 311 311 212 212 212 211 211 211 211 221 112 112 112 111 111 111 121 121 111 Unit 7 411 411 322 312 312 311 311 311 311 311 311 222 212 212 211 211 211 221 221 112 112 112 121 121 121 121 121 111 Unit 8 411 411 322 312 312 311 311 311 311 311 311 222 212 212 211 211 221 221 133 112 122 112 121 121 121 121 121 111 Unit 9 411 411 322 312 312 311 321 311 311 311 311 222 222 212 211 221 221 221 133 112 122 122 121 121 121 121 121 121 Unit 10 421 411 322 312 322 321 321 321 311 321 311 222 222 212 211 221 221 221 133 112 122 122 121 121 121 121 121 121 Unit 11 421 411 322 312 322 321 321 321 311 321 321 222 222 212 211 221 221 221 133 112 122 122 121 121 121 131 121 121 Unit 12 421 411 322 322 322 321 321 321 311 321 321 222 222 222 221 221 221 221 133 112 122 122 121 121 131 131 121 121 Unit 13 421 421 322 322 322 321 321 321 321 321 321 222 232 222 221 221 221 221 133 112 122 122 121 121 131 131 121 121 Unit 14 421 421 322 322 322 321 321 321 321 321 321 222 232 222 231 221 221 221 133 112 122 122 121 121 131 131 121 121 Unit 15 421 421 322 322 322 321 321 321 321 321 321 222 232 232 231 231 231 221 133 112 122 122 121 131 131 131 121 131 Unit 16 421 421 322 322 322 321 321 321 321 321 321 232 232 232 231 231 231 231 133 112 132 132 131 131 131 131 131 131 Unit 17 421 421 322 322 322 321 321 321 321 321 321 232 232 232 231 231 231 231 133 112 132 132 131 131 131 131 131 131

Table A.3: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Location of Mortar Batches in the Wall

134

A.1 Newcastle Wall Results

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 0.40 0.00 0.25 0.00 0.18 0.00 0.18 0.00 0.27 0.00 0.23 0.00 0.52 0.00 0.03 0.00 0.31 0.00 0.31 0.00 0.49 0.00 0.17 0.00 0.19 0.00 0.29 0.00

Unit 2 0.62 0.42 0.51 0.63 0.56 0.39 0.47 0.41 0.54 0.49 0.48 0.35 0.23 0.55 0.04 0.58 0.33 0.65 0.12 0.44 0.14 0.19 0.09 0.07 0.18 0.17 0.24 0.00

Unit 3 0.59 0.24 0.45 0.50 0.59 0.33 0.40 0.32 0.45 0.38 0.36 0.19 0.42 0.59 0.11 0.60 0.78 0.34 0.27 0.32 0.49 0.07 0.07 0.26 0.20 0.18 0.15 0.00

Unit 4 0.53 0.40 0.44 0.58 0.45 0.24 0.54 0.40 0.15 0.36 0.48 0.11 0.45 0.54 0.17 0.44 0.54 0.68 0.14 0.10 0.22 0.22 0.09 0.08 0.18 0.26 0.25 0.00

Unit 5 0.44 0.38 0.44 0.59 0.31 0.16 0.48 0.43 0.56 0.55 0.48 0.28 0.03 0.37 0.37 0.39 0.42 0.26 0.18 0.33 0.26 0.10 0.34 0.14 0.24 0.20 0.19 0.00

Unit 6 0.57 0.44 0.39 0.39 0.43 0.49 0.28 0.47 0.40 0.52 0.61 0.38 0.03 0.15 0.31 0.44 0.41 0.52 0.40 0.16 0.43 0.20 0.05 0.20 0.05 0.44 0.20 0.00

Unit 7 0.41 0.44 0.28 0.65 0.54 0.53 0.47 0.40 0.60 0.29 0.47 0.46 0.06 0.46 0.10 0.33 0.47 0.47 0.19 0.49 0.34 0.34 0.07 0.14 0.15 0.30 0.30 0.00

Unit 8 0.57 0.58 0.51 0.56 0.50 0.26 0.47 0.42 0.46 0.24 0.61 0.09 0.05 0.50 0.07 0.34 0.50 0.48 0.23 0.27 0.09 0.30 0.20 0.35 0.22 0.10 0.23 0.00

Unit 9 0.50 0.60 0.28 0.40 0.42 0.53 0.33 0.49 0.52 0.38 0.31 0.11 0.39 0.49 0.07 0.35 0.72 0.58 0.11 0.18 0.37 0.11 0.24 0.33 0.22 0.13 0.28 0.00

Unit 10 0.54 0.50 0.33 0.58 0.65 0.59 0.25 0.34 0.52 0.45 0.62 0.11 0.43 0.60 0.45 0.50 0.51 0.54 0.32 0.05 0.26 0.08 0.29 0.25 0.34 0.16 0.25 0.00

Unit 11 0.41 0.38 0.38 0.71 0.48 0.55 0.30 0.57 0.46 0.49 0.39 0.31 0.06 0.82 0.23 0.32 0.42 0.40 0.37 0.29 0.17 0.10 0.13 0.21 0.20 0.20 0.29 0.00

Unit 12 0.68 0.47 0.57 0.67 0.54 0.37 0.12 0.49 0.38 0.46 0.42 0.47 0.40 0.58 0.44 0.42 0.49 0.41 0.17 0.42 0.12 0.32 0.25 0.16 0.24 0.11 0.32 0.00

Unit 13 0.22 0.38 0.50 0.68 0.35 0.42 0.55 0.49 0.33 0.53 0.76 0.30 0.50 0.41 0.63 0.37 0.47 0.17 0.44 0.12 0.07 0.12 0.40 0.16 0.23 0.14 0.36 0.00

