Journal of Constructional Steel Research 56 (2000) 175197

www.elsevier.com/locate/jcsr

The behaviour of locally-supported cylindrical

shells: unstiffened shells
W. Guggenberger a, R. Greiner a, J.M. Rotter b,*

a
Institut fur Stahlbau, Holzbau und Flachentragwerke, Technical University of Graz, Graz, Austria
b
Division of Engineering, University of Edinburgh, Edinburgh EH9 3JN, UK

Received 2 July 1996; received in revised form 20 October 1999; accepted 15 November 1999

Abstract

Many metal silos and tanks are locally supported on columns or other discrete supports to
permit easy access beneath the vessel. The discrete supports introduce local forces into the
cylindrical shell, which in turn produces zones of local high axial compression. The strength
of compressed cylindrical shells has long been known to be governed by buckling consider-
ations, but the buckling strength of discretely supported cylinders has only been investigated
in recent years. Design rules which pre-date this study were empirical in nature and not based
on rigorous buckling calculations or tests. In the present study, the theoretical buckling behav-
iour in the elasticplastic range has been found to be so complicated that much work is needed
to define the buckling strength for the purposes of structural design. This paper presents a
description of the linear and nonlinear behaviour of isotropic unstiffened cylinders which are
discretely supported at the lower edge. Much of the information given here is vital to the
development of a more reliable design rule for shells under these loading conditions. 2000
Elsevier Science Ltd. All rights reserved.

Keywords: Cylinders; Shells; Buckling; Collapse; Local supports; Non-uniform stress; Imperfections;
Elastic-plastic; Nonlinear analysis

1. Introduction

Elevated metal silos and tanks are often supported on columns or other discrete
supports to provide access for the gravity withdrawal of the contents of the vessel.
Each local support induces local high axial compression stresses in the cylindrical

* Corresponding author. Tel.: +44-031-650-7170; fax: +44-031-667-5719.

E-mail address: michael@civ.ed.ac.uk (J.M. Rotter).

0143-974X/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 9 ) 0 0 1 0 2 - 9
176 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Nomenclature
d support width
E Youngs modulus
F factored value of vertical reaction force on a support (=Fo)
Fo reference value of vertical reaction force on a support (=300 kN)
Fcl classical elastic critical reference reaction force on a support
(=2pRscl/n)
L cylinder height
n number of discrete supports
t wall thickness
R cylinder radius
d amplitude of the imperfection
load factor on applied loads
m proportion of the circumference which is supported (Eq. (2))
n Poissons ratio
scl classical elastic critical stress for a perfect cylinder under uniform
axial compression (Eq. (1))
smm mean axial membrane stress immediately above a support at
buckling
s mm mean stress ratio above support at buckling (=smm/scl)
su,e-p axial compression stress at elasticplastic buckling for a uniformly
supported imperfect cylinder
sxx axial membrane stress within the shell
sxxo axial membrane stress immediately above each support
s xx mean axial stress ratio above each support (=sxx/scl); and
sy unixial yield stress
GMNL geometrically and materially nonlinear analysis
GMNLI geometrically and materially nonlinear analysis including
imperfections
GNL geometrically nonlinear elastic analysis
GNLI geometrically nonlinear elastic analysis including imperfections
GNL-max maximum strength from geometrically nonlinear elastic analysis
GNL-min post-buckling minimum strength from geometrically nonlinear
elastic analysis
MNL materially nonlinear analysis (small displacement theory)

shell just above the support. It has long been known that the strength of an axially
compressed thin-walled cylindrical shell is controlled by buckling, and that buckling
occurs at low absolute stresses and is very sensitive to geometric imperfections.
However, until recently, almost all the studies in the literature related only to uniform
compression or a slight variation from it, such as global bending [1,2]. Very little
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 177

