PLATES AND SHELLS

Lecture 8.1 : Introduction to Plate Behaviour and Design

Lecture 8.2 : Behaviour and Design of Unstiffened Plates

Lecture 8.3 : Behaviour and Design of Stiffened Plates

Lecture 8.4.1 : Plate Girder Behaviour and Design I

Lecture 8.4.2 : Plate Girder Behaviour and Design II

Lecture 8.4.3 : Plate Girder Design - Special Topics

Lecture 8.5.1 : Introduction to Design of Box Girders

Lecture 8.5.2 : Advanced Design of Box Girders

Lecture 8.6 : Introduction to Shell Structures

Lecture 8.7 : Basic Analysis of Shell Structures

Lecture 8.8 : Design of Unstiffened Cylinders

Lecture 8.9 : Design of Stringer-Stiffened Cylindrical Shells

ESDEP WG 8

PLATES AND SHELLS

Lecture 8.1: Introduction to Plate

Behaviour and Design
OBJECTIVE/SCOPE

To introduce the series of lectures on plates, showing the uses of plates to resist in-plane
and out-of-plane loading and their principal modes of behaviour both as single panels and as
assemblies of stiffened plates.

PREREQUISITES

None.

RELATED LECTURES

Lecture 8.2: Behaviour and Design of Unstiffened Plates

Lecture 8.3: Behaviour and Design of Stiffened Plates

SUMMARY

This lecture introduces the uses of plates and plated assemblies in steel structures. It
describes the basic behaviour of plate panels subject to in-plane or out-of-plane loading,
highlighting the importance of geometry and boundary conditions. Basic buckling modes and
mode interaction are presented. It introduces the concept of effective width and describes the
influence of imperfections on the behaviour of practical plates. It also gives an introduction to
the behaviour of stiffened plates.

1. INTRODUCTION

Plates are very important elements in steel structures. They can be assembled into
complete members by the basic rolling process (as hot rolled sections), by folding (as cold
formed sections) and by welding. The efficiency of such sections is due to their use of the high
in-plane stiffness of one plate element to support the edge of its neighbour, thus controlling the
out-of-plane behaviour of the latter.

The size of plates in steel structures varies from about 0,6mm thickness and 70mm width
in a corrugated steel sheet, to about 100mm thick and 3m width in a large industrial or offshore
structure. Whatever the scale of construction the plate panel will have a thickness t that is much
smaller than the width b, or length a. As will be seen later, the most important geometric
parameter for plates is b/t and this will vary, in an efficient plate structure, within the range 30 to
250.

2. BASIC BEHAVIOUR OF A PLATE PANEL

Understanding of plate structures has to begin with an understanding of the modes of

behaviour of a single plate panel.

2.1 Geometric and Boundary Conditions

The important geometric parameters are thickness t, width b (usually measured transverse
to the direction of the greater direct stress) and length a, see Figure 1a. The ratio b/t, often called
the plate slenderness, influences the local buckling of the plate panel; the aspect ratio a/b may
also influence buckling patterns and may have a significant influence on strength.
 In addition to the geometric proportions of the plate, its strength is governed by its
boundary conditions. Figure 1 shows how response to different types of actions is influenced by
different boundary conditions. Response to in-plane actions that do not cause buckling of the
plate is only influenced by in-plane, plane stress, boundary conditions, Figure 1b. Initially,
response to out-of-plane action is only influenced by the boundary conditions for transverse
movement and edge moments, Figure 1c. However, at higher actions, responses to both types of
action conditions are influenced by all four boundary conditions. Out-of-plane conditions
influence the local buckling, see Figure 1d; in-plane conditions influence the membrane action
effects that develop at large displacements (>t) under lateral actions, see Figure 1e.

2.2 In-plane Actions

As shown in Figure 2a, the basic types of in-plane actions to the edge of a plate panel are
the distributed action that can be applied to a full side, the patch action or point action that can be
applied locally.
 When the plate buckles, it is particularly important to differentiate between applied
displacements, see Figure 2b and applied stresses, see Figure 2c. The former permits a
redistribution of stress within the panel; the more flexible central region sheds stresses to the
edges giving a valuable post buckling resistance. The latter, rarer case leads to an earlier collapse
of the central region of the plate with in-plane deformation of the loaded edges.

2.3 Out-of-plane Actions

Out-of-plane loading may be:

 uniform over the entire panel, see for example Figure 3a, the base of a water tank.
 varying over the entire panel, see for example Figure 3b, the side of a water tank.
 a local patch over part of the panel, see for example Figure 3c, a wheel load on a bridge deck.
2.4 Determination of Plate Panel Actions

In some cases, for example in Figure 4a, the distribution of edge actions on the panels of
a plated structure are self-evident. In other cases the in-plane flexibilities of the panels lead to
distributions of stresses that cannot be predicted from simple theory. In the box girder shown in
Figure 4b, the in-plane shear flexibility of the flanges leads to in-plane deformation of the top
flange. Where these are interrupted, for example at the change in direction of the shear at the
central diaphragm, the resulting change in shear deformation leads to a non-linear distribution of
direct stress across the top flange; this is called shear lag.

In members made up of plate elements, such as the box girder shown in Figure 5, many
of the plate components are subjected to more than one component of in-plane action effect.
Only panel A does not have shear coincident with the longitudinal compression.
If the cross-girder system EFG was a means of introducing additional actions into the box, there
would also be transverse direct stresses arising from the interaction between the plate and the
stiffeners.

2.5 Variations in Buckled Mode

i. Aspect ratio a/b

In a long plate panel, as shown in Figure 6, the greatest initial inhibition to buckling is the
transverse flexural stiffness of the plate between unloaded edges. (As the plate moves more into
the post-buckled regime, transverse membrane action effects become significant as the plate
deforms into a non-developable shape, i.e. a shape that cannot be formed just by bending).
As with any instability of a continuous medium, more than one buckled mode is possible, in this
instance, with one half wave transversely and in half waves longitudinally. As the aspect ratio
increases the critical mode changes, tending towards the situation where the half wave length
a/m = b. The behaviour of a long plate panel can therefore be modelled accurately by considering
a simply-supported, square panel.

ii. Bending conditions

As shown in Figure 7, boundary conditions influence both the buckled shapes and the critical
stresses of elastic plates. The greatest influence is the presence or absence of simple supports, for
example the removal of simple support to one edge between case 1 and case 4 reduces the
buckling stress by a factor of 4,0/0,425 or 9,4. By contrast introducing rotational restraint to one
edge between case 1 and case 2 increases the buckling stress by 1,35.
iii. Interaction of modes

Where there is more than one action component, there will be more than one mode and therefore
there may be interaction between the modes. Thus in Figure 8b(i) the presence of low transverse
compression does not change the mode of buckling. However, as shown in Figure 8b(ii), high
transverse compression will cause the panel to deform into a single half wave. (In some
circumstances this forcing into a higher mode may increase strength; for example, in case 8b(ii),
predeformation/transverse compression may increase strength in longitudinal compression.)
Shear buckling as shown in Figure 8c is basically an interaction between the diagonal,
destabilising compression and the stabilising tension on the other diagonal.
Where buckled modes under the different action effects are similar, the buckling stresses under
the combined actions are less than the addition of individual action effects. Figure 9 shows the
buckling interactions under combined compression, and uniaxial compression and shear.
2.6 Grillage Analogy for Plate Buckling

One helpful way to consider the buckling behaviour of a plate is as the grillage shown in Figure
10. A series of longitudinal columns carry the longitudinal actions. When they buckle, those
nearer the edge have greater restraint than those near the centre from the transverse flexural
members. They therefore have greater post buckling stiffness and carry a greater proportion of
the action. As the grillage moves more into the post buckling regime, the transverse buckling
restraint is augmented by transverse membrane action.

2.7 Post Buckling Behaviour and Effective Widths

Figures 11a, 11b and 11c describe in more detail the changing distribution of stresses as a plate
buckles following the equilibrium path shown in Figure 11d. As the plate initially buckles the
stresses redistribute to the stiffer edges. As the buckling continues this redistribution becomes
more extreme (the middle strip of slender plates may go into tension before the plate fails). Also
transverse membrane stresses build up. These are self equilibrating unless the plate has clamped
in-plane edges; tension at the mid panel, which restrains the buckling is resisted by compression
at the edges, which are restrained from out-of-plane movement.
An examination of the non-linear longitudinal stresses in Figures 11a and 11c shows that it is
possible to replace these stresses by rectangular stress blocks that have the same peak stress and
same action effect. This effective width of plate (comprising beff/2 on each side) proves to be a
very effective design concept. Figure 11e shows how effective width varies with slenderness (p
is a measure of plate slenderness that is independent of yield stress; p = 1,0 corresponds to
values of b/t of 57, 53 and 46 for fy of 235N/mm2, 275N/mm2 and 355N/mm2 respectively).

Figure 12 shows how effective widths of plate elements may be combined to give an effective
cross-section of a member.
2.8 The Influences of Imperfections on the Behaviour of Actual Plates

As with all steel structures, plate panels contain residual stresses from manufacture and
subsequent welding into plate assemblies, and are not perfectly flat. The previous discussions
about plate panel behaviour all relate to an ideal, perfect plate. As shown in Figure 13 these
imperfections modify the behaviour of actual plates. For a slender plate the behaviour is
asymptotic to that of the perfect plate and there is little reduction in strength. For plates of
intermediate slenderness (which frequently occur in practice), an actual imperfect plate will have
a considerably lower strength than that predicted for the perfect plate.

Figure 14 summarises the strength of actual plates of varying slenderness. It shows the reduction
in strength due to imperfections and the post buckling strength of slender plates.
2.9 Elastic Behaviour of Plates Under Lateral Actions

The elastic behaviour of laterally loaded plates is considerably influenced by its support
conditions. If the plate is resting on simple supports as in Figure 15b, it will deflect into a shape
approximating a saucer and the corner regions will lift off their supports. If it is attached to the
supports, as in Figure 15c, for example by welding, this lift off is prevented and the plate
stiffness and action capacity increases. If the edges are encastre as in Figure 15d, both stiffness
and strength are increased by the boundary restraining moments.
Slender plates may well deflect elastically into a large displacement regime (typically where d >
t). In such cases the flexural response is significantly enhanced by the membrane action of the
plate. This membrane action is at its most effective if the edges are fully clamped. Even if they
are only held partially straight by their own in-plane stiffness, the increase in stiffness and
strength is most noticeable at large deflections.

Figure 15 contrasts the behaviour of a similar plate with different boundary conditions.

Figure 16 shows the modes of behaviour that occur if the plates are subject to sufficient load for
full yield line patterns to develop. The greater number of yield lines as the boundary conditions
improve is a qualitative measure of the increase in resistance.
3. BEHAVIOUR OF STIFFENED PLATES

Many aspects of stiffened plate behaviour can be deduced from a simple extension of the basic
concepts of behaviour of unstiffened plate panels. However, in making these extrapolations it
should be recognised that:

 "smearing" the stiffeners over the width of the plate can only model overall behaviour.
 stiffeners are usually eccentric to the plate. Flexural behaviour of the equivalent tee section
induces local direct stresses in the plate panels.
 local effects on plate panels and individual stiffeners need to be considered separately.
 the discrete nature of the stiffening introduces the possibility of local modes of buckling. For
example, the stiffened flange shown in Figure 17a shows several modes of buckling. Examples
are:

(i) plate panel buckling under overall compression plus any local compression arising from the
combined action of the plate panel with its attached stiffening, Figure 17b.

(ii) stiffened panel buckling between transverse stiffeners, Figure 17c. This occurs if the latter
have sufficient rigidity to prevent overall buckling. Plate action is not very significant because
the only transverse member is the plate itself. This form of buckling is best modelled by
considering the stiffened panel as a series of tee sections buckling as columns. It should be noted
that this section is monosymmetric and will exhibit different behaviour if the plate or the
stiffener tip is in greater compression.

(iii) overall or orthotropic bucking, Figure 17d. This occurs when the cross girders are flexible. It
is best modelled by considering the plate assembly as an orthotropic plate.

 
 
4. CONCLUDING SUMMARY

 Plates and plate panels are widely used in steel structures to resist both in-plane and out-of-
plane actions.
 Plate panels under in-plane compression and/or shear are subject to buckling.
 The elastic buckling stress of a perfect plate panel is influenced by:

 plate slenderness (b/t).

 aspect ratio (a/b).

 boundary conditions.

 interaction between actions, i.e. biaxial compression and compression and shear.

 The effective width concept is a useful means of defining the post-buckling behaviour of a plate
panel in compression.
 The behaviour of actual plates is influenced by both residual stresses and geometric
imperfections.
 The response of a plate panel to out-of-plane actions is influenced by its boundary conditions.
 An assembly of plate panels into a stiffened plate structure may exhibit both local and overall
modes of instability.

5. ADDITIONAL READING

1. Timoshenko, S. and Weinowsky-Kreiger, S., "Theory of Plates and Shells" Mc Graw-Hill, New
York, International Student Edition, 2nd Ed.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.2: Behaviour and Design of

Unstiffened Plates
OBJECTIVE/SCOPE

To discuss the load distribution, stability and ultimate resistance of unstiffened plates
under in-plane and out-of-plane loading.

PREREQUISITES

Lecture 8.1: Introduction to plate behaviour and design

RELATED LECTURES

Lecture 8.3: Stiffened Plates

Lectures 8.4: Plate Girder Behaviour and Design I and II

Lecture 8.6: Introduction to Shell Structures

SUMMARY

The load distribution for unstiffened plate structures loaded in-plane is discussed. The
critical buckling loads are derived using Linear Elastic Theory. The effective width method for
determining the ultimate resistance of the plate is explained as are the requirements for adequate
finite element modelling of a plate element. Out-of-plane loading is also considered and its
influence on the plate stability discussed.

1. INTRODUCTION

Thin-walled members, composed of thin plate panels welded together, are increasingly
important in modern steel construction. In this way, by appropriate selection of steel quality,
geometry, etc., cross-sections can be produced that best fit the requirements for strength and
serviceability, thus saving steel.

Recent developments in fabrication and welding procedures allow the automatic

production of such elements as plate girders with thin-walled webs, box girders, thin-walled
columns, etc. (Figure 1a); these can be subsequently transported to the construction site as
prefabricated elements.
 Due to their relatively small thickness, such plate panels are basically not intended to
carry actions normal to their plane. However, their behaviour under in-plane actions is of specific
interest (Figure 1b). Two kinds of in-plane actions are distinguished:

a) those transferred from adjacent panels, such as compression or shear.

b) those resulting from locally applied forces (patch loading) which generate zones of highly
concentrated local stress in the plate.

The behaviour under patch action is a specific problem dealt with in the lectures on plate
girders (Lectures 8.5.1 and 8.5.2). This lecture deals with the more general behaviour of
unstiffened panels subjected to in-plane actions (compression or shear) which is governed by
plate buckling. It also discusses the effects of out-of-plane actions on the stability of these
panels.

2. UNSTIFFENED PLATES UNDER IN-PLANE LOADING

2.1 Load Distribution

2.1.1 Distribution resulting from membrane theory

The stress distribution in plates that react to in-plane loading with membrane stresses
may be determined, in the elastic field, by solving the plane stress elastostatic problem governed
by Navier's equations, see Figure 2.
where:

u = u(x, y), v = v(x, y): are the displacement components in the x and y directions

eff = 1/(1 + ) is the effective Poisson's ratio

G: is the shear modulus

X = X(x, y), Y = Y(x, y): are the components of the mass forces.

The functions u and v must satisfy the prescribed boundary (support) conditions on the
boundary of the plate. For example, for an edge parallel to the y axis, u= v = 0 if the edge is
fixed, or x = xy = 0 if the edge is free to move in the plane of the plate.

The problem can also be stated using the Airy stress function, F = F(x, y), by the
following biharmonic equation:

4F = 0

This formulation is convenient if stress boundary conditions are prescribed. The stress
components are related to the Airy stress function by:

; ;

2.1.2 Distribution resulting from linear elastic theory using Bernouilli's hypothesis

For slender plated structures, where the plates are stressed as membranes, the application
of Airy's stress function is not necessary due to the hypothesis of plane strain distributions,
which may be used in the elastic as well as in the plastic range, (Figure 3).
 However, for wide flanges of plated structures, the application of Airy's stress function
leads to significant deviations from the plane strain hypothesis, due to the shear lag effect,
(Figure 4). Shear lag may be taken into account by taking a reduced flange width.
2.1.3 Distribution resulting from finite element methods

When using finite element methods for the determination of the stress distribution, the
plate can be modelled as a perfectly flat arrangement of plate sub-elements. Attention must be
given to the load introduction at the plate edges so that shear lag effects will be taken into
account. The results of this analysis can be used for the buckling verification.

2.2 Stability of Unstiffened Plates

2.1.1 Linear buckling theory

The buckling of plate panels was investigated for the first time by Bryan in 1891, in
connection with the design of a ship hull [1]. The assumptions for the plate under consideration
(Figure 5a), are those of thin plate theory (Kirchhoff's theory, see [2-5]):

a) the material is linear elastic, homogeneous and isotropic.

b) the plate is perfectly plane and stress free.

c) the thickness "t" of the plate is small compared to its other dimensions.

d) the in-plane actions pass through its middle plane.

e) the transverse displacements w are small compared to the thickness of the plate.

f) the slopes of the deflected middle surfaces are small compared to unity.

g) the deformations are such that straight lines, initially normal to the middle plane, remain
straight lines and normal to the deflected middle surface.

h) the stresses normal to the thickness of the plate are of a negligible order of magnitude.

 
 Due to assumption (e) the rotations of the middle surface are small and their squares can
be neglected in the strain displacement relationships for the stretching of the middle surface,
which are simplified as:

x = ∂u/∂x ,         2E/2 xy∂u/∂y + ∂v/∂x                 (1)

An important consequence of this assumption is that there is no stretching of the middle surface
due to bending, and the differential equations governing the deformation of the plate are linear
and uncoupled. Thus, the plate equation under simultaneous bending and stretching is:

D4w = q-kt{x ∂2w/∂x2 + 2xy ∂2w/∂x∂y + y ∂2w/∂y2}                (2)

where D = Et3/12(1 - 2) is the bending stiffness of the plate having thickness t, modulus of
elasticity E, and Poisson's ratio ; q = q(x,y) is the transverse loading; and k is a parameter. The
stress components, x, y, xy are in general functions of the point x, y of the middle plane and
are determined by solving independently the plane stress elastoplastic problem which, in the
absence of in-plane body forces, is governed by the equilibrium equations:

 ∂x/∂x + ∂xy/∂y  = 0,     ∂xy/∂x + ∂y/∂y  = 0                (3)

supplemented by the compatibility equation:

2 (x + y) = 0 (4)

Equations (3) and (4) are reduced either to the biharmonic equation by employing the Airy stress
function:

4 F = 0 (5)

defined as:

x =  ∂2F/∂y2 ,   y = ∂2F/∂x2 ,   xy = -∂2F/∂x∂y            

or to the Navier equations of equilibrium, if the stress displacement relationships are employed:

2 + [1/(1- )] ∂/∂x {∂u/∂x + ∂v/∂y} = 0               

2 + [1/(1- )] ∂/∂y {∂u/∂x + ∂v/∂y} = 0                (6)

where = /(1 + ) is the effective Poisson's ratio.

Equation (5) is convenient if stress boundary conditions are prescribed. However, for
displacement or mixed boundary conditions Equations (6) are more convenient. Analytical or
approximate solutions of the plane elastostatic problem or the plate bending problem are possible
only in the case of simple plate geometries and boundary conditions. For plates with complex
shape and boundary conditions, a solution is only feasible by numerical methods such as the
finite element or the boundary element methods.

Equation (2) was derived by Saint-Venant. In the absence of transverse loading (q = 0),
Equation (2) together with the prescribed boundary (support) conditions of the plate, results in an
eigenvalue problem from which the values of the parameter k, corresponding to the non-trivial
solution (w  0), are established. These values of k determine the critical in-plane edge actions
(cr, cr) under which buckling of the plate occurs. For these values of k the equilibrium path has
a bifurcation point (Figure 5b). The edge in-plane actions may depend on more than one
parameter, say k1, k2,...,kN, (e.g. x, y and xy on the boundary may increase at different rates). In
this case there are infinite combinations of values of k i for which buckling occurs. These
parameters are constrained to lie on a plane curve (N = 2), on a surface (N = 3) or on a
hypersurface (N > 3). This theory, in which the equations are linear, is referred to as linear
buckling theory.

Of particular interest is the application of the linear buckling theory to rectangular plates,
subjected to constant edge loading (Figure 5a). In this case the critical action, which corresponds
to the Euler buckling load of a compressed strut, may be written as:

cr = k E or cr = k E (7)

where E = (8)

and k, k are dimensionless buckling coefficients.

Only the form of the buckling surface may be determined by this theory but not the
magnitude of the buckling amplitude. The relationship between the critical stress cr, and the
slenderness of the panel  = b/t, is given by the buckling curve. This curve, shown in Figure 5c,
has a hyperbolic shape and is analogous to the Euler hyperbola for struts.