Unit 14 0.52 0.37 0.37 0.48 0.77 0.49 0.28 0.51 0.48 0.40 0.54 0.37 0.54 0.40 0.27 0.46 0.30 0.25 0.44 0.28 0.24 0.38 0.21 0.27 0.16 0.29 0.31 0.00

Unit 15 0.83 0.39 0.40 0.42 0.52 0.50 0.27 0.66 0.42 0.54 0.55 0.62 0.59 0.09 0.76 0.09 0.42 0.50 0.51 0.06 0.27 0.09 0.24 0.28 0.17 0.19 0.44 0.00

Unit 16 0.77 0.55 0.56 0.43 0.45 0.45 0.23 0.26 0.29 0.50 0.82 0.10 0.45 0.19 0.43 0.42 0.40 0.46 0.51 0.03 0.11 0.06 0.26 0.23 0.38 0.28 0.34 0.00

Unit 17 0.00 0.42 0.00 0.38 0.00 0.29 0.00 0.36 0.00 0.53 0.00 0.10 0.00 0.64 0.00 0.44 0.00 0.53 0.00 0.03 0.00 0.36 0.00 0.16 0.00 0.23 0.00 0.00

Table A.4: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Bond Strength Values (MPa)
Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Unit 1 512 511 511 511 413 413 413 413 412 411 411 332 311 311 311 212 212 212 211 211 112 112 112 112 112 112 112 111 Unit 2 512 511 511 511 511 413 413 413 412 411 411 332 311 311 311 212 212 212 211 211 211 112 112 112 112 112 112 111 Unit 3 512 511 511 511 511 413 413 413 412 411 411 332 311 311 311 212 212 212 211 211 211 211 123 112 112 112 112 111 Unit 4 512 511 511 511 413 413 413 413 412 411 411 332 311 311 311 212 212 212 211 211 211 211 123 123 123 112 112 111 Unit 5 512 511 511 511 413 413 413 413 412 411 411 332 311 311 311 212 212 212 221 211 211 211 123 123 123 123 122 111 Unit 6 512 511 511 511 413 413 413 413 412 411 411 332 321 321 321 212 212 212 221 211 221 221 123 123 123 123 122 111 Unit 7 512 511 511 511 413 413 413 413 412 411 411 332 321 321 321 212 212 222 221 211 221 221 123 123 123 123 122 121 Unit 8 512 511 511 511 413 413 413 422 422 411 411 332 321 321 321 212 212 222 221 211 221 221 132 123 123 123 122 121 Unit 9 521 511 511 511 413 413 422 422 422 411 411 332 321 321 321 212 222 222 221 221 221 221 132 123 123 123 122 121 Unit 10 521 521 521 521 423 423 422 422 422 421 421 332 321 321 321 222 222 222 221 221 221 221 132 123 132 132 122 121 Unit 11 521 521 521 521 423 423 422 422 422 421 421 332 321 321 321 222 222 222 221 221 221 221 132 132 132 132 122 121 Unit 12 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 222 232 232 221 231 221 221 132 132 132 132 132 131 Unit 13 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 222 232 232 231 231 221 221 132 132 132 132 132 131 Unit 14 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 232 232 232 231 231 231 221 132 132 132 132 131 131 Unit 15 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 232 232 232 231 231 231 123 132 132 131 131 131 131 Unit 16 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 232 232 232 231 231 231 123 131 131 131 131 131 131 Unit 17 521 521 521 521 423 423 422 422 422 421 421 332 331 331 331 232 232 232 231 231 231 131 131 131 131 131 131 131

Table A.5: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Location of Mortar Batches in the Wall

135

A.1 Newcastle Wall Results

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 0.57 0.00 0.74 0.00 0.28 0.00 0.32 0.00 0.77 0.00 0.76 0.00 0.61 0.00 0.58 0.00 0.36 0.00 0.81 0.00 0.84 0.00 1.09 0.00 0.68 0.00 0.95 0.00

Unit 2 0.48 0.67 0.52 0.38 0.28 0.26 0.24 0.21 0.82 0.65 1.00 0.54 0.93 0.69 0.89 0.41 0.53 0.59 1.00 0.75 0.56 0.46 1.24 0.32 0.57 1.13 0.80 0.00

Unit 3 0.68 0.57 0.47 0.26 0.17 0.27 0.16 0.15 0.62 0.89 1.01 0.55 0.85 0.56 0.63 0.40 0.46 0.59 0.70 0.66 0.57 0.52 0.33 0.35 0.54 1.35 0.67 0.00

Unit 4 0.73 0.73 0.70 0.20 0.27 0.27 0.17 0.16 0.71 0.96 0.70 0.62 0.62 0.70 0.59 0.43 0.40 0.56 0.54 0.66 0.52 0.38 1.06 0.43 0.64 0.67 0.75 0.00

Unit 5 0.49 0.47 0.37 0.67 0.25 0.26 0.23 0.15 0.70 0.68 0.71 0.57 0.57 0.64 0.72 0.47 0.42 0.43 0.59 0.63 0.67 0.27 0.42 0.35 0.30 0.98 0.67 0.00

Unit 6 0.33 0.36 0.29 0.27 0.22 0.27 0.16 0.18 0.64 0.66 0.69 0.56 0.72 0.78 0.51 0.41 0.38 0.47 0.55 0.37 0.77 0.57 0.54 0.53 0.78 0.48 0.57 0.00

Unit 7 0.33 0.54 0.46 0.24 0.22 0.47 0.22 0.19 0.98 1.06 0.99 0.64 0.74 0.56 0.78 0.51 0.39 0.47 0.54 0.56 0.54 0.50 0.52 0.55 0.45 0.97 0.48 0.00