attention has been paid to local loads and the buckling strength of a shell above a
local support. No rationally-based design proposal has yet been presented in the open
literature, though there have been a number of simplified design concepts [24]. In
recent studies [511] the buckling behaviour has been found to be very complicated,
with many parameters influencing the buckling strength. Thus, much work is needed
to define the elastic and plastic buckling strength for the purposes of structural design.
This paper presents a description of the linear and nonlinear behaviour of isotropic
unstiffened cylinders which are discretely supported at the lower edge. Much of
the information in this paper is vital to the development of useful design rules for
these shells.
The geometry of a typical silo or tank on discrete supports is shown in Fig. 1.
There are many ways in which the silo can be supported, arising from different
designers attempts to reassure themselves that their structure would be adequately
strong in the absence of a rational design method. In particular, many different stiff-
ening arrangements have been proposed (e.g. [12]), with the intention of eliminating
potential buckling failures, but it is difficult for a designer to know if such stiffening
arrangements are needed for a particular structure, as many satisfactory designs have
been produced without stiffening. A number of studies have developed analysis tech-
niques for designers to determine the linear stress distribution in a locally ring-stiff-
ened locally supported shell [1315], but none of these were able to propose a buck-
ling strength assessment to show when the stress redistribution achieved by the ring
stiffeners was adequate.
This study concentrates on the simpler case of the unstiffened cylinder. When the
buckling strength of an unstiffened cylinder can be assessed with confidence, it will
be possible to identify cylinders which need stiffening, and to study the changes in
buckling behaviour and the increases in buckling strength which stiffening provides.

Fig. 1. Typical geometries for silos or tanks on discrete supports. (a) Unstiffened, (b) strengthened with
bottom course wall thickening, (c) with longitudinal stiffeners, (d) additional ring stiffener.
178 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

2. Finite element modelling

The cylindrical shell was studied here using non-linear finite element analyses in
the ABAQUS system [16]. This program has been extensively verified against test
results for locally supported cylinder buckling [1719]. Its capacity to follow geo-
metrically nonlinear paths and detect bifurcations has been demonstrated for this
case.
The cylindrical shell was treated as supported on four discrete axially flexible
supports, and symmetry was exploited so that only one eighth of the shell was ana-
lysed. Teng and Rotter [7,8] and She and Rotter [10] have shown that the buckling
strength at a single discrete support is not sensitive to the number of supports around
the circumference provided that this number is not large, so the limitation to only
four supports is not restrictive. Further, the buckling mode above each support is
relatively local, so the question of interaction between buckles above adjacent sup-
ports, and the adequacy of a finite element model which assumes symmetry at the
support and mid-span is not in question.
A complete silo, tank or pressure vessel structure has closures at its ends: these
are frequently conical shells, with or without a ring at the junction, but they may
also be tori-spherical or tori-conical. All these types of end closure apply a stiff
radial restraint to the ends of the cylinder: the analysis of an isolated cylinder with
completely rigid radial restraint at each end is thus relevant to most practical struc-
tures.
The loading conditions to which the cylindrical shell is subjected vary somewhat
from one application to another: the dominant forces may arise from pressure on the
closure, from wall friction due to stored solids, or from more uniform loads applied
at a distant boundary. Whatever the global loading case, the local stress regime close
to the support is similar, and since the buckling behaviour is found to be quite
local, the details of the loading which induces the high forces in the supports are
not important.
The modelling used in this study is shown in Fig. 2. A single radius-to-thickness
ratio of R/t=500 and a single height to radius ratio of L/R=2.0 were used throughout.
The material was assumed to be elasticperfectly plastic with a modulus E=206 000
MPa, a Poissons ratio n of 0.30 and a yield stress sy of 235 MPa. Conditions of
symmetry were assumed at the cylinder half height, and the lower edge had four
supports of varying width d=2b. For reference purposes, a shell thickness of 10 mm
was used, and a reference uniform axial stress of 3.820 MPa (=1200 kN applied to
the complete circumference) was applied at the top of the shell. For comparison
purposes, it may be noted that this reference stress induces a mean axial membrane
stress of 1.53% of the classical elastic critical stress scl [20,21], which is given by
E t t
scl 0.605E

(1)
3(1n ) R
2 R

The behaviour shown in the following description refers to the load factor applied
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 179

Fig. 2. Simplified structural model used in this study.

to this reference loading. This loading was increased by different load factors in the
following calculations. The coordinate system and boundary conditions are shown in
Fig. 3.
The support is treated as axially flexible, applying a uniform stress to the edge

Fig. 3. Typical finite element mesh and boundary conditions (S8R5 or S4R5 shell elements).
180 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

of the cylinder. The lower boundary was restrained against radial and tangential
displacements, but not against rotation, modelling a ring which would keep it circu-
lar. Although the following discussion relates to flexible supports which apply a
uniform stress to the shell, other studies (e.g. [10]) have found similar effects when
rigid supports were used.

