fat JP TS 3) t y $

133! Sty pari 1 eit +1

: bins ated st : ness Rea
i xt } ‘ aT) i °

int haat et Mara aty ql i pears)

f ue ateParatty ig ;
rH ti PPA abtetat ine teaattaesy |4 {
} 14 sh i i As? t i +

Behaviour of
Thin-Walled
Structures

| | Edited by |
| nN picchmed

14, SPENCE}, Le

ELSEVIER APPLIED SCIENCE PUBLISHERS

Meena ata Oe aE HOSE LURE poet cer a EH a
 Digitized by the Internet Archive
in 2023 with funding from
Kahle/Austin Foundation

https://archive.org/details/behaviourofthinw0000unse
 BEHAVIOUR OF
THIN-WALLED STRUCTURES
 FRONTISPIECE

James McQuistan Harvey obtained his PhD degree from the University of Glasgow
in 1953 for research in the field of cold formed steel sections. After five years spent in
the cold formed steel industry, he returned to academic life in 1958 as a senior
lecturer at the University of Strathclyde (then the Royal College of Science and
Technology), Glasgow. He later became Reader in Mechanics of Materials and in
1969 was appointed the first Trades House of Glasgow Professor of Mechanics of
Materials. He occupied this position until his early retirement in September 1982
and for the last five years was also Deputy Principal of the University of
Strathclyde.
Throughout his academic life Professor Harvey was involved in thin-walled
structures research and played a significant role in the area of cold formed sections
at both national and international level. In the light of this involvement it is fitting
that his retirement should be marked by a Conference devoted to the behaviour of
thin-walled structures.
 BEHAVIOUR OF
THIN-WALLED
STRUCTURES

Edited by

J. RHODES

and

J. SPENCE
Department of Mechanics of Materials,
University of Strathclyde, Glasgow, UK

ELSEVIER APPLIED SCIENCE PUBLISHERS

LONDON and NEW YORK
 ELSEVIER APPLIED SCIENCE PUBLISHERS LTD
Ripple Road, Barking, Essex, England

Sole Distributor in the USA and Canada

ELSEVIER SCIENCE PUBLISHING CO., INC.
52, Vanderbilt Avenue, New York, NY 10017, USA

British Library Cataloguing in Publication Data

Behaviour of thin-walled structures.

1. Thin-walled structures—Congresses
I. Rhodes, J. II. Spence, J.
624.1'77 TA660.T5

ISBN 0-85334-246-6

WITH 14 TABLES AND 212 ILLUSTRATIONS

© ELSEVIER APPLIED SCIENCE PUBLISHERS LTD 1984

The selection and presentation of material and the opinions expressed in

this publication are the sole responsibility of the authors concerned

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the copyright owner,
Elsevier Applied Science Publishers Ltd, Ripple Road, Barking, Essex, England

Printed in Great Britain by Galliard (Printers) Ltd, Great Yarmouth

 Preface

This volume contains the papers presented at the Retiral Conference for
Professor James M. Harvey, held at the University of Strathclyde on 29-30
March 1983 and organised by the Department of Mechanics of Materials.
Attendance at the Conference was by invitation and was limited to about 50
delegates, all specialists in the thin-walled structures field.
The Conference was sponsored by:

The University of Strathclyde

The Cold Rolled Sections Association
Conoco (UK) Ltd

We should like to acknowledge our indebtedness to these organisations for

their assistance which helped make the Conference highly successful from
both technical and social points of view. We should also like to record our
gratitude to the various members of the Department who devoted a great
deal of time and effort to make the Conference run smoothly.
The years since 1969, when James Harvey first became Trades House
Professor, have seen a considerable diversification in the field of thin-walled
structures, both in the types of structure involved and in the design and
analysis philosophies adopted to deal with these structures.
Many of the problems encountered in thin-walled structures arise
because ofthe effects of local buckling. The knowledge of various facets of
this phenomenon has increased dramatically since the 1960s. Problem
areas which were hitherto either too complex for rigorous analysis or whose
Vv
vi Preface

subtleties were not fully realised have in these years been subjected to
intensive study. As a result, great advances have been made in the areas of
elasto-plastic buckling and post-buckling behaviour, interaction of local
buckling with other forms of buckling, and the effects of imperfections on
the behaviour of plate and shell structures.
Also, during this time the trend towards lighter, thinner structures has
led to the desire to use extremely light gauge metal, for example in roof
sheeting, and has necessitated a substantial amount of research and
appraisal of various aspects of design at loads far beyond local buckling.
The growth in use oflightweight strong materials, such as fibre-reinforced
plastics, has also been a contributory factor towards this need for advances
in the knowledge of the far post-buckling range.
New problems in the field of buckling have also been posed with the
emergence of offshore oil platforms. Problem sources have arisen both in
the platform structures and in the associated subsea pipelines. This has
resulted in significant research input in the area of buckling and collapse of
stiffened cylinders, and on the various problems of pipeline buckling and
buckle propagation in pipelines.
The present papers cover some aspects of all of the above problems.
Twenty papers deal with the general topics of: compression members,
column behaviour, cold formed sections, lightweight structures, shells and
pressure vessels and plates. Of these, nine are devoted to new applied
research within these topics, five present surveys of theoretical, experimen-
tal and industrial aspects of past research and six are concerned with design
implications.
We hope that readers will find the papers interesting and informative.

J. RHODES
J. SPENCE
 Contents

MedCLIC AeWe Hee aD c 5 AC, Meets ecb ates ae oss, Sod I ats

Bulbswmipsmnd pcade unc: 6c a Geto Ww umnl eas

P. S. BuLson (Military Vehicles and Engineering
Establishment, Christchurch, Dorset BH23 2BB, UK)

Promoting the Use of Thin-Walled Hollow Rectangular

Sections by Increasing Steel Grade..... Reees ees ee 21
R. Maquol, C. MASSONNET and J. RONDAL (Jnstitut du
Génie Civil, Université de Liege, 4 Quai Banning, 400B
Liége, Belgium)

. Amplitude Modulation of Short-Wave Buckling Modes . . 35

W. T. Koirer (Laboratory for Engineering Mechanics,
Delft University of Technology, Mekelweg 2, PO Box
5033, 2600GA Delft, The Netherlands)

. Locally Imperfect Plain Channel Columns............. 47

J. LOUGHLAN (College of Aeronautics, Cranfield Institute
of Technology, Cranfield, Bedford MK430AL, UK)
and A.R. UpapuyaA (Structures Division, National
Aeronautical Laboratory, Bangalore 560017, India)
vil
Vili Contents

An Examination of Shakeout in Stiffened Plates........ 69

P. A. Frieze and E. DryMakis (Department of Naval
Architecture and Ocean Engineering, University of
Glasgow, Glasgow G12 8QQ, UK)

he Cold, Formed Industryescene eee

ee oe 93
R. D. Homer (Metal Sections Ltd, Oldbury, West
Midlands B69 4HE, UK)

UMR Research on Cold Formed Steel Structures ....... 111

WEI-WEN YU (Department of Civil Engineering,
University of Missouri—Rolla, Rolla, Missouri 65401,
USA)

Some Thoughts on Future Cold Formed Steel Design Rules 125

J. RHODES (Department of Mechanics of Materials,
University of Strathclyde, Glasgow G1 1XJ, UK)

Design of Profiled Steel Sheeting and Decking.......... 143

E. R. BRyAN and J. M. Davies (Department of Civil
Engineering, University of Salford, Salford MS 4WT,
UK)

The Instability of Composite Structures with Initial

JINPCRLECEIONSR Genie 25 Bain, Pe eee ae ee, ee ee 165
I. H. MArsHALL (Department of Mechanical and
Production Engineering, Paisley College of
Technology, Paisley PAI 2BE, UK) and W. M. Banks
(Department of Mechanics of Materials, University of
Strathclyde, Glasgow G1 1XJ, UK)

1 The Background to Forthcoming Design Proposals for Two

Shell. Buckling ?Problem sae: unwiani k er ee ee bz
G. D. GALLETLY (Department ofMechanical Engineering,
University of Liverpool, Liverpool L69 3BX, UK)

125 Axisymmetric Collapse of Cylinders Including

Deformations of Ring Stitlenimga. 2p eee) eee noes 211
J. G. A. CroLt_ (Department of Civil and Municipal
Engineering, University College London, London
WCIE 6BT, UK)
 Contents

Iya Local Loads on Cylindrical Vessels: A Fourier Series

SOMMON Sree tee ee ee Wh ey PALE | oy Pipa
G. Dutuik (British Gas, Midlands Research Station,
Solihull B91 2JW, UK) and A. 8. Tootu (Department
of Mechanics of Materials, University of Strathclyde,
Glasgow G1 1XJ, UK)

14. In-Plane Bending Behaviour of Flanged Pipe Bends under

SICADVa State ENCED eerie prep e nn ateaar ang sient IAG
G. THOMSON (Ferranti PLC, Robertson Avenue,
Edinburgh, UK)

15. Plastic Design of Submarine Pipelines’................ 287

J. W. B. STARK and P. E. DE WINTER (Jnstitute TNO for
Building Materials and Building Structures, PO Box 49,
2600AA Delft, The Netherlands)

16. The Inelastic Response of Pipes under External Pressure

Aid DENGINGe teem eee oe ee en ones ems ys
C. J. Tay (National University of Singapore, Singapore),
W. J. M. STEEL (Britoil PLC, Glasgow, UK) and J.
SPENCE (Department of Mechanics of Materials, Univer-
sity of Strathclyde, Glasgow G1 1XJ, UK)

EE The Strength of Plates Subjected to Biaxial Forces........ o29

A. F. Drier and P. J. DOWLING (Department of Civil
Engineering, Imperial College of Science and Technology,
Prince Consort Road, London SW7, UK)

18. Elastic Buckling of Stiffened Steel Plates of High Aspect

Raromnderm Uniaxial Compression. : 0 = ..42. sea nwas 255
W. Katzer and N. W. Murray (Department of Civil
Engineering, Monash University, Clayton, Victoria
3168, Australia)

A Brief Review of Plate Buckling Research ............ 375

A. C. WALKER (Department of Mechanical Engineering,
University of Surrey, Guildford, Surrey GU25XH,
UK)
x Contents

20. Buckling of Aluminium Plates in Compression ......... 399

D. S. Morrtin and J. B. Dwicnut (Engineering
Department, University of Cambridge, Cambridge
CB2IPZ, UK)
 Bulbs, Lips and Beads

P. S. BULSON
Military Vehicles and Engineering Establishment,
Christchurch, Dorset,
and University of Southampton, UK

ABSTRACT

The history of research and analysis to establish the behaviour of plates and
sections reinforced by bulb and lip stiffeners along free edges is reviewed.
Simple design formulae are given to establish the elastic critical stress in
uniform compression. The treatment of ultimate compressive strength is
discussed with particular reference to tests on aluminium alloy extrusions,
and the paper ends with random thoughts on the research/design interface.

1. INTRODUCTION

The author’s first experience of thin-walled structures was as the recipient

of a research grant to pursue the experimental and theoretical aspects of
local buckling, with particular reference to the effect of edge reinforcements
on the flanges of aluminium alloy extruded sections.
At the time, the early 1950s, the practice in cold formed steel sections was
to turn the edges of the flanges inwards or outwards to form ‘lips’. In
aluminium alloy extrusions, however, there was much more scope, for it
was possible to design edge reinforcement of almost any imaginable shape.
Circular bulbs, semi-circular beads or rectangular lips could be formed,
and it was even possible to stiffen the edges of reinforcing lips with further
reinforcing lips to form stiffened stiffeners!
There were two analytical difficulties in establishing the elastic local
1
D) P. S. Bulson

buckling stress of a lipped flange. The first was the problem of

asymmetrically placed reinforcement, in which the neutral axis of bending
of the reinforcement did not coincide with the neutral axis of bending of the
parent flange. The second was that the complex determinantal solution that
took account of the interaction between the flange and the reinforcement
was of academic interest, but of little use to designers.
Two ways of overcoming the problem were evolved with regard to
buckling. The first was merely to specify the size of the lip or bulb that
would be sufficiently stiff to offer simple support to the edge of the flange, a
method used in the cold formed steel codes of practice. The second was to
establish a reasonably accurate approximation that was simple to apply,
and would give the buckling stress of any flange/bulb combination. This
method was used in evolving the design rules for the UK code of practice
for structural aluminium at the end of the 1960s. The size of the edge
reinforcements also influenced the behaviour of open sections subject to
torsional instability, and design codes gave rapid ways of calculating the
torsion constants of a wide range of cross-sectional geometries.
Matters rested there, in Britain at least, until the adoption of limit state
design, and the consequent rise in the importance of ultimate compressive
strength of thin-walled sections. It was not acceptable to assume that the
edge reinforcement required to ensure elastic buckling in a simple-
supported mode would be sufficient to ensure continuity of this mode until
collapse occurred. This had been demonstrated clearly in experiments on
the shear buckling of stiffened plate girder webs, where it was necessary to
consider the failure mechanism in order to establish ultimate strengths. The
change in design philosophy meant a review was needed of the influence of
stiffening members on collapse strength, and this is still in progress in
connection with the writing of new design codes for steel and aluminium
construction.
The purpose of this paper is to rewalk the paths of research of the past
thirty years and to examine existing design rules with particular reference to
lipped sections in uniform compression.

2. ELASTIC BUCKLING

2.1. Lipped Flanges

The elastic buckling of thin flat plates with edge reinforcement, or
having edges ‘supported by elastic beams’, was first examined in Russia
 Bulbs, Lips and Beads 3

by Casilev’ in 1914, but his work was not summarised or made widely
available to English-speaking structural engineers until the publication of
Timoshenko’s monograph? on the theory of elastic stability in 1936. It was
also in this book that the results of analytical studies by A.J. Miles of the
University of Michigan were presented in graphical form, and these gave a
first insight into the fundamentals of the problem to many a designer and
research student. Timoshenko also reported that the problem had been
discussed in the early 1930s by Melan? and Rendulic.*
The designers of aircraft had begun to use edge reinforced sections
before this theoretical work was reported, but behaviour in compression
was almost entirely judged from the results of large numbers of strut tests.
Thus we find Sechler and Dunn? in their 1942 book on airplane structural
analysis and design stating that ‘for sections having bulbs no rational
methods of analysis are available, and recourse must be had to testing
enough lengths of the section to determine the coefficients in the short
column equation’. They suggested that design methods which assumed that
the bulb supplied simple support to the edge of the sheet should be used
with caution.
When rational theories became available in the 1950s and 1960s, it was
shown that thin plates in compression having one longitudinal edge simply
supported and the other reinforced by a symmetrically placed stiffener
equal in thickness to the plate behaved as indicated in Fig. |. The critical
stress in end compression, o,,, 1s given by the familiar equation

2 kn?E Ne
Per 19(1 — v2) \b (1)
where k is the buckling coefficient, ¢ is plate thickness, b is plate width, and
E and vy are the elastic modulus and Poisson’s ratio respectively. If the
length of the plate =a, and the width of the reinforcement = b,, Fig. 1
shows the relationship between k and a/b for various values of b, /b. The
analysis assumes that when buckling occurs the simply supported
longitudinal edge remains straight, whereas the edge between the stiffener
and the plate bends outwards. However, if b, is made wide enough, the
section is transformed into three thin component plates, and at buckling
the junction between the plates will remain sensibly straight. In fact, there is
a gradual transition from one mode to the other as b, grows wider, but let
us for the moment concentrate on the outward bending mode, associated
with lips having a relatively small cross-sectional area compared to the
component plate.
4 P. S. Bulson

a/b
Fig. 1. Symmetric lips on hinged flanges.

Two important features are apparent in Fig. 1:

(a) Long plates simply supported on one longitudinal edge buckle into
only a single longitudinal half wave. This is true whether the other
edge is free, or whether it is supported by a small lip.
(b) As the ratio a/b increases, the value of & for a lipped plate falls
below that for a plate with a free edge. At the value of a/b where the
buckling coefficients are equal, the critical stress of the lip as an
isolated strut constrained to bend in a plane normal to the plate is
equal to that of a plate with a free edge.

If the plate is fixed along one longitudinal edge, and reinforced by a lip
along the other, the picture changes somewhat, and a typical solution for a
lip equal in thickness to the plate is shown in Fig. 2. The curves linking k
and a/b now take the familiar ‘garland’ form, with &k reaching a minimum
value (k,,:,) aS a/b increases. We can make the following observations:

(c) Adding asmalllipcancause k,,;, to fall slightly below its value for a
plate fixed along one edge, free along the other.
 Bulbs, Lips and Beads 5

Fig. 2. Symmetric lips on built-in flanges.

(d) As the width of the lip is increased, k,,,,, approaches its value for a
plate fixed along one edge, simply supported along the other. When
this value is reached the buckling mode changes to several
longitudinal half waves.

In the calculations leading to Figs. 1 and 2 the torsional stiffness ofthe lip
was neglected, and the error in so doing is small. The author®’’ has shown
that it is about 5%. A close approximation to the value of k,,;,, in Fig. 2 can
be obtained by employing the energy method of solution, and examples are
given in ref. 6. The parameters that influence the value of k,,;, for plates
fixed along one edge are EJ/bD and A/bt, where E/ is the flexural rigidity of
the lip bending in its own plane, D is the unit flexural stiffness of the plate,
and A is the cross-sectional area of the lip. The variation ofk,,;, for a range
of values of EJ/bD and A/bt is shown in Fig. 3.
The critical stress of a uniform thickness thin-walled channel or I section
can be expressed in terms of a buckling coefficient applied to the web plate.
6 P. S. Bulson

EI/bD
Fig. 3. Variation of k,,;, with area and stiffness parameters.

byt s 50

0 02 0.4 io ee 10
b,/y
Fig. 4. Buckling coefficients for lipped channel sections of uniform thickness.
 Bulbs, Lips and Beads 7

Thus,

wE fave
Gor = (nin) 12(1 p= vy?) (5) (2)

where b,, is the width ofthe web and f its thickness. If the flange width = b,,
it can be shown that for a channel of uniform thickness (k,,;,,), Varies with
b,/b,, according to the dashed curve in Fig. 4. If a symmetric lip, equal in
thickness to the flange, is added at the extremity of each flange, the value of
(K min)w Will increase by a factor («) that is governed by the ratio 5, /b,,. The
exact solution for a lipped channel is also shown in Fig. 4, which is drawn
for a section with 6, /t = 50.
In the range 0 < b, /b,, <0.25, the lip factor («) can be expressed quite
accurately by the approximate relationship
b 2
a=1+40 (Fe) (3)
b,
as indicated in Fig. 5. For lipped channel sections with 6,t= 20, the factor
is
b 2
=1+8(—
a=1+ (5) 4
(4)

For the purposes ofinitial design, a simpler relationship can be formulated,

applicable to both channel and I sections with webDeemer in the range
20 <b ft = 90, This 1s

iS
ee a ==
(5)
60 \ t
so that, in general,
n?E pave
O oy
he)
( ee
A. Ot 12(1 =) (+)
6
( )

applicable in the range 0 < b, /b, <0.25, 20 < b,,/t < 50.
So far, the discussion has been limited to symmetrical lips, but most
practical sections, whether rolled, folded or extruded, have the reinforce-
ment located on one side of the plate, to form inward or outward turning
lips. A method ofdealing with this was outlined by the author in ref. 8, and
shown to give reasonable agreement with tests by Dwight,” but a simpler
and equally accurate procedure is to replace Jin the expression EJ/bD by an
effective value equal to double the moment ofinertia of the symmetrical lip.
8 P. S. Bulson

3-0

[Keir ty

ZO

ee)

ro) 005 al OIG O02 oOv5

Fig. 5. Comparison of exact and approximate values of (k,,;,).-

This is equivalent to using an effective value of b, equal to ./2 x the

measured value in eqns (3), (4) and (5). Equation (5) now becomes, to a near
approximation,
to hbe
~1+—|(—
i +5 (*) )7
If the ends of the lips have reinforcing beads, as shown in Fig. 6, the
simplest method of taking these into account is to calculate a new value for
b, that makes the flexural stiffness of the beaded lip equal to that ofa plain
lip having the same thickness as the flange.