Unit 8 0.65 0.47 0.34 0.57 0.17 0.50 0.13 0.12 0.95 1.17 1.00 0.68 0.81 0.62 0.88 0.56 0.57 0.62 0.50 0.57 0.41 0.28 0.83 0.49 0.69 1.09 0.67 0.00

Unit 9 0.77 0.59 0.34 0.25 0.12 0.42 0.12 0.10 1.09 1.49 0.77 0.80 0.85 0.76 0.61 0.56 0.43 0.62 0.55 0.71 0.46 0.36 0.41 0.99 0.56 0.40 0.68 0.00

Unit 10 0.67 0.36 0.27 0.22 0.11 0.53 0.11 0.11 1.09 1.79 0.92 0.82 0.82 0.50 0.68 0.54 0.56 0.53 0.38 0.73 0.58 0.55 0.61 0.68 0.58 0.71 0.69 0.00

Unit 11 0.57 0.59 0.51 0.16 0.11 0.58 0.21 0.13 0.75 0.59 0.82 0.71 0.71 0.61 0.61 0.55 0.59 0.61 0.36 0.39 0.49 0.37 0.31 0.29 0.46 0.51 0.51 0.00

Unit 12 0.65 0.63 0.22 0.17 0.15 0.54 0.24 0.13 0.65 1.03 1.15 0.87 0.74 0.81 0.83 0.54 0.73 0.58 0.50 0.56 0.29 0.67 0.66 0.77 0.79 0.60 0.63 0.00

Unit 13 0.63 0.74 0.19 0.21 0.25 0.58 0.28 0.12 0.85 0.99 0.98 0.74 0.75 0.75 0.71 0.73 0.49 0.47 0.58 0.84 0.42 0.47 0.76 0.78 0.80 0.54 0.72 0.00

Unit 14 0.33 0.92 0.22 0.22 0.29 0.44 0.16 0.25 0.62 1.27 0.92 0.58 0.58 0.71 0.71 0.54 0.52 0.33 0.58 0.26 0.66 0.80 0.51 0.40 0.76 0.48 0.75 0.00

Unit 15 0.57 0.54 0.38 0.24 0.29 0.36 0.23 0.13 1.06 0.83 0.83 0.63 0.50 0.68 0.81 0.64 0.63 0.36 0.48 0.47 0.49 0.29 0.33 0.68 0.96 0.42 0.86 0.00

Unit 16 0.59 0.72 0.16 0.18 0.28 0.55 0.28 0.17 0.93 0.99 0.76 0.48 0.43 0.55 0.57 0.55 0.47 0.27 0.73 0.52 0.67 0.66 0.45 1.35 0.12 0.66 1.01 0.00

Unit 17 0.00 0.50 0.00 0.13 0.00 0.37 0.00 0.16 0.00 0.91 0.00 0.33 0.00 0.51 0.00 0.56 0.00 0.47 0.00 0.75 0.00 0.66 0.00 1.09 0.00 0.95 0.00 0.00

Table A.6: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Bond Strength Values (MPa)
Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Unit 1 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 112 112 112 112 112 112 111 Unit 2 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 112 112 112 112 112 111 Unit 3 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 112 112 112 111 Unit 4 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 114 113 112 111 Unit 5 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 114 113 113 111 Unit 6 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 114 113 113 111 Unit 7 411 411 411 322 312 312 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 114 113 113 111 Unit 8 411 411 411 322 322 322 311 311 311 311 311 311 211 211 211 211 211 211 211 211 125 115 115 114 114 113 113 111 Unit 9 411 421 411 322 322 322 311 311 311 311 311 311 211 221 211 211 211 211 211 211 125 115 115 123 123 122 113 111 Unit 10 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 211 125 124 124 123 123 122 113 111 Unit 11 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 124 124 123 123 122 122 121 Unit 12 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 124 124 123 123 122 122 121 Unit 13 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 124 124 123 123 122 122 121 Unit 14 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 124 124 123 123 121 121 121 Unit 15 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 124 124 121 121 121 121 121 Unit 16 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 125 121 121 121 121 121 121 121 Unit 17 421 421 421 322 322 322 321 321 321 321 321 321 221 221 221 221 221 221 221 125 121 121 121 121 121 121 121 121

Table A.7: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Location of Mortar Batches in the Wall

136

A.1 Newcastle Wall Results

Course 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Unit 1 0.00 0.24 0.00 0.50 0.00 0.43 0.00 0.48 0.00 0.43 0.00 0.44 0.00 0.09 0.00 0.07 0.00 0.39 0.00 0.39 0.00 0.03 0.00 0.26 0.00 0.15 0.00 0.00

Unit 2 0.46 0.27 0.35 0.70 0.58 0.49 0.43 0.76 0.39 0.57 0.56 0.07 0.08 0.08 0.19 0.13 0.20 0.42 0.32 0.51 0.64 0.34 0.40 0.46 0.30 0.50 0.27 0.00

Unit 3 0.39 0.40 0.58 0.73 0.32 0.57 0.33 0.51 0.48 0.60 0.45 0.51 0.05 0.07 0.21 0.41 0.32 0.46 0.31 0.44 0.40 0.54 0.48 0.50 0.41 0.43 0.44 0.00

Unit 4 0.37 0.29 0.40 0.61 0.36 0.48 0.46 0.79 0.65 0.55 0.54 0.51 0.29 0.11 0.24 0.37 0.39 0.25 0.32 0.22 0.29 0.38 0.28 0.47 0.62 0.51 0.54 0.00

Unit 5 0.21 0.21 0.48 0.61 0.19 0.67 0.33 0.42 0.51 0.69 0.44 0.46 0.37 0.10 0.42 0.44 0.12 0.32 0.33 0.21 0.36 0.14 0.29 0.36 0.57 0.42 0.41 0.00