3. Axial membrane stress distributions in the shell

Although the strength predictions described below derive almost entirely from
nonlinear analysis, a linear stress analysis is shown first to provide an understanding
of the key phenomena.
The pattern of axial compression membrane stress sxx above the support, which
varies in both the meridional and circumferential directions, is illustrated in Fig. 4
at a load factor of =10. The highest stress is naturally immediately above the sup-
port, but it declines rapidly as the force is dispersed into the shell. The bell-shaped
distribution in the circumferential direction is important in understanding the buck-
ling behaviour, since a finite area of highly stressed shell is needed for a buckle to
occur, and the rate at which the axial stress falls with height is closely related to
the rate at which the highly stressed zone increases in circumferential width. Because
symmetry has been adopted at the cylinder half-height, the axial stress remains quite
non-uniform around the circumference at this position, but (as noted above) this does
not influence the buckling behaviour markedly, since the buckle is so local.
When the support is narrow, very high local axial stresses can occur above it, and
buckling may be preceded by yielding. The critical point for yielding is marked in

Fig. 4. Axial membrane stress distribution above a local support (critical zones for local membrane
yielding or local buckling are marked).
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 181

Fig. 4 as immediately above the support. Although high bending stresses develop a
little higher above the support, and first surface yield occurs at this point, they have
little influence on the plastic buckling strength, so they are not described in detail
here.
The bell-shaped circumferential variation of axial stress plays a key role in the
buckling behaviour and strength. The shape and amplitude of the bell at a height of
L/20 above the support are shown in Fig. 5a and b for different support widths.
The support width can be expressed in terms of either the supported proportion
of the circumference, m, or the ratio of the support width to the radius d/R, which
indicates the localisation of the support. The two are related through

m
nd

n d
2pR 2p R
(2)

in which n is the number of supports. Theoretically, the support width can vary from
zero to a condition where the entire circumference is supported, which is given by
d 2p
m1 or (3)
R n
which, for four columns, means (d/R)=p/2. The normalisation with respect to the
radius (d/R) is preferred over the supported circumference m because the buckling
strength for practical supports is independent of the number of supports. If the sup-
ported circumference is used, a different curve is required when the number of sup-

Fig. 5. (a) Axial membrane stress distributions sxx at the height x=L/20 above the support (axial stress
immediately above the support smm is constant). (b) Axial membrane stress distributions sxx at the height
x=L/20 above the support (total force at the support Fo=dtsmm is constant).
182 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 5. (continued).

ports is changed, making the description much more complicated. Moreover, the
supported circumference draws attention away from the width of the individual sup-
port, which has been shown above to be important in determining whether the
strength is force or stress controlled.
Fig. 5a shows the effect of support width, for a constant mean stress immediately
above the support (=10). For the narrower supports (small d/R or m), the stress
pattern at L/20 above the support is very similar, and the peak stress magnitude rises
almost linearly with the force applied to the support (compare d/R=0.033 and 0.066
or m=0.021 and 0.042). By contrast, the widest support, (d/R=0.589, m=0.375) has
a large zone in which the axial stress is still at the value applied to the support, and
the peak stress at this height is independent of support width. Beyond the edge of
the support, the rate of decay around the circumference is similar for wide and narrow
supports, but it is clear that a wide support produces a large zone of reasonably
uniform stress, which can easily accommodate a buckle, whilst a buckle above a
narrow support must form in a zone of rapidly varying stress. Thus, lower buckling
stresses arise with wide supports, and yielding often precedes buckling above nar-
row supports.
It should be noted that the integral under each axial stress distribution in Fig. 5a
is equal to the total applied load at the support, but this load is smaller for narrow
supports than for wide ones because a constant applied stress has been assumed.
The stress patterns at L/20 for different support widths in the narrow range are
similar (Fig. 5b). Here, the same force has been applied to each width, and the
spreading of the stresses just above the support leads to almost identical stresses
above narrow supports (d/R=0.033 and 0.066 or m=0.021 and 0.042). It is reasonable
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 183

to deduce from this that the force at buckling will be independent of support widths
in narrow supports, but that the stress at buckling will be quite independent of sup-
port width for wide supports. The boundary between these two behaviours may be
expected to occur around d/R=0.20.3 (m=0.130.19). This is, however, a rather wide
support in practical terms.