2.2. Bulbed Flanges

Torsional stiffness becomes important for edge stiffeners in the form of
circular bulbs or heavy square and triangular reinforcements, and to
neglect it would lead to over-conservative design. Methods of taking
torsional stiffness into account are given in refs. 6 and 7.
If a single, long plate is fixed along one longitudinal edge and reinforced
by a symmetrically placed circular bulb, diameter d, along the other, values
 Bulbs, Lips and Beads 9

Fig. 6. Beaded lips on a thin-walled channel section.

of k,;, can be accurately computed, and the results for a plate having
b/t = 20 are given in Fig. 7. The effect of neglecting torsion is shown by the
dashed curve, and this indicates that at higher values of d/b, k,...min would be
reduced by almost 50%.
Thin-walled channel and I sections reinforced by heavy bulbs have been
examined in refs.6 and 7, and the results for a bulbed channel with
b,,/t = 20 are given in Fig. 8. For uniform sections with 20 < b,/t < 50 and
0 <d/t <6, it is approximately true to use the relationship
ipa

although it gives somewhat conservative values as b,/t approaches 50.

For bulbs attached to one side of the plate or flange, the procedures for
thin lips may be followed. The equivalent value of dis about 1.2 x the
measured value, so we can write, for unsymmetric bulbs,

Lac
=1+—|- 9
" 20 (“) 0)

2.3. Design Practice

The treatment of local buckling in most design codes calls at some stage for
a calculation of the elastic critical stress of thin-walled sections. The critical
stress is a function of the geometry of the complete section, and can be
determined by taking account of the complete interaction at buckling of all
elements of the section. The analysis of this interaction is forbidding, but
was undertaken by Chilver'® for sections such as channels and top-hats
that could be formed in bending a single sheet of thin material, and by the
author’ for I and box sections, in the early 1950s. The effects of reinforcing
lips were taken into account, and the results were presented in a simple way,
 P. S, Bulson

ed _—_—
_—
—
=
<i ——- _

NEGLECTING TORSIONAL
STIFENESS OF BULB

nee ol A/b Or 03

(+)
Fig. 7. Buckling coefficient for bulbed flanges.
 Bulbs, Lips and Beads 11

b,/t = 20

be/b
Fo -W
Fig. 8. Buckling coefficients for bulbed channel sections of uniform thickness.

usually in the form of graphs linking the buckling coefficient for one of the
component plates in terms of geometrical ratios of plate widths, thicknesses
and lip dimensions. However, the writers of codes of practice for structures
other than aircraft have, since those days, always considered that the
graphs were unnecessarily complex, particularly for initial design
calculations. At best, a few were relegated to the appendices to interest
those who wished to make academically accurate calculations of critical
stress.
A slight simplification was proposed by Rhodes and Harvey'! in the
early 1970s, who devised critical loads by considering the stability of
component plates with rotational restraint along their longitudinal edges,
and matching these restraints at the common junctions. However, as
Walker’? has pointed out in a book that indicated the sources ofthe design
recommendations for cold formed steel sections in building (British
Standard 449 Addendum No. 1) 1974, the calculation of o,,, and hence
12 P. S. Bulson

maximum stress, can be approximated by synthesising the effectiveness of

the component plates either as ‘stiffened’ (simply supported on both
longitudinal edges) or ‘unstiffened’ (simply supported along one edge, free
along the other). By specifying minimum dimensions for a reinforcing lip,
flanges supported by lips were taken as ‘stiffened’. It was considered that a
lip greater in width than 5,/5, where 6, is flange width, was sufficiently
strong to restrain the flange—lip junction from lateral deflection at buckling,
and if the lip was less than 5,./5, the edge of the plate was considered to be
free. The ratio b,./5 was settled on from test information.
As the lip is increased in width beyond 5,,/5 it eventually behaves as a flat
plate in the ‘unstiffened’ state, and if the lip width, 5, ,gets much beyond 10t
this can cause a reduction in the overall critical stress of the section. In the
UK, lipped flanges in cold rolled steel were normally designed to be greater
than 5,/5, and where possible less than 10¢ in width.
This ‘broad brush’ treatment had been refined somewhat when the code
of practice for structural aluminium (CP 118) was written at the end of the
1960s. An attempt was made to take the effect of the lip into account when it
was not large enough to convert the flange into a ‘stiffened’ element. The
code was written in terms of ‘effective slenderness ratios’ in its treatment of
plate buckling, so that the buckling coefficient was not k from eqn (1), but
m, where

2
_ 12(1—v?)
ay? (10)

Using this approach, the design clause for a thin flange reinforced with a lip
of the same thickness as the flange reads as follows:
The value of mb/t to be used is the largest of those obtained by

(1) calculation for the flange as a web element, where m = 1.6,

(2) calculation for the lip as a flange element, where m=5S.1,
(3) calculation for the flange—lip combination, where m= 5.1 x
[1 — (62/8072).
The clause for a thin flange reinforced with an unsymmetric bulb was:
The value of mb/t to be used is the larger of those obtained by

(1) calculation for the flange as a web element, where m = 1.6,

(2) calculation for the flange—bulb combination, where m = 5.1 x
[1 — (d?/25617)].
 Bulbs, Lips and Beads 13

The expressions for the increase in buckling coefficient to allow for the
flange—lip combination were similar to those given in eqns (7) and (9), but
they were adjusted to allow for the fact that m decreases as k increases, in
proportion to syle. They were also slightly more conservative than eqns (7)
and (9). Further, in order to get a meaningful set of design rules,
it was assumed that the minimum critical stress of a plate simply
supported along one edge, lipped along the other, would increase as the lip
was increased in size, whereas we know from Fig. 1 that the buckling
coefficient for long hinged plates does not increase when the plate is
reinforced by an edge stiffener. The ‘lip factor’ [1 — (b?7/80t7)], which
should strictly be applied to values of m corresponding to plates with edge
restraint along the hinged edge, has been applied to m = 5.1, which is the
value for a plate simply supported along this edge, free on the other. This
notion allows the idea of ‘web’ elements and ‘flange’ elements to be used,
and at the same time takes account of the increase in elastic buckling stress
of the section due to the presence of a lip.
In the revision of CP 118, taking place at the time of writing, there is a
possibility that the critical stress of web or flange elements will be calculated
on the basis of an increase in the plate buckling coefficient, k, using eqns (7)
and (9). The precise nature of the design clauses is still under discussion.
For the flange—lip combination, it is interesting to judge the value of 5, /t
that will produce a buckling coefficient value for the flange equal to that for
a simply supported plate. As an example, consider a channel with equal
flange and web widths. The value of A, is 0.9, and the flange-lip
combination to produce a ‘stiffened’ flange is given by
2 b
0.9 make fu =4 so that= 11.7
40 \ ¢t t

For a flange with b,,/t = 50, the value of 5, /b, is just over 1/5, which agrees
with the figure used in the design of cold formed steel sections. The required
value, however, varies with the 5,./t ratio of the flange, and with b,/b,, for
the section.

3. ULTIMATE COMPRESSIVE STRENGTH

3.1. Lipped Cruciform Sections

Tests by the author on aluminium alloy cruciform sections with
asymmetric lips, carried out many years ago, were publicised at the time in
a research report of limited distribution, and it is worth examining the
14 P. S. Bulson

386

COMPKESSIVE
MAXIMUM
(N/mm?)
STRECS

re) a5 0:30 0-45 0: 60

LENGTH

Fig. 9. Ultimate compressive strength of lipped cruciform sections in high

strength aluminium alloy.

results again. The research supported a move to standardise a series of

lipped I sections that were more structurally efficient than extrusions based
on conventional joists with heavy flanges. Lipped flanges having b,./t ratios
between 12 and 20 were proposed, with lip dimensions varying from
b,/bg = 0 to 0.5. The results for cruciform extrusions in a high strength alloy
with a 0.2°% proof stress of 490 N/mm? are shown in Fig. 9, taken from
ref. 13. They show that for long sections compressed between parallel
platens, the maximum compressive stress was not increased by the flange
reinforcement. Since the roots of the flanges are effectively hinged, this
suggests that the conclusions drawn from Fig: | can be applied to ultimate
strengths as well as elastic buckling.
The ultimate strength of the cruciforms was nevertheless much higher
than their theoretical elastic buckling strength—about double. This
suggests a large measure of post buckling strength in asymmetrically
reinforced hinged flanges with b,/t ratios of 16.

3.2. Lipped I Sections

Figure 10 shows more test results by the author, taken from the same
 Bulbs, Lips and Beads 15

443
462

43)

400

370

(N
COMPLESSIVE
(mm?)
MAXIMUM
STRESS

. Fig. 10. Ultimate compressive strength of lipped I sections in high strength

aluminium alloy.

period. The presence of flange reinforcements on I sections raised the

maximum compressive stress by as much as 40% over that for a plain
section when b,/b, = 0.66 and b,/b, =0.25.
We know that the value of A, for the section shown is 0.76, and
applying eqn (7) gives a buckling coefficient for the lipped flange of
k,.(1 + (1/40)4*) = 0.76 x 1.4 = 1.06. The value of o,, for the section is then
given by
a 1.0677E 71 \2
Go eye) 13
and [oy >/o,,]'/? = 1.13, where ao , is the 0.2% proof stress of the alloy.
16 P. S. Bulson

Using the well-known relationship

1/3

2a.=0.66 (2) (11)

90.2 90.2

the maximum compressive stress (a,,) is calculated as 301 N/mm7?, which is

considerably lower than the measured value of about 464 N/mm’. In fact,
the lipped cruciform and lipped-I tests show that the expression for ¢,,/d9 >
is very conservative for accurate tests on high strength aluminium sections,
particularly if the b,/t ratios are low. Perhaps we should consider linking
our design curves to the manufacturing accuracy of component plates, as
well as to the ratio of the elastic critical stress to the proof stress of the
material.

3.3. Lipped Channel Sections

The test results for cold rolled steel lipped channels, originally reported by
Chilver,!* have been compared with proposed design curves by Walker.?
He shows that eqn(11l) and a relationship employing an assumed
imperfection parameter, ¢/t, both give a good indication of the lower
scatterband of the results. He also compared the maximum compressive
stress from the design curves with the approximate method of synthesising
the maximum stress from a knowledge of the collapse stress of the plate
elements of the section. In this example it was assumed as before that the
lips give simple support to the flanges as long as their width is greater than
one-fifth of the flange width. For the geometry chosen to illustrate the
technique (an inwardly lipped channel 100mm x 50mm x 3mm thick,
reinforced by lips 20mm in width) the approximate method gives a
maximum compressive stress about 9% higher than that given by the
design curves. Further, in this example, the approximate procedure seems
to be more complicated than the accurate method.
Thin-walled lipped channels have been tested more recently by
Thomasson,'° and his results were discussed by Rhodes!® in a paper in
1982. The work was particularly aimed at the interaction between local
collapse and overall column buckling in struts with plate width to thickness
ratios of the order of 150. These interactive effects can noticeably affect
column strength in very thin sections of high yield material. The analysis
for elastic critical stress used by Rhodes gives values of buckling coefficient,
k, that agree closely with values obtained earlier by Kloppel and Shiedel.!”
Rhodes goes on to explore the post buckling stiffness of single lipped plates
 Bulbs, Lips and Beads 17

(N/mm)
STRESS
FAILING

re) 30 6O Le)
SLENDERNESS RATIO
Fig. 11. Strut tests on a lipped I section in high strength aluminium alloy.

when this is influenced by the collapse of the lip in a torsional mode. The
reduction in axial stiffness is between 55 % and 65 %, depending on lip size,
and the reduction in flexural stiffness of the section is very severe—over
90 %. This has a significant effect on the column buckling stress of members
containing such elements. With regard to lipped channels having
asymmetrically stiffened flanges, Rhodes suggests that there is a high
sensitivity to local imperfections, particularly of the lip. He applies his
theory to lipped channels with 5, /t = 50 and symmetrical reinforcement
and shows, for example, that with a geometry given by b,/b, =0.6,
b, /b, =0.4, the reduction in axial stiffness at buckling is about 30%.
The distribution of compressive stresses around a lipped channel
immediately after buckling is given in ref. 18. The non-uniformity results in
a shift of the neutral axis in bending and a decrease in column buckling
load. The ultimate load relationship obtained in this way agrees reasonably
well with experiments conducted by Harvey’? of similar vintage to the
author’s tests on aluminium sections.
For sections having lower values of 6,,/t, the loss of buckling strength is
much less, and in a series of strut tests on lipped I sections, reported by the
author in ref. 13, the use of the maximum local buckling stress as the upper
18 P. S. Bulson

limit to a Perry-Robertson type of strut curve was shown to give close

agreement. A typical test result is given in Fig. 11.

4. DISCUSSION

This paper has not reviewed the problem of lipped or bulbed flanges
forming the compressive material of thin-walled beams subjected to
bending stress; neither has it explored the behaviour of large thin-walled
sections with edge reinforcements welded into position. Both of these
aspects are frequently met in design, and are important subjects.
Nevertheless, within the limited scope of uniform end compression of
folded, rolled or extruded sections, there are enough problems left to keep
research workers locked in discussion for some years.
In looking back three decades or so, the author has been aware of
features of the research/design interface that perhaps have a wider
background. The first is the timeless value of well conducted experiments.
Theories come and go and treatments in codes of practice are subject to
cycles of fashion, but good test results are indestructible. It is fitting, in view
of the purpose of this conference, that the penultimate of the references
quotes the PhD thesis of Professor Harvey, from 31 years ago; and perhaps
it is more than coincidence that the tests in support of the treatment of
lipped aluminium flanges come from the author’s postgraduate work ofthe
same period; and those in support of lipped cold rolled steel channels
quoted in ref. 12 were first recorded in the 1950 PhD thesis of Sir Henry
Chilver. Great emphasis was placed in those days on the importance of
productive experimental work as a vital part of engineering postgraduate
experience. In one of the research investigations mentioned above the
student machined, prepared and tested 377 specimens. There was no
productivity bonus, but a lifetime grasp of the fundamentals and difficulties
of structural testing was guaranteed.
The treatment of bulbs, lips and beads in design codes is a representative
illustration of the difficulties faced by code committees. If the author’s own
experience may be quoted, the static design of thin-walled, lipped
aluminium alloy sections for general structural purposes was carried out in
the 1950s and 1960s by using ref. 2, a collection of aeronautical data sheets,
a text book or two on aircraft structures, and a few reports issued by the
Royal Aircraft Establishment in this country and by the National Agencies
in the USA. As everything was written in a different notation and in
contrasting styles, the designer had to come to grips with the fundamentals
 Bulbs, Lips and Beads 19

to be sure that he understood the problem and could evolve his own design
procedures. Moreover, he was using the ‘allowable stresses’ design method
and limiting his needs to a knowledge of elastic critical stress and plasticity
reduction factors. Over the years he became very familiar with the plate
buckling coefficient, effective widths, and formulae like eqn (11).
In 1969 the first Code of Practice for the structural use of aluminium was
published by the British Standards Institution. It was a splendid attempt at
codification, but written in a new language. In the local buckling sections
there was no mention of the conventional plate buckling coefficient, k,
because the equivalent slenderness ratio was used to give a concise method
of using a small family of design curves for local, lateral, torsional and
column buckling. It was simple to use, but not easy for the designer to
retain the grip on fundamentals that he had developed in the preceding
years. The effect of lips and bulbs on buckling strength could be easily
calculated, but without thoroughly understanding the principles of
interaction between a plate element and its edge stiffener.
The 1969 Code will probably be replaced by another in the mid-1980s. It
will no doubt be a most useful document, but the language will have
changed again. The policy now is to use ‘limit state’ design and the
reinforcing lips and bulbs will be attached to ‘compact’, ‘semi-compact’ or
‘non-compact’ sections. The old-fashioned ‘plate buckling coefficient’ is
likely to be hidden away again in new, concise methods of summarising the
relationship between elastic buckling and ultimate strength. Since research
leaders in civil and structural engineering departments of the universities
are still rightly content to use the traditional approach to buckling
problems, as illustrated in refs. 16 and 18, designers find themselves having
to master several approaches to the same subject when they wish to practice
according to the code and at the same time keep abreast of progressive
research. In searching for simplicity we seem to complicate the divergence.
This results in those who can find time having to add to the strain of their
lives by writing books and articles in an attempt to bridge the gap.
The third feature of the research/design interface illustrated by the
subject of this paper was touched upon by Professor Harvey himself in
writing the introduction to the proceedings of a conference held at the
University of Strathclyde four years ago.”° He pointed out that work used
in design recommendations is like the visible part of an iceberg and that
designers should be aware of the vast unseen substructure of fundamental
work beneath. This is important because the lack of a clear understanding
of the physics of structural behaviour can lead to accidents. A quick review
of published work on lipped sections suggests that the properties of his
20 P. S. Bulson

iceberg are about 500 pages of research and analysis to 16 lines in a design
code. In wishing Professor Harvey a long and rewarding retirement we
naturally wonder whether he will find a similar relationship between
practice swings and match-winning iron shots.

REFERENCES

. CASILEV, K., Mem. Inst. Engrs. Ways of Commun., St Petersburg, 1914.