Unit 6 0.15 0.33 0.35 0.52 0.64 0.58 0.33 0.29 0.41 0.52 0.59 0.52 0.25 0.09 0.40 0.43 0.24 0.39 0.22 0.24 0.46 0.48 0.40 0.27 0.64 0.26 0.23 0.00

Unit 7 0.29 0.25 0.68 0.71 0.42 0.57 0.24 0.42 0.27 0.53 0.67 0.42 0.32 0.08 0.13 0.06 0.32 0.29 0.35 0.22 0.56 0.46 0.59 0.33 0.59 0.22 0.21 0.00

Unit 8 0.30 0.38 0.64 0.58 0.14 0.55 0.43 0.28 0.52 0.58 0.56 0.38 0.14 0.29 0.08 0.16 0.24 0.34 0.20 0.37 0.69 0.64 0.25 0.32 0.48 0.51 0.54 0.00

Unit 9 0.69 0.36 0.72 0.57 0.60 0.44 0.20 0.56 0.40 0.59 0.56 0.58 0.42 0.33 0.32 0.16 0.13 0.45 0.33 0.36 0.47 0.59 0.59 0.61 0.59 0.40 0.37 0.00

Unit 10 0.81 0.59 0.42 0.70 0.49 0.37 0.30 0.56 0.58 0.65 0.70 0.61 0.51 0.18 0.47 0.37 0.20 0.70 0.39 0.58 0.41 0.58 0.55 0.35 0.59 0.41 0.37 0.00

Unit 11 0.43 0.43 0.49 0.59 0.42 0.35 0.39 0.46 0.55 0.50 0.39 0.57 0.18 0.19 0.39 0.11 0.17 0.35 0.43 0.48 0.80 0.65 0.40 0.29 0.59 0.62 0.58 0.00

Unit 12 0.79 0.44 0.46 0.60 0.58 0.52 0.47 0.61 0.53 0.43 0.47 0.42 0.42 0.25 0.43 0.24 0.26 0.47 0.45 0.60 0.54 0.41 0.34 0.38 0.65 0.50 0.50 0.00

Unit 13 0.34 0.26 0.19 0.57 0.54 0.44 0.61 0.61 0.39 0.67 0.57 0.47 0.31 0.22 0.23 0.21 0.18 0.31 0.41 0.80 0.27 0.42 0.42 0.41 0.64 0.31 0.31 0.00

Unit 14 0.44 0.36 0.80 0.66 0.40 0.50 0.43 0.60 0.40 0.57 0.49 0.36 0.13 0.20 0.14 0.18 0.09 0.34 0.39 0.36 0.30 0.49 0.37 0.38 0.45 0.14 0.16 0.00

Unit 15 0.58 0.53 0.58 0.53 0.62 0.59 0.45 0.36 0.47 0.57 0.44 0.03 0.13 0.32 0.12 0.21 0.17 0.48 0.30 0.66 0.40 0.49 0.42 0.16 0.42 0.27 0.27 0.00

Unit 16 0.79 0.63 0.65 0.60 0.59 0.47 0.56 0.68 0.48 0.43 0.43 0.03 0.03 0.34 0.31 0.48 0.32 0.34 0.31 0.64 0.35 0.22 0.18 0.11 0.16 0.23 0.23 0.00

Unit 17 0.29 0.00 0.59 0.00 0.25 0.00 0.33 0.00 0.24 0.00 0.40 0.00 0.03 0.00 0.33 0.00 0.05 0.00 0.28 0.00 0.05 0.00 0.14 0.00 0.20 0.00 0.16 0.00

Table A.8: Newcastle Wall 4, 1:1.5:6 + a.e. , Fair Quality Mason, Bond Strength Values (MPa)

137

A.2 Greyscale Plots for Unit Flexural Bond Strengths in Each of the Newcastle, Australia Walls

A.2

Greyscale Plots for Unit Flexural Bond Strengths in Each of the Newcastle, Australia Walls

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 1

1.2

Course Number

0.8

0.6

0.4

0.2

10

11

12

13

14

15

16

17

Figure A.1: Newcastle Wall 1, 1:1:6, High Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa)

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 1

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Course Number

10

11

12

13

14

15

16

17

Figure A.2: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa)

138

A.2 Greyscale Plots for Unit Flexural Bond Strengths in Each of the Newcastle, Australia Walls

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 1

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0

Figure A.3: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa)

Course Number

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 1

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Course Number

10

11

12

13

14

15

16

17

Figure A.4: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Flexural Bond Strenghts Dipicted in a Grayscale (MPa)

139

A.3 Unit Flexural Bond Strengths Plotted for Each Course in Each Newcastle, Australia Wall

A.3

Unit Flexural Bond Strengths Plotted for Each Course in Each Newcastle, Australia Wall

Figure A.5: Newcastle Wall 1, 1:1:6, High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa)

Figure A.6: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa)

140

A.3 Unit Flexural Bond Strengths Plotted for Each Course in Each Newcastle, Australia Wall

Figure A.7: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa)

Figure A.8: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Flexural Bond Strengths Plotted for Each Course (MPa)

141

A.4 Average Unit Flexural Bond Strength for Each Course in Each Newcastle, Australia Wall

A.4

Average Unit Flexural Bond Strength for Each Course in Each Newcastle, Australia Wall

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0.1

Batch 4

Batch 3

Course Number

Batch 2

Batch 1

0.15

0.2

0.25

0.3

0.35

0.4 0.45 0.5 0.55 0.6 0.65 0.7 Average Unit Flexural Bond Strength (MPa)