4. Nonlinear behaviour and buckling in the perfect shell

The nonlinear, bifurcation and post-buckling behaviour of elastic shells under uni-
form axial compression has long been studied [2126]. In particular, the reversed
direction of incremental axial deformations immediately after bifurcation in uni-
formly loaded perfect and near-perfect cylinders under quasi-static conditions is well
known. Since significant bending deformations occur above a local support, it was
not clear whether the same behaviour might be expected in a locally supported cylin-
der or not.
Loadaxial displacement curves are shown in Fig. 6a for a perfect elastic cylinder

Fig. 6. (a) Loadaxial displacement curves for a perfect elastic locally supported cylinder with a narrow
support width of d/R=0.066 (m=0.042). (b) Loadradial displacement curves for a perfect elastic locally
supported cylinder with a narrow support width of d/R=0.066 (m=0.042). (c) Loadradial displacement
curves for a perfect elastic locally supported cylinder with a wide support width of d/R=0.2 (m=0.125).
184 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 6. (continued).

on local supports of width d/R=0.066 (m=0.042) (refer to Fig. 3b for node

numbering). The two curves show the displacements at the middle (larger values)
and the edge of the support. Whilst the reversal of axial deformations is not marked,
it is clear that the buckling phenomenon above a local support may be sudden and
imperfection-sensitive, though less so than a uniformly compressed cylinder. The
response well into the post-buckling range is very stable: this is a particularly useful
observation when severe imperfections are under consideration.
The relationships between load factor and radial displacement for the same cylin-
der are shown in Fig. 6b. The buckle profile can be identified by comparing curves
at different heights, and the rather stable shape of the buckle can be inferred. Local
prebuckling deformations due to bending above the support precede snap-through
buckling. Much more complicated curves, indicating changes of buckling mode in
the post-buckling region, are obtained for wider supports, as indicated in Fig. 6c.
The smooth changes in radial deformation within the buckle (Fig. 6c) indicate that
this is a snap-through response, not a bifurcation, and that local supports lead to a
unique path in place of the multiple clustered bifurcations of a uniformly sup-
ported shell.
The prebuckling deformations and the development of buckling displacements are
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 185

Fig. 6. (continued).

shown in Fig. 7, which illustrate the transformation from a long-wave bending pre-
buckling deformation into a local snap-through buckle.

5. Changes in buckling strength with support width for perfect cylinders

The variation of the elastic buckling strength with support width for the full range
of possible support widths is shown in Fig. 8. The strengths are displayed in terms
of the ratio of the mean axial membrane stress immediately above the support sxxo
to the classical elastic critical stress scl. The latter is shown as a bold line for refer-
ence. Several alternative methods of calculation have been used, and are marked
with simple abbreviations for clarity: GNL-max is the maximum load (limit load)
found using a geometrically non-linear elastic calculation (=buckling strength); GNL-
min is the minimum of the elastic post-buckling response, which has often been used
in the past as an approximate lower bound on the strength of imperfect cylinders.
At small support widths, the perfect cylinder buckling strength (GNL-max) falls
progressively, as was foreshadowed by examining the pre-buckling stress distri-
butions: at larger widths (though impractical), an interesting rise in strength is found
186 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 7. Development of buckling deformations above a narrow support (d/R=0.066). (a) At load factor
=3.74, shortly before buckling; (b) at load factor =3.80, shortly after buckling.