The buckling coefficients, "k", may be determined either analytically by direct integration
of Equation (2) or numerically, using the energy method, the method of transfer matrices, etc.
Values of k and k for various actions and support conditions are shown in Figure 6 as a
function of the aspect ratio of the plate  =a/b. The curves for k have a "garland" form. Each
garland corresponds to a buckling mode with a certain number of waves. For a plate subjected to
uniform compression, as shown in Figure 6a, the buckling mode for values of  < , has one
half wave, for values  <  < , two half waves, etc. For  =  both buckling modes, with
one and two half waves, result in the same value of k  . Obviously, the buckling mode that gives
the smallest value of k is the decisive one. For practical reasons a single value of k  is chosen for
plates subjected to normal stresses. This is the smallest value for the garland curves independent
of the value of the aspect ratio. In the example given in Figure 6a, k is equal to 4 for a plate
which is simply supported on all four sides and subjected to uniform compression.
Combination of stresses x, y and 

For practical design situations some further approximations are necessary. They are
illustrated by the example of a plate girder, shown in Figure 7.
 The normal and shear stresses, x and  respectively, at the opposite edges of a subpanel
are not equal, since the bending moments M and the shear forces V vary along the panel.
However, M and V are considered as constants for each subpanel and equal to the largest value
at an edge (or equal to the value at some distance from it). This conservative assumption leads to
equal stresses at the opposite edges for which the charts of k  and k apply. The verification is
usually performed for two subpanels; one with the largest value of x and one with the largest
value of . In most cases, as in Figure 7, each subpanel is subjected to a combination of normal
and shear stresses. A direct determination of the buckling coefficient for a given combination of
stresses is possible; but it requires considerable numerical effort. For practical situations an
equivalent buckling stress creq  is found by an interaction formula after the critical stresses creq
and cro , for independent action of  and  have been determined. The interaction curve for a
plate subjected to normal and shear stresses, x and  respectively, varies between a circle and a
parabola [6], depending on the value of the ratio ψ of the normal stresses at the edges (Figure 8). 
This relationship may be represented by the approximate equation:

(9)

For a given pair of applied stresses (, ) the factor of safety with respect to the above curve is
given by:

 = (10)

The equivalent buckling stress is then given by:

creq = creq 2 + 32} (11)

where the von Mises criterion has been applied.

For simultaneous action of x, y and  similar relationships apply.

2.2.2 Ultimate resistance of an unstiffened plate

General

The linear buckling theory described in the previous section is based on assumptions (a)
to (h) that are never fulfilled in real structures. The consequences for the buckling behaviour
when each of these assumptions is removed is now discussed.

The first assumption of unlimited linear elastic behaviour of the material is obviously not
valid for steel. If the material is considered to behave as linear elastic-ideal plastic, the buckling
curve must be cut off at the level of the yield stress y (Figure 9b).
 When the non-linear behaviour of steel between the proportionality limit p and the yield
stress y is taken into account, the buckling curve will be further reduced (Figure 9b). When
strain hardening is considered, values of cr larger than y, as experimentally observed for very
stocky panels, are possible. In conclusion, it may be stated that the removal of the assumption of
linear elastic behaviour of steel results in a reduction of the ultimate stresses for stocky panels.

The second and fourth assumptions of a plate without geometrical imperfections and
residual stresses, under symmetric actions in its middle plane, are also never fulfilled in real
structures. If the assumption of small displacements is still retained, the analysis of a plate with
imperfections requires a second order analysis. This analysis has no bifurcation point since for
each level of stress the corresponding displacements w may be determined. The equilibrium path
(Figure 10a) tends asymptotically to the value of cr for increasing displacements, as is found
from the second order theory.
 However the ultimate stress is generally lower than cr since the combined stress due to
the buckling and the membrane stress is limited by the yield stress. This limitation becomes
relevant for plates with geometrical imperfections, in the region of moderate slenderness, since
the value of the buckling stress is not small (Figure 10b). For plates with residual stresses the
reduction of the ultimate stress is primarily due to the small value of p (Figure 9b) at which the
material behaviour becomes non-linear. In conclusion it may be stated that imperfections due to
geometry, residual stresses and eccentricities of loading lead to a reduction of the ultimate stress,
especially in the range of moderate slenderness.

The assumption of small displacements (e) is not valid for stresses in the vicinity of cr as
shown in Figure 10a. When large displacements are considered, Equation (1) must be extended
to the quadratic terms of the displacements. The corresponding equations, written for reasons of
simplicity for a plate without initial imperfections, are:

    (12)

This results in a coupling between the equations governing the stretching and the bending
of the plate (Equations (1) and (2)).

  (13a)

  (13b)

where F is an Airy type stress function. Equations (13) are known as the von Karman equations.
They constitute the basis of the (geometrically) non-linear buckling theory. For a plate without
imperfections the equilibrium path still has a bifurcation point at cr, but, unlike the linear
buckling theory, the equilibrium for stresses  > cr is still stable (Figure 11). The equilibrium
path for plates with imperfections tends asymptotically to the same curve. The ultimate stress
may be determined by limiting the stresses to the yield stress. It may be observed that plates
possess a considerable post-critical carrying resistance. This post-critical behaviour is more
pronounced the more slender the plate, i.e. the smaller the value of cr.
Buckling curve

For the reasons outlined above, it is evident that the Euler buckling curve for linear
buckling theory (Figure 6c) may not be used for design. A lot of experimental and theoretical
investigations have been performed in order to define a buckling curve that best represents the
true behaviour of plate panels. For relevant literature reference should be made to Dubas and
Gehri [7]. For design purposes it is advantageous to express the buckling curve in a
dimensionless form as described below.
The slenderness of a panel may be written according to (7) and (8) as:

p = (b/t) kcr (14)

If a reference slenderness given by:

y = fy (15)

is introduced, the relative slenderness becomes:

p = p/y = ycr  (16)

The ultimate stress is also expressed in a dimensionless form by introducing a reduction factor:

k = u /y (17)

Dimensionless curves for normal and for shear stresses as proposed by Eurocode 3 [8] are
illustrated in Figure 12.
 These buckling curves have higher values for large slendernesses than those of the Euler
curve due to post critical behaviour and are limited to the yield stress. For intermediate
slendernesses, however, they have smaller values than those of Euler due to the effects of
geometrical imperfections and residual stresses.

Although the linear buckling theory is not able to describe accurately the behaviour of a
plate panel, its importance should not be ignored. In fact this theory, as in the case of struts,
yields the value of an important parameter, namely p, that is used for the determination of the
ultimate stress.

Effective width method

This method has been developed for the design of thin walled sections subjected to
uniaxial normal stresses. It will be illustrated for a simply-supported plate subjected to uniform
compression (Figure 13a). 
 

The stress distribution which is initially uniform, becomes non-uniform after buckling,
since the central parts of the panel are not able to carry more stresses due to the bowing effect.
The stress at the stiff edges (towards which the redistribution takes place) may reach the yield
stress. The method is based on the assumption that the non-uniform stress distribution over the
entire panel width may be substituted by a uniform one over a reduced "effective" width. This
width is determined by equating the resultant forces:

b u = be y (18)

and accordingly:

be =  u.b/y = kb                (19)

which shows that the value of the effective width depends on the buckling curve adopted. For
uniform compression the effective width is equally distributed along the two edges (Figure 13a).
For non-uniform compression and other support conditions it is distributed according to rules
given in the various regulations. Some examples of the distribution are shown in Figure 13b. The
effective width may also be determined for values of  < u. In such cases Equation (19) is still
valid, but p, which is needed for the determination of the reduction factor k, is not given by
Equation (16) but by the relationship:

p cr)  (20)

The design of thin walled cross-sections is performed according to the following procedure:

For given actions conditions the stress distribution at the cross-section is determined. At each
subpanel the critical stress cr, the relative slenderness p and the effective width b e are
determined according to Equations (7), (16) and (19), respectively. The effective width is then
distributed along the panel as illustrated by the examples in Figure 13b. The verifications are
finally based on the characteristic Ae, Ie, and We of the effective cross-section. For the cross-
section of Figure 14b, which is subjected to normal forces and bending moments, the verification
is expressed as:

(21)

where e is the shift in the centroid of the cross-section to the tension side and m the partial safety
factor of resistance.
 

The effective width method has not been extended to panels subjected to combinations of
stress. On the other hand the interaction formulae presented in Section 2.2 do not accurately
describe the carrying resistance of the plate, since they are based on linear buckling theory and
accordingly on elastic material behaviour. It has been found that these rules cannot be extended
to cases of plastic behaviour. Some interaction curves, at the ultimate limit state, are illustrated in
Figure 15, where all stresses are referred to the ultimate stresses for the case where each of them
is acting alone. Relevant interaction formulae are included in some recent European Codes - see
also [9,10].
Finite element methods

When using finite element methods to determine the ultimate resistance of an unstiffened plate
one must consider the following aspects:

 The modelling of the plate panel should include the boundary conditions as accurately as
possible with respect to the conditions of the real structure, see Figure 16. For a conservative
solution, hinged conditions can be used along the edges.
 Thin shell elements should be used in an appropriate mesh to make yielding and large
curvatures (large out-of-plane displacements) possible.
 The plate should be assumed to have an initial imperfection similar in shape to the final collapse
mode.
 The first order Euler buckling mode can be used as a first approximation to this shape. In
addition, a disturbance to the first order Euler buckling mode can be added to avoid snap-through
problems while running the programme, see Figure 17. The amplitude of the initial imperfect
shape should relate to the tolerances for flatness.

 The program used must be able to take a true stress-strain relationship into account, see Figure
18, and if necessary an initial stress pattern. The latter can also be included in the initial shape.
 The computer model must use a loading which is equal to the design loading multiplied by an
action factor. This factor should be increased incrementally from zero up to the desired action
level (load factor = 1). If the structure is still stable at the load factor = 1, the calculation process
can be continued up to collapse or even beyond collapse into the region of unstable behaviour
(Figure 19). In order to calculate the unstable response, the program must be able to use more
refined incremental and iterative methods to reach convergence in equilibrium.
3. UNSTIFFENED PLATES UNDER OUT-OF-PLANE ACTIONS
3.1 Action Distribution

3.1.1 Distribution resulting from plate theory

If the plate deformations are small compared to the thickness of the plate, the middle
plane of the plate can be regarded as a neutral plane without membrane stresses. This assumption
is similar to beam bending theory. The actions are held in equilibrium only by bending moments
and shear forces. The stresses in an isotropic plate can be calculated in the elastic range by
solving a fourth order partial differential equation, which describes equilibrium between actions
and plate reactions normal to the middle plane of the plate, in terms of transverse deflections w
due to bending.
 4w =

where:

q = q(x, y) is the transverse loading

D = Et3/12(1- 2
) is the stiffness of the plate having
thickness t, modulus of elasticity E,
and Poisson's ratio  .

is the biharmonic operator

In solving the plate equation the prescribed boundary (support) conditions must be taken into
account. For example, for an edge parallel to the y axis, w = w/n = 0 if the edge is clamped, or
w = w2/n2 = 0 if the edge is simply supported.

Some solutions for the isotropic plate are given in Figure 20.
An approximation may be obtained by modelling the plate as a grid and neglecting the twisting
moments.

Plates in bending may react in the plastic range with a pattern of yield lines which, by
analogy to the plastic hinge mechanism for beams, may form a plastic mechanism in the limit
state (Figure 21). The position of the yield lines may be determined by minimum energy
considerations.

If the plate deformations are of the order of the plate thickness or even larger, the
membrane stresses in the plate can no longer be neglected in determining the plate reactions.

The membrane stresses occur if the middle surface of the plate is deformed to a curved
shape. The deformed shape can be generated only by tension, compression and shear stains in the
middle surface.
 This behaviour can be illustrated by the deformed circular plate shown in Figure 22b. It is
assumed that the line a c b (diameter d) does not change during deformation, so that a c b is
equal to the diameter d. The points which lie on the edge "akb" are now on a k b , which must
be on a smaller radius compared with the original one.
 Therefore the distance akb becomes shorter, which means that membrane stresses exist in
the ring fibres of the plate.

The distribution of membrane stresses can be visualised if the deformed shape is frozen.

It can only be flattened out if it is cut into a number of radial cuts, Figure 22c, the gaps
representing the effects of membrane stresses; this explains why curved surfaces are much stiffer
than flat surfaces and are very suitable for constructing elements such as cupolas for roofs, etc.

The stresses in the plate can be calculated with two fourth order coupled differential equations, in
which an Airy-type stress function which describes the membrane state, has to be determined in
addition to the unknown plate deformation.

In this case the problem is non-linear. The solution is far more complicated in
comparison with the simple plate bending theory which neglects membrane effects.

The behaviour of the plate is governed by von Karman's Equations (13).

where F = F(x, y) is the Airy stress function.

3.1.2 Distribution resulting from finite element methods (FEM)

More or less the same considerations hold when using FEM to determine the stress
distribution in plates which are subject to out-of-plane action as when using FEM for plates
under in-plane actions (see Section 2.1.3), except for the following:

 The plate element must be able to describe large deflections out-of-plane.

 The material model used should include plasticity.

3.2 Deflection and Ultimate Resistance

3.2.1 Deflections

Except for the yield line mechanism theory, all analytical methods for determining the
stress distributions will also provide the deformations, provided that the stresses are in the elastic
region.
 Using adequate finite element methods leads to accurate determination of the deflections
which take into account the decrease in stiffness due to plasticity in certain regions of the plate.
Most design codes contain limits to these deflections which have to be met at serviceability load
levels (see Figure 23).
3.2.2 Ultimate resistance

The resistance of plates, determined using the linear plate theory only, is normally much
underestimated since the additional strength due to the membrane effect and the redistribution of
forces due to plasticity is neglected.

An upper bound for the ultimate resistance can be found using the yield line theory.

More accurate results can be achieved using FEM. The FEM program should then
include the options as described in Section 3.1.2.

Via an incremental procedure, the action level can increase from zero up to the desired
design action level or even up to collapse (see Figure 23).

4. INFLUENCE OF THE OUT-OF-PLANE ACTIONS ON THE STABILITY OF

UNSTIFFENED PLATES

The out-of-plane action has an unfavourable effect on the stability of an unstiffened plate
panel in those cases where the deformed shape due to the out- of-plane action is similar to the
buckling collapse mode of the plate under in-plane action only.

The stability of a square plate panel, therefore, is highly influenced by the presence of out-of-
plane (transversely directed) actions. Thus if the aspect ratio  is smaller than , the plate
stability should be checked taking the out-of-plane actions into account. This can be done in a
similar way as for a column under compression and transverse actions.

If the aspect ratio  is larger than the stability of the plate should be checked
neglecting the out-of-plane actions component.

For strength verification both actions have to be considered simultaneously.

When adequate Finite element Methods are used, the complete behaviour of the plate can
be simulated taking the total action combination into account.

5. CONCLUDING SUMMARY

 Linear buckling theory may be used to analyse the behaviour of perfect, elastic plates under in-
plane actions.
 The behaviour of real, imperfect plates is influenced by their geometric imperfections and by
yield in the presence of residual stresses.
 Slender plates exhibit a considerable post-critical strength.
 Stocky plates and plates of moderate slenderness are adversely influenced by geometric
imperfection and plasticity.
 Effective widths may be used to design plates whose behaviour is influenced by local buckling
under in-plane actions.
  The elastic behaviour of plates under out-of-plane actions is adequately described by small
deflection theory for deflection less than the plate thickness.
 Influence surfaces are a useful means of describing small deflection plate behaviour.
 Membrane action becomes increasingly important for deflections greater than the plate
thicknesses and large displacement theory using the von Karman equations should be used for
elastic analysis.
 An upper bound on the ultimate resistance of plates under out-of-plane actions may be found
from yield live theory.
 Out-of-plane actions influence the stability of plate panels under in-plane action.

6. REFERENCES

[1] Bryan, G. K., "On the Stability of a Plane Plate under Thrusts in its own Plane with
Application on the "Buckling" of the Sides of a Ship". Math. Soc. Proc. 1891, 54.

[2] Szilard, R., "Theory and Analysis of Plates", Prentice-Hall, Englewood Cliffs, New Jersey,
1974.

[3] Brush, D. O. and Almroth, B. O., "Buckling of Bars, Plates and Shells", McGraw-Hill, New
York, 1975.

[4] Wolmir, A. S., "Biegsame Platten und Schalen", VEB Verlag für Bauwesen, Berlin, 1962.

[5] Timoshenko, S., and Winowsky-Krieger, S., "Theory of Plates and Shells", Mc Graw Hill,
1959.

[6] Chwalla, E., "Uber dés Biégungsbeulung der Langsversteiften Platte und das Problem der
Mindersteifigeit", Stahlbau 17, 84-88, 1944.

[7] Dubas, P., Gehri, E. (editors), "Behaviour and Design of Steel Plated Structures", ECCS,
1986.

[8] Eurocode 3: "Design of Steel Structures": ENV 1993-1-1: Part 1.1: General rules and rules
for buildings, CEN, 1992.

[9] Harding, J. E., "Interaction of direct and shear stresses on Plate Panels" in Plated Structures,
Stability and Strength". Narayanan (ed.), Applied Science Publishers, London, 1989.

[10] Linder, J., Habermann, W., "Zur mehrachsigen Beanspruchung beim"

Plattenbeulen. In Festschrift J. Scheer, TU Braunschweig, 1987.

ESDEP WG 8

PLATES AND SHELLS

Lecture 8.3: Behaviour and Design of

Stiffened Plates
OBJECTIVE/SCOPE

To discuss the load distribution, stability and ultimate resistance of stiffened plates under in-
plane and out-of-plane loading.

PREREQISITES

Lecture 8.1: Introduction to Plate Behaviour and Design

Lecture 8.2: Behaviour and Design of Unstiffened Plates

RELATED LECTURES

Lecture 8.4.1: Plate Girder Behaviour & Design I

Lecture 8.6: Introduction to Shell Structures

SUMMARY

The load distribution for in-plane loaded unstiffened plate structures is discussed and the
critical buckling loads derived using linear elastic theory. Two design approaches for
determining the ultimate resistance of stiffened plates are described and compared. Out-of-plane
loading is also considered and its influence on stability discussed. The requirements for finite
element models of stiffened plates are outlined using those for unstiffened plates as a basis.

1. INTRODUCTION

The automation of welding procedures and the need to design elements not only to have
the necessary resistance to external actions but also to meet aesthetic and serviceability
requirements leads to an increased tendency to employ thin-walled, plated structures, especially
when the use of rolled sections is excluded, due to the form and the size of the structure. Through
appropriate selection of plate thicknesses, steel qualities and form and position of stiffeners,
cross-sections can be best adapted to the actions applied and the serviceability conditions, thus
saving material weight. Examples of such structures, shown in Figure 1, are webs of plate
girders, flanges of plate girders, the walls of box girders, thin-walled roofing, facades, etc.
Plated elements carry simultaneously:

a) actions normal to their plane,

b) in-plane actions.

Out-of-plane action is of secondary importance for such steel elements since, due to the
typically small plate thicknesses involved, they are not generally used for carrying transverse
actions. In-plane action, however, has significant importance in plated structures.

The intention of design is to utilise the full strength of the material. Since the slenderness
of such plated elements is large due to the small thicknesses, their carrying resistance is reduced
due to buckling. An economic design may, however, be achieved when longitudinal and/or
transverse stiffeners are provided. Such stiffeners may be of open or of torsionally rigid closed
sections, as shown in Figure 2. When these stiffeners are arranged in a regular orthogonal grid,
and the spacing is small enough to 'smear' the stiffeners to a continuum in the analysis, such a
stiffened plate is called an orthogonal anisotropic plate or in short, an orthotropic plate (Figure
3). In this lecture the buckling behaviour of stiffened plate panels subjected to in-plane actions
will be presented. The behaviour under out-of-plane actions is also discussed as is the influence
of the out-of-plane action on the stability of stiffened plates.
Specific topics such as local actions and the tension field method are covered in the lectures on
plate girders.

2. STIFFENED PLATES UNDER IN-PLANE LOADING

2.1 Action Distribution

2.1.1 Distribution resulting from membrane theory

The stress distribution can be determined from the solutions of Navier's equations (see
Lecture 8.2 Section 2.1.1) but, for stiffened plates, this is limited to plates where the longitudinal
and transverse stiffeners are closely spaced, symmetrical to both sides of the plate, and produce
equal stiffness in the longitudinal and transverse direction, see Figure 4. This configuration leads
to an isotropic behaviour when the stiffeners are smeared out. In practice this way of stiffening is
not practical and therefore not commonly used.
All deviations from the "ideal" situation (eccentric stiffeners, etc.) have to be taken into account
when calculating the stress distribution in the plate.
2.1.2 Distribution resulting from linear elastic theory using Bernouilli's hypothesis

As for unstiffened plates the most practical way of determining the stress distribution in
the panel is using the plane strain hypothesis. Since stiffened plates have a relatively large width,
however, the real stress distribution can differ substantially from the calculated stress distribution
due to the effect of shear lag.

Shear lag may be taken into account by a reduced flange width concentrated along the edges and
around stiffeners in the direction of the action (see Figure 5).
2.1.3 Distribution resulting from finite element methods

The stiffeners can be modelled as beam-column elements eccentrically attached to the

plate elements, see Lecture 8.2, Section 2.1.3.

In the case where the stiffeners are relatively deep beams (with large webs) it is better to
model the webs with plate elements and the flange, if present, with a beam-column element.

2.2 Stability of Stiffened Plates

2.2.1 Linear buckling theory

The knowledge of the critical buckling load for stiffened plates is of importance not only
because design was (and to a limited extent still is) based on it, but also because it is used as a
parameter in modern design procedures. The assumptions for the linear buckling theory of plates
are as follows:

a) the plate is perfectly plane and stress free.

b) the stiffeners are perfectly straight.

c) the loading is absolutely concentric.

d) the material is linear elastic.

e) the transverse displacements are relatively small.

The equilibrium path has a bifurcation point which corresponds to the critical action (Figure 6). 
 

Analytical solutions, through direct integration of the governing differential equations

are, for stiffened plates, only possible in specific cases; therefore, approximate numerical
methods are generally used. Of greatest importance in this respect is the Rayleigh-Ritz approach,
which is based on the energy method. If o, and I represent the total potential energy of the
plate in the undeformed initial state and at the bifurcation point respectively (Figure 6), then the
application of the principle of virtual displacements leads to the expression:

(I) = (o + o) = (o + o + 2o + ....) = 0 (1)

since I is in equilibrium. But the initial state is also in equilibrium and therefore o = 0. The
stability condition then becomes:

(2o) = 0 (2)

2o in the case of stiffened plates includes the strain energy of the plate and the stiffeners and
the potential of the external forces acting on them. The stiffeners are characterized by three
dimensionless coefficients , , υ expressing their relative rigidities for extension, flexure and
torsion respectively.