. TIMOSHENKO, S. P., Theory of elastic stability, New York, McGraw-Hill, 1936.
. MELAN, E., Repts. Intern. Congr. Appl. Mech., 3 (1930) 59.
. RENDULIC, L., Ingr.-Arch., 3 (1932) 447.
ABWN. SECHLER, E. E. and DunNn, L. G., Airplane structural analysis and design, New
York, Wiley, 1942, p. 194.
. Butson, P. S., ‘The local instability of structural sections with flange
reinforcements’, Thin-walled steel structures, London, Crosby Lockwood,
1969, pp. 81-101.
. Butson, P. S., The stability of flat plates, London, Chatto & Windus, 1970.
. BuLson, P. S., ‘Local stability and strength of structural sections’, Thin-walled
structures, London, Chatto & Windus, 1967.
. Dwicut, J. B., ‘Aluminium sections with lipped flanges and their resistance to
local buckling’, Symp. on Aluminium in Structural Engineering, London,
Aluminium Federation, June 1964.
. CHILVeER, A. H., Aeronautical Quarterly, 4 (1953) 251.
. RHopes, J. and Harvey, J. M., Int. Journal of Mechanical Science, 13 (1971)
787, 867.
. WALKER, A. C. (ed.), Design and analysis of cold formed sections, London,
Intertext, 1975.
. BuLson, P. S., ‘Local instability problems of light alioy struts’, Aluminium
Development Association Research Report No. 29, December 1955.
. CHILVER, A. H., The Engineer, August 7 (1953) 180.
. THOMASSON, P. O., ‘Thin-walled C-shaped panels in axial compression’,
Swedish Council for Building Research, Document DI: 1978, Stockholm.
. RHODES, J., ‘Buckling and failure of edge stiffened plates’, IUTAM Symposium
on Collapse, London, August 1982.
. KLoppet, V. K. and SCHIEDEL, E., Der Stalbau, No. 12, 1968.
. LOUGHLAN, J. and Ruopes, J., ‘Interaction buckling of lipped channel
columns’, Stability problems in engineering structures and components,
London, Applied Science Publishers, 1979.
. Harvey, J. M., Studies on the interaction of plate components of structural
sections under selected load conditions, Ph.D. Thesis, University of Glasgow,
1952.
20. Harvey, J. M., ‘Introduction’, Thin-walled structures, J. Rhodes and A. C.
Walker (eds), London, Granada, 1980.
 e

Promoting the Use of Thin-Walled Hollow

Rectangular Sections by Increasing Steel Grade

R. MAquol, C. MASSONNET and J. RONDAL

University of Liége, Belgium

ABSTRACT

On the basis of a computation method developed at Liége as the outcome of

an experimental research on the behaviour, in axial compression, of thin-
walled profiles with rectangular cross-section, the authors propose optimal
production ranges of this type of profile. Thin-walled tubes find, in the
optimal ranges, a place whose importance increases with increasing yield
stress of the steel.

NOTATION

Area of the cross-section.

Section width, measured between external forces.
3 Section width, measured between median planes of the walls.
Young’s modulus.
Collapse stress for local plate buckling.
Yield stress of the steel.
Section depth, measured between external faces.
SS
ty
ape
3 Section depth, measured between median planes of the walls.
Radius of gyration.
>>
~~ Buckling length.
Collapse load in axial compression.
21 Buckling reduction coefficient.
=
21
nNto R. Maquoi, C. Massonnet and J. Rondal

Reduction coefficient for interaction between wall and general

buckling.
Yield load in axial compression (squash load).
Wall local buckling collapse load in axial compression.
Shee 4c)
<sy
SetWall local buckling reduction coefficient.
=< Es Mean radius of the corner.
Wall thickness.
Parameter of the buckling curves.
Parameter of the wall buckling curves.
Usefulness parameter of thin-walled profiles.
Slenderness ratio.
Reduced slenderness.
SS
Le
ee
ot
Plate buckling reduced slenderness.
Reduced slenderness in wall—general buckling interaction.
Eulerian slenderness.
Buckling critical stress.
SS
CE >
te 3 Wall buckling critical stress.
a Mean collapse stress.
mS) Efficiency parameter.

1. INTRODUCTION

Because of the excellent distribution of the material around their

longitudinal axes, hollow profiles with square or rectangular cross-section
show a behaviour in axial compression which surpasses substantially that
of open profiles. Moreover, by comparison with round tubes, the presence
of four plane walls greatly simplifies the conception of the connections. For
these various reasons, hollow square or rectangular sections are structural
elements that are very appropriate for use as columns in buildings or as
chords, diagonals or posts in trusses.
Because of buckling, both moments of inertia must be as large as
possible. Economy requires one thus to promote profiles which, at constant
cross-section, have large cross-sectional dimensions b and h and small wall
thickness ¢. However, such a trend is rapidly restricted because of the local
buckling of the walls as soon as their thinness 6/t or h/t exceeds a certain
limit. The interaction between the local buckling of the walls and the overall
buckling of the compressed column generates a decrease of the collapse
load, that must be reflected in the design formulae.
In the course of a theoretical and experimental research sponsored by
 The Use of Thin-Walled Hollow Rectangular Sections 23

ECSC (European Coal and Steel Community) and CIDECT (Centre

International pour le Développement et lEtude de la Construction
Tabulaire), the authors have developed a design method for hollow profiles
with square or rectangular cross-section, which takes account of the
aforementioned interaction between both buckling modes.!~? This
method is based on the substitution, in the column buckling equations, of
the stub column collapse load Ny to the column squash load N,, in axial
compression. Figure 1 summarises the calculation process for the square
sections. For the rectangular sections, the procedure is hardly more
complicated; it is described in refs.2 and 3. Thus a design method is
available which is simple and accurate, for the theoretical results are in
good agreement with test results. In addition, it is, advantageously, fully
described by a set of explicit equations and, therefore, it is tempting to
exploit it systematically in order to analyse the optimal design of such
profiles.
In ref. 4, two of the authors have already shown the following, on the
basis of this method.

The optimal efficiency of a hollow profile with square section is never

reached by a profile with thick walls, which means a profile whose plate
buckling slenderness

peeenagy
ON
———
ES FU
f 1
(1)
is less than the 0.8 limit value as adopted by ECSC in its
Recommendations. In this definition, the width b,, of the profile is
measured between the median planes of the opposite walls and 1¢
represents the wall thickness; the yield stress is f, and E is Young’s
modulus.

The optimal efficiency corresponds to the limit plate buckling

slenderness A, = 0.8, if the condition below is fulfilled:
7 \ 3/4

JA\E
/is the column buckling length and A the cross-sectional area. The value
of parameter y depends on the manufacturing process of the section:
(a) hot finished profiles: Ve So 55
(b) cold finished profiles: y=1.014.
24 R. Maquoi, C. Massonnet
and J. Rondal

Interaction between local and

Local plate buckling overall buckling

= VFy/ocry = (b,/1.9t) y N

NSB 1/204) (VB CAROSS

ea,

Sarl He (N=
le Geen anak ]

= ot heen. = (Wai) vf JE

=iCU/22"2) 41 4a(a* =O0.2)2

“Ai ea(he = 0.2 )t eA

Manufacturing process

Hot finished profiles

Cold finished profiles

Notation:

A cross-sectional area 1 radius of gyration

N ultimate load yield stress
buckling length

Fig. 1. Design procedure for square hollow compression members.

 The Use of Thin-Walled Hollow Rectangular Sections 25

The optimal efficiency corresponds to a thin profile, which means

ae > 0-8, if:
i BNO A:

== (4) =y (3)

For rectangular sections, the optimal thickness must be computed as a

function of the width of the largest wall.
Though interesting, the above conclusions do not permit one to evaluate
the effective saving of material that could result from the use of thin walls.
Therefore the authors have established charts that give, as functions of the
cross-sectional width, the range of thickness values ensuring an efficiency
larger than 95% of the optimal efficiency.°

2. OPTIMAL RANGES OF HOLLOW SECTIONS WITH

SQUARE OR RECTANGULAR CROSS-SECTIONS

A range of hollow sections is completely defined by a ‘fan’ of thickness

values for each specified width 5. In order to guarantee a ‘nearly optimal’
character to the investigated ranges, the authors have used the following
briefly described procedure.

Choice of a discrete series of values of width b, for example those adopted

in the ISO Specifications (ISO 657/14 for hot finished sections and
ISO 4019 for cold finished sections).

Computation, for each specified width b, on the one hand, of the

optimum thickness /,,,, and, on the other hand, of the efficiency (that is
the ratio between the mean collapse stress o, of the section and the yield
stress f,) for various values of the thickness ¢ near to the optimal value.

Choice of lower and upper limit values of thickness 7, enabling one to

comply with the condition:
p =0.95popy (4)
which preserves the ‘nearly optimal’ character of the profile series.

Such a technique has produced sets of optimal ranges (Figs. 2-5). Figures 2
and 3 are devoted to hot finished profiles the yield stresses of which are
equal to 240 and 360 N/mm/ respectively, while Figs. 4 and 5 concern cold
26 R. Maquoi, C. Massonnet and J. Rondal

t(mm)

30

=10m

25
8m

6m
20

1SO_657/XIV

AON
10 Begin

b(mm)

Ie 100 200 300 400

Fig. 2. Hot finished profiles (f, = 240 N/mm’).

finished profiles with similar values of the yield stress. Calculations leading
to the aforementioned figures show that:

The optimal thickness corresponds ‘practically’ always to the limit plate

buckling thinness, 1.e.
I, = 08
The loss of efficiency is less pronounced with regard to the optimal
 The Use of Thin-Walled Hollow Rectangular Sections 24

t(mm)

100 200 300 400

Fig. 3. Hot finished profiles (f, = 360 N/mm’).

thickness for an increase of thickness than for a decrease; this explains

that the thickness ranges are substantially larger above the line 1, = 0.8
than below it (Figs. 2-5).

Figures 2 and 3 (hot finished profiles) present in addition the extreme

thickness values as recommended by ISO and those produced by the BSC
(British Steel Corporation—Tubes Division). If it is assumed that, for
obvious reasons, the width of a hollow section is generally chosen with
28 R. Maquoi, C. Massonnet and J. Rondal

t (mm)

30

E
S
25 E a

E
wo

20

i: os
5S
~—wt

D
15 E =
a
N Oo
vp)

/
be)
=| / /| ioe

| WYa

EEN vise
b(mm)

100 200 300 400

Fig. 4. Cold finished profiles (f, = 240 N/mm’).

respect to the column length, these figures demonstrate especially that:

The ISO range seems rather well calibrated, although the thicknesses
seem to be a little too large for small sections.

The BSC range is less well calibrated; it uses excessively large thickness
values for small sections; for large sections, it does not propose a
sufficiently wide fan of thicknesses, especially towards the large
thicknesses.
 The Use of Thin-Walled Hollow Rectangular Sections 29

t(mm)

30

E
25 L is)
Bh ae
i

ao

€
wo

20
S a
a
g
dy
155 Ae =
(je)
“

l= 2-4-6-8-10m b (mm)
0
100 200 300 400

Fig. 5. Cold finished profiles (f, = 360 N/mm’).

Figures 4 and 5 (cold finished sections) show that the ISO range is here less
well calibrated than for hot finished sections; indeed, the sections are too
thin for the large widths.
One should keep in mind that, in the range of the small sections, various
problems—such as the requirements for the connections, the technological
conditions due to fabrication and corrosion—require that the walls should
not be too thin. This justifies the manufacturing of sections whose wall
thickness exceeds the optimal value computed in axial compression.
30 R. Maquoi, C. Massonnet and J. Rondal

It has been stressed above that, in the vicinity of the optimal thickness,
the change of efficiency on the side of the large thicknesses is substantially
less than on the side of the small thicknesses. This remark enables a simple
extension of the results obtained for square hollow sections to the case of
rectangular ones. One has only to relate the choice of the optimal thickness
to the width of the largest wall and use directly the results obtained for
square tubes. This procedure yields only a very small error because, on the
one hand, the area of the narrow walls is always less than that of the wide
ones and, on the other hand, the efficiency of the narrow walls is only
slightly changed.

t (mm)

30

25

20

657/X1V
1S0
15

b(mm)

AS PMye 100 200 300 400

Fig. 6. Hot finished profiles (f, = 700 N/mm?).

 The Use of Thin-Walled Hollow Rectangular Sections 31

3. INFLUENCE OF THE YIELD STRESS ON THE USE OF

THIN-WALLED HOLLOW SECTIONS

From the consideration of Figs. 2—5, it is found that, for the common steel
grades of hollow sections, the thin-walled sections have a rather marginal
position in the optimal ranges. However, Figs. 6 and 7, which are drawn for
a yield stress of 700 N/mm7?, show that the situation becomes quite different
for higher steel grades.
Such a high value of the yield stress could look strange; it must, however,
be stressed that, in Japan, it is presently planned to fabricate welded boxes

t(mm)

30

E
ro)
u
25

20

x
15 ,

10+

Y {=2-4-6-8-10m
VE: 3b(mm)

: 100 200 300 400

Fig. 7. Cold finished profiles (f, = 700 N/mm?).
32 R. Maquoi, C. Massonnet and J. Rondal

made from rolled plates whose yield stress is nearly 700 N/mm? or even
greater.°
Progress recently achieved in steel metallurgy should soon enable one to
put on the market, with excellent economic conditions, plates with very
high yield stress that could be used in the fabrication of cold finished tubes.
For hot finished tubes, the situation does not favour such an important
increase of the yield stress; nevertheless, some manufacturers are already
able to roll H profiles the yield stress of which exceeds 500 N/mm?.
Such a viewpoint is also supported by the fact that the detrimental
influence of the residual stresses on the ultimate strength capacity in axial
compression decreases relatively when the yield stress increases. This is due
to the fact that the intensity of residual stresses increases less than in
proportion to the yield stress. Proposals in this way have already been
presented by one of the authors’ but would require a more extended
experimental confirmation.
The authors are convinced that the use of high steel grades may lead to
substantial savings, even when the buckling phenomenon is governing.

4. CONCLUSIONS

For the steel grades that are presently used in the fabrication of hollow
profiles, the thin-walled sections occupy only a marginal place in the
optimal range established in axial compression. Various factors will
apparently lead, in the near future, to an increase of the yield stress of rolled
plates, which constitute the basic material for hollow sections. As a
corollary, an increased use of thin-walled sections may be predicted.

REFERENGES

1. BRAHAM, M., RONDAL, J. and MAssonnetT, C., ‘Large size buckling tests on steel
columns with thin-walled rectangular hollow sections. Comparison with design
methods’, Thin-walled structures, J. Rhodes and A.C. Walker (eds), London,
Granada, 1980, pp. 3-13.
2. BRAHAM, M., GRIMAULT, J. P. and RONDAL, J., “Flambement des profils creux a
parois minces’, Commission des Communautés Européennes, rapport eur 6730 fr,
décembre 1979.
3. BRAHAM, M., GRIMAULT, J. P., MASSONNET, C., Mouty, J. and RoNDAL, J.,
‘Buckling of thin-walled hollow sections. Cases of axially loaded rectangular
sections’, Acier-Stahi-Steel, No. 1, 1980, pp. 30-6.-
 The Use of Thin-Walled Hollow Rectangular Sections 33

. RONDAL, J. and MaAquol, R., ‘On the optimum design of square hollow
compression members’, I[UTAM Symposium on Collapse, University College
London, 31 August—3 September 1982.
. RONDAL, J. and MAQuotl, R., Etude d’une gamme optimale de profils creux carreés
et rectangulaires, Annales de Il’Institut Technique de Batiment et des Travaux
Publics, No. 409, novembre 1982.
. Usami, T. and Fukumoto, Y., Local and overall buckling of welded box
columns, Journal of the Structural Division, ASCE, 108, No. ST 3 (March 1982),
525-42.
. Maaquol, R., ‘Some improvements to the buckling design of centrally loaded
columns’, Structural Stability Research Council, Annual Technical Session and
Meeting, New Orleans, 30-31 March, 1982.
 ¥

tere
 Amplitude Modulation of Short-Wave Buckling
Modes

W. T. KOITER
Delft University of Technology, The Netherlands

ABSTRACT

Amplitude modulation of short-wave local buckling modes has proved to bea

useful tool in the approximate analysis of anumber of buckling problems for
plates and shells. This concept seems to have been applied first in a
modification of Hutchinson's analysis of a complete spherical shell under
external pressure.*> More or less localised imperfections and the associated
buckling analysis have improved our understanding of the effect of a local
dent.'® Finally, the non-linear interaction between local sheet buckling and
overall buckling of stiffened plate and shell structures may be described
effectively as a modulation of the amplitude of the local short-wave mode due
to the long-wave overall mode.*"'!~ '3

1. INTRODUCTION

The concept of amplitude modulation of short-wave buckling modes seems

to have been employed first in a modification of Hutchinson’s ingenious
analysis of the buckling problem of a complete spherical shell under
external pressure.* In this case amplitude modulation served to continue
the deflection pattern analytically over the entire surface of the shell.°
Closely related to the amplitude modulation employed in the spherical
shell problem is the approximate analysis of the influence of more or less
localized imperfections.'*® The exponential modulating factor, eqn (19),
serves here to reduce a short-wave periodic pattern to a more or less
ep)
36 W. T. Koiter

localized dimple. The result of the approximate analysis is in this case that
more or less localized imperfections of the type considered here are equally
harmful as periodic imperfections with an amplitude reduced by a factor of
PIER
A different type of amplitude modulation of short-wave buckling modes
may occur in stiffened plate or shell structures due to the interaction of local
panel buckling with overall instability. The possibly disastrous conse-
quences of this interaction in some cases has been generally recognised
since Van der Neut’s pioneering paper.’’ The possibility of an approximate
analysis by means of amplitude modulation is here based on the property
that overall instability is characterised by a relatively large wavelength of
the deformation pattern.'!~ '*
Our discussion of amplitude modulation of short-wave buckling modes
in previous papers and also here is admittedly of a somewhat rough and
approximate character. It is essentially an application of the Rayleigh—Ritz
method, starting from assumed deflection patterns and neglecting ‘small’
terms in the subsequent evaluation of the energy. It may therefore be
worthwhile to note that a more mathematical approach, employing
multiple scaling in an asymptotic analysis, may serve as some justification
of our engineering analysis (cf. e.g. refs. 15, 16, 18).