Fig. 8. Summary of critical stress ratios s mm for perfect elastic cylinders for the full range of support
widths with four supports.
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 187

(due to twin buckle interference above the support), followed by a later fall, and
finally a rise to the condition for a perfect uniformly supported cylinder. The changes
in the postbuckling minimum (GNL-min) shadow those in the buckling strength for
practical supports, but do not entirely match them for really wide supports: the con-
dition in which half the circumference is supported is seen to give a higher buckling
strength than adjacent widths, though it appears to have no postbuckling mini-
mum load.
A more detailed picture of the changes at practical support widths is shown in
Fig. 9 (d/R0.2). The progressive rise in elastic buckling strength is a direct conse-
quence of the narrower compression zone in which the buckle must form. Results
from other studies show similar buckling strength predictions even when rigid sup-
ports and distributed axial loading are used [10].

6. Imperfections in locally supported cylinders

Although the minimum of the post-buckling response for a perfect elastic cylinder
offers a traditional estimate of the worst effects of serious imperfections, there is no

Fig. 9. Summary of critical stress ratio s mm for perfect elastic cylinders for the practical range of support
widths (d/R0.2).
188 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

formal guarantee of this, and no indication is given of the tolerances which may be
required in design. The next studies examined the effect of a local inward imperfec-
tion immediately above the support. The behaviour of shells with a support width
of d/R=0.130 (m=0.083) is described in detail here to permit comparisons between
different analyses. However, similar calculations were performed on other widths
with similar outcomes. For this geometry, the perfect elastic shell buckles at a load
factor of =5.59, and the minimum of the post-buckling response is at =3.04.
Several different forms of imperfection were examined: the first two were in the
classical eigenmode (Fig. 10a) and the snap-through incremental deformation pattern
(Fig. 10b). The resulting non-linear loadaxial displacement curves are shown in Fig.
11a and b for the classical linear bifurcation mode, and the snap-through incremental
deformation pattern respectively. In the former, an imperfection amplitude of d/t=0.5
brings the peak strength down to 3.28, though the larger imperfection (d/t=1.0) shows
a subsequent increase in strength of 3.36. In Fig. 11b, the snap-through mode shows
a similar response, though the strength reductions are not so large, so this imperfec-
tion form is less critical here. The classical eigenmode has often been recommended
in the past as a worst imperfection mode. It may be that it produces low strengths
here because it is closely related to the post-buckling shape: this idea needs
further investigation.
A further imperfection form was then studied, using an approximation (Fig. 12)
to the perfect shell elastic postbuckling pattern (Fig. 10c), as recommended in the
German shell buckling code DIN 18800 [27]. Hermite cubic spline surfaces were
used and various amplitudes were examined. These amplitudes are expressed as pro-
portions of the shell wall thickness d/t. A comparison of the shell response with
several alternative imperfection forms is shown in Fig. 13, where the imperfection
amplitudes and dimensions recommended for nonlinear finite element analysis by
DIN 18800 [27] were used, with both outward and inward amplitudes in the form
of Fig. 12.
The DIN code inward imperfection produces the lowest calculated elastic strength

Fig. 10. Modal forms from which imperfection patterns were derived. (a) Classical eigenmode (linear
eigenvalue analysis); (b) incremental snap-through mode; (c) post-buckling mode (at load factor =2.74).
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 189

Fig. 11. (a) Loadaxial displacement curves for elastic cylinders with d/R=0.13, with various imperfec-
tion amplitudes in the classical eigenmode. (b) Loadaxial displacement curves for elastic cylinders with
d/R=0.13, with various imperfection amplitudes in the snap-through incremental mode.

in the present study (=3.10) and the value is close to the postbuckling minimum.
Similar calculations were performed for much wider supports, with similar results
[6].

7. Material nonlinearity: elasticplastic behaviour in perfect and imperfect

shells

When the supports are wide and the shell is thin, the elastic buckling load provides
an accurate measure of the strength. However, yielding affects the prebuckling and
buckling behaviour when narrower supports are used, and practical geometries tend
to be in this range.
The effect of yielding on both a perfect and an imperfect cylinder with a support
width of d/R=0.130 (m=0.083) is shown in Fig. 14. The curves are marked to indicate
the form of analysis leading to each. The simplest analysis is geometrically linear
with material nonlinearity, leading to a classical limit load, and is marked MNL
190 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 11. (continued).