0.75

0.8

0.85

0.9

0.95

Figure A.9: Newcastle Wall 1, 1:1:6, High Quality Mason, Average Unit Flexural Bond Strength for Each Course (MPa)

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Average Unit Flexural Bond Strength (MPa)

Batch 4

Batch 3

Course Number

Batch 2

Batch 1

0.8

0.9

Figure A.10: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strenght for Each Course (MPa)

142

A.4 Average Unit Flexural Bond Strength for Each Course in Each Newcastle, Australia Wall

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Average Unit Flexural Bond Strength (MPa) 0.8

Batch 5

Batch 4

Course Number

Batch 3

Batch 2

Batch 1

0.9

Figure A.11: Newcastle Wall 1, 1:1:6, High Quality Mason, Average Unit Flexural Bond Strength for Each Course (MPa)

28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Average Unit Flexural Bond Strength (MPa) 0.8

Batch 4

Batch 3

Course Number

Batch 2

Batch 1

0.9

Figure A.12: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Average Flexural Bond Strenght for Each Course (MPa)

143

A.5 Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls

A.5

Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls
3 Truncated Normal Lognormal Weibull Gamma Gumbel

2.5

Probability Density

1.5

0.5

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4

Unit Flexural Bond Strength (MPa)

Figure A.13: Newcastle Wall 1, 1:1:6, High Quality Mason, Probability Distribution Fits

1.2

(Predicted Unit Flex. Bond Strength (MPa)

Perfect Fit Truncated Normal Lognormal Weibull Gamma Gumbel

0.8

0.6

0.4

(1)

CDF

0.2

0.2

0.4

0.6

0.8

1.2

Unit Flexural Bond Strength (MPa)

Figure A.14: Newcastle Wall 1, 1:1:6, High Quality Mason, CDF1

144

A.5 Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls

3.5 3 Truncated Normal Lognormal Weibull Gamma Gumbel

Probability Density

2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6

0.7

0.8

0.9

Unit Flexural Bond Strength (MPa)

Figure A.15: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Probability Distribution Fits

0.8

(Predicted Unit Flex. Bond Strength (MPa)

0.7

0.6

Perfect Fit Truncated Normal Lognormal Weibull Gamma Gumbel

0.5

0.4

0.3

CDF

(1)

0.2

0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Unit Flexural Bond Strength (MPa)

Figure A.16: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, CDF1

145

A.5 Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls

3.5 3 Truncated Normal Lognormal Weibull Gamma Gumbel

Probability Density

2.5 2 1.5 1 0.5 0 0 0.2 0.4 0.6 0.8 1 1.2

1.4

1.6

1.8

Unit Flexural Bond Strength (MPa)

Figure A.17: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Probability Distribution Fits

(Predicted Unit Flex. Bond Strength (MPa)

1.6

1.4

Perfect Fit Truncated Normal Lognormal Weibull Gamma Gumbel

1.2

0.8

0.6

(1)

0.4

CDF

0.2 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Unit Flexural Bond Strength (MPa)

Figure A.18: Newcastle Wall 3, 1:1:6, Fair Quality Mason, CDF1

146

A.5 Probability Distribution Fits and Associated Inverse of the Cumulative Distribution Function (CDF1 ) for Each of the Newcastle, Australia Walls

4 3.5 3 Truncated Normal Lognormal Weibull Gamma Gumbel

Probability Density

2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Unit Flexural Bond Strength (MPa)

Figure A.19: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Probability Distribution Fits

0.8

(Predicted Unit Flex. Bond Strength (MPa)

0.7

0.6

Perfect Fit Truncated Normal Lognormal Weibull Gamma Gumbel

0.5

0.4

0.3

CDF

(1)

0.2

0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Unit Flexural Bond Strength (MPa)

Figure A.20: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, CDF1

147

A.6 Autocorrelation function values plotted for each course in all walls

A.6

Autocorrelation function values plotted for each course in all walls

1

Autocorrelation Coefficient (k)

0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4

Lag k

10

Figure A.21: Newcastle Wall 1, 1:1:6, High Quality Mason, Autocorrelation Function Values for All Courses in the Wall

Autocorrelation Coefficient (k)

0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4

Lag k

10

Figure A.22: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for All Courses in the Wall

148

A.6 Autocorrelation function values plotted for each course in all walls

Autocorrelation Coefficient (k)

0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4

Lag k

10

Figure A.23: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall

Autocorrelation Coefficient (k)

0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4

Lag k

10

Figure A.24: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall

149

A.6 Autocorrelation function values plotted for each course in all walls

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.25: Sao Carlos Wall 1,1:2:9, High Quality Mason, Autocorrelation Function Values for All Courses in the Wall

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.26: Sao Carlos Wall 2,1:2:9 , Fair Quality Mason, Autocorrelation Function Values for All Courses in the Wall

150

A.7 Autocorrelation function values plotted for between courses in all walls

A.7

Autocorrelation function values plotted for between courses in all walls

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Autocorrelation Coefficient (k)

Lag k

Figure A.27: Newcastle Wall 1, 1:1:6, High Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall

Autocorrelation Coefficient (k)

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.28: Newcastle Wall 2, 1:1.5:6 + a.e., High Quality Mason, Autocorrelation Function Values for Between All Courses in the Wall

151

A.7 Autocorrelation function values plotted for between courses in all walls

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.29: Newcastle Wall 3, 1:1:6, Fair Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.30: Newcastle Wall 4, 1:1.5:6 + a.e., Fair Quality Mason, Autocorrelation Function Values for Between All Courses in the Wall

152

A.7 Autocorrelation function values plotted for between courses in all walls

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.31: Sao Carlos Wall 1,1:2:9, High Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall

(k)