2. APPROXIMATE DESCRIPTION OF SHORT-WAVE

BUCKLING MODES

Shallow or short-wave buckling modes are characterised by a minimum

wavelength L which is small in comparison with the smallest principal
radius of curvature R. In a region of linear dimensions of the order ./LR
(large in comparison with L and small in comparison with R) we introduce
orthogonal Cartesian coordinates x, y in a tangent plane in the direction of
the principal curvatures 1/R,, 1/R,. Let w denote the normal deflection.
The displacement components tangent to the lines of curvature are denoted
by u and v. The middle surface strains and the changes of curvature are
then given by the formulae®

Wee.
ii 9x eel DW Vy Vy,
ail ah Fe
a5 2W,)
R, 1 o

aye Sy tO Waele

Kies = Waxy Kyy = Woyy Kxy = Wyxy (1)

where subscripts x or y preceded by a comma denote partial differentiation

with respect to the coordinate in question.
 Amplitude Modulation of Short-Wave Buckling Modes 37

We restrict our attention to linear pre-buckling states in which the initial

stresses may be represented by o, = Ao, o, = Aoy and t,,, =At2,, where of
etc. are the stresses due to a unit load system and / is a non-dimensional
load factor. The increment in potential energy in the shell portion under
consideration, the energy functional, is then given by®
: : 4 Eh
Plu] = |] es Wat Op Wa 2 oh at oie92)
2 Pe
Ww Ww 4 WwW
«4(1. + v2) +(0,- 2 +403) + 2v («..- 2 +402)
1 2 1

x oy Wied
+40 eat +3(1 —v)(u,, +0, + W,.Wsy ,

Eh (2)
+ 2(1 — v)W,3y |dA
24(1 — v2) {Wy24 + Wyoy + 2VW,..Woyy»yy

This energy functional presupposes conservative external loads, but for the
shallow buckling modes under consideration it is immaterial whether the
loads are so-called dead loads or an external fluid pressure.’ According to
the general theory of elastic stability,*”’ initial imperfections Kw°(x, y),
where w° specifies the shape and x is a non-dimensional measure of the
magnitude, are taken into account to a first approximation by adding to
eqn (2) the integral

Q[u] = |Ah(olw,w,,. + TyWyWey + TeyWoxWoy + TryWsyW,x)dA (3)

The sign of the potential energy functional eqn (2) is governed primarily
by the quadratic terms. Equilibrium is at the stability limit, if this second
variation is semi-definite positive, i.e. ifithas a minimum value zero for one
or more non-vanishing displacement fields u, v, w, the so-called buckling
mode(s). These buckling modes satisfy the equations of neutral
equilibrium, in our present approximation given by®
] v
U,xx +3(1 2 a
yy
v)u, ry +3(1
12 Ar Ves,
y ~
ee
(ara
ike Woo =.0)

1 y
a(t a5 V)U, xy as ne om V Urey + Vy yy a ( an 7] Wsy = 0
: R, R 1 /

ae (3 +i) (; Ar Jo +(3 ay z +a)

12 Ran: Re uR en AR RER WR
—® (62Wge OPWyy+280) = 0 (4)
38 W. T. Koiter

where A is the two-dimensional Laplacian operator. In the derivation of

eqn (4) free use has been made of the approximation that the curvatures
1/R, and 1/R,, and the stresses of, a) and tY, in the pre-buckling state, may
be dealt with as constant factors in the case of short-wave buckling modes.
In the absence of pre-buckling shear stresses (t,,=0) eqns (4) have
doubly periodic solutions of the form

u = Asin px/R, cos qy/R,

v = Bcospx/R, sin qy/R,

w = Ccos px/R,cos qy/R, (5)

where at least one of the wave numbers p and q is large in shallow buckling
modes. The ratios A/C and B/C are obtained from the first pair of eqns (4):

2} Pauhel¢ zee a 2 Mbt.

Ota [ ee {oem R, He *
ia eI pa een as 2 phar?
BCS ep y E eke \o+y R, ae Maha)
and the condition for a non-vanishing value of C reads

h? v)
——,
4c?R3
(p* +q’)* + (p? +497)? + = (p?+9°)(pPopt+q?ae) (7)

where « = R,/R, and c = [3(1 — v’)]'””.

In a circular cylindrical shell of radius R we have R, = — Rand «=0. In
the case of axial compression with a unit load system o° = — Eh/cR the
critical load factor is unity. In a spherical shell we have Rj = R, = R, a =1
and in the case of external pressure with a unit load system of =o) =
— Eh/cR we have again a critical load factor unity.
A necessary condition for stability of the equilibrium at the critical load
factor is that the cubic terms in the energy functional eqn (2), evaluated for
the buckling modes, vanish indentically. Since the integral of a cube of a
sine or cosine over a period is zero each separate buckling mode eqns (5)
satisfies this requirement. In the case of cylindrical shells under axial
compression, however, and also in the case of spherical shells under
external pressure, the conditions for the critical wave numbers, p* + q* =
px/2cR/hand p* + q? = 2cR/h respectively, are satisfied for a large number
of values of p and gq, and there are combinations of pairs p,, 91; Po, 923
P3, 43 such thatp,;+p, +p, =0 and simultaneously g, + q, + q3 =0. The
 Amplitude Modulation of Short-Wave Buckling Modes 39

linear combination of such a triple of buckling modes then results in a non-

vanishing cubic term in eqn (2) and the equilibrium at the critical load
factor is (highly) unstable.*°
In the case of a circular cylindrical shell we consider a linear combination
of two modes
w =h[b,cos2mx/R + c,,cos mx/R cos my/R] (8)

where m=,./cR/2h. Retaining only quadratic and cubic terms in the

functional eqn (2) and averaging the energy per unit area the result is

ie
Eh?
|HC A068 + ded) +553c bu | 0)
The post-buckling paths are obtained by putting the partial derivatives
with respect to the non-dimensional amplitudes by and c,, equal to zero.
The result is
bo = —2(1 — A)/3e C= —e §(l—A)/3e (10)
and the deflection in the unstable post-buckling path is given by

w= [—2(1 — A)h/3c][cos 2mx/R + 4cos mx/R cos my/R] CEE)

Geometric imperfections in the shape of the periodic post-buckling path

are described by
Wo, = kh[cos 2mx/R+4cos mx/R cos my/R] (12)
and they result for k < 0,a dominantly inward imperfection, in a limit point
A* <1 of the load factor given by
(1 — 2%)? = —6cKA* (13)
The similar analysis for a spherical shell under external pressure employs
the linear combination of buckling modes

w =h[a,cos2mx/R + a, cos mx/R cos m,/3 y/R] (14)

and the associated post-buckling path is given by

8 16
ay =— (1-A) a, = +—(1-A) (15)
FASC } 9c

Geometric imperfections in the shape of the periodic post-buckling paths

Wy = kh[cos 2mx/R + 2cos mx/R cos m/3 y/R] (16)

40 W. T. Koiter

now result in a limit point / for k > 0

9¢

(1A)? => lt (17)

3. MODULATION OF HUTCHINSON’S MODES FOR

SPHERICAL SHELLS?

Let x = 0 denote the equator in a sphere and y/R the longitude along the
equator. Hutchinson’s mode,°*

w =a,h[cos 2mx/R+2cos mx/Rcosm,/3 ¥/R] (18)

may give an adequate description near the equator but it fails completely
when the poles are approached. This defect may be removed by the
introduction of an additional modulating factor (sin 6)‘ in eqn (18), and
also in the associated in-plane displacement components, where @ is the
polar angle and & is an integral number not less than 2 and small in
comparison with m. All singularities in middle surface strains and changes
of curvature near the poles which are present in the deflection pattern
eqn (18) disappear by the introduction of the factor (sin 6)*. Our
assumption that k is small in comparison with m implies that all strains and
changes of curvature, obtained from eqn (18) and the associated in-plane
components, are also multiplied approximately by a factor (sin 6)*. The
relative error in the evaluation of the energy is of order k?/m*. We consider
in particular the case that k is large in comparison with unity but small in
comparison with m, say k is of order m'/*. The buckling pattern is now
confined to an equatorial zone. The integrals of the quadratic and cubic
terms in the energy functional now have a factor

|(sin @)7*"*
dé and |(sin 6)3**1d@
(0) 10)

respectively. For large values ofk the ratio of these two integrals is x) hee In
other words, the cubic term in the energy is reduced by a factor J 2/3 in
comparison with the quadratic term. As a consequence the limit load factor
for imperfections described by eqn (16) multiplied by a factor (sin 6)* is
given by
 Amplitude Modulation of Short-Wave Buckling Modes 41

and imperfections of the modulated type with a dimensionless amplitude x

at the equator are equally harmful as imperfections of Hutchinson’s short-
wave type with an amplitude ./2/3 = 0.816x. This represents only a minor
reduction in imperfection sensitivity.

4. MORE OR LESS LOCALISED IMPERFECTIONS!°

More or less localised imperfections may be obtained from the doubly

periodic shapes eqns (12) or (16) by means of a Gaussian exponential
modulating factor in both directions:

E*(x, y).=exp[—4n7
(x? + y7)/R?] (19)
where pu’ is a fairly large number (in order that the modulated periodic
shape decays sufficiently rapidly) but at the same time small in comparison
with m?. The latter condition enables us to treat the modulating factor
again as a constant factor for the purpose of differentiation of the mode.
The behaviour of the shell with the more or less localised imperfections
described above is analysed again by the Rayleigh—Ritz method. The
deflections are assumed in the same shape as the imperfections, in the
circular cylindrical shell in the form

W =a,)E*(x, y)h[cos 2mx/R + 4cos mx/R cos my/R] (20)

and for the spherical shell in the form

W =a E*(x, y)h[cos 2mx/R + 2cos mx/R cos m,/3 y/R] (21)

The quadratic and cubic terms in the energy functional in the modulated
modes eqns (20) or (21) are now obtained from the similar terms in the
periodic modes by a multiplication of their integrands by the square and
cube of the modulating factor E*(x, y) respectively. In view of the rapid
decay of the exponential factor the integration may now be carried out
from — © to + «© in both xand y directions. In both directions we arrive at
quadratic integrals of type

| e-® cos na da _ve exp (—n?/4u2) (22)

ap kl
where nis either a large number (of order m) or zero. It follows that we need
only consider integrals of type eqn (22) with n = 0, and their value is J t/t.
42 W. T. Koiter

The cubic integrals are of the same type with p? replaced by (3/2)? and
their value for n=0 is thus \/2n/y,/3, a factor ./2/3 smaller.
The essential part of the energy expression eqn (9) and its counterpart for
the spherical shell is the expression between the brackets. The single
modification of the expression between the brackets in the case of
modulated modes is that the coefficient of the cubic term is decreased by a
factor 2/3 in comparison with the quadratic term. The critical load factor
remains unity, and the post-buckling paths in the absence of imperfections
are described for the cylindrical shell by (cf. eqn (10))
bb = —(—-A)j/e Cn = £41 —A)/e (23)
and for the spherical shell by (cf. eqn (15))
4 8
a= (1-4) = 45 (1-4) (24)
Imperfections of types eqn (12) or eqn (16), modulated by the factor
eqn (19), now result in a reduced limit load factor A* given by
(1 — A*)? = —4ckd*
for the cylindrical shell, and by
(l= 7"). ScK At
for the spherical shell. More or less localised imperfections of the type
considered here are equally harmful, in the case of an inward dent, as
periodic imperfections of an amplitude reduced by a factor 2/3.
It may be worthwhile to repeat that our present Rayleigh—Ritz type
analysis is quite approximate in character. One of the most doubtful
aspects is perhaps the treatment of the modulating factor E*(x, y) defined
by eqn (19) as a constant factor for the purpose of differentiation of
displacement components. The effect of derivatives of E*(x,y) has been
investigated in ref. 1 in the case of a cylindrical shell under axial
compression. For the value y?/m* =0.1, representing a fairly localised
dent, Gristchak has established that our simple first approximation has an
error of no more than about 11 %.

5. INTERACTION BETWEEN SHORT-WAVE LOCAL

BUCKLING AND LONG-WAVE OVERALL BUCKLING

The imperfection-sensitive interactions between short-wave local buckling

and long-wave overall buckling in the case of (nearly) coincident critical
 Amplitude Modulation of Short-Wave Buckling Modes 43

loads were exhibited most clearly in Van der Neut’s pioneering paper at the
1968 International Congress of Theoretical and Applied Mechanics.!7 In
his analysis the load-shortening relations for buckled flat plate strips with
imperfections were applied to a column built-up from such plate strips. A
mathematically entirely different, but physically nearly equivalent
approach consists of amplitude modulation of the basic short-wave local
mode.*'!!~!3 In this alternative approach the amplitude of the basic local
mode is taken to be a slowly varying function f(x) of the axial coordinate.
The errors introduced by this assumption and its various consequences are
of order /*/L?, where / is the wavelength of the short-wave local mode and L
is the wavelength of the overall mode.
Amplitude modulation of a single basic short-wave local mode yields
quite satisfactory results, as described in the references given above, and
exhibited in the oral presentation, whenever the deflections in the local mode
are largely confined to one side of the neutral axis of the overall mode. This
happens to be the case for a number of well-designed stiffened panels with
rather solid or hat-shaped stiffeners. In such cases the effect of
imperfections in the local mode is quite significant for very small
imperfections but it decays quite rapidly for larger imperfections. The
result is that the reduction in critical load due to such imperfections may be
bounded above by some 10%.
The sensitivity to local imperfections may even be reduced further, if the
basic local mode involves significant deflections on both sides of the neutral
axis for overall buckling. The reason is that in this case the overall mode
tends to increase the magnitude of the deflection in the local mode only on
one side of the neutral axis and to decrease it, approximately equally, on the
other side. An adequate analysis of this case, however, requires that at least
one additional local mode of the same short wavelength should be taken
into accountin order that a combination of the two local modes may
involve dominating deflections on only one side of the neutral axis. The
analysis becomes here more complicated, and we refer to a first attempt to
cover such situations in a previous paper.'*

REFERENCES

1. GristcHaK, V. Z., ‘Asymptotic formula for the buckling stress of axially

compressed circular cylindrical shells with more or less localized short-wave
imperfections’, Report WTHD-88, Department of Mechanical Engineering,
Delft University of Technology, 1976.
44 W. T. Koiter

. Hut, D., Tennyson, R. C. and HANSEN, J. S., Mode interaction of axially

stiffened cylindrical shells: effects of stringer axial stiffness, torsional rigidity
and eccentricity, J. App/. Mech., 48 (1981) 915-22.
. HuTcHINSON, J. W., Imperfection-sensitivity of externally pressurized
spherical shells, J. Appl. Mech., 34 (1967) 49-SS.
. Korver, W. T., ‘Over de stabiliteit van het elastisch evenwicht’, Proefschrift
Delft, 1945. English translations ‘On the stability of elastic equilibrium’. (a)
NASA Technical translation TT F-10, 833, 1967. (b) Air Force Flight
Dynamics Laboratory TR 70-25.
. Korter, W. T., The nonlinear buckling problem of a complete spherical shell
under uniform external pressure, Proc. Kon. Ned. Ak. Wet. Amsterdam, B72
(1969) 40-123.
. Korter, W. T., ‘The influence of more or less localised imperfections on the
buckling of circular cylindrical shells under axial compression’. (Distributed in
report form to participants at the IUTAM Symposium on Buckling of
Structures, Harvard University, June 1974.) Complex Analysis and its
Applications, the I. N. Vekua Anniversary Volume, Akademia Nauk, Moscow,
pp. 242-4.
. Korter, W. T., ‘General theory ofshell stability. Thin shell theory’, New trends
and applications, W. Olszak (ed), CISM Courses and Lectures 240, New York,
Springer Verlag Wien, 1980, pp. 65-87.
. Korrer, W. T., ‘The application of the initial post-buckling analysis to shells’,
Buckling of shells, E. Ramm (ed), Berlin usw., Springer Verlag, 1982, pp.
sa7)
. Korrer, W. T., ‘General equations of elastic stability for thin shells’, Proc.
Symp. Theory of Shells in honor of Lloyd Hamilton Donnell, D. Muster (ed),
University of Houston Press, 1967, pp. 187-227.
. Korrer, W. T., ‘Current trends in the theory of buckling’, Proc. IUTAM
Symposium Buckling of Structures, Harvard University, 1974, B. Budiansky
(ed), New York, Springer Verlag, 1976, pp. 1-16.
. Korter, W. T. and KuIkEN, G. D. C., ‘The interaction between local buckling
and overall buckling on the behaviour of built-up columns’, Delft University of
Technology, Report WTHD-23, 1971.
. Korrer, W. T. and PIGNATARO, M., ‘A general theory for the interaction
between local and overall buckling of stiffened panels’, Delft University of
Technology, Report WT HD-83, 1976. Also in Problemi attuali di meccanica
teorica e applicate, Torino 1977, pp. 179-222.
. Korter, W. T., “General theory of mode interaction in stiffened plate and shell
structures’, Delft University of Technology, Report WTHD-91, 1976.
. Koirer, W. T. and VAN DER NEut, A., ‘Interaction between local and overall
buckling of stiffened compression panels’, International Conference on Thin-
walled Structures, University of Strathclyde, 1979.
. LANGE, C. G. and NEwELL, A. C., The post-buckling problem for thin elastic
shells, SIAM Journal Appl. Math., 21 (1971) 605-29.
. LANGE, C. G. and KRIEGSMANN, G. A., The axisymmetric branching behaviour
of complete spherical shells, Quart. Appl. Math., 39 (1981) 145-78.
. VAN DER NEuT, A., The interaction of local buckling and column buckling of
thin-walled compression members’, Proc. 12th Int. Congr. Theor. & Appl.
 Amplitude Modulation of Short-Wave Buckling Modes 45

Mech., M. Hetenyi and W.G. Vincenti (eds), Berlin, Springer Verlag, 1969,
pp. 389-99.
. POTIER-FERRY, M., ‘Amplitude modulation, phase modulation and locali-
zation of buckling patterns’, Paper presented at IUTAM Symposium on
Collapse: The Buckling of Structures, 1982. To be published by Cambridge
University Press, 1983.
 4

Locally Imperfect Plain Channel Columns

J. LOUGHLAN
Cranfield Institute of Technology, UK

and

A. R. UPADHYA
National Aeronautical Laboratory, Bangalore, India

ABSTRACT

A study is made of the effects of local imperfections on the equilibrium

behaviour of plain channel columns in compression and bending. A semi-
energy method of analysis is used and the effects of local deformations in all
constituent plate elements of the section are taken into consideration. It is
shown that plain channels are extremely sensitive to small local
imperfections particularly when their local buckling effects are predominant
in the plate elements of the section which are perpendicular to the axis of
overall column bending. A study is also made of the effects of allowing the
locally buckled form to change with loading and it is shown that the
asymptotic approach to a reduced Euler load based on an unchanging locally
deflected form can be seriously in error for some cross-sectional geometries.
Although the local imperfection analysis is outlined in the paper with
reference to acolumn whose material is orthotropic the results presented are
for isotropic columns.

NOTATION

A, Ao Magnitude coefficients for local deflection and local

imperfection, respectively.
47
48 J. Loughlan and A. R. Upadhya

4145412, 4225 466 Elements of compliance matrix [a] for orthotropic

plate;
a,,=1/E,, Ay, = —V42/Ey,
= —V21/E22
a,,=1/E,, 466 =1/G,2
433 (2a,, + G66).
Ey iE Modulus of elasticity in the x and y directions
respectively (see the | and 2 directions in Fig. 1).
é Non-dimensional load eccentricity = e/d.
Gis Elastic shear modulus in x—y plane.
i Suffix relating to particular plate element of section
({= 1 for web, = 2, 3 for flanges).
NNN In-plane stress resultants.
ae Maximum load.
RoR Radius of curvature of overall deflected form, and
overall imperfection, respectively.
u Web compression.
W, Wo Local deflection and local imperfection, respectively.
WeaW ae Amplitude of w and w, respectively at web mid-point.
x Coordinate along length of column.
4 Compression eccentricity.
é Web compressive strain = 2u/A.
0,05 Overall deflection and imperfection respectively.
OO: Amplitude of 6 and 6, respectively at column centre.
Vio Poisson’s ratio in the x and y directions respectively (see
the 1 and 2 directions in Fig. 1).