(materially nonlinear analysis). A geometrically nonlinear analysis of an elastic per-

fect cylinder is marked GNL, and that including yielding is marked GMNL. In this
example, the perfect cylinder elastic buckling load is similar to the plastic limit load,
placing it in a region of interactive behaviour. The imperfect cylinders are marked
GNLI for the elastic calculation and GMNLI for full material and geometric nonlin-
earity. The adopted imperfection was the approximation to the elastic postbuckling
form (Fig. 10c). A comparison of these curves shows several things: the prebuckling
path is not strongly affected by geometric nonlinearity, though the buckling strength
is; the perfect cylinder buckling strength is only slightly altered by yielding (8%
reduction), but the imperfect cylinder is much more sensitive (28% reduction); the
elastic imperfect cylinder has a maximum strength which is close to the post-buckling
minimum of the perfect cylinder curve; and the fully nonlinear analysis GMNLI is
far below all other strength predictions.
This example shell has a practical thickness (R/t=500) and rests on supports of
practical width (d/R=0.13). The reduction in strength due to plasticity is very signifi-
cant. Also, the post-peak behaviour shows that the catastrophically unstable behav-
iour of the perfect shell becomes much milder in imperfect cylinders, and these
effects are seen whether yielding occurs or not.
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 191

Fig. 12. Local imperfection, similar in form to the perfect shell elastic postbuckling pattern, as a bi-
cubic Hermite spline function.

Elasticplastic strength predictions for imperfect shells on a wide range of support

widths are shown in Fig. 15. The buckling strength is presented as the total force
carried by the four supports relative to the force in the cylinder at uniform axial
compression membrane yield. For widths in excess of d/R=0.30 (a wide support),
the elastic and plastic calculations produce effectively the same result (Fig. 15a), in
spite of some local yielding. For narrower supports (Fig. 15b), elastic calculations
lead to an effectively constant force at buckling, irrespective of the support width.
By contrast, the plastic calculation leads to lower strengths, which are ultimately
bounded by the condition that the axial membrane stress immediately above the
support cannot exceed the stress limit of the von Mises envelope (sxxmax=1.155 sy).
For R/t=500, elasticplastic interaction occurs in the range 0.03d/R0.30. A sum-
mary of the materially and geometrically nonlinear calculated strengths is also shown
in Fig. 8.

8. Design recommendations

Together with results presented elsewhere [10], the above calculations have led
to the following design rule recommendation for elasticplastic imperfect unstiffened
cylinders [11]. This rule involves a plastic squash load limitation, an imperfect elastic
buckling load assessment and an interaction rule between them.
The buckling strength of an imperfect cylinder can be described in terms of the
mean axial membrane stress in the shell immediately above the support smm at buck-
192 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 13. A comparison of loadaxial displacement curves for cylinders with d/R=0.13, with various
imperfection modes but the same amplitude (d/t=1 inward).

ling. The elasticplastic buckling strength can be assessed by combining the elastic
buckling strength smme, which can be approximated by

smme 0.010.06
R
d
t
65 scl
R
(4)

with the plastic limit strength, given by

smmp1.155sy (5)

Since the latter is higher than is permissible in squat columns, it seems appropriate
to reduce it to the yield stress to ensure that the shell cannot be deemed stronger
than the support beneath it, so
smmpsy (6)
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 193

Fig. 14. Comparison of the effect of geometric nonlinearity, plasticity and imperfections on the load
axial displacement curves for cylinders with d/R=0.13.

The proposed elasticplastic interaction [11] is given by

smmmin smmp, 1.15 smmesmmp
smme+smmp
, smme (7)

Eq. (7) represents a linear interaction similar to the Rankine formula.