For rectangular plates simply supported on all sides (Figure 6) the transverse
displacements in the buckled state can be approximated by the double Fourier series:

(3)

which complies with the boundary conditions. The stability criterion, Equation (2), then
becomes:

(4)

since the only unknown parameters are the amplitudes a mn, Equations (4) form a set of linear and
homogeneous linear equations, the number of which is equal to the number of non-zero
coefficients amn retained in the Ritz-expansion. Setting the determinant of the coefficients equal
to 0 yields the buckling equations. The smallest Eigenvalue is the so-called buckling coefficient
k. The critical buckling load is then given by the expression:

cr = kE or cr = kE (5)

with E =
 The most extensive studies on rectangular, simply supported stiffened plates were carried
out by Klöppel and Scheer[1] and Klöppel and Möller[2]. They give charts, as shown in Figure
7, for the determination of k as a function of the coefficients  and , previously described, and
the parameters  = a/b and  =2/1 as defined in Figure 6a. Some solutions also exist for
specific cases of plates with fully restrained edges, stiffeners with substantial torsional rigidity,
etc. For relevant literature the reader is referred to books by Petersen[3] and by Dubas and
Gehri[4].

When the number of stiffeners in one direction exceeds two, the numerical effort required
to determine k becomes considerable; for example, a plate panel with 2 longitudinal and 2
transverse stiffeners requires a Ritz expansion of 120. Practical solutions may be found by
"smearing" the stiffeners over the entire plate. The plate then behaves orthotropically, and the
buckling coefficient may be determined by the same procedure as described before.

An alternative to stiffened plates, with a large number of equally spaced stiffeners and the
associated high welding costs, are corrugated plates, see Figure 2c. These plates may also be
treated as orthotropic plates, using equivalent orthotropic rigidities[5].

So far only the application of simple action has been considered. For combinations of
normal and shear stresses a linear interaction, as described by Dunkerley, is very conservative.
On the other hand direct determination of the buckling coefficient fails due to the very large
number of combinations that must be considered. An approximate method has, therefore, been
developed, which is based on the corresponding interaction for unstiffened plates, provided that
the stiffeners are so stiff that buckling in an unstiffened sub-panel occurs before buckling of the
stiffened plate. The critical buckling stress is determined for such cases by the expression:

vcr = k Z1s E (6)

where E has the same meaning as in Equation (5).

s is given by charts (Figure 8b).

Z1 =

k , k are the buckling coefficients for normal and shear stresses acting independently

For more details the reader is referred to the publications previously mentioned.

Optimum rigidity of stiffeners

Three types of optimum rigidity of stiffeners *, based on linear buckling theory, are
usually defined[6]. The first type I*, is defined such that for values  > I* no further increase of
k is possible, as shown in Figure 9a, because for  = I* the stiffeners remain straight.
 The second type II*, is defined as the value for which two curves of the buckling
coefficients, belonging to different numbers of waves, cross (Figure 9b). The buckling
coefficient for  < II* reduces considerably, whereas it increases slightly for  > II*. A stiffener
with  = II* deforms at the same time as the plate buckles.

The third type III* is defined such that the buckling coefficient of the stiffened plate
becomes equal to the buckling coefficient of the most critical unstiffened subpanel (Figure 9c).

The procedure to determine the optimum or critical stiffness is, therefore, quite simple.
However, due to initial imperfections of both plate and stiffeners as a result of out of straightness
and welding stresses, the use of stiffeners with critical stiffness will not guarantee that the
stiffeners will remain straight when the adjacent unstiffened plate panels buckle.

This problem can be overcome by multiplying the optimum (critical) stiffness by a factor,
m, when designing the stiffeners.

The factor is often taken as m = 2,5 for stiffeners which form a closed cross-section together with
the plate, and as m = 4 for stiffeners with an open cross-section such as flat, angle and T-
stiffeners.

2.2.2 Ultimate resistance of stiffened plates

Behaviour of Stiffened Plates

Much theoretical and experimental research has been devoted to the investigation of
stiffened plates. This research was intensified after the collapses, in the 1970's, of 4 major steel
bridges in Austria, Australia, Germany and the UK, caused by plate buckling. It became evident
very soon that linear buckling theory cannot accurately describe the real behaviour of stiffened
plates. The main reason for this is its inability to take the following into account:

a) the influence of geometric imperfections and residual welding stresses.

b) the influence of large deformations and therefore the post buckling behaviour.

c) the influence of plastic deformations due to yielding of the material.

d) the possibility of stiffener failure.

Concerning the influence of imperfections, it is known that their presence adversely

affects the carrying resistance of the plates, especially in the range of moderate slenderness and
for normal compressive (not shear) stresses.

Large deformations, on the other hand, generally allow the plate to carry loads in the
post-critical range, thus increasing the action carrying resistance, especially in the range of large
slenderness. The post-buckling behaviour exhibited by unstiffened panels, however, is not
always present in stiffened plates. Take, for example, a stiffened flange of a box girder under
compression, as shown in Figure 10. Since the overall width of this panel, measured as the
distance between the supporting webs, is generally large, the influence of the longitudinal
supports is rather small. Therefore, the behaviour of this flange resembles more that of a strut
under compression than that of a plate. This stiffened plate does not, accordingly, possess post-
buckling resistance.
 As in unstiffened panels, plastic deformations play an increasingly important role as the
slenderness decreases, producing smaller ultimate actions.

The example of a stiffened plate under compression, as shown in Figure 11, is used to
illustrate why linear bucking theory is not able to predict the stiffener failure mode. For this plate
two different modes of failure may be observed: the first mode is associated with buckling
failure of the plate panel; the second with torsional buckling failure of the stiffeners. The overall
deformations after buckling are directed in the first case towards the stiffeners, and in the second
towards the plate panels, due to the up or downward movement of the centroid of the middle
cross-section. Experimental investigations on stiffened panels have shown that the stiffener
failure mode is much more critical for both open and closed stiffeners as it generally leads to
smaller ultimate loads and sudden collapse. Accordingly, not only the magnitude but also the
direction of the imperfections is of importance.
 Due to the above mentioned deficiencies in the way that linear buckling theory describes
the behaviour of stiffened panels, two different design approaches have been recently developed.
The first, as initially formulated by the ECCS-Recommendations [7] for allowable stress design
and later expanded by DIN 18800, part 3[8] to ultimate limit state design, still uses values from
linear buckling theory for stiffened plates. The second, as formulated by recent Drafts of ECCS-
Recommendations [9,10], is based instead on various simple limit state models for specific
geometric configurations and loading conditions. Both approaches have been checked against
experimental and theoretical results; they will now be briefly presented and discussed.

Design Approach with Values from the Linear Buckling Theory

With reference to a stiffened plate supported along its edges (Figure 12), distinction is
made between individual panels, e.g. IJKL, partial panels, i.e. EFGH, and the overall panel
ABCD. The design is based on the condition that the design stresses of all the panels shall not
exceed the corresponding design resistances. The adjustment of the linear buckling theory to the
real behaviour of stiffened plates is basically made by the following provisions:

a) Introduction of buckling curves as illustrated in Figure 12b.

b) Consideration of effective widths, due to local buckling, for flanges associated with stiffeners.

c) Interaction formulae for the simultaneous presence of stresses x, y and  at the ultimate limit
state.

d) Additional reduction factors for the strut behaviour of the plate.

e) Provision of stiffeners with minimum torsional rigidities in order to prevent lateral-torsional

buckling.
 
Design Approach with Simple Limit State Models

Drafts of European Codes and Recommendations have been published which cover the
design of the following elements:

a) Plate girders with transverse stiffeners only (Figure 13a) - Eurocode 3 [11].

b) Longitudinally stiffened webs of plate and box girders (Figure 13b) - ECCS-TWG 8.3, 1989.

c) Stiffened compression flanges of box girders (Figure 13c) - ECCS [10].

 
 Only a brief outline of the proposed models is presented here; for more details reference
should be made to Lectures 8.4, 8.5, and 8.6 on plate girders and on box girders:

The stiffened plate can be considered as a grillage of beam-columns loaded in

compression. For simplicity the unstiffened plates are neglected in the ultimate resistance and
only transfer the loads to the beam-columns which consist of the stiffeners themselves together
with the adjacent effective plate widths. This effective plate width is determined by buckling of
the unstiffened plates (see Section 2.2.1 of Lecture 8.2). The bending resistance Mu, reduced as
necessary due to the presence of axial forces, is determined using the characteristics of the
effective cross-section. Where both shear forces and bending moments are present
simultaneously an interaction formula is given. For more details reference should be made to the
original recommendations.

The resistance of a box girder flange subjected to compression can be determined using
the method presented in the ECCS Recommendations referred to previously, by considering a
strut composed of a stiffener and an associated effective width of plating. The design resistance
is calculated using the Perry-Robertson formula. Shear forces due to torsion or beam shear are
taken into account by reducing the yield strength of the material according to the von Mises yield
criterion. An alternative approach using orthotropic plate properties is also given.

The above approaches use results of the linear buckling theory of unstiffened plates
(value of Vcr, determination of beff etc.). For stiffened plates the values given by this theory are
used only for the expression of the rigidity requirements for stiffeners. Generally this approach
gives rigidity and strength requirements for the stiffeners which are stricter than those mentioned
previously in this lecture.

Discussion of the Design Approaches

Both approaches have advantages and disadvantages.

The main advantage of the first approach is that it covers the design of both unstiffened
and stiffened plates subjected to virtually any possible combination of actions using the same
method. Its main disadvantage is that it is based on the limitation of stresses and, therefore, does
not allow for any plastic redistribution at the cross-section. This is illustrated by the example
shown in Figure 14. For the box section of Figure 14a, subjected to a bending moment, the
ultimate bending resistance is to be determined. If the design criterion is the limitation of the
stresses in the compression thin-walled flange, as required by the first approach, the resistance is
Mu = 400kNm. If the computation is performed with effective widths that allow for plastic
deformations of the flange, Mu is found equal to 550kNm.
 The second approach also has some disadvantages: there are a limited number of cases of
geometrical and loading configurations where these models apply; there are different
methodologies used in the design of each specific case and considerable numerical effort is
required, especially using the tension field method.

Another important point is the fact that reference is made to webs and flanges that cannot
always be defined clearly, as shown in the examples of Figure 15.

For a box girder subjected to uniaxial bending (Figure 15a) the compression flange and
the webs are defined. This is however not possible when biaxial bending is present (Figure 15b).
Another example is shown in Figure 15c; the cross-section of a cable stayed bridge at the
location A-A is subjected to normal forces without bending; it is evident, in this case, that the
entire section consists of "flanges".
Finite Element Methods

In determining the stability behaviour of stiffened plate panels, basically the same
considerations hold as described in Lecture 8.2, Section 2.2.2. In addition it should be noted that
the stiffeners have to be modelled by shell elements or by a combination of shell and beam-
column elements. Special attention must also be given to the initial imperfect shape of the
stiffeners with open cross-sections.

It is difficult to describe all possible failure modes within one and the same finite element
model. It is easier, therefore, to describe the beam-column behaviour of the stiffeners together
with the local and overall buckling of the unstiffened plate panels and the stiffened assemblage
respectively and to verify specific items such as lateral-torsional buckling separately (see Figure
16). Only for research purposes is it sometimes necessary to model the complete structure such
that all the possible phenomena are simulated by the finite element model.
3. STIFFENED PLATES UNDER OUT-OF-PLANE ACTION APPLICATION
3.1 Action Distribution

3.1.1 Distribution resulting from plate theory

The theory described in Section 3.1.1 of Lecture 8.2 can only be applied to stiffened
plates if the stiffeners are sufficiently closely spaced so that orthotropic behaviour occurs. If this
is not the case it is better to consider the unstiffened plate panels in between the stiffeners
separately. The remaining grillage of stiffeners must be considered as a beam system in bending
(see Section 3.1.2).

3.1.2 Distribution resulting from a grillage under lateral actions filled in with unstiffened sub-
panels

The unstiffened sub-panels can be analysed as described in Section 3.1.1 of Lecture 8.2.

The remaining beam grillage is formed by the stiffeners which are welded to the plate,
together with a certain part of the plate. The part can be taken as for buckling, namely the
effective width as described in Section 2.2.2 of this Lecture. In this way the distribution of forces
and moments can be determined quite easily.

3.1.3 Distribution resulting from finite element methods (FEM)

Similar considerations hold for using FEM to determine the force and moment
distribution in stiffened plates which are subject to out-of-plane actions as for using FEM for
stiffened plates loaded in-plane (see Section 2.1.3) except that the finite elements used must be
able to take large deflections and elastic-plastic material behaviour into account.

3.2 Deflection and Ultimate Resistance

All considerations mentioned in Section 3.2 of Lecture 8.2 for unstiffened plates are valid
for the analysis of stiffened plates both for deflections and ultimate resistance. It should be noted,
however, that for design purposes it is easier to verify specific items, such as lateral-torsional
buckling, separately from plate buckling and beam-column behaviour.

4. INFLUENCE OF OUT-OF-PLANE ACTIONS ON THE STABILITY OF STIFFENED

PLATES

The points made in Section 4 of Lecture 8.2 also apply here; that is, the stability of the
stiffened plate is unfavourably influenced if the deflections, due to out-of-plane actions, are
similar to the stability collapse mode.
5. CONCLUDING SUMMARY

 Stiffened plates are widely used in steel structures because of the greater efficiency that the
stiffening provides to both stability under in-plane actions and resistance to out-of-plane
actions.
 Elastic linear buckling theory may be applied to stiffened plates but numerical techniques such
as Rayleigh-Ritz are needed for most practical situations.
 Different approaches may be adopted to defining the optimum rigidity of stiffeners.
 The ultimate behaviour of stiffened plates is influenced by geometric imperfections and yielding
in the presence of residual stresses.
 Design approaches for stiffened plates are either based on derivatives of linear buckling theory
or on simple limit state models.
 Simple strut models are particularly suitable for compression panels with longitudinal stiffeners.
 Finite element models may be used for concrete modelling of particular situations.

6. REFERENCES

[1] Klöppel, K., Scheer, J., "Beulwerte Ausgesteifter Rechteckplatten", Bd. 1, Berlin, W. Ernst u.
Sohn 1960.

[2] Klöppel, K., Möller, K. H., "Beulwerte Ausgesteifter Rechteckplatten", Bd. 2, Berlin, W.
Ernst u. Sohn 1968.

[3] Petersen, C., "Statik und Stabilität der Baukonstruktionen", Braunschweig: Vieweg 1982.

[4] Dubas, P., Gehri, E., "Behaviour and Design of Steel Plated Structures", ECCS, 1986.

[5] Briassoulis, D., "Equivalent Orthotropic Properties of Corrugated Sheets", Computers and
Structures, 1986, 129-138.

[6] Chwalla, E., "Uber die Biegungsbeulung der langsversteiften Platte und das Problem der
Mindeststeifigeit", Stahlbau 17, 1944, 84-88.

[7] ECCS, "Conventional design rules based on the linear buckling theory", 1978.

[8] DIN 18800 Teil 3 (1990), "Stahlbauten, Stabilitätsfalle, Plattenbeulen", Berlin: Beuth.

[9] ECCS, "Design of longitudinally stiffened webs of plate and box girders", Draft 1989.

[10] ECCS, "Stiffened compression flanges of box girders", Draft 1989.

[11] Eurocode 3, "Design of Steel Structures": ENV 1993-1-1: Part 1.1: General rules and rules
for buildings, CEN, 1992.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.4.1: Plate Girder Behaviour and

Design I
OBJECTIVE

To introduce basic aspects of the behaviour and design of plate girders. To explain how the
typical proportions employed influence the types of behaviour that must be addressed in design,
and to identify the various buckling considerations involved, as a preparation for subsequent
consideration of the design approaches of Eurocode 3 [1].

PREREQUISITES

None

RELATED LECTURES
Lectures 3.2: Erection
Lecture 7.2: Cross-section Classification
Lecture 7.3: Local Buckling
Lecture 8.1: Introduction to Plate Behaviour and Design
Lecture 8.4.2: Plate Girder Behaviour and Design - II
Lecture 8.4.3: Plate Girder Design - Special Topics
Lecture 11.8: Splices in Buildings
Lecture 14.4: Crane Runway Girders
Lecture 15B.3: Plate Girder and Beam Bridges
SUMMARY
Modern plate girders are introduced by explaining typical usage, types and the reasons for their
inherent slender proportions. Their behaviour is described with particular emphasis on the
different forms of buckling that can occur. The general basis of plate girder design is discussed
in a simplified way as a prelude to a more detailed presentation in Lectures 8.4.2 and Lecture
8.4.3. Post-buckling and tension field action are introduced and the roles of the main components
in a plate girder identified.
1. INTRODUCTION

Modern plate girders are normally fabricated by welding together two flanges and a web
plate, as shown in Figure 1. Such girders are capable of carrying greater loads over longer spans
than is generally possible using standard rolled sections or compound girders. Plate girders are
typically used as long-span floor girders in buildings, as bridge girders, and as crane girders in
industrial structures.

Plate girders are at their most impressive in modern bridge construction where main
spans of well over 200m are feasible, with corresponding cross-section depths, haunched over
the supports, in the range of 5-10m. Because plate girders are fabricated separately, each may be
designed individually to resist the applied actions using proportions that ensure low self-weight
and high load resistance.

For efficient design it is usual to choose a relatively deep girder, thus minimising the
required area of flanges for a given applied moment, Msd. This obviously entails a deep web
whose area will be minimised by reducing its thickness to the minimum required to carry the
applied shear, Vsd. Such a web may be quite slender (i.e. a high d/t w ratio) and may be prone to
local buckling (see Lecture 7.3) and shear buckling (see below). Such buckling problems have to
be given careful consideration in plate girder design. One way of improving the load carrying
resistance of a slender plate is to employ stiffeners (Lecture 8.1); the selection of appropriate
forms of stiffening is an important aspect of plate girder design.

1.1 Types

There are several forms of plate girder; Figure 2 illustrates three different types -
unstiffened, transversely stiffened, and transversely and longitudinally stiffened. The three
girders shown have bisymmetric I-profile cross-sections, although flanges of different size are
sometimes used, as already shown in Figure 1. Other types of cross-section (see Figure 3) are
monosymmetric I-profiles, which are popular in composite construction with the smaller flange
on top (see Lecture 10.2), or as crane girders (see Lecture 14.4) with the larger flange on top.
Figure 3 also shows two other (less common) variations - the "delta girder" and the tubular-top-
flange girder - both being possible solutions in cases of long laterally-unsupported top
compression flanges prone to lateral-torsional buckling (see Lecture 7.9.1 and 7.9.2).
 There is also considerable scope for variation of cross-section in the longitudinal
direction. A designer may choose to reduce the flange thickness (or breadth) in a zone of low
applied moment, especially when a field-splice facilitates the change. Equally, in a zone of high
shear, the designer might choose to thicken the web plate (see Figure 4). Alternatively, higher
grade Fe E355 steel might be employed for zones of high applied moment and shear, while
standard grade Fe E235 would be used elsewhere. So-called "hybrid" girders with different
strength material in the flanges and the web offer another possible means of more closely
matching resistance to requirements. More unusual variations are adopted in special
circumstances, such as bridgework (see Lecture 15B.4) e.g. tapered girders, cranked girders,
haunched girders (see Figure 5), and of course, plate girders with web holes to accommodate
services, see Figure 6.
1.2 Proportions

Since the designer, in principle, is quite free to choose all the dimensions of a plate
girder, some indication of the more usual proportions is now given (see also Figure 7):

Depth: Overall girder depth, h, will usually be in the range L o/12  h  Lo/8, where Lo is the
length between points of zero moment. However, for plate girder bridges the range will extend to
approximately Lo/20.

Flange breadth: The breadth, b, will usually be in the range h/5  b  h/3, b being in multiples
of 25mm. 'Wide flats' may be used unless the flange is very wide.
Flange thickness: The flange thickness, tf, will usually at least satisfy the requirements of
Eurocode 3 (Table 5.3.1) for Class 3 (semi-compact) sections, i.e. c/t f  14. The thickness will
usually be chosen from the standard plate thicknesses.

Web thickness: Web thickness, tw, will determine the exact basis for web design, depending on
whether the web is classified with regard to shear buckling as "thick" or "thin" (see later). Thin
webs will often require stiffening; this may take the form of transverse stiffeners, longitudinal
stiffeners or a combination, see Figure 2. Longitudinally stiffened girders are more likely to be
found in large bridge construction where high d/t w ratios are appropriate, e.g. 200  d/tw  500,
due to the need to minimise self-weight.

Clearly, depending on the particular loading pattern, and on depth and breadth restrictions, one
can expect wide variations within all the above limits which should be regarded as indicative
only.

2. DESIGN CONCEPTS

Under static loading, ultimate limit states such as strength and stability will normally
govern most plate girder design, with serviceability limit states such as deflection or vibration
being less critical. Some absolute limits on plate slenderness are advisable so as to ensure
sufficient robustness during erection. A generally accepted method [2] for designing plate girders
(which is permitted by Eurocode 3) subject to a moment M ad and a coincident shear Vad is to
proportion the flanges to carry all the moment with the web taking all the shear. This provides a
particularly convenient means for obtaining an initial estimate of girder proportions.

Thus, at any particular cross-section along a laterally-restrained plate girder, subject to

specific values of bending moment and shear force, the flange and web plates can be sized
separately. The required flange plate area may readily be obtained as follows:

Af = M/[(h - tf)fy/MO]  M/(hfy/MO) (1)

(An iteration or two may be required depending on an assumed value of tf and its
corresponding fy value from Table 3.1, Eurocode 3). Because the (normally) slender web will
prevent the plastic moment of resistance of the cross-section from being attained, the flange b/t f
ratio need only comply with the Eurocode 3 (Table 5.3.1) requirements for a Class 3 (semi-
compact) flange. The cross-sectional moment of resistance may then be checked using:

Mf.Rd = b tf (h - tf)fy/MO (2)

Unfortunately, economic sizing of the web plate is not quite as straightforward, although
where a thick web (defined later) is acceptable it can be rapidly sized by assuming uniform shear
stress y over its whole area. The web-to-flange fillet welds must be designed to transmit the
longitudinal shear at the flange/web interface.
3. INFLUENCE OF BUCKLING ON DESIGN

Provided that the individual plate elements in a girder are each kept sufficiently stocky,
the design may be based on straightforward yield strength considerations. Economic and
practical considerations will, however, dictate that not all of these conditions will always be met.
In most cases various forms of buckling must be taken into account. Figure 8 lists the different
phenomena.
3.1 Shear Buckling of the Web

Once the d/tw value for an unstiffened web exceeds a limiting figure (69 in Eurocode 3)
the web will buckle in shear before it reaches its full shear capacity Awy. Diagonal buckles, of
the type shown in Figure 9(a), resulting from the diagonal compression associated with the web
shear will form. Their appearance may be delayed through the use of vertical stiffeners, see
Figure 9(b) since the load at which shear buckling is initiated is a function of both d/t w and panel
aspect ratio a/d.
3.2 Lateral-Torsional Buckling of the Girder

This topic is covered in Lecture 7.9.1 and 7.9.2.