1. INTRODUCTION

The compressional and flexural stiffnesses of a thin-walled section are

known to be significantly altered by local buckling. The tangent stiffnesses
reduce instantaneously at buckling and continue to reduce in the post-
buckling range due, initially, to changes in the locally deflected form and
then to the combined effects of local form change and plasticity well into
the post-buckling range. Whereas the stiffnesses of a geometrically perfect
section reduce instantaneously at buckling, those of a locally imperfect
section reduce in a continuous manner, starting from the onset of loading
and extending over the entire loading range.
Due to the change in section shape caused by local buckling the neutral
 Locally Imperfect Plain Channel Columns 49

axis of the section effectively shifts thus causing an initially concentrically

loaded section to become effectively eccentrically loaded and an
eccentrically loaded section to effectively change its eccentricity. As with
the compressional and flexural stiffnesses for the geometrically perfect
section the tangent shift in neutral axis is instantaneous at buckling and the
shift continues in the post-buckling range whereas for the locally imperfect
section the shift in neutral axis is continuous from the onset of loading.
The compressional and flexural stiffnesses of a locally imperfect section
are known to be considerably lower, depending on imperfection
magnitude, than those of its geometrically perfect counterpart at loads less
than the local buckling load, and this, coupled with the continual shift in
neutral axis of the section, can result in the elastic failure of a compression
member and in unstable equilibrium at failure. Behaviour ofthis nature will
occur, of course, in a member designed such that its local and Euler
buckling loads are simultaneous, or nearly so, whereby interaction between
the independent buckling modes leads to an imperfection sensitive
structure and to failure of the member at a significantly reduced load from
that which could be attained by the geometrically perfect structure.
The interaction between local buckling and overall member behaviour
has been studied by many investigators in recent years and the effects of
imperfections on equilibrium and failure behaviour has been considered for
a number of problems. Some typical reports on stiffened plating are those
by Van der Neut,’ Koiter and Pignataro,? Walker,’ and Fok et al.* to
name a few, and in these the complexities associated with the phenomenon
of interaction are reflected.
Typical reports concerned with other structural shapes are those by
Graves Smith,° Rhodes and Harvey,° Loughlan and Rhodes,’:? and
Hancock” who studied the behaviour of thin-walled box, plain channel,
lipped channel and I section columns respectively. In ref. 5 consideration is
given to the weakening effects of plasticity and to the effects of residual
stresses and local imperfections on the strength of box columns. A
changing post-buckled shape in the elastic range, from that of the buckling
mode, is considered in refs. 6, 7, 8 for locally perfect plain® and lipped’”’®
channel columns in which the local form change is associated with column
bending after local buckling, while the effect of local imperfections on the
behaviour ofI section columns is considered in ref. 9 for the case of pure
compression up to the bifurcation load for overall buckling.
An approach to the analysis of orthotropic plain channel columns has
been outlined in a paper by Upadhya and Loughlan’® and in this, due to
space limitations, only brief mention has been made of the effects of local
50 J. Loughlan and A. R. Upadhya

imperfections. In the present paper the authors give a detailed outline of a

local imperfection analysis for plain channel columns and present results
from this showing the effects of local imperfections on various aspects of
column behaviour. Although the theory is presented with reference to
orthotropic channels, the results given are for isotropic columns and in
these the effects of allowing the locally deflected shape to change with
loading is clearly shown.

2. THEORETICAL APPROACH

A plain channel column loaded in compression and bending is shown in

Fig. 1. The length of the column is Z and pinned conditions are assumed at
its ends. The cross-sectional notation of the plain channel is shown in
Fig. 2. It is assumed that the column contains an overall geometrical
imperfection described by the expression
TX
On0 =10 0c COS —a 1
(1)

The axial load P is applied on the axis of symmetry of the section and at a
distance e from its minor principal axis thus causing a moment M about the
web centre line given by

M =P(e+d+6) (2)
By considering a representative section of the column situated between
two local buckling nodes and treating this as a short strut of length 4 which
has its ends compressed according to the displacement system shown in
Fig. 3, we can obtain expressions for the moment M and axial load P in
terms of the displacement parameters u and « of the compression system at
the buckle nodes.
To obtain these, consider first the cross-section shown in Fig. 4 with
initial imperfections wy) and total deflections w. The local buckling
deflections are described by the expression
1X
w; = AY;(y;)
cos He (3)

and the initial imperfections by

TX
Wo, = AotW eos a. (4)
 Locally Imperfect Plain Channel Columns 51

BUCKLED _FORM_AT
COLUMN CENTRAL
SECTION

P(APPLIED LOAD)

WEB BUCKLES

PRINCIPAL STIFFNESS
DIRECTIONS BUCKLE NODES
Fig. 1. Central portion of column showing local buckles.

FLANGE

b=,

WEB
Fig. 2. Cross-sectional notation of plain channel.
52 J. Loughlan and A. R. Upadhya

+. U(1-%)

Fig. 3. Compression system at nodes of central buckle.

where A and Ay are magnitude coefficients for the local deflections and
initial imperfections respectively and Y,(y;) is the locally deflected shape of
the cross-section. It is assumed that the shape Y;();) is the same as the
buckling mode of the locally perfect section and this is obtained very
accurately from the multi-term Rayleigh—Ritz analysis outlined in ref. 10.
Although an infinite number oflocally imperfect shapes are possible for the
section it is convenient in analysis to assume that the imperfections are
similar in form to the buckling mode. This not only leads to simplicity in
analysis but has been shown by Rhodes et a/.'’ to give results for plates
which closely approximate actual plate behaviour.

LOCAL IMPERFECTION Wo
ToTAL DEFLECTION W

Fig. 4. Coordinate system for local deflections.

 Locally Imperfect Plain Channel Columns 53

To obtain the stresses within the section which comply with the
deformations that take place during loading use is made of von Karman’s
compatibility equation written in the following form to take due account of
imperfections:
ot 4 ot
F, F, ;
(43>); l aay.
ex Ht (ds5); ax?
Ax dy2 + (414); oye

=
62w, l
\? 2, 62, L u =
07Wo.\2
1 ay
07Wo, 07 Wo,
1 1 5

je Ox at (ee Ox MOY, | e
where F; is a middle surface force function for the ith plate such that
0?F, 0° F, 0?F,
Nx), => Ny) See Nxy), = — 6

Substituting for w, and Wo, from eqns (3) and (4) into eqn (5) we can now
solve for the stresses and, by applying the principle of minimum strain
energy as outlined in ref. 10 for the locally perfect column, we may write an
expression relating end displacements at the buckle nodes and out of plane
deflections as

a Ao
C3

where € = 2u/A and the constants C, to C, are given in ref. 10.

Integration of the longitudinal membrane forces around the cross-
section and of the moments of these forces about an axis through the web
mid-surface now gives the axial load P on the section and the moment M
about the web mid-surface as
b
P =éX, — ¢ —*— — (A? — A2)C, (8)
(411)2
b3 2b3
M=e-——— —eo a (An ACs (9)
(441). 3(41;)2 oe kas
where the constant X, is given in ref. 10. The product é« in eqns (7), (8) and
(9) can be shown from geometrical considerations to be related to the
longitudinal curvature of the section as
l _@ (0 — do) £0
= 9" =— 10
Rae dx? De ey
For a column with local imperfections the local deflections start to grow
54 J. Loughlan and A. R. Upadhya

as soon as load is applied to the column and, due to this, as mentioned in

the Introduction, the compressional and flexural stiffnesses of the section
and the effective position of the section neutral axis will alter from the onset
of loading and continue to do so over the entire loading range. In view of
this and by analogy with the solution of the locally perfect column in the
post-buckling range (see ref. 10) the behaviour of the locally imperfect
column will be described by

6, = ctla = anya( - aN axae VP —_ - 1a%cees

P 2 (P*), pee P
(PE).
(11)
where (Px), = 17([E/]*),/L? and parameters thus ( ), are current values
associated with load level P. In eqn (11) d* defines the effective position of
the section neutral axis, [E/]* is the flexural stiffness of the locally
deformed section, P# is a reduced Euler load for the column and the
parameter Popyc is a local buckling load for the section under uniform
compression. Expressions for ({E/]*),., (d*), and (Peryc), have been
developed by Upadhya!? and these are stated as follows:

(ET). = (12)

where
: = b3 = ( bo G ~) ca
: (441)2 L3 X, \C@y1)2_ 33C C3 2C;
and

ee ee

eer eles
2D OG (aa types (dia

X, (ao) s 2C, A 2C3 (4;1)2

(d*), = are (13)

i Ao jeAe Ae
(Perucle =
ame hera ear Ae) 89
 Locally Imperfect Plain Channel Columns 55

When A, =0 and A £0 eqns (12), (13) and (14) are identical to those
developed in ref. 10 for a locally perfect column which has undergone local
buckling. It is also of note that when A, #0 the values of ([E/]*)., (d*),
and (Pcruc), tend towards those obtained from the equations developed in
ref. 10 for large values of A. This is due, of course, to the fact that the
equilibrium behaviour of a locally imperfect column tends towards that of
its geometrically perfect counterpart well into the post-buckling range.
When A, = 0 and the effects of local buckling are not present in eqns (12),
(13) and (14) then ([E7]*), becomes (£7), the flexural stiffness of the locally
perfect unbuckled column, (d*), becomes d, the neutral axis position ofthe
locally perfect unbuckled column and, as expected, (Pcryc). =9 since
Pecrucis only associated with a column which has undergone local buckling.
It is evident from eqns (12) and (13) that due to the presence of even a small
value of local imperfection the flexural stiffness and neutral position of the
column cross-section are always different from those of its locally perfect
counterpart, even at zero load, 1.e. when A = Ao.

2.1. Overall Equilibrium

To obtain the load-deflection equilibrium path of the locally imperfect
column it is first of all necessary to obtain the appropriate equilibrium
relationship between the applied load P, column deflection 6 and local
deflection A. The first step in achieving this relationship is to rearrange
eqn (8) so that € may be written in the following form after utilising
eqn (10):

z= P 5” b?
2 A2 — 42
2¢
Resid iS

: ee CRSb ae 0) xX.

Consideration is now given to the equilibrium of internal and external

moments at the column central cross-section. Substituting for éa in terms of
6” from egn (10) into eqn (9) and then substituting for M from eqn (9) into
the moment equilibrium expression described by eqn (2) gives the following
relationship for 6” after utilising eqn (15):

b2 Cab?
Ple+d+6,-—~~ |+ (A?
— A2)| C,b, -—*+-
s"= (a,,)2 X, (4, 1)2X, (16)
E b3 E b,
(4;,)23 reel

Substituting for é«, €é and 6” from eqns (10), (15) and (16) respectively
56 J. Loughlan and A. R. Upadhya

into eqn (7) now gives the final relationship as

Palio A
on (2, +258) -C, (1-2) =0 (17)
E

where

NO oe
3x.)
F 2CP , C2X(ay1)2
=a
ees 4©:
_—— —Q _ |NN)
aIs
ae
oa

ne ; aor
mds 2G) G oy i
1 ae —C 2(e+d
é anal 3 2|+* oe (4,1).

na {oe 6
; b, (441)2X,

Since the parameters ({EJ]*),, (d*), and (Peruc). vary with the local
deflection magnitude coefficient A then eqn (11) cannot be used directly to
obtain the load—deflection equilibrium path of the deforming column.
Using eqn (17), however, in conjunction with eqn (11) the equilibrium path
can be obtained quite easily by solving these numerically in a step-by-step
procedure using iterative techniques.
Knowing that the initial state of the system 1s P = 0, A = A, and 6, = 6,
then, for a value of 6, slightly in excess of 0,,, a realistic ‘guess’ at the
equilibrium value of P corresponding to this can be made and hence A can
be obtained numerically from eqn (17). Knowing A the values of ({E/]*)
(d*), and (Pcryc), can now be obtained from eqns (12), (13) and (14) and
by substituting for these in eqn (11) a numerical evaluation of Pcan now be
made for the given 6,. If the value of P obtained from eqn (11) is the same as
that used in eqn (17) then obviously this is the required solution for the
given 6,. If itis not, however, as is more likely, then the value of P obtained
from eqn (11) may be substituted into eqn (17) for the same 6, and the
process repeated until the difference between two consecutive values of P is
zero. The solution will then be complete for the given 6, and to establish the
complete load—deflection equilibrium path of the column the procedure
outlined is repeated for a number of small increments in 6,.

3. CHANGE IN BUCKLED SHAPE

So far in the analysis it has been considered that the locally buckled shape
of the cross-section remains constant with change in load and that it is of
 Locally Imperfect Plain Channel Columns 57

the same form as the buckling mode of the locally perfect section. Since the
parameters C, to C, are functions of the locally deflected shape and are
therefore constant for a given shape, the implication of this assumption is
that changes in the values of ([E/]*),, (d*), and (Peruc), occur only with
change in the local deflection magnitude A for any given imperfection
amplitude A,. During loading, however, changes in the locally deflected
shape do in fact occur, caused principally by column bending, and the
coefficients C, to C, are variable which means that changes in the values of
({EJ]*)., (d*), and (Peryuc). depend, significantly, on the current locally
deflected shape. The degree of this significance can be considerable for
certain cross-sectional geometries and an analysis which precludes the
possibility of a changing locally deflected shape can give results which are
seriously misleading.
In this paper local form change is accounted for in an approximate
manner and a study ofits effects on column equilibrium behaviour is made.
The indirect approach considered in analysis first makes use of the readily
available facility of being able to obtain a range of buckled shapes
corresponding to different compression eccentricities from the multi-term
Rayleigh—Ritz analysis described in ref. 10. Using these shapes it is now
possible to obtain a series of solutions describing the load—deflection
equilibrium behaviour of a column, each solution being based on a
different buckled shape and hence on different values of ({E/]*),, (d*), and
(Peruc)e: Lhe lowest envelope of the equilibrium paths obtained in this way
is then considered to be the true solution since for a given column deflection
the solution which gives the lowest axial load is that which prescribes the
least value of the strain energy of the system.
The approach outlined gives results for locally perfect columns which
sufficiently describe the effects of local form change on column behaviour.
The method can also be used, quite easily, to account for these effects for
columns with local imperfections. In the case of the imperfection analysis
presented in this paper, however, the implication of using this approach is
that as well as the locally deflected form changing with load the form of the
local imperfections also changes. Allowing for this fictitious change in the
local imperfection shape with load, however, gives results which are
considered by the authors to be realistically representative of the practical
situation for the case of columns which are sensitive to the combined effects
of local imperfections and local form change.

3.1. Variation in Local Buckle Magnitude

It has been observed in tests carried out by many investigators in recent
years that the simultaneous presence of column and local deflections is
58 J. Loughlan and A. R. Upadhya

associated with a variation in the magnitude of the local buckles along the
column. In practice the local buckle magnitudes are more pronounced at
the centre of the column and diminish along the length of the member. This,
of course, is intuitively what one would expect and in fact was evident some
thirty years ago in the work by Bijlaard and Fisher'*:'* who conducted a
series of tests on centroidally loaded I section and square tube members for
which local buckling occurred prior to overall column buckling. In this
work there is clear photographic evidence relating to an I section column
which highlights the proportioning of the local buckle magnitudes along
the length of the column due to overall bending. Due to centroidal loading
and the symmetry of the cross-section, column bending, in this case, is
initiated by overall bifurcation of the locally buckled member and the
equilibrium behaviour after bifurcation is shown in ref. 14 to be fairly
unstable for I sections. In this instance, therefore, the variation of local
deflections along the column does not affect the ultimate carrying
capability of the member but has a direct influence on post-failure
equilibrium behaviour only.
In the case of pin-ended, centroidally loaded, singly symmetric sections
such as lipped and plain channels for which local buckling occurs before
Euler buckling, local deflection variation along the column occurs from the
onset of local buckling and in this case will influence the behaviour of the
column prior to failure. In this paper the approach used considers
essentially the behaviour of a single local buckle along the column and from
this consideration the governing differential equation for the lateral
equilibrium of the locally buckled column is developed (see eqn (20) in
ref. 10). This method of approach is considered to lead to a somewhat
elegant solution to the interactive buckling problem and has been shown to
give results which are in good agreement with experiment. Typical
comparisons showing this agreement are to be found in refs. 6 and 8 for
plain and lipped channel columns respectively.
The use of the differential equation developed in the analysis
automatically introduces a form of local deflection variation along the
column and this can be examined by suitable manipulation of the relevant
equations in the analysis. On examination it has been found that the local
deflection magnitudes do in fact diminish along the column in the presence
of column bending as has been observed in practice.
For geometrically perfect eccentrically loaded columns, depending on
eccentricity magnitude, the analysis can lead to a discontinuity of the
column deflections between the pre- and post-local buckling solutions at
the local buckling load. The reason for this is obtained from an
 Locally Imperfect Plain Channel Columns 59

examination of the local deflection variation along the column at local

buckling. From this it is found that local buckling occurs over the central
portion of the column and the outer portions remain unbuckled. To
account for this it is necessary to set up two differential equations for the
column, one pertaining to the locally buckled portion and the other
associated with the unbuckled portions. It is then necessary to satisfy the
compatibility requirements at the junction between the two portions and
the first author is currently working on this particular aspect of the
problem.
This discontinuity at the local buckling load was encountered in ref. 4 in
relation to stiffened plating and the approach used in that paper was to
introduce an effective length in the post-buckling range to enforce
compatibility. Since the writer has not yet completed his analysis of the
two-differential-equation approach to the problem the same concept of
effective length has been utilised in this paper for the eccentric loading
results given.