A well-defined rule might need to meet two further restrictions: first, the evaluated
buckling stress at the support should not be lower than the elasticplastic buckling
stress for uniform compression of an imperfect cylinder su,e-p. Second, the evaluated
buckling stress at the support should not produce a higher strength assessment than
uniform compression in the case where the entire circumference consists of local
supports touching each other edge to edge. These two requirements may be written as
smmsu,e-p (8)
and
194 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Fig. 15. (a) Summary of critical load ratios for imperfect elasticplastic cylinders for the full range of
support widths with four supports. (b) Summary of critical load ratios for imperfect elasticplastic cylin-
ders for the practical range of support widths (d/R0.2).

smm
nd
2pR
su,e-p (9)

Although these two restrictions should be stated, they seldom influence the result
because the bounds are widely separated because the value of (nd/2pR) is small in
normal practice (typically 0.050.3).

9. Conclusions

The elastic and elasticplastic buckling behaviour of upright cylinders on discrete

flexible supports has been examined for a geometry in the practical range. Both
perfect and imperfect cylinders have been studied using a nonlinear finite element
analysis, and the response well into the post-buckling range has been explored.
The dispersal of local high axial stresses introduced by a support has been exam-
ined, and the implications for buckling strength deduced. The buckling behaviour,
with the change from finite prebuckling deformations to localised buckling defor-
mations has been shown. Many calculations of both elastic and plastic buckling
strength have been made, and the results related to the yield stress or the classical
elastic critical stress for the cylinder.
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 195

Fig. 15. (continued).

Narrow supports have been shown to carry much higher stresses at buckling than
wider supports, because the stress regime is less uniform. It has been shown that
the elastic buckling strength of a narrow support is effectively described by the force
which it can carry, which is virtually independent of the support width. By contrast,
the strength of a wide support is strongly affected by the mean stress above the
support. The transition between these two descriptions lies in the width range
0.1d/R0.3.
Yielding has been shown to affect the buckling strength for a wide range of sup-
ports of practical width, with cylinders of practical radius-to-thickness ratio. The
strength of a narrow support in the plastic range is restricted by the maximum stress
on the biaxial von Mises interaction envelope.
Design recommendations, initially presented elsewhere but partly based on these
calculations, have been provided to give designers a simple buckling assessment cri-
terion.
196 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197

Acknowledgements

The authors would like to thank the University of Graz for its financial support
for the cooperative research work between their two institutions. Thanks are also
due to the UK Science and Engineering Research Council and the European CA-Silo
Programme, who funded travel and accommodation expenses for this cooperation.

References

[1] ECCS. European recommendations for steel construction: buckling of shells. 4th ed Brussels: Euro-
pean Convention for Constructional Steelwork, 1988.
[2] Samuelson LA, Eggwertz S, editors. Shell stability handbook. London: Elsevier Applied Science,
1992.
[3] Samuelson LA. Design of cylindrical shells subjected to local loads in combination with axial or
radial pressure. In: Proceedings of an International Colloquium on the Stability of Plate and Shell
Structures, Gent, Belgium, April 68. Brussels, 1987:58996.
[4] Samuelson LA. Effect of local loads on the stability of shells subjected to uniform pressure distri-
bution. In: Proceedings of the IUTAM Symposium on Contact Loading and Local Effects in Thin-
Walled Plated and Shell Structures, August 1990; Prague, Preliminary Report, 1990:348.
[5] Guggenberger W. Buckling of cylindrical shells under local axial loads. In: International Colloquium
on Buckling of Shell Structures on Land, in the Sea and in the Air, 1991 Sept 1719; Villeurbanne,
Lyon, France, 1991:32333.
[6] Guggenberger W. Nichtlineares Beulverhalten von Kreiszylindrischalen unter lokaler Axialbelastung,
Dr.techn dissertation. Technische Universitat Graz, Austria: Institut fur Stahlbau, Holzbau und Flach-
entragwerke, 1992.
[7] Teng JG, Rotter, JM. A study of buckling in column-supported cylinders. Proceedings of the IUTAM
Symposium, Prague, Prelim. Rpt 1990. In: Krupka V, Drdacky M, editors. Contact loading and local
effects in thin-walled plated and shell structures. Prague: Academia Press, 1992:5261.
[8] Teng JG, Rotter JM. Linear bifurcation of column-supported perfect cylinders: support modelling
and boundary conditions. Thin-Walled Structures 1992;14:24163.
[9] Rotter JM, Teng JG, Li HY. Buckling in thin elastic cylinders on column supports. In: International
Colloquium on Buckling of Shell Structures on Land, in the Sea and in the Air, 1991 Sept 1719;
Villeurbanne, Lyon, France, 1991:33443.
[10] She KM, Rotter JM. Nonlinear and stability behaviour of discretely-supported cylinders. In:
Research Report 93-01. UK: Department of Civil Engineering, University of Edinburgh, 1993.
[11] Rotter JM, Greiner R, Guggenberger W, Li HY, She KM. Proposed design rule for buckling strength
assessment of cylindrical shells under local axial loads. ECCS TWG8.4 Buckling of Shells, Sep-
tember 1993; submitted.
[12] O ry H, Reimerdes HG, Tritsch W. Beitrag zur Bemessung der Schalen von Metallsilos. A contri-
bution to the layout of large metallic silo-shells. Internal unnumbered report 1984. Stahlbau.
1984;53(8):24348.
[13] Greiner R. Zur Langskrafteinleitung in stehende zylindrische Behalter aus Stahl. Stahlbau
1984;53(7):2105.
[14] Rotter JM. Analysis and design of ringbeams. In: Rotter JM, editor. Design of steel bins for the
storage of bulk solids. Australia: University of Sydney, 1985:16483.
[15] O ry H, Reimerdes HG. Stresses in and stability of thin walled shells under non-ideal load distri-
bution. In: Proceedings of the International Colloquium on the Stability of Plate and Shell Structures,
1987 Apr 68; Gent, Belgium. Brussels, 1987:55561.
[16] Abaqus. Abaqus theory manual. Users manual. version 4.8 Providence (RI): Hibbitt, Karlsson and
Sorensen Inc, 1989.
[17] Guggenberger W, Greiner R. Local buckling of unstiffened steel cylinders numerical results and
 W. Guggenberger et al. / Journal of Constructional Steel Research 56 (2000) 175197 197