3.3 Local Buckling of the Compression Flange

Provided that outstand proportions c/tf are suitably restricted, local buckling will have no
effect on the girder's load carrying resistance.

3.4 Compression Buckling of the Web

Webs for which d/tw  124 and which are not subject to any axial load will permit the
full elastic moment resistances of the girder to be attained. If this limit of d/t w (or a lower one if
axial compression in the girder as a whole is also present) is exceeded, then moment resistance
must be reduced accordingly. If it is desired to reach the girder's full plastic moment resistance a
stricter limit will be appropriate.

3.5 Flange Induced Buckling of the Web

If particularly slender webs are used, the compression flange may not receive enough
support to prevent it from buckling vertically rather like an isolated strut buckling about its
minor axis. This possibility may be eliminated by placing a suitable limit on d/tw. Transverse
stiffeners also assist in resisting this form of buckling.

3.6 Local Buckling of the Web

Vertical loads may cause buckling of the web in the region directly under the load as for
a vertical strut. The level of loading that may safely be carried before this happens will depend
upon the exact way in which the load is transmitted to the web, the web proportions, and the
level of overall bending present.

4. POST-BUCKLING STRENGTH OF WEB

Owing to the post-buckling behaviour (see Lecture 8.3) plates, unlike struts, are often
able to support loads considerably in excess of their initial buckling load. In plate girder webs a
special form of post-buckling termed "tension field action" is possible. Tension field action
involves a change in the way in which the girder resists shear loading from the development of
uniform shear in the web at low shear loads, to the equivalent truss action, shown in Figure 10, at
much higher loads. In this action the elements equivalent to truss members are: the flanges,
which form the chords; the vertical stiffeners which form the struts; and the diagonal tension
bands which form the ties. The compressive resistance of the other diagonal of each web panel is
virtually eliminated by the shear buckling. The way in which this concept is utilized in design is
explained in Lecture 8.4.2.
5. DESIGN CONSIDERATIONS

The principal functions of the main components found in plate girders may be summarised as
follows:

Flanges   resist moment

Web   resists shear
Web/flange welds   resist longitudinal shear at interface
Vertical stiffeners   improve shear buckling resistance
Longitudinal stiffeners   improve shear and/or bending resistance.
6. CONCLUDING SUMMARY

 The main components in a plate girder have been identified and their principal functions noted.
 Initial sizing may be made on the basis that the flanges carry all of the moment and the web
takes all of the shear.
 Shear buckling is likely to prevent the full web shear resistance from being attained in slender
webs. Its appearance need not imply failure since additional load may be carried through
tension field action.
 Web stiffeners (transverse and/or longitudinal) enhance both initial buckling and post-buckling
resistance.

7. REFERENCES

[1] Eurocode 3: "Design of Steel Structures": European Prestandard ENV1993-1-1: Part 1,

General rules and rules for buildings, CEN, 1992.

[2] Narayanan, R. (ed)., "Plated Structures; Stability and Strength", Applied Science Publishers,
London, 1983.

Chapter 1 covers basic aspects of plate girder behaviour and design.

8. ADDITIONAL READING

1. Dubas, P. and Gehri, E. (eds), "Behaviour and Design of Plated Steel Structures", Publication No
44, ECCS, 1986.

Chapters 4 and 5 provide more detailed accounts of the main features of plate girder behaviour
and design.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.4.2: Plate Girder Behaviour and

Design II
OBJECTIVE/SCOPE

To present the basic design methods for plate girders subjected to either shear or moment, or a
combination of both.

PREREQUISITES

Lecture 8.4.1: Plate Girder Behaviour and Design I

RELATED LECTURES

Lecture 7.3: Local Buckling

Lecture 7.8.1: Restrained Beams I

Lecture 7.8.2: Restrained Beams II

Lecture 7.9.1: Unrestrained Beams I

Lecture 7.9.2: Unrestrained Beams II

Lecture 8.4.3: Plate Girder Design - Special Topics

SUMMARY

The design methods for plate girders subject to bending and shear, according to the methods of
Eurocode 3[1], are presented. For shear loading two methods are described: the "simple post-
critical method", and the "tension field method"; interaction diagrams can be used with both
methods to allow for the effect of coincident moments.
1. INTRODUCTION

Any cross-section of a plate girder is normally subjected to a combination of shear force

and bending moment. The primary function of the top and bottom flange plates of the girder is to
resist the axial compressive and tensile forces arising from the applied bending moment. The
primary function of the web plate is to resist the applied shear force.

Plate girders are normally designed to support heavy loads over long spans in situations
where it is necessary to produce an efficient design by providing girders of high strength to
weight ratio. The search for an efficient design produces conflicting requirements, particularly in
the case of the web plate. To produce the lowest axial flange force for a given bending moment,
the web depth (d) must be made as large as possible. To reduce the self weight, the web
thickness (tw) must be reduced to a minimum. As a consequence, in many instances the web plate
is of slender proportions and is therefore prone to buckling at relatively low values of applied
shear. A similar conflict may exist for the flange proportions. The required flange area is defined
by the flange force and material yield stress. The desire to increase weak axis second moment of
area encourages wide, thin flanges. Such flanges are prone to local buckling.

Plate elements do not collapse when they buckle; they can possess a substantial post-
buckling reserve of resistance. For an efficient design, any calculation relating to the ultimate
limit state should take the post-buckling action into account. This is particularly so in the case of
a web plate in shear where the post-buckling resistance arising from tension field action can be
very significant.

Thus, in designing a plate girder it is necessary to evaluate the buckling and post-
buckling action of webs in shear, and of flange plates in compression. The design of plate girder
flanges largely follows procedures already discussed in Lecture 7.8, Lecture 7.9.1, and Lecture
7.9.2 for beams. However, the design of web plates operating in the post-buckling range is very
different and will be discussed here in some detail. The lecture will start by concentrating upon
the resistance of plate girders to predominantly shear loading. The effects of high co-existent
bending moments will be considered.

The lecture will concentrate only on the main aspects of girder design assuming a basic
cross-section. In particular, it is assumed that:

1. Only transverse web stiffeners are present (i.e. there are no longitudinal stiffeners).
2. Transverse web stiffeners possess sufficient stiffness and strength to resist the actions
transmitted to them by the web.
3. An appropriate means is available to anchor the tension field.
4. No vertical patch loads are applied between the positions of the transverse web stiffeners.
5. Only solid webs are considered (i.e. there are no web openings or holes).

Lecture 8.4.3 considers other important cases that do not comply with the above assumptions.
2. SHEAR BUCKLING RESISTANCE

A typical transversely stiffened plate girder is shown diagrammatically in Figure 1, which

also defines the notation used. The shear buckling resistance of the web depends mainly on the
depth to thickness ratio (d/tw), and upon the spacing (a) of the transverse web stiffeners.

Intermediate transverse stiffeners are normally employed to increase the shear buckling
resistance of the web, although designers may sometimes choose to use a thicker web plate rather
than incur the additional fabrication costs arising from the use of intermediate stiffeners. Girders
without intermediate stiffeners are normally termed "unstiffened" girders, even though they will
normally have stiffeners at points of support and possibly at the position of load application.

Web buckling should be checked in all cases where the depth to thickness ratio, (d/t w), of
the web exceeds 69 . Eurocode 3 then offers the choice of 2 methods for plate girder design.
The methods are:

a) the simple post-critical method, which may be applied to both stiffened and unstiffened
girders and is therefore of general application.

b) the tension field method, which may only be applied to girders with intermediate transverse
stiffeners. Even for such girders its range of application is limited to a range of stiffener spacing
defined by:

1,0  a/d  3,0

There is now considerable evidence [2] that tension field action does develop in girders
where the stiffener spacing lies outside this range, and also in unstiffened girders; such evidence,
however, has yet to be presented in a form that is suitable for inclusion in a design code.

The simple post-critical method is seen as a general-purpose method which can be

applied to the design of all girders. The tension field method, on the other hand, can be applied to
a certain range of girders only, but will lead to considerably more efficient designs for these
girders, because it takes full account of the post-buckling reserve of resistance. Each method will
now be discussed.
2.1 Calculation of the Shear Buckling Resistance by the Simple Post-Critical Method

This simple approach allows the design shear buckling resistance (Vba.Rd) to be
determined directly as follows:

Vba.Rd = d tw ba/M1 (1)

where all the terms in the expression are familiar, except the post-critical shear strength, ba. The
calculation of this term depends upon the slenderness of the web which may be conveniently
expressed by the following parameter:

(2)

Here, k is a shear buckling factor calculated from elastic buckling theory [3]. For
simplicity, it is conservatively assumed in this calculation that the boundaries of the web panel
are simply supported, since the true degree of restraint offered by the flanges and adjacent web
panels is not known. The resulting expression obtained for the shear buckling factor is dependent
upon the spacing of the transverse web stiffeners as follows:

for closely spaced intermediate stiffeners (a/d < 1,0) :

k = 4 +

for widely spaced intermediate stiffeners (a/d  1,0) :

k = 5,34 +

for unstiffened webs: k = 5,34

Knowing the shear buckling factor, the slenderness parameter is determined from
Equation (2) and the calculation of the post-critical shear strength then depends, as illustrated in
Figure 2, upon whether the web is:

a) stocky or thick ( w  0,8 , region AB in Figure 2) in which case the web will not buckle and
the shear stress at failure will reach the shear yield stress of the web material:

ba = fyw/

where fyw is the tensile yield strength

b) intermediate (0,8 < w < 1,2, region BC in Figure 2) which represents a transition stage from
yielding to buckling action with the shear strength being evaluated empirically from the
following:

ba = [1 - 0,625 ( w - 0,8)] (fyw/ )

c) slender or thin ( w  1,2, region CD in Figure 2) in which case the web will buckle before it
yields and a certain amount of post-buckling action is taken into account empirically:

ba =

The calculation of the shear buckling resistance by the simple post-critical method is then
completed by substitution of the appropriate value of ba into Equation (1).

 
2.2 Calculation of the Shear Buckling Resistance by the Tension Field Method

For transversely stiffened girders where the transverse stiffener spacing lies within the
range 1,0  a/d  3,0, full account may be taken of the considerable reserve of post-buckling
resistance. This reserve arises from the development of "tension field action" within the girder.

Figure 3a shows the development of tension field action in the individual web panels of a
typical girder. Once a web panel has buckled in shear, it loses its resistance to carry additional
compressive stresses. In this post-buckling range, a new load-carrying mechanism is developed,
whereby any additional shear load is carried by an inclined tensile membrane stress field. This
tension field anchors against the top and bottom flanges and against the transverse stiffeners on
either side of the web panel, as shown. The load-carrying action of the plate girder than becomes
similar to that of the N-truss in Figure 3b. In the post-buckling range, the resistance offered by
the web plates is analogous to that of the diagonal tie bars in the truss.

The total shear buckling resistance for design (V bb.Rd) is calculated in Eurocode 3, by
superimposing the post-buckling resistance upon the initial elastic buckling resistance as follows:
Total shear resistance = elastic buckling resistance + post-buckling resistance:

Vbb.Rd = (d tw bb)/M1 + 0,9 (gtw bb sin )/M1 (3)

The basis for this assumed behaviour is shown diagrammatically in Figure 4.

Figure 4a shows the situation prior to buckling, as represented by the first term in
Equation (3). At this stage, equal tensile and compressive principal stresses are developed in the
web. The shear buckling strength, bb, is calculated from elastic buckling theory and leads to
equations similar, but not identical, to those given earlier in Section 2.1 for the simple post-
critical shear strength ba. Thus, the calculation of the shear buckling resistance again depends, as
shown in Figure 5, upon whether the web is:

a) stocky or thick ( w = 0,8, region AB in Figure 5) in which case the web will not buckle and
the shear yield stress is again taken:

bb = fyw/

where fyw is the tensile yield strength

b) intermediate (0,8 < w < 1,25, region BC in Figure 5) where, in the transition from yield to
buckling:

bb = [1 - 0,8 ( w - 0,8)] (fyw/ )

c) slender or thin ( w  1,25, region CD in Figure 5) where the web will buckle and, from
elastic buckling theory:

bb = [1/ w
2
][fyw/3]

Thus, knowing bb, the first term of the expression in Equation (3) can be evaluated.

The evaluation of the second term, corresponding to the post-buckling action, is more
complex although it may still be reduced to a convenient design procedure, as described below.

In the post-buckling range, as shown in Figure 4b, an inclined tensile membrane stress
field is developed, at an inclination  to the horizontal. Since the flanges of the girder are
flexible, they will begin to bend inwards under the pull exerted by the tension field.
 Further increase in the load will result in yield occurring in the web under the combined
effect of the membrane stress field and the shear stress at buckling. The value of the tension field
stress (bb) at which yield will occur, termed the "strength of the tension field" in Eurocode 3,
may be determined by applying the Von Mises-Hencky yield criterion [2]. This results in the
following expression for the strength of the tension field:

bb = [fyw2 - 32bb + 2] 0.5 - 

where the term  = 1,5 bb sin 2 is introduced for convenience only.

Once the web has yielded, final failure of the girder will occur when the mechanism
comprising 4 plastic hinges has formed in the flanges, as shown in Figure 4c. A detailed analysis
of this collapse mechanism, by considering the internal forces developed in the web and imposed
upon the flanges (see [2]) allows the width (g) of the tension field, which appears in the second
term of Equation (3), to be evaluated:

g = d cos  - (a - sc - st) sin 

where, as in Figure 4c, sc and st denote the positions at which the plastic hinges form in the
compression and tension flanges respectively.

The hinge positions are calculated [2] from the knowledge that the hinges will form at the
point of maximum moment, and therefore zero shear, in the flanges; the appropriate expression is
as follows:

S = [2/sin ][MNf.Rk/tw bb]1/2  a                              (4)

where MNf.Rk is the plastic moment of resistance of the flange, i.e. 0,25 bt f2 fyf. When high
bending moments are applied to the girder, in addition to shear, then axial forces (N f.Sd) will be
developed in the flanges. Such axial forces will, of course, reduce the plastic moment of
resistance of the flanges. Their effects can be calculated from standard plasticity theory as:

MNf.Rk = 0,25 btf2 fyf {1 - [Nf.Sd / (btf fyf /mo  )]2}  (5)

All the terms required for the calculation of the total shear resistance from Equation (3),
other than the inclination  of the tension field, are now known. Unfortunately, the value of 
cannot be determined directly and an iterative procedure has to be adopted in which successive
values of  are assumed and the corresponding shear resistance evaluated in each case. The
process is repeated until the value of  providing the maximum, and therefore the required, value
of the shear resistance has been established. The variation of the shear resistance with  is not
very rapid. The correct value of  lies between a minimum of /2 and a maximum of , where 
is the slope of the panel diagonal tan-1(d/a), as shown in Figure 6. A parametric study [2] has
established that, for girders of normal proportions, the value of  which produces the maximum
value of shear resistance is approximately given by:
 =  /1,5

The assumption of this value of  will lead either to the correct value or to an
underestimation of the shear resistance. It will therefore give a safe approximation and will also
give a good starting value of  if a more accurate process of iteration is to be carried out. The
correct value of  is that which gives the maximum value of Vbb.Rd.
3. INTERACTION BETWEEN SHEAR AND BENDING

In general, any cross-section of a plate girder will be subjected to bending moment in

addition to shear. As discussed in [2], this combination makes the stress conditions in the girder
web considerably more complex. In the first place, the stresses from the bending moment will
combine with the shear stresses to give a lower buckling load. Secondly, in the post-buckling
range, the bending stresses will influence the magnitude of the tension field membrane stresses
required to produce yield in the web. Finally, as already discussed with reference to Equation (5),
the axial flange forces arising from the bending moment will reduce the plastic moment of
resistance of the flanges.

The proper evaluation of all these effects is complex but, as discussed in [2], certain
assumptions may be made about the interaction of moment and shear to produce a simple and
effective design procedure. In Eurocode 3, the procedure for allowing for moment/shear
interaction naturally depends upon whether the simple post-critical method of Section 2.1 or the
tension field method of Section 2.2 is being used to calculate the shear buckling resistance. Each
case will now be considered separately.

3.1 Interaction between Shear and Bending in the Simple Post-Critical Method

The interaction between shear and bending can be conveniently represented by the
diagram shown in Figure 7a (Figure 5.6.4a of Eurocode 3) where the shear resistance of the
girder is plotted on the vertical axis and the moment resistance is plotted horizontally. The
interaction represents a failure envelope, with any point lying on the curve defining the co-
existent values of shear and bending that the girder can just sustain.
 The interaction diagram can be considered in 3 regions. In region AB, the applied
bending moment Mad is low and the girder can then sustain a shear load V ad that is equal to the
full value of the shear buckling resistance calculated from the simple post-critical method, as in
Equation (1). Thus, in this region:

Msd  Mf.Rd

Vsd  Vba.Rd (6)

The moment that defines the end of the range at point B (M f.Rd) is the plastic moment of
resistance of the cross-section consisting of the flanges only, i.e. neglecting any contribution
from the web. In this calculation it is necessary to appreciate that the plates of the compression
flange may buckle and, if necessary, to take this into account by adopting an effective width b eff
for the flange. The calculation of this effective width is as described in Lecture 7.3 for an
outstand element in compression.

At the other extreme of the interaction diagram in region CD, the applied shear V ad is
low. Provided it does not exceed the limiting value of 0,5 V ba at point C then the plastic moment
of resistance of the complete cross-section Mpl.Rd need not be reduced to allow for shear.

In the intermediate region BC the co-existent applied moment M Sd and shear VSd values must
satisfy the following relationship:

MSd  Mf.Rd + (Mpl.Rd - Mf.Rd) [1 - (2VSd/Vba.Rd - 1)2] (7)

The complete range of moment/shear interaction has thus been defined for the simple post-
critical method.

3.2 Interaction between Shear and Bending in the Tension Field Method

The procedure for the tension field method follows that described above for the simple
post-critical method. It leads to the construction of a similar, though not identical, interaction
diagram, see Figure 7b (Figure 5.6.4b of Eurocode 3).

In the low moment region AB, again defined by values of the applied moment less than Mf.Rd, the
girder can sustain a shear load Vad that is equal to the "web only" shear resistance V bw.Rd
calculated from tension field theory. Thus:

MSd  Mf.Rd

VSd  Vbw Rd

The "web only" shear resistance is the specific value of the total shear resistance Vbw.Rd
calculated from Equation (1), for the case when MNf.Rk in EC3=0 in Equation (5). This is, in
effect, a conservative approach which neglects the contribution of the flanges to the tension field
action.
 At the other extreme in region CD, the procedure remains as for the simple post-critical
method except that the limiting value of shear at point C is now taken as 0,5V bw. Similarly, the
procedure for the intermediate region BC remains as before except that the substitution of the
tension field value Vbw for Vba in Equation (7) gives:

MSd  Mf.Rd + (Mpl.Rd - Mf.Rd) [1 - (2VSd / Vbw.Rd - 1)2] (8)

The complete range of moment/shear interaction is thus defined for the tension field method.

4. CONCLUDING SUMMARY

 Procedures for the design of plate girders subject to shear utilize varying degrees of post-
buckling resistance and correspond to either the "simple post-critical" or "tension field"
methods of Eurocode 3.
 Moment resistance of plate girders may normally be based on the plastic moment resistance of
a cross section consisting of the flanges only.
 Design for coincident shear and moment should be undertaken using an interaction diagram.
The simplest approach consists of designing the web to carry the whole of the shear, with the
flanges resisting the moment.
 The "tension field" method is more restricted in application than the "simple post-critical"
method, but gives higher strengths.
 Other aspects of design (stiffeners, etc.) are discussed in Lecture 8.4.3.

5. REFERENCES

[1] Eurocode 3 "Design of Steel Structures": European Prestandard ENV1993-1-1: Part 1.1,
General rules and rules for buildings, CEN, 1992.

[2] Narayanan, R. (ed), "Plated Structures; Stability and Strength", Applied Science Publishers,
London 1983.

Chapter 1 covers basic aspects of plate girder behaviour and design.

[3] Bulson, P. S. "The Stability of Flat Plates", Chatto & Windus, London, 1970.

General coverage of plate buckling and explanation of k values for numerous cases.

6. ADDITIONAL READING
1. Dubas, P. and Gehri, E. (eds)., "Behaviour and Design of Plated Steel Structures", Publication No.
44, ECCS, 1986.
Chapters 4 and 5 provide a detailed coverage of plate girder design, taking the reader well
beyond the content of this lecture. They also refer to numerous original sources.
2. Galambos, T. V. (ed)., "Guide to Stability Design Criteria for Metal Structures", 4th Edition, John
Wiley, 1988.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.4.3: Plate Girder Design -

Special Topics
OBJECTIVE/SCOPE

To extend the coverage of plate girder design previously given in Lectures 8.4.1 and Lecture
8.4.2. To include the design of transverse web stiffeners and end posts and consideration of patch
loading. To outline design procedures for longitudinally stiffened girders and for girders with
large web openings.