4. RESULTS AND DISCUSSION

The numerical work involved in the analysis was carried out using the
VAX-11 computer system at Cranfield Institute of Technology and the
main scientific programming language FORTRAN 4 was used to write
the computer program. Although an extensive amount of data is available
from the computer the authors have elected to present results for isotropic
columns and have chosen these to highlight, in their view, the most
significant effects of local form change and local imperfections.
Figure 5 shows load-deflection equilibrium paths and reduced Euler
load curves P*/P,. for a centroidally loaded perfect column with a flange-
to-web ratio of 0.4. The broken curves are based on an unchanging locally
deflected shape with load and the full lines take local form change into
consideration in the approximate or indirect manner described earlier. The
length of the column is such that its local and Euler buckling loads are
simultaneous and due to this the nature of the equilibrium at buckling is
unstable. From the figure it can be seen that the immediate change of
flexural stiffness for this section at local buckling is given by P£/P, ~ 0.71.
If post-local buckling behaviour is now based on the deflected shape
obtained at buckling then the load—deflection equilibrium curve for the
column would tend asymptotically towards the P/P, value of 0.71 as
indicated. The actual behaviour, however, is seen to be quite different from
60 J. Loughlan and A. R. Upadhya

b¢
ie
Ar
DOCHcud,
a7 WC e.pian Didare anOMe nieee a
t =0.0, ‘la =0-4, +-= 100, bw 72h 49

Fig. 5. Effect of local form change on equilibrium behaviour and reduced Euler
load (é = 0.0).

this due solely to the high sensitivity of this particular cross-sectional

geometry to changes in the locally deflected shape after buckling. The
change in local form, caused principally by column bending, results in a
substantial reduction in section flexural stiffness and the reduced Euler load
of the column is seen to drop very rapidly from 0.71 P, to about 0.22P, at
larger column deflections. In this instance the load—deflection equilibrium
path of the column is tending towards a reduced Euler load which is
continually becoming smaller and due to this the rate of unloading is seen
to be markedly higher than that of the equilibrium curve based on
unchanging deflected shape.
Figure 6 shows the load—deflection equilibrium paths of columns which
are loaded eccentrically towards the flange free edges. The columns are
geometrically perfect and have the same cross-sectional details as those
considered in Fig.5. The applied eccentricity causes positive overall
deflections to occur from the onset of loading and consequently the overall
bending of the column leads to higher compression towards the flange free
 Locally Imperfect Plain Channel Columns 61

0.8

Gee
doc _-99, Woc
“96-00,
_ Pf
bie-0.4, ¥-100
bw _

Fig. 6. Effect of local form change on equilibrium behaviour (é= +0.1).

edges. As in Fig. 5 the broken curves are based on an unchanging locally

deflected shape and the full lines take local form change into account. It will
be noted from Fig. 6 that for the column with L/bw = 20.35 the difference
between the two solutions is small but as L/bw reduces the difference
becomes increasingly more noticeable. The reason for this is due to the fact
that for the given eccentricity of loading the local buckling modes for
longer columns correspond to higher compression eccentricities due to
column bending and that after local buckling these shapes are less sensitive
to local form change with further column deflection than are those for
shorter columns which correspond to lower compression eccentricities. The
curves in Fig. 6 signifying an unchanging locally deflected form after
buckling suggest that for the three lower values of L/bw stable post-local
buckling behaviour will occur. The curves which allow for local form change,
however, show that for the L/bw values of 11.75 and 15.17 the behaviour
after local buckling is initially stable but for loads not much in excess ofthe
local buckling load the behaviour changes from being stable to unstable.
Figure 7 shows the effect of local form change on the equilibrium
behaviour of columns loaded eccentrically towards the web. In this case
negative overall deflections are induced to occur from the onset of loading
and column bending reduces the compression at the flange free edges. At
the point of local buckling, however, for the four L/bw values considered,
local buckling is initiated by the flanges and the neutral axis of the section
moves towards the web thus altering the eccentricity towards the flange free
62 J. Loughlan and A. R. Upadhya

+e © LcAD PoinT
10 i; ale bs X NEUTRAL AXIS

08
aes
Dit ell a
oa 13-57

ie
Fe

ad @) 2 4 6 8 10 oe
00¢
t
0.0, Woe
t
=0.0, Bw
bf =0.4, BW
t
=100
Fig. 7. Effect of local form change on equilibrium behaviour (é= —0.1).

edges and causing a change in direction of the column deflection as

indicated. From Fig. 7 it is seen that the locally deflected shapes obtained at
buckling for these columns are highly sensitive to local form change after
buckling for the four L/bw values considered and that for the L/bw values of
9.59, 11.75 and 13.57 unstable load—deflection behaviour occurs at loads
slightly above the local buckling load.
Figures 8 and 9 show the effect of local imperfections on the flexural
stiffness and section neutral axis position respectively for a centroidally
loaded column with a flange-to-web ratio of 0.5. For given local
imperfection magnitudes the variations in these parameters are shown with
respect to change in the web local deflection magnitude. The curves shown
do not allow for local form change with loading. It is seen from Fig. 8,
however, that the immediate loss of flexural stiffness for this section at local
buckling is of the order of 73.5 % and due to this its sensitivity to local form
change is low. The high loss in flexural stiffness indicated is due to the very
low subsequent effectiveness of the flanges after local buckling and a
measure of this is shown in Fig. 9 where it is seen that the immediate shift in
neutral axis is towards the web and that its position relative to the web
centre line is only 4% of its initial value for the locally unbuckled section.
 Locally Imperfect Plain Channel Columns 63

10

Bt os , &=0.0
Bw
Woc Pw = 833 , = 0.0
0-57 ‘
N WwW Wo y

Vie |a
Dy
WceMoc-web cent deflection

5 ( We “oc)/t
0 1 2 3 4
Fig. 8. Reduction in flexural stiffness for locally imperfect section.

°9 1 2 3 4 Bam cr Oct
Fig. 9. Shift of neutral axis for locally imperfect section.
64 J. Loughlan and A, R. Upadhya

When local imperfections are present the local deflections begin to grow
from the onset of loading and these are accompanied by a gradual
reduction in the flexural stiffness and a gradual change in the neutral axis
position as shown in Figs. 8 and 9. When the imperfection magnitudes are
small the initial values of flexural stiffness (E/)* and neutral axis position
d*, at zero load, are close to those of the locally unbuckled perfect column,
but the rate of change ofthese values with the growth of local deflections is
high. For larger values of imperfection magnitude, (E/)* and d* change at
a smaller rate with increasing local deflections but their initial values at zero
load are progressively smaller, with increasing imperfection magnitude,
than those of the locally unbuckled perfect column. As the local deflections
increase both (E/)* and d* approach the corresponding values for the
locally buckled perfect column for all imperfection magnitudes. It is seen
from Figs.8 and 9 that for the unloaded column an imperfection
magnitude of w,,/t =0.5 results in a 20 % loss of section flexural stiffness
and a 26 % shift in neutral axis position towards the web centre line. For an
imperfection magnitude of wo,/t = 1.0 the loss in flexural stiffness is 43 °%
and the shift in neutral axis is 57° towards the web centre line.
Figure 10 shows the effect of local imperfections and local form change

0 4 8 12 16 20 éc

bt = 0°45 bw = O10) t =2145 e =0.0, d0¢ _g¢

bw t bw t
Fig. 10. Effect of local imperfections on equilibrium behaviour.
 Locally Imperfect Plain Channel Columns 65

with load on the equilibrium behaviour of the centroidally loaded column

considered earlier in Fig. 5. As before, the broken curves are based on an
unchanging locally deflected shape while the full lines take local form
change into account. Equilibrium curves for two local imperfection
magnitudes of wo,/t =0.1 and 0.2 respectively are shown as well as those
for the locally perfect column. The imperfection curves signifying an
unchanging locally deflected shape with load are seen to reach ultimate
load values and then to unload very gradually with increasing column
deflection whereas the curves which allow for local form change are seen to
have lower ultimate loads and to unload at a much higher rate with
increasing deflection. It is clear from this that the equilibrium curves
obtained from a consideration of the combined effects of local
imperfections and local form change are quite different, for this design,
from those obtained from the imperfection analysis which precludes local
form change.
Figure 11 shows the ultimate load sensitivity of this column to local

1.0

0.9

bt
ey =0-4,
bw
t = 100,
L
bw = 2t49,
a
e =(:0)
Soc_
t =00

Fig. 11. Reduction in maximum load due to local imperfections.

66 J. Loughlan and A. R. Upadhya

imperfections. Again the broken curve signifies an unchanging locally

deflected form whereas the full line curve takes this into account. Since the
column is a simultaneous mode design it will obviously be imperfection
sensitive; however, the local imperfection analysis which includes local
form change is seen to have a higher degree of sensitivity by an amount
which depends of course on the local imperfection magnitude. A local
imperfection magnitude of w,,/t =0.1 is seen to reduce the ultimate load of
the perfect column by 23 % when consideration is given to local form change
whereas the reduction is 19.5 °% when this is precluded. The corresponding
reductions for an imperfect magnitude of wy,/t =0.2 are 31.5% and 24%
respectively. It is seen, therefore, that the difference between the two
solutions with regard to ultimate load is more noticeable as the local
imperfection magnitude increases although the resulting equilibrium
behaviour from the two solutions is quite different for all imperfection
magnitudes.

5. CONCLUSIONS

A theoretical local imperfection analysis has been presented for plain

channel columns in which the local imperfections are considered to be
similar in shape to the local buckling mode of the perfect column. It is
assumed in the analysis that the material of the column is orthotropic and
that its principal stiffness directions are along the length of the column and
orthogonal to this as shown in Fig. 1.
The results presented in the paper are for isotropic columns and, in
particular, for sections whose local buckling effects are predominant in the
flanges. For columns with a flange-to-web ratio of b,/b, =0.4 it has been
shown that the effects of allowing the local deflections to change shape with
load are considerable and that the asymptotic approach to a constant
reduced Euler load gives erroneous results, which, depending on column
length and loading eccentricity, can be seriously misleading.
Local imperfections have been shown to substantially alter section
flexural stiffness and neutral axis position. In particular local imperfection
magnitudes of the order of the material thickness have been shown to result
in values for the flexural stiffness and neutral axis position for an unloaded
section which are considerably different from the full section values.
For a simultaneous mode design a consideration of the combined effects
of local imperfections and local form change has been shown to result in
quite different load—deflection equilibrium behaviour of the column from
that obtained from an analysis which does not allow for local form change.
 Locally Imperfect Plain Channel Columns 67

The ultimate loads obtained from the solution which takes this into
account have also been shown to be lower.

REFERENCES

. VAN DER NEUuT, A., ‘Mode interaction with stiffened panels’, Proc. [UTAM
Symp. on Buckling of Structures, Harvard University, 17-21 June 1974,
pp. 117-32.
. Korter, W. T. and PIGNATARO, M., ‘An alternative approach to the interaction
between local and overall buckling in stiffened panels’, Proc. 1UTAM Symp. on
Buckling of Structures, Harvard University, 17-21 June 1974, pp. 133-48.
. WALKER, A. C., Interactive buckling of structural components, Science
Progress, 62 (1975) 579.
. Fox, W. C., WALKER, A. C. and RuHopes, J., Buckling of locally imperfect
stiffeners in plates, Proc. of the ASCE, J. of the Eng. Mechs. Div., 103,
No. EMS, October 1977.
. GRAVES SMITH, T. R., The effect of initial imperfections on the strength ofthin-
walled box columns, J. Mech. Sci., 13 (1971) 911-25.
. RHopes, J. and Harvey, J. M., ‘Interaction behaviour of plain channel
columns under concentric or eccentric loading’, Second Int. Colloquium on the
Stability of Steel Structures, Liege, 13-15 April 1977, pp. 439-44.
. LOUGHLAN, J. and Ruopes, J., ‘Interaction buckling of lipped channel
columns’, Stability problems in engineering structures and components, T.H.
Richards and P. Stanley (eds), London, Applied Science Publishers, 1979,
pp. 179-98.
. LOUGHLAN, J. and Ruopes, J., ‘The interactive buckling of lipped channel
columns under concentric or eccentric loading’, International Conference on
Thin-walled Structures, University of Strathclyde, Glasgow, Scotland, 3-6
April 1979.
. Hancock, G. J., Interaction buckling in I-section columns, Proc. of the
ASCE, J. of the Structural Div., 107, No. ST1, January 1981.
. Upapuya, A. R. and LOUGHLAN, J., ‘The effect of mode interaction in
orthotropic fibre reinforced composite plain channel section columns’, In:
Composite Structures, 1. H. Marshall (ed.), London, Applied Science Publishers,
1981.
. Ruopes, J., HARvey, J. M. and Fox, W. C. The load-carrying capacity of
initially imperfect eccentrically loaded plates, /nt. J. Mech. Sci., 17 (1975)
161-75.
. Upapnya, A. R., ‘A study of the buckling behaviour of composite reinforced
metal panels and composite plain channel section columns in compression’,
Ph.D. Thesis, Cranfield Institute of Technology, Cranfield, Bedford, England,
September 1980.
. BIJLAARD, P. P. and FisHeEr, G. P., ‘Interaction of column and local buckling in
compression members’, NACA Technical Note 2640, 1952.
. BULAARD, P. P. and FIsHER, G. P., ‘Column strength of H-sections and square
tubes in post-buckling range of component plates’, NACA Technical Note
2994, 1953.
 An Examination of Shakeout in Stiffened Plates

P. A. FRIEZE and E. DRyYMAKIS

University of Glasgow, UK

ABSTRACT

A numerical procedure for the elasto-plastic large deflection analysis of

axially loaded stiffened plates treated as beam-columns is described. The
procedure is used to examine the effect on compressive behaviour of welded
flat-bar stiffened plates of varying degrees of tensile pre-load. Design curves
for a practical range of plate panels and flat-bar stiffened plates is then
presented.

NOTATION

Cross-sectional area.
Smallest and largest eigenvalues.
ok
a o>
oO Plate length and width.
Tangential axial rigidity.
S Gy z Damping coefficients for u and w displacements.
Tangential flexural rigidity.
oh
ao
eke™~7) Flat-bar stiffener dimensions.
Modulus of elasticity.
>= Damping factors for u and w displacements.
Column length.
Cross-sectional moment.
TE
Ah Axial load.
69
> Oo P. A. Frieze and E. Drymakis

Radius of gyration of cross-section.

Time, increment Af.
Plate thickness.
Axial displacement.
= r=) Under-load and initial out-of-plane deflections.
Axial direction.
Out-of-plane direction.
Ratio of stiffener area to plate area.
Non-dimensional plate slenderness.
oO
Magnitude of column initial lack-of-straightness.
ttm
ett
te
eae
eg
ah Node spacing (Fig. 2).
lo)ok Magnitude of plate initial out-of-plane deflection.
Axial strain.
Curvature.
Non-dimensional column slenderness.
QArvoo Stress.
Cr Plate compressive welding residual stress (Fig. 3).
Oy Yield stress.
Py» Py Densities for u and w displacements.

Subscripts
a End of a time increment.
b Beginning of a time increment.
i Node reference number (Fig. 2).
Z At distance z.

Superscripts
Value non-dimensionalised with respect to value at yield.
/

Differentation with respect to time.

1. INTRODUCTION

Residual stresses in steel structures arise from a variety of processes some of

which are:

(i) Hot- and cold-rolling as employed in the production of sections

and plates;
(ii) _Cold-bending as used during the formation of curved components,
both sections and plates;
 An Examination of Shakeout in Stiffened Plates 71

(iii) Flame-cutting as adopted extensively in the preparation of plates

for bridge, ship, platform and similar structures; and
(iv) Welding as used almost exclusively for the jointing of plates and
sections in most steel construction.

Both the pattern and magnitude of these stresses vary enormously. Their
effect on the response of structures also demonstrates a wide variation
which is, however, probably more dependent on the structural action in the
member concerned than on the source of the stresses themselves.
The effect of hot-rolling is probably most noticeable in structural
sections of thicker proportions. This process tends to produce stresses
which show little variation along the length of the member but significant
changes across the section. The latter arise from differential cooling of the
section following rolling which generally results in tensile stresses being
developed in the thickest portions, such as the flange—web junctions, and
equilibrating compressive stresses elsewhere. These compressive stresses
can reach 50% of yield and since they occur at the tips of flanges have
proven to have a severe weakening influence on such sections subjected to
compression loading.
Cold-rolling of plates tends to produce stresses again uniform along the
length and varying through the thickness. These arise from plastic working
of the material. They are generally not large in magnitude but usually result
in a reduction in stiffness before yield stress is reached under both tensile
and compressive loading. In coupon tests this produces a ‘rounded’ stress—
strain curve. From a practical viewpoint, if the plating is subjected to
further processing, such as cold-bending or welding, the stresses introduced
by these are usually of greater magnitude so that the cold-rolling ones can
be considered as secondary.
Cold-bending particularly of sections can result in significant residual
stresses. These exhibit a ‘zig-zag’ pattern through the height of the section
and in particular circumstances can seriously reduce the strength of
members in flexure.' The magnitude of these stresses and their
consequences can be reduced by the use of double-bending where the
member is over-bent before bending back to the required curvature.
Flame-cutting of plates tends to produce a pattern of stresses similar to
the longitudinal ones which arise from welding, viz. tensile along the line of
the heat input with equilibrating compressive stresses across the remainder
of the plate. As flame-cutting is generally precursory to welding, its residual
stresses tend to be secondary in their effect on structural response.
Welding produces complex patterns of stresses which are manifested in
12 P. A. Frieze and E. Drymakis

two ways. First, there are those directly associated with the weld and the
immediately surrounding parent material, the ‘heat affected zone’. These
are usually tensile in nature and of a magnitude at least equal to the yield
stress of the material (property changes sometimes occur during welding to
produce a material of higher yield stress). Second, there are those which
arise as a consequence of maintaining equilibrium with the weld-associated
tensile stresses. These are obviously compressive in nature, unless the plate
is unrestrained when shrinkage will occur instead, and may be of a
magnitude up to 40% of yield if the welding is predominantly uni-
directional. Considerably higher values may result if orthogonal weld
patterns are adopted. As in the case of hot-rolling of sections, it is the
compressive stresses that generally are of concern from a structural
strength point of view since, in conjunction with compressive loading, they
precipitate either premature buckling or yielding both of which tend to lead
to some form of general or overall instability.
In attempting to deal with residual stresses, designers have normally used
one of the following:

(1) Strength formulations which account for their detrimental

influence. This approach has been used, for example, in the design
of columns’ and of welded plating under complex loading.*
(ii) Stress-relief, where by the heating of welded components with the
consequential reduction in yield stress, plastic flow occurs to relieve
particularly the higher levels of built-in stresses. This approach is
recommended for complex configurations involving thick sections
of plates such as pressure vessels, joints etc., and is as much an asset
from the fatigue viewpoint as it is for strength.

Both of these approaches are perfectly satisfactory, with the former

probably now tending to predominate as sophisticated numerical
procedures allow most residual stress patterns to be incorporated within
the analysis and their effects on strength to be determined directly.
An alternative approach is available, and it is probably one that is used
more by default than by active consideration although the structural
engineering profession is aware of its potential and certainly expresses
taking implicit, if not explicit, advantage of it. This is the concept of pre-
loading in which a structure is deliberately, or inadvertently, loaded in the
direction opposite to that which it will experience in practice. This is done in
order that residual stresses of a high value but of opposite sign to the in-
service stresses will be strained into yield and, as a result of plastic flow, be
 An Examination of Shakeout in Stiffened Plates 73

of smaller magnitude upon unloading. More importantly, however, the

equilibrating stresses which are in the same sense as the in-service ones will
also be relaxed. This would appear to be of particular benefit if the in-
service stresses are compressive since it would tend to reduce the possibility
of failure arising from the onset of buckling either at a Jocal or an overall
level.
Inadvertent pre-load usually occurs sometime during the construction
phase of a structure such as in fabrication, transportation or erection. It
may even occur in-service when, for example, alternating loading of the
structure takes place as in a ship at sea. Under these conditions, the
relaxation of residual stresses that occurs is usually referred to as
‘shakeout’.
Explicit advantage cannot usually be taken of shakeout except in simple
structures through lack of data. Although a number of studies have been
performed on structures in compression which examine in detail the effect
of residual stresses of various origins, none appears to have considered
tension pre-load in an effort to quantify the benefits that may be gained
from shakeout.
In an attempt to shed some light on this phenomenon in welded stiffened
plates typical of those used in the construction of bridges, ships and decks
of platforms, the work described in this paper was undertaken. It was a
numerical investigation into the longitudinal compressive strength of
stiffened plates following the application of tensile loading in which the
beam-column approach was adopted to model the stiffened panel, and the
residual stresses arising from welding were considered explicitly in a
separate plate panel analysis used to derive load—end shortening relations
to serve as ‘stress-strain’ curves for the plate section of the beam-column.
Initial imperfections were considered in both the plate panel and stiffened
panel analyses. The magnitudes selected were appropriate for the
structures in question.