comparison with test results. In: ECSC Research Report G/G No. 1.1. Austria: Institute for Steel,
Timber and Shell Structures, Technical University of Graz, 1994.
[18] Guggenberger W, Greiner R. Numerical analysis of unstiffened steel cylinders with artificial imper-
fections and comparison with test results. In: ECSC Research Report G/G No. 2.A. Austria: Institute
for Steel, Timber and Shell Structures, Technical University of Graz, 1994.
[19] Rathe J, Greiner R. Local loads in cylindrical structures. Subproject B, ECSC Contract No. 7210-
SA/208: Enhancement of ECCS design recommendations and development of Eurocode 3, parts
related to shell buckling, Final Report. Belgium: University of Gent, May 1996.
[20] Timoshenko SP, Gere JM. Theory of elastic stability. 2nd ed New York: McGraw-Hill, 1961.
[21] Yamaki N. Elastic stability of circular cylindrical shells. Amsterdam: North Holland/Elsevier Applied
Science, 1984.
[22] Donnell LH, Wan CC. Effect of imperfections on buckling of thin cylinders and columns under
axial compression. J Appl Mech, Trans ASME 1950;17(1):7383.
[23] Hoff NJ. The perplexing behaviour of thin circular cylindrical shells in axial compression. Israel J
Technol 1960;4(1):128.
[24] Thielemann FW, Esslinger ME. On the postbuckling equilibrium and stability of thin-walled circular
cylinders under axial compression. In: Proceedings of the 2nd IUTAM Symposium on the Theory
of Thin Shells, Sept 1967; Copenhagen. Berlin: Springer, 1968.
[25] Esslinger M. Hochgeschwindigkeitsaufnahmen vom Beulvorgang dunnwandiger axialbelasteter Zyl-
inder. Der Stahlbau 1970;3:736.
[26] Hillmann J. Grenzlasten und Tragverhalten axial gestauchter Kreiszylinderschalen im Vor- und
Nachbeulbereich. Bericht Nr 85/45. Germany: Institut fur Statik, Technical University of
Braunschweig, 1985.
[27] DIN 18800. Stahlbauten: stabilitatsfalle, schalenbeulen. DIN 18800 Part 4. Berlin: Deutsches Institut
fur Normung, November 1990.