Autocorrelation Coefficient

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0 1 2 3 4 5 6 7 8 9 10

Lag k

Figure A.32: Sao Carlos Wall 2,1:2:9 , Fair Quality Mason, Autocorrelation Function Values for All Between Courses in the Wall

153

Appendix B

Appendix to Chapter 4
B.1 Summary of specications for testing the non-linear 2D and 3D FEA model
Horizontal and vertical mortar joint non-linear tension softening interface elements Expanded brick elements Linear normal stiness mod. Linear tangential stiness mod. Tensile bond strength Tensile fracture energy 303 (N/mm3 ) 124 (N/mm3 ) variable (N/mm2 ) variable (N/mm)

Brick Youngs Modulus Bricks poissions ratio Brick density

5884 (N/mm2 ) 0.15 1800 (kg/mm3 )

Table B.1: Summary of material parameter values to be used for the FEA wall analysis using the non-linear tension softening interface elements

154

B.1 Summary of specications for testing the non-linear 2D and 3D FEA model

Horizontal and vertical mortar joint non-linear tension softening interface elements

Linear normal stiness mod. Tangential normal stiness mod. Tensile bond strength Tensile fracture energy Cohesion Tan frict. angle Tan dilatency angle Tan residual frict. angle Conning normal stress Exponential degradation coecient Capped critical compressive strength Shear traction control factor Compressive fracture energy Equivalent plastic relative displacement Shear fracture energy factor Brick Youngs Modulus Bricks poissions ratio Brick density Linear normal stiness mod. Tangential normal sti mod. Tensile strength Tensile fracture energy

303 (N/mm3 ) 124 (N/mm3 ) variable (N/mm2 ) variable (N/mm) 0.65 (N/mm2 ) 0.75 0.6 0.75 -1.2 (N/mm2 ) 5 20 (N/mm2 ) 9 15 (N/mm2 ) 0.012 0.065 5884 (N/mm2 ) 0.15 1800 (kg/mm3 ) 1000 (N/mm3 ) 1000 (N/mm3 ) 2 (N/mm2 ) 0.5 (N/mm)

Expanded brick elements Potential brick cracks (all values articially high to force cracking in mortar joints and not brick joints)

Table B.2: Summary of material parameter values to be used for the FEA wall analysis using the composite interface elements

155

B.2 Doherty 2000 static push test wall load versus deection results

B.2

Doherty 2000 static push test wall load versus deection results

Figure B.1: Load Deection Curve for Static Push Test Wall Specimen 3 from (Doherty 2000)

156

B.2 Doherty 2000 static push test wall load versus deection results

Figure B.2: Load Deection Curve for Static Push Test Wall Specimen 12 from (Doherty 2000)

Figure B.3: Load Deection Curve for Static Push Test Wall Specimen 13 from (Doherty 2000)

157

B.3 Summary of specications for modelling Doherty (2000) walls using the 3D non-linear FEA model

B.3

Summary of specications for modelling Doherty (2000) walls using the 3D non-linear FEA model
Horizontal and vertical mortar joint interface elements Linear normal stiness mod. Linear tangential stiness mod. Tensile bond strength Tensile fracture energy Cohesion Tan frict. angle Tan dilatency angle Tan residual frict. angle Conning normal stress Exponential degradation coecient Capped critical compressive strength Shear traction control factor Compressive fracture energy Equivalent plastic relative displacement Shear fracture energy factor Brick Youngs Modulus Bricks poissions ratio Brick density Linear normal stiness mod. Tangential normal sti mod. Tensile strength Tensile fracture energy 303 (N/mm3 ) 124 (N/mm3 ) 0.328 (N/mm2 ) 0.0056 (N/mm) 0.65 (N/mm2 ) 0.75 0.6 0.75 -1.2 (N/mm2 ) 5 20 (N/mm2 ) 9 15 (N/mm2 ) 0.012 0.065 5884 (N/mm2 ) 0.15 1800 (kg/mm3 ) 1000 (N/mm3 ) 1000 (N/mm3 ) 2 (N/mm2 ) 0.5 (N/mm)

Expanded brick elements Potential brick cracks (all values articially high to force cracking in mortar joints and not brick joints)

Table B.3: Summary of material parameter values to be used in the 3D FEA analysis of the walls tested by Doherty (2000)

158

B.3 Summary of specications for modelling Doherty (2000) walls using the 3D non-linear FEA model

Self-weight

Executed rst in one load step Self weight applied as gravity x density of expanded brick elements user specied load step size iteration Executed after self-weight has been applied Load step method Load Step size Unloading determination Arc-length control

1x(self-weight) newton regular, rst tangent energy based step sizes min: 0.001 max: 0.005 negative pivots sperical path DOF in the y-direction on 1 node (6883) node at midheight of wall newton regular, rst tangent 5000 N but model makes increases above this load as needed to facilitate failure peak load

Lateral load

Iteration Token value for load applied at 10 courses from bottom of wall to facilitate easy interpretation of failure load

Table B.4: Wall loading and analysis specications for uncracked static-push tests from Doherty (2000)

159

Appendix C

Appendix to Chapter 5
C.1 Summary of FEA specications for the full size wall in vertical bending
3D Model see also Figure 4.3(b) Bricks HE20 CHX60: A twenty-node, 3D isoparametric solid brick element based on quadratic interpolation and Gauss integration IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation IS88 CQ48I: A sixteen-node, 3D interface element based on quadratic interpolation 2x4x1 2x4x1 (interface elements are only allowed to be one element thick) 1x4x1 (interface elements are only allowed to be one element thick)