PREREQUISITES

Lecture 8.4.1: Plate Girder Behaviour and Design I

Lecture 8.4.2: Plate Girder Behaviour and Design II

RELATED LECTURES

Lectures 3.1: Fabrication

Lectures 3.2: Erection

Lecture 3.5: Fabrication/Erection of Buildings

SUMMARY

The detailed design of particular elements of plate girders is considered in this lecture.
The structural action of web panels, designed as described in earlier lectures, imposes stringent
requirements on adjacent boundary elements. This lecture considers the design of transverse web
stiffeners and end posts according to Eurocode 3 [1] and also considers the particular problems
caused by patch loading. Two other aspects of design, not currently covered by Part 1.1 of
Eurocode 3, viz. the design of longitudinally stiffened girders and girders with large web
openings, are also discussed.
1. INTRODUCTION

The two previous lectures on plate girders, Lectures 8.4.1 and Lecture 8.4.2, have
concentrated upon the main aspects of the structural behaviour upon which the design principles
are based. The two design approaches proposed in Eurocode 3 [1] have been outlined; these are
the "simple post-critical" method, which is generally applicable, and the "tension field" method
which gives significantly higher load resistances by taking the post-buckling resistance of the
girders into account.

This lecture seeks to complete the discussion of plate girder design by considering further
aspects of detailed design. For example, the development of post-buckling action in a web plate,
assumed in the previous lectures, can only occur when the elements at the boundary of that web
plate are able to provide an adequate anchorage for the tension field forces developed within the
plate. This lecture will consider the design of these boundary elements. These elements may be
in the form of intermediate transverse stiffeners or end posts.

Girders may be subjected to high loads in localised regions, away from stiffener
positions, creating a possibility that crippling of the web plate may occur. An example of this
occurs in crane gantry girders subjected to a vertical loading which travels along the flange. The
effects of such "patch loading" must be carefully taken into account in design. This aspect is very
thoroughly treated in Eurocode 3 [1]. This lecture outlines the relevant design principles.

Two other important aspects of plate girder design are the treatment of girders with
longitudinal web stiffeners, and of girders with large openings in the web plates. Openings are
frequently required, particularly in building construction, to allow access for service ducts, etc.
Neither of these two situations is covered in Part 1.1 of Eurocode 3. This lecture discusses good
practice in relation to the two situations.

2. TRANSVERSE WEB STIFFENERS

To achieve an effective design, i.e. a plate girder of high strength/weight ratio, it is

usually necessary to provide intermediate transverse web stiffeners. Eurocode 3 [1] only allows
the application of the tension field method, which has been shown in earlier lectures to give a
significantly enhanced load resistance, when the web is stiffened. The Eurocode also specifies
that such stiffeners must be spaced such that the stiffener spacing/web depth ratio (a/d) is within
the following range:

1,0  a/d  3,0

Transverse stiffeners play an important role in allowing the full ultimate load resistance
of a plate girder to be achieved. In the first place they increase the buckling resistance of the
web; secondly they must continue to remain effective after the web buckles, to provide
anchorage for the tension field; finally they must prevent any tendency for the flanges to move
towards one another.
 The satisfactory performance of a transverse stiffener can best be illustrated by
comparing the girders shown, after testing, in Slides 1 and 2. In Slide 1 the stiffeners have
remained straight and have clearly fulfilled the function of vertical struts in the simplified N-
truss model of the post-buckling action discussed in Lecture 8.4.2, see Figure 1. In Slide 2 the
stiffener has failed and has been unable to limit the buckling to the adjacent sub-panels of the
girder; instead, the buckle has run through the stiffener position extending over both panels.
Consequently, significant reduction in the failure load of the girder occurred.
 Slide 1

Slide 2

The requirements to ensure adequate stiffener performance are given in Section 5.6.5 of
Eurocode 3. First, the stiffener must be of adequate rigidity in the direction perpendicular to the
plane of the web to prevent web buckling. This condition is satisfied provided the stiffener has a
second moment of area Is that satisfies the following empirical formulae:
Is  1,5 d3tw3 /a2 when a/d <

Is  0,75 dtw3 when a/d 

Secondly, the buckling resistance of the vertical "stiffener strut" must be sufficient to
support the tension field forces shown in Figure 2a (which has been reproduced from Fig. 5.6.3a
of Eurocode 3). It must also resist the resultant axial compressive force N s that is imposed upon
it. This force is calculated as follows:

Ns = Vsd - dtw bb /M1

where bb is the initial shear buckling resistance of the web panels, calculated as given in Lecture
8.4.2. When the two web panels adjacent to the particular stiffener being designed are not
identical, the lower value of bb for the two panels should be taken. As previously, V sd is the
design value of the shear force.
 

The buckling resistance of the stiffener strut to this axial compressive force is then
calculated using Section 5.7.6 of Eurocode 3. Since the stiffener is attached to the web plate, a
portion of the web acts effectively with the stiffener in resisting the axial compression. It is
difficult to calculate the extent of this portion of the web but experimental observations have
allowed an empirical effective web width of 30tw to be established, as shown in Figure 3 (which
has been reproduced from Figure 5.7.4 of Eurocode 3). Having established the effective cross-
section of the stiffener strut in this way, its buckling resistance is determined as for any other
compression member according to Section 5.5.1 of the Eurocode.
 For a "load bearing" stiffener, i.e. a transverse stiffener located at a position where an
external load is applied to the girder, an additional consideration is necessary. The resistance of
the effective cross-section of the load bearing stiffener should also be checked at a position close
to the loaded flange.

3. END PANELS AND POSTS

The requirement for adequate boundary members to support the loading imposed by the
post-buckling tension field is particularly onerous in the case of the end panel of the girder. The
situation of the transverse stiffener at the end of the girder, i.e. the "end post", is very different
from that of an intermediate stiffener, compare Figure 2b to Figure 2a. At the end of the girder,
the forces imposed by the tension field in the end panel have to be resisted entirely by the end
post without support from any further adjacent panels.

Design procedures for end panels and posts are given in Clauses 5.6.4.3 and 5.6.4.4 of Eurocode
3 [1]. Basically, they allow the designer two options.

Firstly, the designer may choose not to design an end post that will provide adequate
anchorage for the tension field. As a consequence, the end panel of the web must be designed
according to the simple post-critical method so that a tension field does not develop within it.
This option offers a simple design procedure but has the disadvantage that the calculated shear
resistance of the end panel will be significantly lower than that of the internal web panels in the
girder. Since it is probable that the applied shear in the end region will be higher than at any
point on the span, this procedure will not provide an effective design solution if the stiffener
spacing remains constant over the complete length of the girder. As shown in Figure 4a, the
designer should then reduce the spacing of the stiffeners bounding the end panel so that the shear
resistance of that panel as calculated by the simple post-critical method becomes equal to that
calculated by the tension field method for the internal panels.
The more effective, but more complex, option is to design the end post to provide an adequate
anchorage for the web tension field. The end panel of the web can then be designed according to
the tension field method, so that the design shear buckling resistance (V bb.Rd ) can be calculated as
described in Lecture 8.4.2 for internal web panels, i.e.

Vbb.Rd = [(d tw bb ) + 0,9 (g tw bb sin )]/M1

The slight difference for the end panel arises in the calculation of the width g of the tension field.
For an internal panel the width is given by:

g = d cos  - (a - sc - st) sin 

where, sc and st denote the lengths over which the tension field anchors onto the compression and
tension flanges, see Figure 2a. For an end panel, the failure mechanism may be different since, as
shown in Slide 3, a plastic hinge may also form in the end post. 

Slide 3

This hinge affects the anchorage length for the compression flange which must now be
calculated as:

sc =

where Mpl.1 is the reduced plastic moment of the flange at the internal hinge position, allowing
for the presence of the axial force (Nf1) at that position.

The other plastic hinge will form either at the end of the flange, as in the case of an internal
panel, or in the end post. The location of the hinge, as defined by s s in Figure 2b, will depend
upon which of these two elements has the lower plastic moment of resistance. M pl.2 takes the
lesser of these two values.

In this way, Clause 5.6.4.3 of Eurocode 3 allows the geometry of the tension field
developed in the end panel to be fully defined, see Figure 2b. The design shear buckling
resistance Vbb.Rd of the panel can then be calculated together with the horizontal component Fbb of
the anchorage force of the tension field imposed on the end post:

Fbb = tw ss  bb cos2 

The end post resists this force by acting as a vertical beam spanning between the two flanges. For
this purpose it must satisfy the following criterion:

Mpl.2 + Mpl.3  0,5 Fbb ss

where the reduced plastic moment of the end post:

Mpl.3 = 0,25 bs ts2 fys {1 - [Ns3 /(bs ts fys )]2}

allows for the effect of the axial force in the end post:

Ns3 = Vsd -  bb tw (d - ss )

If it proves difficult to provide an end post in the form of a single plate to resist these
forces, then the designer may consider providing an end arrangement such as that shown in
Figure 4b. In this case, two transverse stiffeners are used. These two stiffeners and the portion of
the web projecting beyond the end support form a rigid end post to provide the necessary
anchorage for the tension field in the end panel. The disadvantage of such an arrangement is that
adequate space must be available to allow the girder to project beyond its end support.

4. WEB CRIPPLING

There are many situations where it is not possible to provide transverse web stiffeners at
all points where vertical loads are applied to the girder. For example, a crane gantry girder is
subjected to a vertical loading that travels along the flange; also, girders may be launched during
construction so that the flange actually moves over the fixed point of support. In such cases,
special consideration must be given to the design of the unstiffened web in the local region
underneath, or above, the applied point or "patch" loading to prevent "web crippling". The webs
of all beams must be checked for this possible local failure. Plate girders are particularly
susceptible to this form of failure because of the slenderness of the web plates that are normally
used in their construction.

"Web crippling" is discussed in Section 5.7 of Eurocode 3. It distinguishes between the two
different loading cases that are shown in Figure 5 (taken from Figure 5.7.1 of the Eurocode). In
Figure 5a, the force is applied to one flange only and is therefore resisted by shear forces
developed within the web plate. In this case the web plate has to be checked for its "crushing"
and for its "crippling" resistance. In the other case, shown in Figure 5b, the force is applied to
one flange, transmitted directly by compressive forces developed in the web, and resisted by a
reactive force on the other flange. The web must again be checked for its "crushing" resistance.
The "buckling" resistance of the web must also be considered in this case.

There are, therefore, three types of web resistance that must be calculated. In each case the
resistance is dependent upon the length over which the applied force is effectively distributed on
the flange. This is termed the "stiff bearing length" (s s ). It is calculated on the assumption of a
dispersion of load through solid steel material at a slope of 1:1.

The terms "crushing", "crippling" and "buckling" resistance are introduced to differentiate
between the phenomena being considered. In each case the appropriate resistance is calculated
from empirical formulae:

"Crushing" resistance, where crushing is local yielding of the web without any buckling, is given
by:

Ry.Rd = (ss + sy) tw fyw /M1

"Crippling" resistance, where crippling is localised buckling of the web in the presence of
plasticity, is given by:

Ra.Rd = 0,5 tw2 (E fyw )1/2 [tf /tw )1/2 + 3 (tw /tf ) (ss /d)]/M1

"Buckling" of the web occurs with out-of-plane deformation over most of the depth of the web.

The "buckling" resistance (Rb.Rd ) for the compressive loading situation illustrated in Figure 5b is
determined simply by considering the web plate as a vertical compression member. First it is
necessary to determine the breadth of the web "strut" (beff ) that is effective in resisting the
compression. This breadth may be calculated as:

beff = [h2 + ss2 ]1/2

where:

h is the overall depth of the girder.

ss is the stiff bearing length discussed above.

The buckling resistance of this idealised strut is then determined as for any other compression
member according to Section 5.5.1 of Eurocode 3.
5. LONGITUDINAL WEB STIFFENERS

To increase the strength/weight ratio of plate girders, slender webs may be reinforced by
longitudinal, as well as transverse, stiffeners. A typical longitudinally stiffened girder is shown
after failure in Slide 4. The main function of the longitudinal stiffeners is to increase the buckling
resistance of the web with respect of both shear and bending loads. An effective stiffener will
remain straight, thereby sub-dividing the web panel and limiting the buckling to the smaller sub-
panels. The resulting increase in the ultimate resistance of the girder can be significant.

Slide 4

The design of webs with longitudinal stiffeners is not covered in Part 1 of Eurocode 3. It
will be addressed in Part 2, for bridges. Because of the greater need for high strength/weight
ratios in bridges, girders with longitudinal stiffeners are more commonly encountered in bridge
than in building construction.

Design of longitudinal stiffeners is usually based on a number of empirical design curves

derived from the results of a parametric study employing numerical modelling techniques. The
design procedure is relatively straightforward, although somewhat conservative. Additional
information on the behaviour of longitudinally stiffened girders has been presented [2] which
will assist the designer to gain a better understanding of the structural action.
6. GIRDERS WITH OPENINGS IN SLENDER WEBS

Holes often have to be cut in the webs of plate girders used in building construction to
provide access for service ducts, etc. No mention of such openings is made in Part 1 of Eurocode
3. Such holes have a particular influence on the behaviour of slender webs because the hole
interrupts the tension field. (Design methods are available for stocky webs with web openings -
see Reference3).

Detailed fundamental work by Narayanan [2] has shown that girders with slender webs
and web openings possess a post-buckling reserve of resistance. The collapse mechanism for
such girders, illustrated in Slide 5, is similar to the shear sway mechanism that is characteristic of
all plate girders, as discussed in Lecture 8.4.2. However, some codes adopt a conservative
approach. They do not take account of such post-buckling action in any girder which has a web
opening with any dimension exceeding some percentage of the minimum dimension of the web
panel in which it is located. To provide a simple design procedure, the shear resistance of the
perforated panel is calculated as the buckling resistance.

Slide 5

The disadvantage of such a procedure is that the shear resistance of the perforated panel
will then be substantially lower than that calculated allowing for the full post-buckling reserve of
resistance in adjacent unperforated panels. The designer should therefore reduce the spacing of
the transverse stiffeners either side of the web opening so that the initial buckling resistance of
the resulting narrow perforated panel is approximately equal to the full post-buckling resistance
of adjacent panels.
7. CONCLUDING SUMMARY

 To develop post-buckling action, the elements bounding the web plate of a girder must be
designed to provide adequate anchorage for the tension field forces developed in the web
panel.
 The tension field method can only be applied when transverse web stiffeners are provided such
that the stiffener spacing/web depth ratio lies within the range: 10,0  a/d  3,0. The stiffeners
must have adequate rigidity in the direction perpendicular to the plane of the web to prevent
web buckling. The buckling resistance of the vertical stiffener strut must also be sufficient to
support the tension field forces.
 Careful consideration must be given to the design of the end post. The designer has the option
of either not allowing the development of tension field action in the end panel or, for a more
efficient design, providing an end post of sufficient rigidity and strength.
 The possibility of web crippling, web crushing and web buckling must be considered in those
localised areas where patch loads are applied to the girder flange.
 Longitudinal web stiffeners allow girders of higher strength/weight ratios to be designed. They
are particularly relevant in bridge construction. They are not considered in Eurocode 3: Part 1.1.
 Large web openings are frequently necessary in building construction to allow the passage of
services. Plate girders with openings are not considered in Eurocode 3 Part 1.1.

8. REFERENCES

[1] Eurocode 3 "Design of Steel Structures".

European Prestandard ENV1993-1-1: Part 1.1, General rules and rules for buildings, CEN, 1992.

[2] Narayanan, R (Editor), "Plated Structures; Stability and Strength", Applied Science
Publishers, London, 1983.

This reference gives detailed information, including the experimental background to the
structural action covered in the clauses of Eurocode 3 referred to in this lecture.

[3] Lawson, R. M., Design for Openings in the Webs of Composite Beams, SCI Publication 068.
The Steel Construction Institute, 1987.

9. ADDITIONAL READING

1. "European Recommendations for the Design of Longitudinally Stiffened Webs and of Stiffened
Compression Flanges", Publication 60, ECCS, 1990.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.5.1: Introduction to Design

of Box Girders
OBJECTIVE/SCOPE

To describe the main features and advantages of box girders; to introduce the methods of global
analysis; to describe aspects of behaviour particular to box girders.

PREREQUISITES

Lecture 6.1: Concepts of Stable and Unstable Elastic Equilibrium

Lecture 8.1: Introduction to Plate Behaviour and Design

Lectures 8.4: Plate Girder Behaviour and Design

RELATED LECTURES

Lecture 8.5.2: Advanced Design of Box Girders

SUMMARY

A general overview is given of the form and behaviour of box girders. Typical
configurations are illustrated and the advantages of box girders over plate girders are highlighted.
The structural behaviour of box sections is described; global analysis is discussed; the particular
features of the design of webs and flanges are introduced; the function and form of cross
sectional restraints. including diaphragms, are described.

1. INTRODUCTION

A box girder is formed when two web plates are joined by a common flange at both the
top and the bottom. The closed cell which is formed has a much greater torsional stiffness and
strength than an open section and it is this feature which is the usual reason for choosing a box
girder configuration.

Box girders are rarely used in buildings (box columns are sometimes used but these are
axially loaded rather than in loaded in bending). They may be used in special circumstances,
such as when loads are carried eccentrically to the beam axis.
Steel and composite box girders are used for highway bridges, though they are most expensive
than plate girders (because fabrication requires more time and effort).

The use of box girders for bridges offers the following advantages over plate girders:

 Very good torsional rigidity

 Wide flanges may be used
 Clean external surfaces
 A non-rectangular cross-section can be used.

Torsional rigidity is particularly advantageous when the girder needs to be curved in plan. A
box section can carry the torsion resulting from vertical loading without the need for lateral
bracing.

The use of wide flanges facilitates the choice of shallow construction depth (that is, a large
span-to-depth ratio) which may be desirable when space or headroom is restricted. Shallower
girders will be heavier but this may be a lesser consideration in such circumstances.

Any stiffening needed to the webs and flanges can usually be arranged inside the box section.
The clear external surfaces offer a neater, more pleasing, appearance and avoid the corners and
crevices which are difficult to protect adequately against corrosion.

The use of inclined webs (closer together at the bottom than at the top) reduces the width of
the bottom flange plate (often advantageous to structural performance, particularly when the
flange is in compression). Inclined faces are frequently considered to give a better appearance
than vertical faces. Inclined webs also offer better aerodynamic performance. The cross-sectional
shape of the box girders of many long-span cable-stayed and suspension bridges are chosen after
wind-tunnel testing to find the shape which offers minimum drag and optimum dynamic
response.

Box girder bridges are constructed with single, twin or multiple box girders. The deck of the
bridge may be of reinforced concrete or it may be a stiffened steel deck. When reinforced
concrete is used the steel girders may be closed box sections or may be open sections (U-shaped)
which are closed when the slab is cast across the top. A selection of configurations is shown in
Figures 1 - 3.

The method of erection is often an important factor in choosing a cross-section for a box
girder bridge. The torsional stability of the box section avoids the need for temporary bracing
(such as is needed with plate girders). The bridge can be built by lifting consecutive pre-
assembled units of the complete cross section and joining them end-to-end. This is particularly
suitable for cantilever erection of long spans.
2. MAIN FEATURES OF BOX GIRDERS

Many of the features of box girders illustrated in Figures 1, 2 and 3 are similar to those of
plate girders, though the proportions may be different. There are a few features which are
particular to box girders.
 Web plates carry shear forces and bending moments. Thin webs need transverse
stiffeners to achieve adequate resistance, in the same way as they do in plate girders. Inclined
webs are deeper (in their planes) and may therefore require more stiffening.

Flange plates connecting two webs are wider than those of corresponding I-beam girders.
Consideration must be given to shear lag. When the flange plate is in compression its stability
(against out-of-plane buckling) must be considered; longitudinal and transverse stiffeners are
frequently necessary.

When open steel sections are used, made into composite box girders by a concrete deck
slab, separate (and relatively small) flange plates are provided at the top of each web. These
flanges need to be stabilised laterally by bracing during construction. Shear connection on the
top of the flanges is similar to that on the flanges of I-beam girders.

When closed steel boxes are used, overlaid by a reinforced concrete slab to form a
composite section, shear connection is required over the full width of the top flanges.

When the bridge deck is steel, the top flanges are stiffened orthotropically to carry wheel
loads from traffic as well as acting as a box girder flange. This stiffening usually takes the form
of longitudinal trough stiffeners supported at regular intervals by transverse beams.

At supports plated diaphragms are provided. At each support there are one or two
bearings, located between the webs and directly below the diaphragms. The diaphragms serve to
transfer the load from the webs to the bearings (generally acting as a deep beam) and to prevent
distortion of the section.

In all except smaller boxes restraint against distortion is also required at intermediate
positions; this can be achieved with braced cross-frames, stiff ring frames or plated diaphragms.
 Access inside box sections is necessary during construction and during the life of the
structure. Manholes must be provided in support and intermediate diaphragms, to permit passage
along the length of the box; the size and location of the holes is taken into account in the design
of the diaphragm.

3. GLOBAL ANALYSIS

As for I-beam girders, global analysis determines the bending moments and shears in the
main beams due to the applied loading. Since the principal loads are vertical, greatest attention is
given to moments and shears in the vertical plane, though horizontal loading and effects must
also be considered.

However, when box girders are used, two additional effects must be considered, torsion
and distortion. As we shall see later, consideration of distortional effects may be limited to local
regions between intermediate diaphragms. Torsional effects must be determined by the global
analysis.

For a single straight uniform section girder any simple line beam analysis would suffice
for bending, shear and torsional effects, but in general the most appropriate model for evaluating
the effects due to vertical loading is the grillage analogy. Beam elements in a grillage model
have three degrees of freedom - vertical deflection, rotation about a transverse axis and rotation
about a longitudinal axis - and are thus able to determine directly the three principal effects
which are to be considered in the stress analysis. Computer programs are available specifically
for grillage analysis, though many designers make use of general purpose space-frame programs
(in which the elements have the full 6 degrees of freedom).

In very wide flanges shear lag effects must be taken into account. When the axial load is
fed into a wide flange by shear from the webs the flange distorts in its plane; plane sections do
not remain plane. The resulting stress distribution in the flange is not uniform (see Figure 4).
This effect is not taken into account in a grillage analysis and must be separately determined.
 The calculation of distortional effects, which include transverse bending stresses and
longitudinal warping stresses can be carried out by methods based on beam-on-elastic-
foundation analogy. Where stiff intermediate cross-frames or diaphragms are provided the
stresses are usually quite small.