2, -MODEL

2.1. Beam-column Representation

A beam-column model of a stiffened plate consists of a stiffener plus an
attached piece of plating of width equal to the stiffener spacing spanning
between transverse frames. Discretisation of the model can take various
forms depending on the degree of sophistication of the adopted numerical
74 P. A. Frieze and E. Drymakis

procedure. The one used in the present work is finite difference based and
assumes the following:

(i) Plane sections remain plane and normal to the neutral axis;
(ii) Torsional behaviour of the stiffener can be ignored;
(iii) The stress-strain curve for the stiffener is identical to that of the
basic material, here steel;
(iv) The stiffener is initially stress-free;
(v) The axial response of the plating can be determined via a separate
stress-strain curve;
(vi) The plating is initially stress-free although the specification for the
plate from which its stress-strain curve is derived may contain
initial stresses;
(vii) Deflections normal to the longitudinal axis are restricted to the
plane of the stiffener;
(viii) Shear and normal stresses are negligible.

2.2. Beam-column Analysis

The numerical procedure used for the analysis requires the governing
beam-column differential equations to be written in finite difference terms.
The equations used were the large deflection ones in which equilibrium is
governed by
d*M d*w dw
dx? se (oO a it )=0 ”
and
dP
—=0
cE 2
(2)

where M and P are the moment and axial load acting on the cross-section
and w, and w are the initial and ‘under load’ deflections normal to the
longitudinal x-axis, and compatibility by
d*w
as (3)
and
du 1/dw\? dwdw,
: = +3 ($) + ax dx 4

where ¢ and ¢ represent the curvature and axial strain in the section, and wis
the axial displacement.
 An Examination of Shakeout in Stiffened Plates 75

Assuming plane sections remain plane, the strain on a fibre distance z

from the neutral axis (see Fig. 1) is
&,=&+2h (5)

Assuming zero normal stresses, invoking Hooke’s Law leads to

o, = E,(€ +z) (6)

where E, is the modulus which varies over the cross-section because of the
non-linear material behaviour of both the plate and stiffener.
P and M can then be found by integrating the stress and the first
moment of the stress across the section:

Pai o,dA
A

=| E,dA +6 |E,zdA
A A

=Ce+G¢ (7)

m= | o,zdA
A

=) B24 +6 |E,z’dA
A A

=Ge+D¢ 7 (8)
where C, G and D are referred to as tangential rigidities. Since E, is stress
dependent, C, Gand D have to be evaluated frequently as loading proceeds.
This process is facilitated by recasting eqns (7) and (8), and (3) and (4) in
incremental form.
The numerical procedure used is referred to as dynamic relaxation or
DR. It is an iterative procedure which uses eqns (1) and (2) to generate out-
of-balance forces until equilibrium of the solution is achieved. These out-
of-balance forces are applied to equations of motion for the displacements
corresponding to the equilibrium conditions, w and uw respectively, as
follows:

Pw si+ C\,w = [out-of-balance, eqn (1)] (9)

du ;
Py ae + C,u = [out-of-balance, eqn (2)] (10)
76 P. A. Frieze and E. Drymakis

| PLATE WIDTH b |

ts
Fig. 1. Beam-column model.

where (') represents differentiation with respect to time ¢, and p and C are
densities and damping coefficients.
Introducing non-dimensional damping factors of the form
Che
K as (11)
Pe
where At is a time increment, and writing eqns (9) and (10) in finite
difference form, the velocity at the end of an increment (represented by
subscript a) can be found in terms of the velocity at the beginning of an
increment (subscript b) by rearranging:

es [out-of-bala nce,e (ly 2a2

eSDeeieaWw ts a ai)
Ww. = ut-of-
ue
ae el Kee
Lo KP wees 5+
At KD) [out-of-balance, eqn (2)] (13)
 An Examination of Shakeout in Stiffened Plates TAL

Displacements can then be found by integrating velocities:

Ww, =W, + Ww, At (14)

u, =u, +u, At (15)
Through the use of interlacing finite differences in time, the velocities w,
and w, occur at a time interval Az/2 before and after the displacements w,.
In an effort to achieve optimal convergence with this procedure, fictitious
densities are introduced. These are derived by integrating eqns (12) and (13)
with respect to time and invoking the analogy that exists between this
procedure and the second-order Richardson process* to give
Nice
p) atb
(16)
where a and b are the smallest and largest eigenvalues of the stiffness matrix
implicit in the finite difference discretisation of the structure.
On the basis that a< b and that an estimate of b can be found by
application of Gershgoren’s theorem, and assuming At = 1, expressions for
the fictitious densities can be found? as

) (alfa),&
4D |P| G, ( dw dwo d?w d?wo
Pwo = (xy? * (Ax)? | 4Ax \“ fax
—
dx l
dx? dx L

arrl| aaa sacs

C dw dw, dw 3) | (18)
=—— | (2+ 2 a
dx dx dx i+1
where subscripts i and i + | refer to the nodes created by the finite difference
discretisation. Interlacing meshes were used in this as illustrated in Fig. 2.
Strictly C and D in eqns(17) and (18) should be node dependent.
However, as with other applications of DR® use of the initial values for
these variables proves to be adequate for a complete analysis.

| = Nl Ax =

eee Anisre cae ey:

@ us, M,PRNODES
—»u NODES
Fig. 2. Spatial interlacing meshes.
78 P. A. Frieze and E. Drymakis

Damping factors are selected to lie just below critical so that oscillations
can occur about the static values. Apart from providing more rapid
convergence than damping just in excess of critical, it provides a visual
means by which convergence can be evaluated in printed output. The sums
of the square of the w and u velocities are also used to provide a second
means of checking convergence.

2.3. Plate Panel Analysis

The procedure used for the plate analysis was one developed previously by
the first author. It is based on the von Karman large deflection plate
equations as modified by Marguerre to include initial imperfections.
Plasticity is incorporated on a single-layer basis using Ilyushin’s yield
criterion and the Prandtl—Reuss flow rule. Initial stresses are included to
enable welding residual stresses to be modelled according to the idealised
distribution shown in Fig. 3. DR is used to solve the equations. Details of
the procedure can be found in ref. 6.

a OE STRESS LEVEL
Sf,
PLATE WIDTH b
Fig. 3. Idealised welding residual stress distribution.

3. COMPARISONS WITH EXISTING PROCEDURES AND

RESULTS

Before embarking on any studies, comparisons were conducted using the

beam-column analysis in order to establish confidence in the method. The
plate panel analysis had previously been compared with other numerical
procedures’ and with test data® and could be relied upon to give good
estimates of plate response.
Comparisons are only presented for stiffened plates having flat-bar
 An Examination of Shakeout in Stiffened Plates 79

stiffeners since this was the only form of stiffener incorporated within the
procedure.

3.1. Numerical Solutions

Results from numerical studies on stiffened plates are reported in
refs. 9-12. Figure 4 shows a comparison between results predicted by these
methods and by the current procedure. Details of the geometries, initial
deformations etc. can be found in ref. 5.
The figure shows that generally very good agreement is achieved
although compared with the method of ref.9 the current procedure
predicts slightly higher strengths.
It was noted in conducting this exercise that none of the reported
procedures could predict post-buckling response: the present procedure
has this capability.

3.2. Experimental Results

Results from tests on stiffened plates are reported in refs. 13 and 14.
Comparisons between these and the strengths as predicted by the current

10
PRESENT
METHOD

0:8

0°6

0-4

0:2

0 0:2 04 0:6 0:8 1-0

OTHER NUMERICAL METHODS
Fig. 4. Comparison between numerically predicted strengths for stiffened plates.
80 P. A. Frieze and E. Drymakis

1-0
PRESENT
METHOD

0:8

0-6

0-4

0-2

0 0-2 0-4 0-6 0-8 1:0

TEST DATA
Fig. 5. Comparison between experimental results and numerical predictions for
stiffened plates.

analysis are presented in Fig. 5. Details of the models are summarised in

ie) ess
The comparison shows that the analytical procedure generally
underestimates strength, but only slightly. When taken in conjunction with
the correlation demonstrated in Fig. 4 for values at the upper and lower
ends of the slenderness range, where no experimental data were available,
the developed numerical procedure can obviously be used with confidence
for further studies on stiffened plates.

4. PARAMETRIC STUDY

In conducting the study into the effect of shakeout on stiffened panel

strength, an effort was made to select geometries, material properties and
initial imperfections appropriate to stiffened panels used in bridge,
platform deck and ship structures.
Generally the most difficult of these to quantify are the initial conditions,
 An Examination of Shakeout in Stiffened Plates 81

particularly those of shape in relation to plate panels. This arises not

necessarily because of any lack of data but because of the difficulty in
deciding what magnitude of plate initial bow is effective in determining
plate strength when the strength analysis is performed in isolation. It is
generally recognised that it is the longitudinal profile that will have the
greatest effect on longitudinal strength although a lot of data, and
specifications, relate to transverse measurements. In keeping with some
previous studies (for example, see ref. 15), an initial sinusoidal bow of
0.145ft [B =(b/t),/o,/E and a, is the yield stress] was selected: this
corresponds to 5/200 in the case of mild steel.

4.1. Plate Panels

Before embarking on the plate panel study, an initial investigation was
conducted in order to select the aspect ratio which would produce
minimum strength. For this purpose, plates of slenderness £ = 1.04, 1.73,
2.77 and 5.19 (b/t = 30, 50, 80 and 150 for mild steel) and aspect ratio 0.7,
0.8, 0.9 and 1.0 were analysed under longitudinal compressive loading in
the presence of 10% compressive residual stress and an initial bow of
0.145ft(a/b). Making the magnitude of the initial bow a function of the
aspect ratio a/b ensured that the longitudinal curvature did not vary from
one aspect ratio to the next.
In this study, and in all the others relating to plate panels, the boundary
conditions assumed were:

(i) rotationally free, i.e. simply supported; —

(ii) zero normal deflections;
(i11) zero shear stress;
(iv) end-loading by prescribed displacements;
(v) unloaded edges unrestrained in-plane.

The results showed that strength did not vary significantly with aspect
ratio and that an a/b of 0.8 would be most appropriate.
Using this aspect ratio, plates of slenderness 1.04 and 2.07 (b/t = 60 for
mild steel) were analysed in the presence of 25 °% compressive residual stress
and an initial bow of 0.116ft (=0.145ft x 0.8) when loaded first by a
limited amount of tension followed by compression up to failure and into
the post-buckling range. The extent of tensile preload used were strains
equal to 0, 10, 25 and 40% of tensile yield. The results are presented in
Figs. 6 and 7 in the form of average stress-strain curves where both stress
and strain have been non-dimensionalised with respect to their yield values.
 82

50
NOISN3IL OVO1-3Yd
%) 49 NIVYLSOT3IA
(

S:0-
P. A. Frieze and E. Drymakis

YA)

advIDAY UIBIIS—Ssd1]
SIAIND
s JOJ

9
passordu09
sajejd YIM snoueA saazZap
Jo uorsua) ‘peoj-aid
9 = ‘p01 49/2 ‘80=
*2 =

“Sly
75Gc
90= O FONT
 ZZ NOISN3L QVO1-3ud

S°0
%) JO GIZA (NIVYLS

-0-
An Examination of Shakeout in Stiffened Plates

| 0 a “+L BBRIDAY UIR.IIS-SSI}S

SdAIND
10J sayeld poyoalqns
0} SNOLIRA sao1Zap
Jo uoIsud} ‘peoj-aid
J = ‘LOT4/0= 49°80
= ST %
°¢e 0= 19911
83
84 P. A. Frieze and E. Drymakis

From the figures it can be seen that the initial loading into tension has a
slope less than unity. This is partly due to the presence of the initial bow
which tends to reduce the stiffness of plates under any in-plane loading and
will be most pronounced for the more slender plates. It is predominantly
due to the edges ofthe plate being initially at tensile yield (see Fig. 3) so that
any tensile loading will render these strips ineffective.
Upon loading in compression, the initial slope appears to be almost
independent of the initial level of tension. This is not unexpected as this is
dictated by the level of initial bow, and for the degree of tension loading
applied little change in this was noted. In-plane yield as manifested by the
rapid change in slope at a compressive strain ratio of about 0.7 is seen to be
affected by the extent of pre-load. Beyond that point, the differences in the
responses are maintained until they converge upon a single post-buckling
curve. From previous studies > this is known to be the post-buckling curve
for the same plate when initially free of any stresses.
This tendency to converge to the initially stress-free result implies that
the strength of stocky plates will be independent of the degree of tension
applied, and that maximum differences will probably occur in the
slenderness range where, for the level of initial bow considered, residual
stresses have their most pronounced influence, i.e. 6 between 1.7 and 1.8.
Comparing Figs.6 and 7 shows that greater differences are obtained
between the results for different pre-loads when 6 =2.07 than when
B = 1.04.
From acompressive strength viewpoint, the influence of tension pre-load
is not marked. For the f =2.07 case, the maximum enhancement in
strength, due to the 40 % initial tensile loading, is some 7%.

4.2. Stiffened Plates

For the beam-column analysis of the stiffened plates, the curves shown in
Figs. 6 and 7 were used as input for the plate stress-strain behaviour. For
the stiffener, an elastic—perfectly plastic material response was assumed.
Just as for the plates, the beam-columns were subjected to different degrees
of tension before compression was applied. In the analysis, a record had to
be kept on the extent of tensile strain that was generated in the plating at
each node so that when compression was applied the correct curve would be
accessed in order to determine the plate current stiffness. If this was
intermediate to the values 0, 10, 25 and 40 % of yield strain, interpolation
was used. For values of compressive strain in the plating in excess of 2.6
times yield strain, linear extrapolation of the plate curves was used.
In all the beam-column analyses, the ends were assumed simply
 An Examination of Shakeout in Stiffened Plates 85

supported and non-deflecting perpendicular to the column axis. Loading

was applied by means of end displacements and symmetry about the centre-
line invoked. The column initial bow was always sinusoidal in shape and
disposed in the direction of the stiffener. A variety of column slendernesses
was examined in the range 4 =0.3-2.5 where A= (//nr),/o,/E, / is the
column length and r the radius of gyration of the entire cross-section. In
order to satisfy the requirement that torsional buckling of the flat-bar
stiffener should not be a possibility, the slenderness ratio of the outstand,

1:0

A=0:57
TENSION PRE- LOAD
(% OF YIELD STRAIN )

iS 0:4

TENSION PRE-LOAD
(% OF YIELD STRAIN)

= 0:4

Stith
Fig. 8. Average stress-strain curves for stiffened plates subjected to various
degrees of tension pre-load, $ = 1.04, «=0.4, Ay =//1000.
86 P. A. Frieze and E. Drymakis

d/t,, was set to 10. The ratio of stiffener area to plate area was generally
taken as 0.4.
Typical results in the form of non-dimensional stress—strain curves are
presented in Figs. 8 and 9. For Fig. 8, a plate of slenderness # = 1.04 and an
initial bow of magnitude A, =//1000 was used: the corresponding values
for Fig.9 were 2.07 and //500. For clarity, generally only the results
pertaining to tension pre-loads of 0 and 40 % are illustrated.

1-0 > = 0-57

. TENSION PRE-LOAD
oO (% OF YIELD STRAIN)

Neal 0)
Oe TENSION PRE-LOAD
(% OF YIELD STRAIN)

ee 0-4

Fig. 9. Average stress-strain curves for stiffened plates subjected to various

degrees of tension pre-load, 6 = 2.07, «=0.4, Ay = //500.
 Y31NI SAMND
g 29 Yo

31V1d
40-1 4989LL-0 %SZ
L0:2 0 499Lb %SZ

Oe
20

o”7
SENEFEDIES

©e)
/
000b/2
005
7-0 9-0

S-0
NOISN3L OVO1-3Yd
%) 40 CT3IA (NIVYLS
An Examination of Shakeout in Stiffened Plates

5-0 0-1 §-4 0-2 4 =’ jn

“Sly oy y3uens
jo pauaynssajejd pojsefqns
0} snoueA soorZap
Jo uorsusy ‘peoj-o1d
87
88 P. A. Frieze and E. Drymakis

Under tension, the initial response of the beam-columns is similar to that

of the plates alone. Differences between the curves are maintained more or
less until column buckling occurs. This can be seen to occur earlier for the
higher levels of tensile pre-load so that the post-buckling response for these
cases falls below those of the smaller levels of pre-load. For more slender
columns, column behaviour dictates the response completely and virtually
no differences are detected for compressive strains in excess of 0.4 times
yield strain.
The effect on compressive strength of tensile pre-load is summarised in
Fig. 10. It shows that for the range of cases examined the benefits of
shakeout are limited to some 5%.

5. DESIGN RECOMMENDATIONS

Having shown that residual stress shakeout did not have a significant effect
on compressive strength, the analysis was then used to examine the strength
of plates stiffened by flat-bar stiffeners when subjected to compression only.
For stiffened plates subjected to a longitudinal stress gradient, the strength
of the plates at the end of the panel will sometimes dictate strength. Plate
strength is considered in section 5.1. For panels subjected to small or zero
stress gradients, beam-column strength will be paramount: this is discussed
in section 5.2.

5.1. Plate Panel Strength

A range of plates having the same specification as those considered in
section 4.1, but with the addition of residual stress magnitudes of 0, 10 and
40 % of yield stress, were analysed. The unloaded edge boundary condition
of unrestrained in-plane was retained in spite of the use in previous studies
of the constrained edge condition argued as appropriate for plate panels
located away from the edges of stiffened panels. It has been noted from the
computer studies that the constrained edge condition generates very high
transverse in-plane stresses particularly in slender plates, which are unlikely
to be resisted completely by the stiffness of the surrounding structure.
Moreover, unrestrained plate data is needed for the design of edge panels of
stiffened plating and of the faces of unstiffened box beams and columns.
For the sake of simplicity, only one set of data should be provided and the
unrestrained results offer the most generality in this respect.
The average stress-strain curves generated by the analysis are presented
in ref. 5. A summary is presented in Fig. 11 which shows the variation of
 An Examination of Shakeout in Stiffened Plates 89

i a/b = 08
g! 55 = 0-116 Bt
On (%)

08 0
----- 10
en 5
0-6 Saree
a aU

0-4

0-2

0 aaa ee 2 See == fe ee eee

1-0 2:0 3:0 B 4-0

Fig. 11. Plate design curves.

plate strength with slenderness ratio 6. These can be used for determining
the strength in longitudinal compression of a practical range of plates.