Element Types

Mortar joints Mid-length brick interface elements

Mesh Density

For half brick For half mortar joint Mid-length brick interface elements

Table C.1: Summary of FEA element type and mesh selection for the full wall

160

C.1 Summary of FEA specications for the full size wall in vertical bending

Horizontal and vertical mortar joint interface elements

Linear normal stiness mod. Linear tangential stiness mod. Tensile bond strength Tensile fracture energy Cohesion Tan frict. angle Tan dilatency angle Tan residual frict. angle Conning normal stress Exponential degradation coecient Capped critical compressive strength Shear traction control factor Compressive fracture energy Equivalent plastic relative displacement Shear fracture energy factor Brick Youngs Modulus Bricks poissions ratio Brick density Linear normal stiness mod. Tangential normal sti mod. Tensile strength Fracture energy

353 (N/mm3 ) 146 (N/mm3 ) variable (N/mm2 ) variable (N/mm) 0.150 (N/mm2 ) 0.75 0.6 0.75 -1.2 (N/mm2 ) 5 20 (N/mm2 ) 9 15 (N/mm2 ) 0.012 0.150 20000 (N/mm2 ) 0.15 1800 (kg/mm3 ) 1000 (N/mm3 ) 1000 (N/mm3 ) 2 (N/mm2 ) 0.5 (N/mm)

Expanded brick elements Potential brick cracks (all values articially high to force cracking in mortar joints and not brick joints)

Table C.2: Summary of material parameter values to be used in the 3D FEA analysis of the full wall

161

C.1 Summary of FEA specications for the full size wall in vertical bending

Boundary conditions at bottom support

The bottom/base row of interface elements are restricted from moving in negative z direction. However, the interface elements allow cracking between the brick elements and the ground. This assimilates the wall being attached to a concrete slab by a mortar joint On the backside of the wall, the bottom edge of the rst course of bricks are restricted from translation in the y direction. This, in combination with the vertical restricition of movement on the underside of the base interface elements, allows the vertical reactions to be constant but pushed to the compressive zone with mid-height displacement, as would be expected when a brick wall is attached to a concrete slab with a mortar joint

Boundary conditions at top support Other stabilising supports

On the backside of the wall, the top edge of the top course of bricks are restricted from translation in the y direction

At the bottom and top courses, at one point on each end end brick, the points are restricted from translation movement in the x-direction. This is to stabilise the wall from displacing from its xed position in the global x-direction

Table C.3: Boundary conditions for the full wall

Self-weight

Executed rst in one load step Self weight applied as gravity x density of expanded brick elements user specied load step size Iteration Executed after self-weight has been applied Load step method Load Step size Unloading determination Arc-length control

1x(self-weight) newton regular, rst tangent energy based step sizes min: 0.001 max: 0.05 negative pivots sperical path DOF in the y-direction on 1 node (26513) (node located at mid-height of wall) newton regular, rst tangent 1000 N/mm2 but model makes icreases above this load as needed to facilitate failure

Lateral load

Iteration Pressure load on face of wall

Table C.4: Wall analysis specications for the full wall

162

C.2 Results from Spatially Variable Bond Strength Simulations in the Full Sized Masonry Wall

C.2

Results from Spatially Variable Bond Strength Simulations in the Full Sized Masonry Wall

CDF(1) (Predicted Wall Failure Load (kPa)

2.3

Perfect Fit Normal Lognormal Weibull Gamma Gumbel

2.25

2.2

2.15

2.15

2.2

2.25

2.3

2.35

Wall Failure Load (kPa)

Figure C.1: Peak load CDF1 for a ft = 0.4MPa and COV = 0.1

2.4

(Predicted Wall Failure Load (kPa)

2.35 2.3 2.25 2.2 2.15 2.1 2.05 2 1.95 1.9 1.9

Perfect Fit Normal Lognormal Weibull Gamma Gumbel

CDF

(1)

2.1

2.2

2.3

2.4

Wall Failure Load (kPa)

Figure C.2: Peak load CDF1 for a ft = 0.4MPa and COV = 0.3

163

C.2 Results from Spatially Variable Bond Strength Simulations in the Full Sized Masonry Wall

2.2

(Predicted Wall Failure Load (kPa)

2.15 2.1 2.05 2 1.95 1.9 1.85 1.8 1.75 1.7 1.7

Perfect Fit Normal Lognormal Weibull Gamma Gumbel

CDF

(1)

1.8

1.9

2.1

2.2

Wall Failure Load (kPa)

Figure C.3: Peak load CDF1 for a ft = 0.4MPa and COV = 0.5

164

BIBLIOGRAPHY

Bibliography
AS3700 (2001). Australian Standard - Masonry Structures. Standards Australia International Ltd, Sydney. Baker, L. (1981). The Flexural Action of Masonry Structures Under Lateral Load. PhD thesis, School of Engineering and Architecture, Deakin University. Baker, L. (1982). A principal stress failure criterion for brickwork in biaxial bending. In Proceedings of the 6th International Brick Masonry Conference, Rome. Baker, L. and Franken, G. (1976). Variability aspects of the exural strength of brickwork. In Proceedings of the 4th International Brick Masonry Conference, Brussels. Bassetti, F. (2000). A short history of masonry. Internet. Bye, G. (1983). Portland Cement: Composition, Production and Properties. Pergamon Press. Candy, C. (1988). The energy method for masonry panels under lateral loading. In Proceedings 8th International Brick and Block Masonry Conference, pages 921, Dublin. Cornelissen, H., Hordijk, D., and Reinhardt, H. (1986). Experimental determination of crack softening characteristics of normal weight and lightweight concrete. Heron, 31(2). Correa, M., Heer, L., Masia, M., and M.G., S. (2008). Flexural bond strength for masonry walls: An experimental and statistical analysis. In Eighth International Seminar on Structural Masonry. CUR (1994). Structural masonry: An experimental/numerical basis for practical design rules. Technical Report 171, CUR, Gouda, The Netherlands. Doherty, K. T. (2000). An Investigation of the Weak Links in the Seismic Load Path of Unreinforced Masonry Buildings. PhD thesis, Department of Civil and Environmental Engineering, The University of Adelaide. Edgall, G. (1993). Structural masonry steering committee informal working group exural strength database. CERAM Building Technology.