In very complex configurations finite element analysis might be used. Shell elements are
connected to model the complete cross-section along the whole length of the girder. Provided
that the mesh is sufficiently fine and the element behaviour is appropriate, effects such as
warping, distortion and shear lag are determined at the same time as the principal bending, shear
and torsion effects.

Folded plate analysis is sometimes used, but it is only appropriate when the box section is
uniform along its length, when there are no intermediate cross-frames and when the loading can
be represented by harmonic series. It is difficult to apply to ordinary design situations.

4. TORSION AND DISTORTION

The general case of an eccentric load applied to a box girder is in effect a combination of
three components - bending, torsion and distortion. As a first step, the force can be separated into
two components, a pair of symmetric vertical loads and a force couple, as shown in Figure 5(a).
However, torsion is in fact resisted in a box section by a shear flow around the whole perimeter
and the couple should in turn be separated into two parts, representing pure torsion and
distortion, as shown in Figure 5(b).
 The first two components, vertical bending loads and a torsional shear flow, are
externally applied forces, and they must be resisted in turn at the supports or bearings. The third
component, distortional forces, comprises an internal set of forces, statically in equilibrium,
which do not give rise to any external reaction. Distortional effects depend on the behaviour of
the structure between the point of application and the nearest positions where the box section is
restrained against distortion.

4.1 Torsion and Torsional Warping

The theoretical behaviour of a thin-walled box section subject to pure torsion is well
known and is treated in many standard texts. For a single cell box, the torque is resisted by a
shear flow which acts around the walls of the box. This shear flow (force/unit length) is constant
around the box and is given by q = T/2A, where T is the torque and A is the area enclosed by the
box. (In Figure 2 the torque is QB/2 and the shear flow is Q/4D.) The shear flow produces shear
stresses and strains in the walls and gives rise to a twist per unit length,  , which is given by the
general expression:

or,

where J is the torsion constant.

 However, it is less well appreciated that this pure torsion of a thin walled section will also
produce a warping of the cross-section, unless there is sufficient symmetry in the section. This is
illustrated in Figure 6 for a rectangular section that is free to warp at its ends. However, in
practice boxes are not subject to pure torsion; wherever there is a change of torque (at a point of
application of load or at a torsional restraint) there is restraint to warping, because the 'free'
warping displacements due to the different torques would be different (restraint is high, for
example. over intermediate supports where torsion is restrained). Such restraint gives rise to
longitudinal warping stresses and associated shear stresses in each wall of the box.

4.2 Distortion

When torsion is applied directly around the perimeter of a box section, by forces exactly
equal to the shear flow in each of the sides of the box, there is no tendency for the cross section
to change its shape. Torsion can be applied in this manner if, at the position where the force
couple is applied, a diaphragm or stiff frame is provided to ensure that the section remains square
and that torque is in fact fed into the box walls as a shear flow around the perimeter.

Provision of such diaphragms or frames is practical, and indeed necessary, at supports

and at positions where heavy point loads are introduced. But such restraint can only be provided
at discrete positions. When the load is distributed along the beam, or when point loads can occur
anywhere along the beam such as concentrated axle loads from vehicles, the distortional effects
must be carried by other means.
 The distortional forces shown in Figure 5(b) are tending to increase the length of one
diagonal and shorten the other. This tendency is resisted in two ways, by in-plane bending of
each of the wall of the box and by out-of-plane bending. this is illustrated in Figure 7.

In general the distortional behaviour depends on interaction between the two sorts of
bending. The behaviour has been demonstrated to be analogous to that of a beam on an elastic
foundation (BEF), and this analogy is frequently used to evaluate the distortional effects.

If the only resistance to transverse distortional bending is provided by out-of-plane

bending of the flange plates there were no intermediate restraints to distortion, the distortional
deflections in most situations would be significant and would affect the global behaviour. For
this reason it is usual to provide intermediate cross-frames or diaphragms; consideration of
distortional displacements and stresses can then be limited to the lengths between cross-frames.
5. FLANGE DESIGN
5.1 Tension Flanges

Tension flanges are designed mainly on the basis of longitudinal bending stresses, in the
same way as for plate girders. Torsional and distortional effects and the effects of shear lag do
need to be taken into account in some circumstances. The strength is taken as the yield stress.

At the Serviceability Limit State elastic behaviour is normally specified. Then the
stresses due to the restraint of torsional warping, distortional warping stresses and the variation
of stress across the flange due to shear lag must be calculated. Stresses are highest adjacent to the
web.

At the Ultimate Limit State, plastic behaviour is normally accepted. Then the stresses due
to the restraint of torsional warping and shear lag can be neglected, since they are secondary
effects.

5.2 Compression Flanges

In addition to considering the load effects in relation to yield strength, the stability of the
compression flange must also be considered.

Relatively narrow flanges may be unstiffened. The strength of the flange plate then
depends on ordinary panel-buckling resistance. It is convenient to express this in design by the
determination of an effective width of compression flange; this is the width which has the same
resistance, at yield stress, as the buckling resistance of the full panel. Typically a panel is
considered fully effective up to a width of 24t (where t is the flange thickness), but thereafter the
effective width is less than the actual width.

Wider flanges are provided with longitudinal stiffeners to provide stability and these in
turn are supported at intervals by transverse stiffeners, cross-frames, or diaphragms.

Usually the longitudinal stiffeners can be designed using rules which effectively treat
them as struts. For this purpose the transverse members restraining them must be sufficiently
stiff. If, when the flanges are particularly wide, the transverse stiffeners are not sufficiently stiff,
the flange would have to be treated as a panel stiffened in two directions and the overall buckling
strength determined; this is too complex for most design purposes.

5.3 Orthotropic Steel Decks

In some bridges, notably long-span bridges and movable bridges, where minimum weight
is desirable, the roadway is carried on an all-steel deck. This form of deck is stiffened
longitudinally; transverse members or diaphragms support the longitudinal stiffeners.

The design of such a deck, to carry the direct loads from the traffic wheels, is an
extremely complex matter. Connection details are subject to onerous fatigue loading.
Configurations currently in use are the result of many years development, analysis, testing and
feedback from previous designs.

6. WEB DESIGN

The determination of the strength of webs in bending and shear follows the same general
rules as for plate girders.

Shear buckling resistance of thin webs is improved by the presence of intermediate

stiffeners. Tension field action can develop in the web in the same way as in plate girders.
However, the further increase in tension field action on account of the bending stiffness of the
flange plate is not normally achievable.

7. CROSS SECTIONAL RESTRAINTS

7.1 General Function and Description

The main functions of cross sectional restraints are:

 to preserve the shape of the box against distortion

 to transfer an externally applied torque to the box walls through shear flow
 to provide transverse support to longitudinal stiffeners under compression
 to support traffic loads directly (from an orthotropic deck)
 to transmit forces from box walls to the supports.

7.2 Support Diaphragms

At support positions a plate diaphragm is normally provided, see Figure 8, although occasionally
a heavily braced frame can be used. Plate diaphragms may be thick and unstiffened in fairly
small boxes; in larger boxes vertical stiffeners are provided over the bearings and sometimes
horizontal stiffening is also provided.
The main purpose of support diaphragms is to transfer the large shear forces from the webs to the
bearings. They must also transmit horizontal forces when the bearing is guided or fixed in
position into shear flow along the web of the box section.

For large box sections, with large support forces, the use of a finite element program is
recommended for the design of support diaphragms.

7.3 Intermediate restraints

For large deep box girders, intermediate restraint against distortion is usually provided by cross
frames, see Figure 9. The frame comprises a ring of four members, each being an effective
section comprising a width of flange or web and the transverse stiffener attached to it. The
corners of the ring need to be designed to provide moment continuity. Diagonal bracing is
connected to the frame as necessary, usually either across the section diagonal or as a V to the
midpoint of a flange stiffener.

7.4 Load-carrying transverse stiffeners

When a box girder has a stiffened steel deck, or a composite box girder is arranged with cross-
girders, such that the deck slab designed to span longitudinally, the transverse stiffeners must
also carry the traffic loads and transfer them to the webs. Such configurations usually also have
cantilevers on the outer faces. Moment continuity must be provided between the transverse
stiffeners, the cantilevers and the cross-girders.
Transverse flange stiffeners which are required to provide support must usually be arranged to
coincide with transverse web stiffeners, so that the load can be transferred into the web.

8. ARTICULATION

The arrangement of the bearings which support the girder, known as the articulation
arrangement, should take account of the high degree of torsional rigidity provided by the box
girders.

For a relatively short bridge of a few spans it is not necessary to provide two bearings at
each intermediate support. Twin bearings at the ends will restrain the box against twist; single
bearings on the box centreline are sufficient at intermediate supports.

If the bridge is highly curved, single bearings may be sufficient at all supports; restraint
against twist is provided by the combined effects of torsional rigidity and geometrical
arrangement of the group of bearings.

9. CONCLUDING SUMMARY

 Box girders are used because of their good resistance to torsion

 box girders can be designed to have a good aerodynamic shape
 the clear external surfaces and the use of inclined webs gives a good appearance
 the grillage method is sufficiently accurate for global analysis
 shear lag and the stability of the compression flange both require consideration in wide flanges
 design of webs is generally similar to that for plate girders
 distortional effects must be considered

10. ADDITIONAL READING

1. Eurocode 3: Design of steel structures: European Prestandard, ENV1993-1-1: Part 1.1: General
rules and rules for buildings, CEN, 1992.
2. Dubas, P. and Gehri, E., Behaviour and Design of Steel Plated Structures, Technical Committee 8
Group 8.3, ECCS-CECM-EKS No44, 1986.
3. Johnson, R. P. and Buckby, R. J., Composite Structures of Steel and Concrete, Volume 2: Bridges,
Collins, London, 1986.
4. British Standard 5400: Part 3: Steel, Concrete and Composite Bridges, Part 3: Code of Practice
for Design of Steel Bridges, British Standards Institution, 1982.
5. Horne, M.R., CIRIA Guide 3, Structural action in steel box girders, Construction Industry Research
and Information Association, London, 1977
6. Kollbrunner, C. F. and Basler, K.: Torsion, Spes/Bordas Lausanne/Paris, 1955.
7. Stahlbau Handbuch: Stahlbau Handbuck für Studium und Praxis, BandI, Stahbau Verlag, Köln,
1982.
8. Dalton, D. C. and Richmond, B., Twisting of Thin Walled Box Girders, Proceedings of the
Institution of Civil Engineers, January 1968.
9. Iles, D.C., Design Guide for Composite Box Girder Bridges, The Steel Construction Institute,
Ascot, 1994
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.5.2: Advanced Design

of Box Girders
OBJECTIVE/SCOPE

To introduce methods of global analysis, methods of determining cross-section distortion, and

shear lag in box girder bridges.

PREREQUISITES

None.

RELATED LECTURES

Lecture 8.5.1: Introduction to Design of Box Girders

SUMMARY

Global analysis may be made by the grillage, orthotropic plate, folded plate and finite
element methods.

Distortion of the box may have to be controlled by diaphragms or cross frames. Simple or
refined methods are available for the calculation of the forces in the diaphragms or cross frames.

In very wide flanges, shear lag effects have to be taken into account.

1. INTRODUCTION

Although steel or steel-concrete composite box girders are usually more expensive per
tonne than plate girders, because they require more fabrication time, they can lead to a more
economic solution overall.

For bridges, box girders have several advantages over plate girders which make their use
attractive:

 a very high torsional stiffness. (In closed box sections torque is resisted mainly by Saint Venant
shear stresses and the torsional stiffness is normally much greater than that of open sections.)
 closed steel boxes provide torsional stiffness during their erection. (They thus avoid the need for
the expensive temporary bracing which is required with plate girders and which also interferes
 with the construction of the concrete slab. For highly curved spans torsional stiffness is almost
always essential during their construction.)
 wide flanges can be used. (This allows large span to depth ratios without resource to very thick
material.)
 box girders have a neat appearance. (The stiffening can remain unseen inside the box.)
 the facility to choose a good aerodynamic shape. (This is particularly important for large
suspension or cable-stayed bridges.)
 box girders use fewer bearings. (Usually torsional restraint need be provided at only one
position along a continuous box; single bearings can be used at all other supports. Further, with
a highly curved box girder, single bearings at all supports is often sufficient.)

Box girders are sometimes used in building structures, but this is not common. This lecture
deals mainly with box girders as used in bridges, both all-steel construction and composite
construction with a reinforced concrete deck slab; most of the general remarks are applicable to
box girders used in buildings

2. GLOBAL ANALYSIS METHODS

Global analysis determines the load effects, bending moments, shear forces, torsional
moments, etc., which occur in all parts of the structure as a result of the applied load. From this
analysis stresses are determined, for comparison with the calculated strengths.

Methods of analysis for composite bridge decks fall into one of three groups.

 those that treat the bridge as a series of interconnected beams

 those that treat separately the various parts of the box section (flanges, webs, diaphragms)
 those that treat the bridge deck as a continuum

Those in the first group are the simplest to analyse, since beam theory can be used for the
behaviour of the individual elements. For a single straight girder a line-beam analysis can be
used, provided this takes account of torsional effects as well as bending effects, but in general a
grillage model is needed. Such an analysis gives good results for the distribution of moments and
forces in multiple girder structures and when a curved single beam is modelled as a series of
straight elements. However, simple beam theory does not take account of the distortion of the
cross section or of shear lag effects and these must be determined separately.

Analysis in the second group is by use of finite element techniques and inevitably involves
the use of a powerful computer program. Provided suitable elements are available within the
computer program, the analysis is able to give results which include most of the structural
effects, including distortion and shear lag, but choice of element type and size requires much
experience, and interpretation of the results also requires careful consideration.

The third group applies more exact theoretical modelling techniques. Examples are treatment
of the whole deck as an orthotropic plate and analysis of folded plate models. However such
techniques can only be properly applied when there is uniformity throughout the structure and
for distributed loading. They are also only able to represent separately some aspects of the
behaviour: the loading therefore needs to be divided into components such as uniform bending,
uniform torsional, warping torsion, and distortion.

3. GRILLAGE
3.1 General

In a grillage analysis, the structure is idealised as a number of longitudinal and transverse

beam elements in a single horizontal plane, rigidly interconnected at nodes. Transverse beams
may be orthogonal or skewed with respect to the longitudinal beams, so that skew, curved,
tapering or irregular decks can be analysed.

In a simple grillage analysis, each beam is allotted a flexural stiffness in the vertical plane
and a torsional stiffness. Vertical loads are applied only at the nodes. Computer software is used
to carry out a matrix stiffness analysis to determine the displacements (rotations about the two
horizontal axes and the vertical displacement) at each node and the forces (bending moments,
torsional moments and vertical shear forces) in the beams connected to each node.

Grillage analysis does not determine warping and distortional effects, nor the effects of
shear lag. Local effects under point loads (wheel loads) can only be studied with a grillage by the
use of a fine mesh of beams locally to the load; local effects are usually determined separately
and added to global results as required.

3.2 Grillage Modelling for Box Girder Bridges

The global structural action of a box girder bridge can be seen as the essentially separate
actions of a reinforced concrete slab (or an ortotropically stiffened steel deck) which bends
transversely and a series of longitudinal beams which deflect vertically and twist. The slab (or
steel deck) bends as a result of being supported along several lines which deflect by different
amounts and in a manner which varies along the span. The global analysis therefore needs to
model accurately the way in which these support lines deflect, so that the interaction between
longitudinal and transverse bending is properly established.

The slab is effectively supported along each web line. The vertical deflection of each web
line depends on a combination of the vertical and torsional deflections of the box girder of which
it is a part. The best way to model these effects is to create a torsionally stiff beam element along
the centreline of each box (i.e. the shear centre) and to connect it to the slab at the web positions.
To do this, short 'dummy' transverse beams are needed; they do not physically represent any
particular part of the structure and the forces in them do not need to be analyzed, but they must
be given sufficient stiffness that their bending is insignificantly small. This form of model for a
twin-box bridge with cantilevers is illustrated in Figure 1 (note that, for clarity, the dummy
beams and longitudinal beams are shown slightly below the slab, whilst they would actually be
treated in the analysis as co-planar).
3.3 Longitudinal Grillage Elements

The main longitudinal beams are assigned the flexural properties of the full section of
each girder (including the slab or deck). In multi-girder structures it is usual to consider the slab
to be divided midway between boxes and for the full width of the cantilever to be included with
the outer box. Strictly this is not exact, since it would introduce a discontinuity in the level of the
neutral axis, but the inaccuracy is negligible.
 The longitudinal elements representing the slab (shown dotted in Figure 1) are not strictly
necessary, as they are much more flexible than the main girders, though they may be helpful in
the application of distributed loads. They are shown here to illustrate the division of the slab.

The longitudinal edge elements may be added to represent the edge beam. They do not
have a major effect on overall performance but are often helpful in the application of load on the
cantilevers.

3.4 Transverse Grillage Elements

Where there are no transverse beams, the transverse elements simply represent a width of
slab equal to the node spacing. Where there are transverse beams, including cross-beams and
diaphragms inside the box, the elements should represent the stiffness of the effective transverse
member.

The slab elements are supported only on the dummy elements, they are not connected
directly to the longitudinal beams. There is no moment continuity between slab elements and the
dummy beams.

3.5 Torsional Rigidities

For an open box section, the torsional stiffness K is given by the general expression:

Where A is the area of the box and t is the thickness of element ds.

When the section is composite, the concrete slab should be transformed into an equivalent
thickness of steel by dividing by the modular ratio.

For a strip of solid slab the torsional stiffness is given by:

Where t is the thickness and b is the width of the strip.

However, in the grillage model only half this stiffness should be assigned to the
transverse elements, since no Saint Venant shear stress flux goes around the perimeter of the
strip's cross-section. Similarly, for an orthotropic steel deck, only the value H should be used for
the torsional rigidity, not 2H.
3.6 Skew Bridges

Skewed arrangements of multiple boxes can be devised, provided that support

diaphragms can remain essentially square to the box centrelines and that there are either no
cross-girders between boxes or the cross-girders are square to the boxes. Similarly, grillage
analysis with skew cross-members is difficult to interpret and gives uncertain results for all
except small skews.

3.7 Interpretation of the Output of a Grillage Analysis

Computer software usually gives values of the vertical shear, bending moments and
torsional moment for each grillage member at each joint in the grillage. Because the continuous
structure has been idealised into discrete elements this discontinuity is unreal. A slightly 'better'
value of moments in the main longitudinal members can be obtained by smoothing, as shown in
Figure 2, though the difference is usually very small.

4. ORTHOTROPIC PLATE ANALYSIS

In orthotropic plate analysis, the deck structure is 'smoothed' across its length and breadth
and treated as a continuum.
 The elastic properties of an orthotropic plate are defined by the two flexural rigidities D x
and Dy and a plate torsional rigidity H. The governing equation relating deflection w to load P
acting normal to the plane of the plate is:

= p(x, y)

Design charts for decks that can be idealised as orthotropic plates have been derived from
series solutions. They give deflections and longitudinal and transverse moments due to a point
load, and so provide a rapid method for distribution analysis. Their applicability is limited to
simply supported decks of skew not exceeding 20 whose elastic properties can be represented
solely by length, breadth, and the three quantities Dx, Dy and H.

In composite structures, they can be used for beam-and-slab decks with not less than five equally
spaced longitudinal members of uniform diaphragms over the supports.

5. FINITE ELEMENT ANALYSIS

The finite element method is used increasingly in civil engineering. It is the most
versatile of the matrix stiffness methods of elastic analysis, and can, in principle, approach the
solution of almost any problem of global analysis of a bridge deck.

In box girders, the finite element method allows the study of shear lag and the
computation of effective flange breadths. It can also analyse local effects in slabs. To do this the
webs, flanges and diaphragms are each divided into a suitable mesh of elements; the detail of the
effects which can be revealed (for example the variation in stress across a flange due to shear
lag) depends on the fineness of the mesh and the capabilities of the element types provided by
the program.

The disadvantage of finite element analysis is its cost, especially because of the high
level of expert time required for the idealisation of the structure. The expert's know-how is
needed in selecting an appropriate pattern of elements, in selecting the right type of element and
in determining the right limit conditions for boundary nodes along the supports. The
interpretation of results also requires experience. The choice of inappropriate elements can be
misleading in regions of steep stress gradient, because the conditions of static equilibrium are not
then necessarily satisfied. The selection of the discretisation density level, or of the material
behaviour, may have serious repercussions on the accuracy of the results.

Nevertheless, for complex situations, or for complex portions of a major structure, there
is no better substitute for a finite element analysis.

6. FOLDED PLATE ANALYSIS

The folded plate method is normally limited to assemblages of rectangular plates. It is not
applicable to skew decks due to coupling between the harmonics. The orthotropic plates may
extend over several spans but must be simply-supported at the extreme ends, with rigid
diaphragms over the end supports. When folded plate diaphragms are used to represent the
transverse frames, the advantages are that it can give a complete and accurate solution in less
computer time than is needed for the finite element method, and it can accept a wide variety of
types of loading and both displacement and force boundary conditions.

To apply the method to a double cellular box-girder bridge with one single internal web,
the distortion must be divided into symmetric and asymmetric deformations. For boxes with
more internal webs, it is possible to divide the deformations of the cross-section into eigenvalue
functions of deformation.

7. TORSIONAL WARPING

Pure torsion of a thin walled section will also produce a warping of the cross-section,
unless there is sufficient symmetry in the section. To illustrate how warping can occur, consider
what would happen to the four panels of a rectangular box section subject to torsion.

Assume that the box width and depth are B and D respectively, and that the flange and
web thicknesses are tf and tw. Under a torque T, the shear flow is given by q=T/2BD.

Consider first the flanges. The shear stress in the flanges is given by f = q/tf =T/2BDtf.
Viewing the box from above, each flange is sheared into a parallelogram, with a shear angle 
=f/G; if the end sections were to remain plane, the relative horizontal displacement between top
and bottom corners would be L at each end (see Figure 3a), and thus there would be a twist
between the two ends of 2L/D = 2f L/DG = TL/BD2Gtf.
 By a similar argument, viewing the box from the side and considering the shear
displacements of the webs, if the end sections were to remain plane the twist of the section would
be TL/B2DGtw. As the twist must be the same irrespective of whether we consider the flanges or
the webs, it is clear that the end sections can only remain plane if TL/BD2Gtf = TL/B2DGtw, i.e.
Dtf = Btw. If this condition is not met, the end sections cannot remain plane; instead there will be
a slight counter-rotation in their planes of the two flanges and of the two webs, and a consequent
warping of the section. Typical warping for this example is shown in Figure 3b.

Of course, for a simple uniform box section subject to pure torsion this warping is
unrestrained and does not give rise to any secondary stresses. But if, for example, a box is
supported and torsionally restrained at both ends and then subjected to applied torque in the
middle, warping is fully restrained in the middle by virtue of symmetry and torsional warping
stresses are generated. Similar restraint occurs in continuous box sections which are torsionally
restrained at intermediate supports.

This restraint of warping gives rise to longitudinal warping stresses and associated shear
stresses in the same manner as bending effects in each wall of the box. The shear stresses
effectively modify slightly the uniformity of the shear stress calculated by pure torsion theory,
usually reducing the stress near corners and increasing it in mid-panel. Because maximum
combined effects usually occur at the corners, it is conservative to ignore the warping shear
stresses and use the simple uniform distribution. The longitudinal effects are, on the other hand
greatest at the corners. They need to be taken into account when considering the occurrence of
yield stresses in service and the stress range under fatigue loading. But since the longitudinal
stresses do not actually participate in the carrying of the torsion, the occurrence of yield at the
corners and the consequent relief of some or all of these warping stresses would not reduce the
torsional resistance. In simple terms, a little plastic redistribution can be accepted at the ultimate
limit state (ULS) and therefore there is no need to include torsional warping stresses in the ULS
checks.

8. CROSS-SECTION DISTORTION

When torsion is applied directly around the perimeter of a box section, by forces exactly
equal to the shear flow in each of the sides of the box, there is no tendency for the cross section
to change its shape.

If torsion is not applied in this manner, there is effectively a set of forces which is trying
to extend the length of one diagonal across the section and reduce the other (see Figure 4).
Diaphragms or frames can be provided to restrain distortion where large distortional forces
occur, such as at support positions, and at intervals along a box, but in general the distortional
effects must be carried by other means.

To illustrate how distortion occurs and is carried between effective restraints, consider a
simply supported box with diaphragms only at the supports and which is subject to a point load
over one web at midspan. Under the distortional forces, each side of the box bends in its own
plane and, provided there is moment continuity around the corners, out of its plane as well. The
deflected shape is shown in Figure 5.
 The in-plane bending of each side gives rise to longitudinal stresses and strains which,
because they are in the opposite sense in the opposing faces of the box, produce a warping of the
cross section (in the example shown the end diaphragms warp out of their planes, whilst the
central plane can be seen to be restrained against warping by symmetry). The longitudinal
stresses are therefore known as distortional warping stresses. The associated shear stresses are
known simply as distortional shear stresses.

The bending of the walls of a box, as a result of the distortional forces, produces
transverse distortional bending stresses in the box section.

The introduction of stiff intermediate cross-frames will restrict distortional effects to the
lengths between frames (rather than between supports). but they must be stiff enough for this
purpose.

In general the distortional behaviour depends on interaction between the two sorts of
behaviour, the warping and the transverse distortional bending. The behaviour has been
demonstrated to be analogous to that of a beam on an elastic foundation (BEF), with the beam
stiffness representing the warping resistance and the elastic foundation representing the
transverse distortional bending resistance. A comprehensive description of the analogy is given
in a paper by Wright [1].

A diagrammatic illustration of the distortional behaviour of a box with a single

intermediate diaphragm is given in Figure 6.
9. SHEAR LAG

When the axial load is fed into a wide flange by shear from the webs the flange distorts in its
plane; plane sections do not remain plane. The resulting stress distribution in the flange is not
uniform In very wide flanges, shear lag effects have to be taken into account for the verification
of stresses, especially for short spans, since it causes the longitudinal stress at a flange/web
intersection to exceed the mean stress in the flange.

Shear lag can be allowed for in the elementary theory of bending, by using an effective flange
breadth (less than the real breadth) such that the stress in the effective breadth equals the peak
stress in the actual flange (see Figure 7). This effective flange breadth depends on the ratio of
width to span.

For a simply supported beam, for example, the effective breadth of the portion between the webs
is e.b , where e , the effective breadth ratio, is given in Table 1.

b/L Mid-span Quarter-span Support

  =0 =1 =0 =1 =0 =1

0,00 1,00 1,00 1,00 1,00 1,00 1,00

 0,05 0,98 0,97 0,98 0,96 0,84 0,77

0,10 0,95 0,89 0,93 0,86 0,70 0,60

0,20 0,81 0,67 0,77 0,62 0,52 0,38

0,30 0,66 0,47 0,61 0,44 0,40 0,28

0,40 0,50 0,35 0,46 0,32 0,32 0,22

0,50 0,38 0,28 0,36 0,25 0,27 0,18

0,75 0,22 0,17 0,20 0,16 0,17 0,12

1,00 0,16 0,12 0,15 0,11 0,12 0,09

Table 1: Effective breadth ratio e for simply supported beams

where

b is the distance between webs.

L is the span of the beam

=

e is the elastic effective breadth ratio.

Fortunately, in most situations the span/breadth ratio is not sufficiently large to cause more than
10-20% increase in peak stress, on account of shear lag.

10. DIAPHRAGMS

At supports, forces are transferred from the box girder, through bearings, to the
substructure below. Principally, these forces are vertical, though lateral restraint also has to be
provided at certain selected positions. Where there is only a single bearing under the box and it
offers little resistance to transverse rotation (e.g. elastomeric pot bearings), there will be no
torsional restraint; the loads transferred from the two webs will be equal (presuming that the
bearing is on the centreline). When there are two bearings, under or close to each of the webs,
torsional restraint is provided to the box; the load from each web will be different, and there will
be a transfer of torsional shear from the flanges. Whenever there is lateral restraint there will be
an associated torque, because the restraint will not be at the level of the shear centre of the box.
 The principal function of a support diaphragm is to provide an adequate load path to
transfer shear forces from the webs to the bearings below the box. In doing so it also resists
distortional forces.

Plated diaphragms are normally provided at supports, since they provide these functions
most easily, although, strictly, an adequately braced cross-frame could also do so.

Clearly, full diaphragms close the box section, yet access into the box is necessary for
completion of fabrication and for future inspection and maintenance. Openings are usually
provided to permit access along the box, but the effect of these openings on the performance of
the diaphragm has to be carefully considered; the size and position of any opening needs to be
limited. This can be a particular problem with small boxes, because the minimum hole size may
be a large proportion of the diaphragm size.

Diaphragms are usually provided with vertical stiffeners above the bearings because of
the large forces involved, though with small boxes a thick unstiffened diaphragm may on
occasion be appropriate.

A diaphragm behaves essentially as a deep beam, with the diaphragm plate acting as its web and
an effective width of each of the box flanges acting as its top and bottom flange.

An example of an intermediate diaphragm in a large box girder of a cable stayed bridge is shown
in Figure 8.

11. CONCLUDING SUMMARY

 Grillage analysis is most often used for grillage analysis. It allows a simple idealisation of the
structure, and a sure interpretation of the output.
 Finite element analysis can be used for complex situations. It is the most versatile of the matrix
stiffness methods of elastic analysis.
 Orthotropic plate analysis and folded plate analysis have a limited application.
 Eccentric loading of the girder section causes distortion which may have to be controlled by the
provision of intermediate diaphragms or cross frames.
 In very wide flanges shear lag effects have to be taken into account.
12. REFERENCE

1. Wright, R N, Abdel-Samad, S R and Robinson, A R, BEF Analogy for analysis of box girder bridges,
Proc. ASCE, vol 94, ST7, 1968.

13. ADDITIONAL READING

1. Eurocode 3: "Design of Steel Structures", ENV1993-1-1: Part 1.1, General rules and rules for
buildings, CEN, 1992.
2. Dubas, P. and Gehri, E., Behaviour and Design of Steel Plated Structures, Technical Committee 8
Group 8.3, ECCS-CECM-EKS, No44, 1986.
3. Johnson, R. P. and Buckby, R. J., Composite Structures of Steel and Concrete, Volume 2: Bridges,
Collins London, 1986.
4. British Standard 5400: Part 3: Steel, Concrete and Composite Bridges, Part 3: Code of Practice
for Design of Steel Bridges, British Standards Institution, 1982.
5. Horne, M.R., CIRIA Guide 3, Structural action in steel box girders, Construction Industry Research
and Information Association, London, 1977
6. Kollbrunner, C. F. and Basler, K., Torsion in Structures - An Engineering Approach (translated
from the German), Springer Verlag, Berlin 1969.
7. Stahlbau Handbuch: Stahlbau Handbuck für Studium und Praxis, BandI, Stahbau Verlag, Köln,
1982.
8. Dalton, D. C. and Richmond, B., Twisting of Thin Walled Box Girders, Proceedings of the
Institution of Civil Engineers, January, 1968.
9. Iles, D.C., Design Guide for Composite Box Girder Bridges, The Steel Construction Institute,
Ascot, 1994
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.6: Introduction to Shell Structures

OBJECTIVE/SCOPE

To describe in a qualitative way the main characteristics of shell structures and to discuss briefly
the typical problems, such as buckling, that are associated with them.

PREREQUISITES

None.

RELATED LECTURES

Lecture 6.1: Concepts of Stable and Unstable Elastic Equilibrium

Lecture 8.1: Introduction to Plate Behaviour and Design

Lecture 8.4.1: Plate Girder Behaviour and Design I

Lecture 8.5.1: Design of Box Girders

SUMMARY

Shell structures are very attractive light weight structures which are especially suited to
building as well as industrial applications. The lecture presents a qualitative interpretation of
their main advantages; it also discusses the difficulties frequently encountered with such
structures, including their unusual buckling behaviour, and briefly outlines the practical design
approach taken by the codes.

1. INTRODUCTION

The shell structure is typically found in nature as well as in classical architecture [1]. Its
efficiency is based on its curvature (single or double), which allows a multiplicity of alternative
stress paths and gives the optimum form for transmission of many different load types. Various
different types of steel shell structures have been used for industrial purposes; singly curved
shells, for example, can be found in oil storage tanks, the central part of some pressure vessels, in
storage structures such as silos, in industrial chimneys and even in small structures like lighting
columns (Figures 1a to 1e). The single curvature allows a very simple construction process and is
very efficient in resisting certain types of loads. In some cases, it is better to take advantage of
double curvature. Double curved shells are used to build spherical gas reservoirs, roofs, vehicles,
water towers and even hanging roofs (Figures 1f to 1i). An important part of the design is the
load transmission to the foundations. It must be remembered that shells are very efficient in
resisting distributed loads but are prone to difficulties with concentrated loads. Thus, in general,
a continuous support is preferred. If it is not possible to have a foundation bed, as shown in
Figure 1a, an intermediate structure such as a continuous ring (Figure 1f) can be used to
distribute the concentrated loads at the vertical supports. On occasions, architectural reasons or
practical considerations impose the use of discrete supports.
 As mentioned above, distributed loads due to internal pressure in storage tanks, pressure
vessels or silos (Figures 2a to 2c), or to external pressure from wind, marine currents and
hydrostatic pressures (Figures 2d and 2e) are very well resisted by the in-plane behaviour of
shells. On the other hand, concentrated loads introduce significant local bending stresses which
have to be carefully considered in design. Such loads can be due to vessel supports or in some
cases, due to abnormal impact loads (Figure 2f). In containment buildings of nuclear power
plants, for example, codes of practice usually require the possibility of missile impact or even
sometimes airplane crashes to be considered in the design. In these cases, the dynamic nature of
the load increases the danger of concentrated effects. An everyday example of the difference
between distributed and discrete loads is the manner in which a cooked egg is supported in the
egg cup without problems and the way the shell is broken by the sudden impact of the spoon
(Figure 2g). Needless to say, in a real problem both types of loads will have to be dealt with
either in separate or combined states, with the conceptual differences in behaviour ever present
in the designer's mind.
 Shell structures often need to be strengthened in certain problem areas by local
reinforcement. A possible location where reinforcement might be required is at the transition
from one basic surface to another; for instance, the connections between the spherical ends in
Figure 1b and the main cylindrical vessel; or the change from the cylinder to the cone of
discharge in the silo in Figure 1c. In these cases, there is a discontinuity in the direction of the in-
plane forces (Figure 3a) that usually needs some kind of reinforcement ring to reduce the
concentrated bending moments that occur in that area.
 Containment structures also need perforations to allow the stored product (oil, cement,
grain, etc.) to be put in, or extracted from, the deposit (Figure 3b). The same problem is found in
lighting columns (Figure 3c), where it is general practice to put an opening in the lower part of
the post in order to facilitate access to the electrical works. In these cases, special reinforcement
has to be added to avoid local buckling and to minimise disturbance to the general distribution of
stresses.

Local reinforcement is also often required at connections between shell structures, such
as commonly occur in general piping work and in the offshore industry. In these cases additional
reinforcing plates are used (Figure 3d), which help to resist the high stresses produced at the
connections.

In contrast to local reinforcement, global reinforcement is generally used to improve the

overall shell behaviour. Because of the efficient way in which these structures carry load, it is
possible to reduce the wall thickness to relatively small values; the high value of the shell
diameter to thickness ratios can, therefore, increase the possibility of unstable configurations. To
improve the buckling resistance, the shell is usually reinforced with a set of stiffening members.

In axisymmetric shells, the obvious location for the stiffeners is along selected meridians
and parallel lines, creating in this way a true mesh which reinforces the pure shell structure
(Figure 4a). On other occasions, the longitudinal and ring stiffeners are replaced by a
complicated lattice (Figure 4b), which gives an aesthetically pleasing structure as well as
mechanical improvements to the global shell behaviour.
2. POSSIBLE FORMS OF BEHAVIOUR

There are two main mechanisms by which a shell can support loads. On the one hand, the
structure can react with only in-plane forces, in which case it is said to act as a membrane. This
is a desirable situation, especially if the stress is tensile (Figure 5a), because the material can be
used to its full strength. In practice, however, real structures have local areas where equilibrium
or compatibility of displacements and deformations is not possible without introducing bending.
Figure 5b, for instance, shows a load acting perpendicular to the shell which cannot be resisted
by in-plane forces only, and which requires bending moments, induced by transverse deflections,
to be set up for equilibrium. Figure 5c, however, shows that membrane forces only can be used
to support a concentrated load if a corner is introduced in the shell.

It is worthwhile also to distinguish between global and local behaviour, because

sometimes the shell can be considered to act globally as a member. An obvious example is
shown in Figure 6a, where a tubular lighting column is loaded by wind and self-weight. The
length AB is subjected to axial and shear forces, as well as to bending and torsion, and the global
behaviour can be approximated very accurately using the member model. The same applies in
Figure 6b where an offshore jacket, under various loading conditions, can be modelled as a
cantilever truss. In addition, for certain types of vault roofs where the support is acting at the
ends, the behaviour under vertical loads is similar to that of a beam.
 Local behaviour, however, is often critical in determining structural adequacy. Dimpling
in domes (Figure 7a), or the development of the so-called Yoshimura patterns (Figure 7b) in
compressed cylinders, are phenomena related to local buckling that introduce a new level of
complexity into the study of shells. Non- linear behaviour, both from large displacements and
from plastic material behaviour, has to be taken into account. Some extensions of the yield line
theory can be used to analyse different possible modes of failure.
 To draw a comparison with the behaviour of stiffened plates, it can be said that the global
action of shell structures takes advantage of the load-diffusion capacity of the surface and the
stiffeners help to avoid local buckling by subdividing the surface into cells, resulting in a lower
span to thickness ratio. A longitudinally-stiffened cylinder, therefore, behaves like a system of
struts-and-plates, in a way that is analagous to a stiffened plate. On the other hand, transverse
stiffeners behave in a similar manner to the diaphragms in a box girder, i.e. they help to
distribute the external loads and maintain the initial shape of the cross-section, thus avoiding
distortions that could eventually lead to local instabilities. As in box girders, special precautions
have to be taken in relation to the diaphragms transmitting bearing reactions; in shells the
reaction transmission is done through saddles that produce a distributed load.

3. IMPORTANCE OF IMPERFECTIONS

As was explained in previous lectures, the theoretical limits of bifurcation of equilibrium

that can be reached using mathematical models are upper limits to the behaviour of actual
structures; as soon as any initial displacement or shape imperfection is present, the curve is
smoothed [2]. Figures 8a and 8b present the load-displacement relationship that is expected for a
bar and a plate respectively; the dashed line OA represents the linear behaviour that suddenly
changes at bifurcation point B (solid line). The plate has an enhanced stiffness due to the
membrane effect. The dashed lines represent the behaviour when imperfections are included in
the analysis.
 As can be seen in Figure 8c, the post-buckling behaviour of a cylinder is completely
different. After bifurcation, the point representing the state of equilibrium can travel along the
secondary path BDC. Following B, the situation is highly dependent on the characteristics of the
test, i.e. whether it is force-controlled or displacement controlled. In the first case, after the
buckling load is reached, a sudden change from point B to point F occurs (Figure 8c) which is
called the snap-through phenomenon, in which the shell jumps suddenly between different
buckling configurations.

The behaviour of an actual imperfect shell is represented by the dashed line. Compared
with the theoretically perfect shell, it is evident that true bifurcation of equilibrium will not occur
in the real structure, even though the dashed lines approach the solid line as the magnitude of the
imperfection diminishes. The high peak B is very sharp and the limit point G or H (relevant to
different values of the imperfection) refers to a more realistic lower load than the theoretical
bifurcation load.

The difference in behaviour, compared with that of plates or bars, can be explained by
examining the pattern of local buckling as the loading increases. Initially, buckling starts at local
imperfections with the formation of outer and inner waves (Figure 9a); the latter represent a
flattening rather than a change in direction of the original curvature and set up compressive
membrane forces which, along with the tensile membrane forces set up by the outer waves, tend
to resist the buckling effect. At the more advanced stages, as these outer waves increase in size,
the curvature in these regions changes direction and becomes inward (Figure 9b). As a result, the
compressive forces now precipitate buckling rather than resist it, thus explaining why
equilibrium, at this stage, can only be maintained by reducing the axial load.
 The importance of imperfections is such that, when tests on actual structures are carried out,
the difference between theoretical and experimental values produces a wide scatter of results (see
Figure 10). As the imperfections are unavoidable, and depend very much on the quality of
construction, it is clear that only a broad experimental series of tests on physical models can help
in establishing the least lower-bound that could be used for a practical application. Thus it is
necessary to choose:

1. The structural type, e.g. a circular cylinder, and a fixed set of boundary conditions.
2. The type of loading, e.g. longitudinal compression.
3. A predefined pattern of reinforcement using stiffeners.
4. A strict limitation on imperfection values.

In consequence, the experimental results can only be used for a very narrow band of
applications. In addition, the quality control on the finished work must be such that the
experimental values can be used with confidence.

To allow for this, Codes of Practice [3] use the following procedure:

1. A critical stress, cr or cr, or a critical pressure, pcr, is calculated for the perfect elastic shell by
means of a classical formula or method in which the parameters defining the geometry of the
shell and the elastic constants of the steel are used.
 2. cr, cr or pcr is then multiplied by a knockdown factor , which is the ratio of the lower bound of
a great many scattered experimental buckling stresses or buckling pressures (the buckling being
assumed to occur in the elastic range) to cr, cr or pcr, respectively.  is supposed to account for
the detrimental effect of shape imperfections, residual stresses and edge disturbances.  may
be a function of a geometrical parameter when a general trend in the set of available test
points, plotted with that parameter as abscissa, points to a correlation between the parameter
and ; such a trend is visible in Figure 10, where the parameter is the radius of cylinder, r,
divided by the wall thickness, t.

4. CONCLUDING SUMMARY

 The structural resistance of a shell structure is based on the curvature of its surface.
 Two modes of resistance are generally combined in shells: a membrane state in which the
developed forces are in-plane, and a bending state where out-of-plane forces are present.
 Bending is generally limited to zones where there are changes in boundary conditions, thickness,
or type of loads. It also develops where local instability occurs.
 Shells are most efficient when resisting distributed loads. Concentrated loads or geometrical
changes generally require local reinforcement.
 Imperfections play a substantial role in the behaviour of shells. Their unpredictable nature
makes the use of experimental methods essential.
 To simplify shell design, codes introduce a knock-down factor to be applied to the results of
mathematical models.

5. REFERENCES

[1] Tossoji, Ei., "Philosophy of Structures", Holden Day 1960.

[2] Brush, D.O., Almroth, B.O., "Buckling of Bars, Plates and Shells", McGraw Hill, 1975.

[3] European Convention for Constructional Steelwork, "Buckling of Steel Shells", European
Recommendation, ECCS, 1988.
ESDEP WG 8

PLATES AND SHELLS

Lecture 8.7: Basic Analysis

of Shell Structures
OBJECTIVE/SCOPE:

To describe the basic characteristics of pre- and post-buckling shell behaviour and to explain and
compare the differences in behaviour with that of plates and bars.

PREREQUISITES

Lecture 8.6: Introduction to Shell Structures

RELATED LECTURES:

Lecture 6.1: Concepts of Stable and Unstable Elastic Equilibrium

Lecture 8.1: Introduction to Plate Behaviour and Design

Lecture 8.4.1: Plate Girder Behaviour & Design I

Lectures 8.5.1: Introduction to Design of Box Girders

SUMMARY:

The combined bending and stretching behaviour of shell structures in resisting load is discussed;
their buckling behaviour is also explained and compared with that of struts and plates. The effect
of imperfections is examined and ECCS curves, which can be used in design, are given.
Reference is also made to available computer programs that can be used for shell analysis.