165

BIBLIOGRAPHY

Ellingwood, B. (1981). Analysis of reliabilty for masonry structures. Journal of Structural Engineering, 107(ST5):756773. Fenton, G. A. (1999). Estimation for stochastic soil models. Journal of Geotechnical and Geoenvironmental Engineering,, 125(6):470485. Fyfe, A., Middleton, J., and Pande, G. (1999). Assessment of the partial factors of safety for masonry wall panels with workmanship defects. In Proceedings 8th North American Masonry Conference, Austin, Texas, USA. Fyfe, A., Middleton, J., and Pande, G. (2000). Numerical evaluation of the inuence of some workmanship defects on partial factor or safety for masonry. Masonry International, 13(2):4853. Grith, M. and Vaculik, J. (2005). Flexural strength of unreinforced clay brick masonry walls. In Proceedings 10th Masonry Symposium, Ban, Alberta. Grith, M., Willis, C., and Lawrence, S. (2001). Flexural behaviour and design of unreinforced (brick) masonry walls. In Proceedings 6th Australasian Masonry Conference, pages 175184, Adelaide. Grimm, C. (1974). Strength and related properties of brick masonry. Proceedings of the ASCE, Journal of Structural Division, 101:217231. Han, Y. (2007). Structural Behaviour of Masonry in Biaxial Bending. PhD thesis, University of Newcastle. Hendriks, M. and Wolters, B. (2007). Diana Finite Element Users Manual, Release 9.2. TNO DIANA BV, P.O. Box 49, 2600 AA Delft, The Netherlands. Hendry, A. (1976). The eect of site factors on masonry performance. In Proceedings First Canadian Masonry Symposium, pages 182198, Calgary. Hordijk, D. (1991). Local Approach to Fatigue of Concrete. PhD thesis, Delft University of Technology. Internet (2008a). http://www.abovetopsecret.com/forum/thread318417/pg1. Internet. Internet (2008b). http://www.eartharchitecture.org/index.php?/categories/28-india. Internet.

166

BIBLIOGRAPHY

Internet (2008c). www.atpm.com. Internet. Lawrence, S. (1983). Behaviour of Brick Masonry Walls Under Lateral Loading. PhD thesis, The University of New South Wales. Lawrence, S. (1985). Random variations in brickwork properties. In 7th IBMAC, Melbourne. Lawrence, S. (1991). Stochastic analysis of masonry structures. Technical report, Division of Building, Construction and Engineering, CSIRO. Lawrence, S. and Cao, H. (1988). Cracking of non-loadbearing masonry walls under lateral forces. In Proceedings of the 8th International Brick/Block Masonry Conference, pages 11841194, Dublin. Lourenco, P. (1996a). Computational Strategies for Masonry Structures. PhD thesis, Technische Universiteit Delft. Lourenco, P. (1996b). A user/programmer guide for the micro-modeling of masonry structures. Technical report, Delft University of Technology, Faculty of Civil Engineering. McNeilly, T., Scrivener, J., Lawrence, S., and Zsembery, S. (1996). A site survey of masonry bond strength. Australian Civil/Structural Engineering Transactions, IEAust, CE38(2,3,and 4):103109. Petersen, R., Masia, M., and Seracino, R. (2007). Experimental verication of nite element model to predict the shear behaviour of nsm frp strengthened masonry walls. In Proceedings of the 14th IBMAC. Prestandard, E. (1996). Design of masonry structures: Part 1-1, general rules for buildings. ENV. Priestley, M. (1981). Spectral Analysis and Time Series, volume 1. Academic Press, London. Stewart, M. and Lawrence, S. (2002). Structural reliability of masonry walls in exure. Masonry International, 15(2):4852. Stewart, M. and Lawrence, S. (2007). Model error, structural reliability and partial safety factors for structural masonry in compression. Masonry International, 20(3):107116.

167

BIBLIOGRAPHY

Sugo, H. (2001a). The development of the mortar/unit bond. In Proceedings 6th Australasian Masonry Conference, pages pp.367377, Adelaide. Sugo, H. (2001b). The inuence of air entraining agent on bond strength and mortar microstructure. In Proceedings 6th Australasian Masonry Conference, Adelaide. Sugo, H. (2007). Personal communication regarding development of bond strength with retempering and drying of mortar. Van Der Pluijm, R. (1992). Material properties of masonry and its components under tension and shear. In Proceedings of the 6th Canadian Masonry Symposium, pages 675686, Saskatoon, Saskatchewan, Canada. Van Der Pluijm, R. (1997). Non-linear behaviour of masonry under tension. Heron, 42(1):2554. Van Zijl, G. (2004). Modeling masonry shear-compression: Role of dilatancy highlighted. Journal Of Engineering Mechanics, ASCE, pages 1289 1296. Willis, C., Grith, M., and Lawrence, S. (2004a). Horizontal bending of unreinforced clay brick masonry walls. Masonry International, 17(3):109121. Willis, C., Grith, M., and Lawrence, S. (2004b). Implications of recent experimental and analytical studies for the design of face-loaded urm walls. In Proceedings 7th Australasian Masonry Conference, pages 255264, Newcastle. Willis, C. R. (2004). Design of Unreinforced Masonry Walls for Out-of-Plane Loading. PhD thesis, Department of Civil and Environmental Engineering, The University of Adelaide.

168