1. INTRODUCTION

Lecture 8.6 introduced several aspects of the structural behaviour of shells in an essentially
qualitative way. Before moving on to consider design procedures for specific applications, it is
necessary to gain some understanding of the possible approaches to the analysis of shell
response. It should then be possible to appreciate the reasoning behind the actual design
procedures covered in Lectures 8.8 and 8.9.
This Lecture, therefore, presents the main principles of shell theory that underpin the ECCS
design methods for unstiffened and stiffened cylinders. Comparisons are drawn with the
behaviour of columns and plates previously discussed in Lectures 6.6.1, 6.6.2 and 8.1.

2. BENDING AND STRETCHING OF THIN SHELLS

The deformation of an element of a thin shell consists of the curvatures and normal
displacements associated with out-of-surface bending and the stretching and shearing of the
middle surface. Bending deformation without stretching of the middle surface, as assumed in the
small deflection theory for flat plates, is not possible, and so both bending and stretching strains
must be considered.

If the shape and the boundary conditions of a shell and the applied loads are such that the
loads can be resisted by membrane forces alone, then these forces may be found from the three
equilibrium conditions for an infinitely small element of the shell. The equilibrium equations
may be obtained from the equilibrium of forces in three directions; that is, in the two principal
directions of curvature and in the direction normal to the middle surface. As a result, the three
membrane forces can be obtained easily in the absence of bending and twisting moments and
shear forces perpendicular to the surface. An example is an unsupported cylindrical shell
subjected to uniform radial pressure over its entire area (Figure 1). Obviously, the only stress
generated by the external pressure is a circumferential membrane stress. The assumption does
not hold if the cylinder is subjected to two uniform line loads acting along two diametrically
opposed generators (Figure 2). In this case, bending theory is required to evaluate the stress
distribution, because an element of the shell cannot be in equilibrium without circumferential
bending stresses. Circumferential bending stresses are essential to resist the external loads, and
because the wall is thin and has very little flexural resistance, they greatly affect its load carrying
resistance [1].
 Significant bending stresses usually only occur close to the boundaries, or in the zone
affected by other disturbances, such as local loads or local imperfections. Locally, the resulting
stresses may be quite high, but they generally diminish at a small distance from the local
disturbance. Bending stresses may, however, cause local yielding which can be very dangerous
in the presence of repeated loadings, since it can result in a fatigue fracture.

It is normally more structurally efficient if a shell structure can be configured in such a

way that it carries load primarily by membrane action. Simpler design calculations will usually
also result.

3. BUCKLING OF SHELLS - LINEAR AND NON-LINEAR BUCKLING THEORY

Buckling may be regarded as a phenomenon in which a structure undergoes local or

overall change in configuration. For example, an originally straight axially loaded column will
buckle by bowing laterally; similarly a cylinder may buckle when its surface crumples under the
action of external loads. Buckling is particularly important in shell structures since it may well
occur without any warning and with catastrophic consequences [2-4].

The equations for determining the load at which buckling is initiated, through bifurcation
on the main equilibrium path of a cylindrical shell, may be derived by means of the adjacent
equilibrium criterion, or, alternatively, by use of the minimum potential energy criterion. In the
first case, small increments (u1, v1, w1) are imposed on the pre-buckling displacements (u0, v0,
w0)

u = u 0 + u1

v = v0 + v1 (1)

w = w0 + w1
 The two adjacent configurations, represented by the displacements before (u 0, v0, w0) and
after the increment (u, v, w) are analysed. No increment is given to the load parameter. The
function represented by (u1, v1, w1) is called the buckling mode. As an alternative, the minimum
potential energy criterion can be adopted to derive the linear stability equations. The expression
for the second variation of the potential energy of the shell in terms of displacements is
calculated. The linear differential equations for loss of stability are then obtained by means of the
Trefftz criterion. Readers requiring a more detailed coverage of shell buckling are advised to
consult [4].

In practice, for some problems, the results obtained by these analyses are adequate and in
accordance with experiment. In other cases, such as the buckling of an axially compressed
cylinder, the results can be positively misleading as they may substantially overestimate the
actual carrying resistance of the shell. The use of these methods leads to the following value for
the axial buckling load of a perfect thin elastic cylinder of medium length:

(2)

Assuming  = 0,3 for steel gives cr = 0,605 E

This buckling load is derived on the assumption that the pre-buckling increase of the
radius due to the Poisson effect is unrestrained and that the two edges are held against
translational movement in the radial and circumferential directions during buckling, but are able
to rotate about the local circumferential axis. These edge restraints are usually called "classical
boundary conditions".

Equation (2) is of little use to the designer because test results yield only 15-60% of this
value. The reason for the big discrepancy between theory and experimental results was not
understood for a long time and has been the subject of many studies; it can be explained as
follows:

The boundary conditions of the shells have a significant effect and can, if modified, give
rise to lower critical loads. Many authors have investigated the effects of the boundary
conditions on the buckling load of cylindrical shells. The value given by Equation (2) refers to a
real cylinder only if the edges are prevented from moving in the circumferential direction, i.e. v
= 0 (Figure 3). If this last condition is removed and replaced by the condition n xy = 0 (i.e. free
displacement but no membrane stress in the circumferential direction) a critical value of
approximately 50% of the classical buckling load is obtained. This boundary condition is quite
difficult to obtain in practice and cylinders with such edge restraints are much less sensitive to
imperfection than cylinders with classical boundary conditions; they are, therefore, not of
primary interest to the designer. If, instead, the top edge of the cylinder is assumed to be free, the
critical buckling load drops down to 38% of the critical value given by Equation (2). In general,
it can be stated that if a shell initially fails with several small local buckles, the critical load does
not depend to any great extent on the boundary conditions, but, if the buckles involve the whole
shell, the boundary conditions can significantly affect the buckling load.

The critical buckling load may also be reduced by pre-buckling deformations. To take
these deformations into account, the same boundary conditions in both the pre- and post-
buckling range must be included. The consequence is that during the compression prior to
buckling, the top and bottom edges cannot move radially (Poisson's ratio not being zero) and,
therefore, the originally straight generators become curved. The post-buckling deformations are
not infinitely small and the critical stress is reduced.
 The complete understanding of the reason for the large discrepancies between theoretical
and experimental results in the buckling of shells has caused much controversy and discussion,
but now the explanation that initial imperfections are the principal cause of the phenomenon is
generally accepted.

4. POST-BUCKLING BEHAVIOUR OF THIN SHELLS

The starting point for this illustrative study of the post-buckling behaviour of a perfect
cylinder, under axial compression (Figure 3), is Donnell's classical equations [2]. A suitable
function for w (trigonometric) may be assumed and introduced into the compatibility equation,
expressed in terms of w and of an adopted stress function F. The quadratic expressions can be
transformed to linear ones by means of well known trigonometric relations. Then the stress
function F, and as a consequence the internal membrane stresses, may be computed. The
expression for the total potential energy can then be written, and minimized, to replace the
equilibrium equation. The solution is improved by taking more terms for w.

In Figure 4, the results obtained by using only two buckling modes are shown and
compared with the curves obtained later, i.e. with a greater number of modes. The results show
that the type of curve does not change by increasing the number of modes, but the lowest point
of the post-buckling path decreases and can attain a value of about 10% of the linear buckling
load. In the limiting case, i.e. where the number of terms increases to infinity, the lowest value of
the post-buckling path tends to zero, while the buckling shape tends to assume the shape of the
Yoshimura pattern (Figure 5). It is the limiting case of the diamond buckling shape that can be
described by the following combination of axi-symmetric and chessboard modes.

(3)

 
 It is worth noting that the buckling load associated with either the combination or the two
single modes is the same and is given by Equation (2).

A comprehensive overview of post-buckling theory is given in [5]. As will be discussed

later, a realistic theory for shell buckling has to take into account the unavoidable imperfections
that appear in real structures. Figure 6 shows the influence of imperfections on the strength of a
cylinder subject to compressive loading and Figure 7 shows typical imperfections.
5. NUMERICAL ANALYSIS OF SHELL BUCKLING

Simple types of shells and loading are amenable to treatment by analytical methods. The
buckling load of complex shell structures can, however, be assessed only be means of computer
programs, many of which use finite elements and have a stability option. CASTEM, STAGS,
NASTRAN, ADINA, NISA, FINELG, ABAQUS, ANSYS, BOSOR and FO4BO8 are some of
the general and special purpose programs available. Correct use of a complicated program
requires the analyst to be well acquainted with the basis of the approach adopted in the program.

The stability options and the reliability of the numerical results depend on the method of analysis
underlying each specific program, and on the buckling modes considered. Analysis of various
types may be performed:

1. Geometrical changes in the pre-buckling range are ignored, the pre-buckling behaviour of the
structure is thus assumed to be linear, and the buckling stress corresponds to that at the
bifurcation point B which is found by means of an eigenvalue analysis (Figure 8a). Applied to a
simple shell, this procedure yields the classical critical load. w denotes the lateral deflection of
the shell wall at some representative point.
2. Non-linear collapse analysis enables successive points on the non-linear primary equilibrium
path to be determined until the tangent to the path becomes horizontal at the limit point
(Figure 8b). At that stage, assuming weight loading, as is normally the case for engineering
structures, non-linear collapse ("snap-through") occurs.
3. Investigating bifurcation buckling from a non-linear pre-buckling state involves a search for
secondary equilibrium paths (corresponding to different buckling modes, e.g. different numbers
of buckling waves along the circumference of an axi-symmetric shell) that may branch off from
the non-linear primary path at bifurcation points located below the limit point (Figure 8c). The
lowest bifurcation point provides an estimate of the buckling load.
4. General non-linear collapse analysis of an imperfect structure consists of determining the non-
linear equilibrium path and the limit point L for a structure whose initial imperfections and
plastic deformations are taken into account (Figure 8d). The limit load, which is the ordinate of
L, causes the structure to "snap-through".
The four load-deflection diagrams given in Figure 8 may relate, for example, to a spherical cap
subjected to uniform radial pressure acting towards the centre of the sphere; in this case the
critical failure mode depends on the degree of shallowness of the cap.
6. BUCKLING AND POST-BUCKLING BEHAVIOUR OF STRUTS, PLATES AND
SHELLS

Equilibrium paths, for a perfectly straight column, a perfectly flat plate supported along
its four edges, and a perfectly cylindrical shell, presented in the preceding Lecture 8.6, are
repeated here (Figure 9) for completeness.
 In each diagram,  represents the uniformly applied compressive stress, cr its critical
value given by classical stability theory, and U the decrease in distance between the ends of the
members.

Each point on the solid or dashed lines represents an equilibrium configuration which is
at least theoretically possible, in the sense that the conditions for equilibrium between external
and internal forces are met.

Simple elastic shortening, according to Hooke's law, is reflected by the three straight
lines OA. They represent the pre-buckling, primary, or fundamental state of equilibrium, in
which the column, the plate and the shell remain perfectly straight, flat and cylindrical,
respectively.

As long as  < cr, the primary equilibrium is stable, i.e. if a minute accidental
disturbance (a very small lateral force, for example) causes a slight transverse deformation of the
member, the deformation disappears when its cause is removed, and the member returns of its
own accord to its previous configuration. Any point of the line OA, which is located above B,
represents, however, unstable equilibrium, i.e. the effect of a disturbance, even an infinitely
small one, does not disappear with its cause, but instantaneously increases and the member is set
in (violent) motion, deviating further and irreversibly from its previous equilibrium
configuration. Some minor cause of disturbance always exists, for example, in the form of an
initial shape imperfection or of an eccentricity of loading. A state of unstable equilibrium,
therefore, although theoretically possible, cannot occur in real structures.

When the stress reaches its critical value, cr, a new equilibrium configuration appears at
point B. This configuration is quite different from the primary one and features lateral
deflections and bending of the strut, the plate, or the wall of the shell.

If the new configuration is characterised by displacements with respect to the primary

state of equilibrium which increase gradually from zero to high (theoretically infinite) values, the
post-buckling states of equilibrium are represented by points on a secondary equilibrium path
which intersects with the primary path at the bifurcation point B.

In fact, B is the lowest of an infinite number of bifurcation points, but the paths branching
off from all the others represent highly unstable equilibrium and have no practical significance.

The great difference between the strut, the plate and the cylinder is embodied in their
post-buckling behaviour. In the case of the column (Figure 9a), the secondary path, BC, is very
nearly horizontal, but in reality it curves imperceptibly upwards; the equilibrium along BC is
almost neutral (it is, strictly speaking, weakly stable). For the plate (Figure 9b) the secondary
path, BC, climbs above B, although less steeply than before; the plate deflects laterally, more and
more under a gradually increasing load, but the equilibrium at points on BC is stable.

After bifurcation, the point representing the state of equilibrium of an axially loaded
cylinder (Figure 9c), in theory, can travel along the secondary path BDC. The equilibrium at
points located below B on the solid curve is, however, highly unstable and, hence, cannot really
exist. What would happen after point B is reached, if it were possible to manufacture a perfect
cylinder from material of unlimited linear elasticity and to support and load or deform it in the
theoretically correct manner, depends on the loading method.

When displacements, u, of one plate of a supposedly rigid testing machine with respect to
the other plate are imposed in a controlled manner, buckles suddenly appear in the wall of the
cylinder. The compressive stress drops at once from cr to the ordinate of point E (only a fraction
of cr), while the shortening of the cylinder remains equal to u cr, the abscissa of B. In contrast
with bifurcation, finite displacements are involved in the transition between the equilibrium
configurations represented by points B and E; such an occurrence is called snap-through. The
buckling process is further complicated by the existence of different intersecting equilibrium
paths, which correspond to different numbers of circumferential buckling waves and which have
the same general shape as BDC. Some parts of these paths represent stable equilibrium, while
other parts represent unstable equilibrium; after the initial snap-through from state B to state E,
the shell can jump repeatedly from one buckling configuration to another.

When the load, rather than the displacement, u, is controlled a different effect occurs; if,
for example, a load = 2rtcr is imposed, the overall shortening of the cylinder almost instantly
increases from ucr to the abscissa of point F, and its wall suddenly exhibits deep buckles, while
the average compressive stress remains equal to cr. It should be noted that this "snap-through"
has dynamic characteristics which are not considered in this description.

Esslinger and Geier [5] explain the fundamentally different behaviour of columns, plates
and shells by the following argument illustrated by Figures 10, 11 and 12.
The differential equation

expresses the lateral equilibrium of any element of an axially loaded strut when bifurcation
occurs (Figure 10a) by stating that the deflecting force per unit length, due to the external loads,
Fcr, given by the first term, cancels out the restoring force per unit length, due to the internal
bending stresses, given by the second term. Both the deflecting forces (Figure 10b) and the
restoring forces (Figure 10c) are proportional to the lateral deflection. Consequently, equilibrium
of the column is independent of the magnitude of the transverse deformation and of u, for a given
constant axial force Fcr.

The restoring forces which balance the lateral forces (Figure 11b) deflecting a buckling
plate (Figure 11a), are due not only to longitudinal and transverse bending moments (Figure
11c), but also to transverse membrane forces (Figure 11d). The restoring forces due to membrane
action are zero, as long as the plate is flat, but they then increase proportionately to the square of
its lateral deflection. As a result, the compressive external load required for equilibrium increases
together with the lateral deformation and with the plate shortening u.

Figure 12a shows the radial component of the buckling pattern of a compressed cylinder
at the bifurcation point. Outward displacements of a curved surface cause tensile membrane
forces, while inward displacements generate compressive membrane forces. Figure 12b gives a
more accurate picture of an inward buckle of very small amplitude; it is seen that the original
sign of the circumferential curvature of the shell wall is not reversed at the start of buckling. The
radial forces arising from the combination of the membrane forces with the curvature of the
deformed cylinder, which still has its initial sign, are shown in Figure 12c. These radial forces all
tend to counteract buckling. Hence the high resistance of a perfect cylinder to the initiation of
buckling, given by Equation (2). Increasing inward displacements cause the change of
circumferential curvature to exceed the magnitude, 1/r, of the original curvature of the cylinder,
as shown in an exaggerated manner in Figure 12d, and more realistically in Figure 12e. In the
region of the inward buckles, the wall of the cylinder is now curved inwards and, as a result, the
compressive membrane forces in these areas no longer resist the appearance of dents, but
precipitate them (Figure 12f). Hence, the total restoring effect of the membrane forces has now
weakened substantially compared with the state prevailing at the bifurcation point. The upshot is
that, once buckling has started, equilibrium is conceivable only under decreasing axial load.

7. IMPERFECTION SENSITIVITY

The behaviour of actual imperfect components differs from the theoretical behaviour
described above and is represented by the dotted curves in Figure 9. They show that true
bifurcation of equilibrium does not actually occur in the case of real structural members.
However, the solid lines provide an approximate picture - the smaller the initial imperfections,
the truer the picture is - of the behaviour of the component, and therein lies the significance of
the bifurcation buckling concept.
 The dotted lines in Figures 9a and 9b, have been drawn for a column and a plate with
slight initial curvature. It can be seen that the carrying resistance of the strut is not much lower
than the theoretical buckling load, provided that the imperfection is not too great. One can
conclude from Figure 9b that the equilibrium path of an imperfect plate may not exhibit any
discontinuity when the compressive stress increases beyond cr, and also that the plate may
possess a considerable post-buckling strength reserve. If it is thin, this reserve may be
considerably greater than the bifurcation buckling load. Raising the stress beyond cr for the
perfect plate does not bring about immediate ultimate failure. Both the column and the plate
finally fail by yielding caused by excessive bending.

Owing to the imperfection of a real cylinder, the dotted equilibrium path does not display
the very sharp high peak B which is a feature of the theoretical equilibrium path OBDC. The
culminating point G or H (Figure 9c) of the dotted line, called a limit point, is at a much lower
level than the bifurcation point, even when the amplitude of the initial deviations from the
perfect cylindrical shape is minute. The lower dotted curve is the equilibrium path for a cylinder
with somewhat larger imperfections. When the loading is due to weight and happens to
correspond to the limit point, the curve must jump horizontally from G or H towards the right
hand branch of the curve. The concomitant shortening, u, of the steel shell is so large, and due to
buckles which are so deep, that normally part of the wall material is strained into the plastic
range and so the buckling phenomenon, in this case a snap-through or non-linear collapse, is
almost always catastrophic.

One should not infer from the description in the preceding paragraph that only imperfect
structural components display behaviour characterized by a limit point. Due to gradual changes
in the geometry of a perfect structure, its primary equilibrium path may be non-linear from the
outset of loading and, indeed, feature a limit point.

As a summary two points can be established:

1. The real collapse stress, uG or uH (Figure 9c), is much lower than the theoretical critical stress,
cr, for the perfect shell, even though the imperfections may be hardly perceptible.
2. Nominally identical shells collapse under markedly different loads because the unintentional
actual imperfections of such shells, as erected, are different in magnitude and in distribution,
and because an appreciable decrease in ultimate load may result from slightly larger
imperfections.

A sweeping generalization to the effect that all shells are always very sensitive to deviations
from the perfect shape would be unwarranted. The imperfection sensitivity depends on the type
of shell and loading. It may vary from slight to extreme, even for the same kind of shell under
different loading conditions. For example, the imperfection sensitivity of cylindrical shells under
uniform external pressure is quite low, whilst the same shells are highly imperfection sensitive
when they are compressed in the meridional direction. The difference relates to the buckling
mode; under axial load, the buckling modes are characterised by waves which, compared to the
diameter, are short in both the longitudinal and the circumferential direction. Small initial
imperfections, which may occur anywhere on the surface of the cylinder and which are likely to
have roughly the same shape as some of the critical buckles, tend to deepen under increasing
load and to trigger off a snap-through at an early loading stage. The buckling pattern under
external pressure, however, consists of buckles which are long in the meridional direction, and
less numerous in the hoop direction, and therefore probably of considerably larger size than the
principal initial dents and bulges.

Another factor that should be mentioned as contributing to the imperfection sensitivity of

axially loaded cylinders is the multiplicity of different buckling modes associated with the same
bifurcation load. Any realistic theoretical treatment of the buckling problem is complicated
further by the existence of residual stresses due to cold or hot forming and/or due to welding.
Behaviour is also affected by the appearance of plastic deformations in the steel and, in some
cases, by the presence of stiffeners. The non-linear structural behaviour of the shell may be due
to the latter, as well as to changes in the geometry resulting from the deformation of the shell.

In conclusion, imperfections are the main cause of the large difference between the ultimate
load obtained in tests and the theoretical buckling load. A wide scatter of results for nominally
identical shells can be seen in Figure 13a where the ratio of experimental buckling loads, F u,
against the theoretical values, Fcr, for axially loaded cylinders are given for different r/t ratios.
Figure 13b gives the factors proposed by ECCS to reduce the theoretical buckling load to values
appropriate for design.
8. CONCLUDING SUMMARY

 Bending and stretching are the modes by which shell structures carry loads.
 For shell structures, in industrial applications, buckling may be the critical limit state due to
slenderness effects.
 Imperfections are the main cause of the very significant difference between the theoretical and
the experimental buckling load.
 There are fundamental differences in initial buckling behaviour between shells and plates.
 In practice, shell buckling analysis can be applied only to special structures which have been
manufactured/constructed using strict quality control procedures that minimise imperfections.

9. REFERENCES
[1] Timoshenko, S. and Woinowsky-Krieger, S., "Theory of Plates and Shells", McGraw-Hill,
New York and Kogakusha, Tokyo, 1959.
[2] Flügge, W., "Stresses in Shells", Springer-Verlag, New York, 1967.
[3] Bushnell, D., "Computerised Buckling Analysis of Shells", Martinus Nijhoff Publishers,
Dordrecht, 1985.
[4] Timoshenko, S. and Gere, J.M., "Theory of Elastic Stability", McGraw-Hill, New York and
Kogakusha, Tokyo, 1961.
[5] Esslinger, M. T., and Geier, B. M., "Buckling and Post Buckling Behaviour of Thin-Walled
Circular Cylinders", International Colloquium on Progress of Shell Structures in the last 10 years
and its future development, Madrid, 1969.
10. ADDITIONAL READING
1. Koiter, W.T., "Over de Stabilitteit van het Elastisch Evenwicht Diss.", Delft, H.J.Paris, Amsterdam,
1945.