5.2. Stiffened Plate Strength

A range of stiffened plates having a specification similar to that described in
section 4.2 have been analysed. Details are presented in ref.5 and
summarised here in the form of strength versus slenderness 4 curves in
Fig. 12. They can be used directly for the design of welded flat-bar stiffened
plates subjected to longitudinal compression: they correspond to plates

EULER PLATE
a! B 4 Op
nis @ 1:04 o16pt 25%
@ 2:07 ON6Bt 25%
STIFFENER
0:6

0-4

0-2

0 0:5 10 feo 2:0 hE

Fig. 12. Stiffened plate design curves.
90 P. A. Frieze and E. Drymakis

having residual stresses of 25°% of yield stress, and although residual

stresses in the flat-bar stiffeners have been ignored, the effect of these can be
included by using an increased level of column initial bow. For example,
A, = //500 might be appropriate when the initial bow is actually //1000.

6. CONCLUSIONS

A numerical procedure for analysing welded flat-bar stiffened plates

subjected to longitudinal compression is described. Comparisons with
available test data and other numerical procedures show it can be used with
confidence in the prediction of stiffened panel compressive strength. In
contrast with most other numerical analyses, it can also be used to predict
the post-buckling behaviour of such structures.
The procedure has been used to examine a range of practical stiffened
plates subjected to tension pre-load followed by compression. The effect of
this residual stress shakeout was not marked, amounting to some 7% in
relation to plate strength and 5 % in relation to stiffened plate strength.
The procedure was then used to examine the compressive strength of a
range of plates and stiffened panels having specifications appropriate to
these structures as used in bridge, platform deck and ship construction. The
results are presented in the form of non-dimensional strength versus
slenderness curves and can be used directly for design.

REFERENCES

1. SmitH, C. S. and Kirkwoop, W., ‘Influence ofinitial deformation and residual

stresses on inelastic flexural buckling of stiffened plates and shells’, in Stee/
plated structures, Dowling et al. (eds), London, Crosby Lockwood Staples,
1977, pp. 838-64.
2. ECCS Recommendations for Steel Construction, European Convention of
Constructional Steelwork, ECCS-EG, 77-ZE, 1978.
3. BS 5400, Part 3: Code of Practice for Design ofSteel Bridges, British Standards
Institution, 1982.
4. CASSELL, A. C., Shells of revolution under arbitrary loading and the use of
fictitious densities in dynamic relaxation, Proc. Instn Civ. Engrs, 45 (1970)
65-78.
5. Drymakis, E., ‘A numerical elasto-plastic buckling analysis for stiffened
plates’, MSc Thesis, University of Glasgow, 1981.
6. Frieze, P. A., Hosss, R. E. and Dowtina, P. J., Application of dynamic
relaxation to the large deflection elasto-plastic analysis of plates, Comp.
Struct., 8 (1978) 301-10.
 An Examination of Shakeout in Stiffened Plates 91

. BRADFIELD, C. D. and CLapny, E., ‘A review of the elasto-plastic analysis of

steel plates loaded in in-plane compression’, Cambridge University
Engineering Department, Report No. CUED, C-Struct., TR77, 1979.
. JACKSON, R. and Frigze, P. A., Design of deck structures under wheel loads,
The Naval Architect, No. 3, May 1981, pp. 119-44.
. CRISFIELD, M. A., Full range analysis of steel plates and stiffened plating under
uniaxial compression, Proc. Instn Civ. Engrs, 59, Part 2, December 1975.
. MooLanl, F. M. and Dow ina, P. J., ‘Ultimate load behaviour of stiffened
plates in compression’, in Steel plated structures, Dowling et al. (eds), London,
Crosby Lockwood Staples, 1977, pp. 51-88.
. SOREIDE, T. H., MOAN, T. and Norpsve, N. T., On the behaviour and design of
stiffened plates in ultimate limit state, J. Ship Research, 22(4) (December 1978)
238-44.
. CARLSEN, C. A., A parametric study of collapse of stiffened panels in
compression, The Structural Engineer, 58B(2) (June 1980).
. Horne, M. R. and NARAYANAN, R., ‘Ultimate load capacity of longitudinally
stiffened panels’, and ‘Further tests on the ultimate load capacity of
longitudinally stiffened panels’, Simon Engineering Laboratories, University of
Manchester, 1974.
14. Dorman, A. P. and DwiGcut, J. B., ‘Tests on stiffened compression panels and
plate panels’, Proc. Int. Conf. Steel Box Girders, Instn Civ. Engrs, London,
February 1973, paper 5.
1S FriEzE, P. A., DOWLING, P. J. and Hosss, R. E., ‘Ultimate load behaviour of
plates in compression’, in Steel plated structures, Dowling et al. (eds), London,
Crosby Lockwood Staples, 1977, pp. 24-50.
 . pai

7 ae ;

ears —
eae i 3 7
ete C pletene ee yhted haat
aaah ce \ ees 7a om od
tarrite (— “eet Ovencs>
AE
higtc Reon sans RET Aniwr nas tan 8 a
a5 da *) 0 © pot nes wi
HW toje" rA.Yenliey , 2 ine i) Tuite ie
4, 1
oo eer See
AeA treete trian ve
 The Cold Formed Industry Scene

R. D. HOMER
Metal Sections Ltd, Oldbury, UK

ABSTRACT

In common with most manufacturing sectors, the cold formed industry scene
is dominated by the recent dramatic contraction of its markets. After
consolidation and retrenchment, the industry is currently seeking ways to
offset reduced and replaced lost markets. A study of past product development
can often reveal helpful leads or pointers and so the paper covers three
decades of such work. Whilst the developments are related to Metal Sections
Ltd, they are intended to represent the industry in general. Factory buildings
with three designs of main frames are described; an aircraft hangar door, a
product from a special variation of the cold formed process which found
unexpected major use in civil engineering and a composite building panel
with a particular example of use are also featured. In conclusion, the paper
deduces two factors valid for consideration. The reader is recommended not
to be put off by size and to include fields previously thought inappropriate.

1. INTRODUCTION

In common with most manufacturing sectors, the cold formed industry

scene is dominated by the recent dramatic contraction of its market. The
general consensus of opinion is that the market has shrunk some 40%
compared with the end of the 1970 period (which was by no means particularly
prosperous). A period of rapid consolidation and retrenchment set in once
93
94 R. D. Homer

it was realised that this was likely to be a long-term trend and currently the
industry is urgently seeking ways to increase its penetration into existing,
and to find replacements for those lost markets.
A study of past product development can often reveal helpful leads and
pointers for formulating plans of action and so the paper is concerned with
this aspect. Since it is only possible to describe work in which I have been
involved, the paper will review some three decades of work at Metal
Sections Ltd. It is intended, however, that the developments described
should be regarded as indicative of the industry as a whole.
Appropriately, the decade of the 1950s—the period in which my story
begins—found Professor Harvey in industry and he took part in some of
the activities described.
The period immediately after the Second World War was an exciting
time, as finally the fetters of a war economy were being loosened. There
was a grave shortage of many things, even essentials, and a large unfilled
demand for a wide variety of industrial products. The cold formed industry
of those times was enterprising and entrepreneurial, even as it is now, and
willing to push out the frontiers of its markets as far as it could. One
area where a major effort was made was the building industry and it is
mainly, although not exclusively, in this field that I wish to describe a few
developments.

2. STORAGE BUILDING AT BURTONWOOD

In November 1953, a successful tender was confirmed by Sir Alfred

MacAlpine, who were the main contractors for a very large storage
building to be erected at Burtonwood in Lancashire. The customer was the
Air Ministry acting on behalf of the United States Air Force, who required
a central stores complex to serve the European operational field.
Our tender was to include for main frames, eaves girders, purlins,
sheeting rails and all secondary steelwork for a single storey industrial
building, covering some 1.25 x 10° square feet (approximately 115000 m7?
or 11.5 ha) (see Figs. 1, 2 and 3). In our design provisions we had gone for
three-pinned portal main frames and latticed purlins and sheeting rails. The
building was split into two basic areas, known to us as the main warehouse
and the Header building. For the main warehouse, the three-pin portal had
a span of 50’0” (15.24 m, was 15’0” (4.6m) to the eaves, had a pitch of 22°
and was spaced at 202” (6.15m) centres. For the Header building, the
spacing for the portals was doubled to 40’ 0” (12.3 m) and lattice rafters at
 The Cold Formed Industry Scene 95

Fig. 1. General view of building at Burtonwood made from cold formed

sections.

ans AN

<<
“Spence

Fig. 2. View of main warehouse steelwork.

 Fig. 3. View of Header building steelwork.

10’0” (3.05 m) were supported at eaves and apex by Warren girders which
spanned the 40’ 4” between portals (a much greater unobstructed area was
specified for the Header building).
Specified design loads were as follows:

Roof covering (corrugated asbestos) 3.0 lb/ft?

Insulation (under purlin) 1.0 lb/ft?
Superload (for snow) 15.0 lb/ft?
Service 6.0 lb/ft?

TOTAL 25.1 lb/ft?

= 1.21kN/m?
In addition, a point load of 1120 1b (SKN) was to be carried at the apex of
the portal. Wind loads to CP3 Chapter V were also considered but found
not to be critical in the design.
 The Cold Formed Industry Scene 97

Three-pinned arches are, of course, completely determinate statistically,

but material usage had to be carefully controlled if the contract was to be
economically viable. Accordingly, it was decided to base the main frame
design using C sections (boxed channels) and to reinforce locally where
required with channels welded inside. For the main warehouse a
x” x 34” x 0.08” (89 x 89 x 2mm) C was chosen as main members, whilst
for the Header building, the C was increased to 4” x 35” x 0.13”
(101 x 89 x 3.25mm). Figure 4 shows the method of building up the C for
local high loads. Spot welding was generally used to attach the stiffening
channels to the C main section and so the effectiveness of the welding was
the key to success. Studies were carried out in the factors governing the
consistency, quality and control of the welding. From these, it was found
that, generally speaking, welding on compression members was not too
critical and spots at regular spacing were satisfactory, but welding at the
ends of tension members had to be very carefully done for the effective
transfer of forces. Fortunately, the need for stiffness in the tensile areas was
not very great.
In Europe (though of course much less so in the USA) the building design
involved here was considered very new and, notwithstanding the fact that
the Air Ministry inspectors examined the steelwork at each stage of
manufacture, the Air Ministry gave us the option of giving a 10 year
guarantee for the structure or carrying out full scale loading tests. Since the
10 year guarantee was thought to be a negative approach, which could
mean different things at different times, the test option was accepted. It was
specified that the test frame should be loaded up to a 50% overstress.
Representative structures for both the main warehouse and the Header
building were erected at the works and tested using calibrated springs as the
load source. Figure 5 shows the set-up with loading mechanism attached.
The tests, in fact, were taken beyond 50% overstress without failure,
reaching 50.7 lb/ft? in one case, and thus the design was validated.
To conclude on this project, a few facts and figures might be of interest:

(1) The tender value accepted by the main contractor for the main
warehouse and the Header building’s structural framing was
£193,000.
(2) Atthe time the building was the largest in Europe by a considerable
margin.
(3) The time from the go-ahead to completing the weather proofing
was 14 months which was an all-time record for a building of this
size and which earned the main contractor considerable acclaim.
 ‘OSNOYIIVM UIPU JO SUOTIIIG “fp “BI

Is™GGy
6S) SOE]GO-Z
ys zLoLE = AAT

+e"“Sr ys"
"SN DEC / = xx>

«
=J
y"GrDE ‘sm ZZ) = “my

w
srcle} = A>
Pd lsoe = XXT

"SMIOS
yINGIG-F = KAT

SIG-O
SN'OS IPE T = YI “WN OSIGSE/ = yay
| =|
ie 2¢
x y id
A A
a dh 6 f
€
ize

ay
 ‘SUIP[INg JopeoH{ “(pluor—p “B14
sw ges} = XS 'SNIIQQ) = ALY
POOL ai AT aan: OSepy = AAT
SMOSODE|

SNES Zig) no me SIG Dey ARM ‘SN! @ZE] = xxy

pr Vesa AAT ~ $692 = xxT
V3RY

eli eon! -z = ~xI

-=OSGIS
SM‘SNOS
avy
Ger]
Yaoy
py
=

7 y
of
ra
&
y
BS , T
|
“
=6 22
w
or
a i ==
=)
a] hee J)
t A
aS
Ins v7 =|
“ie

4
100 R. D. Homer

iia

561,
WE.)=
ae

Fig. 5. Three-pinned arches awaiting test.

(4) Cold formed sections were used entirely in the construction and the
greatest individual thickness of any section was 7,” (4.76mm),
although the gusset plates joining the member were, of course,
thicker.
(5) In spite of there being some corrosion problems (caused by the
proximity to the ICI chemical plants at Widnes and Runcorn),
Metal Sections made a ?% profit on the project.

3. SUBSEQUENT DESIGN STRATEGY

In the aftermath of this project, several conclusions were drawn by the

company and whilst the fact that the contract had proved profitable
encouraged commercial exploitation of the product, it was recognised that
three-pinned arches were not the most economical of frames—even
accepting that with hindsight they were over-engineered. Also, lattice joist
construction was not the most economical for purlins and side rails.
However, it was decided to go ahead with what were called standard
 The Cold Formed Industry Scene 101

buildings, using the same construction but, meanwhile, research studies

were put in hand to develop two-pinned arches and alternative purlin and
sheeting rail designs.
As a result of this work, a range of two-pinned arches were developed to
cover 40’0” (12.2m), 50’0” (15.2 m), 60’0” (18.3 m) and 75’0” (22.86 m)
spans all at 20’0” (6.1 m) centres and with various heights to eaves. These
were of lattice construction but instead of using C shaped sections for the
base profile 33” (89mm) x 2” (51 mm), U sections (plain channels) were
used, again reinforced locally where necessary. Bracing members were box
channels and Corten high tensile low alloy steel was employed. Figures 6
and 7 show two buildings using this type of frame. Again, the design was
tested and Fig. 8 shows the test frame in position, testing-actually being
carried out in April 1957.
Following a visit to the USA, it was realised that a Z section was the ideal
and most economical purlin and side rail design; accordingly, in about
1956 this section replaced the lattice joists previously used. In fact, this was
the original reason for the introduction of the Z purlin into this country,
though its potential for general use was quickly realised.
The third and final phase for these types of buildings was for two pinned

Fig. 6. A building in Dundee using 40’0” two-pinned arch.

102 R. D. Homer

Fig. 7. A building at Oldbury, West Midlands, using cold formed section two-
pinned arches.

LOADING SHACKLE

Fig. 8. Two-pinned arch awaiting test.

 The Cold Formed Industry Scene 103

portals made from two cold formed channels welded back to back. Figure 9
shows a three 30’ (9.1 m) span bay x 80’ (24.4m) building. Designs were
made for 40’0” (12.2m) span frame with the same construction though
none were actually made.
Although the foregoing remarks principally concern the main frame,
complete buildings, clad, insulated and erected, were offered and cold

Fig. 9. View of building using solid web two-pinned arches.

formed section solutions were used for gable framing, wind bracing and
other items of secondary steelwork. In addition, some conventional
triangular trusses were made using similar construction. Figures 10 and 11
show two examples.
Another project undertaken during the same decade was the design of
aircraft hangar doors (Fig. 12). Made entirely from Corten steel cold
formed sections each door measured 32’0” (9.75m) in height, 23’ 6”
(7.16m) long and 12” (0.30m) thick and the specification called for a
tolerance level of +4” (6mm) across diagonals. The doors were erected in a
hangar at London Airport.
104 R. D. Homer

Fig. 11. Cold formed section truss on a school.

 The Cold Formed Industry Scene 165

Fig. 12. Framing for hangar door.

4. DEVELOPMENTS SINCE 1960

During the 1960s, the most noticeable achievement was the development of
the Z purlin which is, of course, still a major product within the industry.
Early in this decade an interesting special development of the cold formed
process lead to the development of a machine for producing lock seamed
spiral tube. Figure 13 illustrates the product which was produced in
diameters from 3” (76mm) to 5’ (1.5m) and in a range of materials and
thicknesses. Originally, it was thought that the tube would find its main
outlet in high velocity air conducting but it eventually found its largest
volume use as voids and piling casings in civil engineering concrete
constructions (Fig. 14).
Towards the end of the decade when, yet again, there seemed a market
for industrial buildings, a composite polyurethane panel, together with a
comprehensive building system for its use, was developed. This found a use
in many types of buildings, including hospitals, offices, as curtain walling
for concrete buildings and was considered for domestic houses, particularly
for hot climates.
106 R. D. Homer

Fig. 13. Manufacture of spiral lock seam tube.

 The Cold Formed Industry Scene 107

Fig. 14. Lock seam tube as voids in road work.

Figures 15, 16 and 17 show different aspects of such a building. It is

particularly interesting because for the project to meet cost limitations it
was decided to use the basic panel as load bearing and so accordingly the
University of Strathclyde was asked to conduct tests regarding bending
stiffness, capacity for carrying axial loads and racking stiffness. A trial
house was erected in South London, having planning permission for a
temporary building from the Local Authority. Due to difficulties with the
customer, the building was there for a much longer period than anticipated
and full building regulations requirements were imposed. Using the
information provided by the University of Strathclyde, we were able to
obtain consent and the only additional steelwork required was four angle
braces at ceiling level to the corners.
The 1970 decade was not conspicuous from a development point of view,
though the Z purlin was further developed and associated components for
complete systems and other forms of cold formed purlins and sheeting rails
made their debut.

5. CONCLUDING REMARKS

Having concentrated on products developed by Metal Sections, it is not

implied that other cold rollers were doing nothing because it is known that
108 R. D. Homer

oh S583 a

Fig. 15. Composite foam filled panel being used in house construction.
 The Cold Formed Industry Scene 109

Fig. 16. Further view of foam filled panels in house construction.

110 R. D. Homer

Ba a:

Fig. 17. View of finished house using foam filled panel walls.

several were also working along the same lines—the Brockhouse ‘Clasp’
system, for example, being a particularly successful project.
In conclusion, are there any pointers from the pastto helpus formulate
plans for the 1980 decade? Two pieces of advice seem relevant: one is not to
be put off by size of the project and, second, not to be afraid to look for cold
formed solutions in areas which it was previously thought impossible or
inappropriate.

Acknowledgement
The Author wishes to thank the Directors and Management of Metal
Sections Ltd, for permission to use material and data from their archives.
 UMR Research on Cold Formed Steel Structures

WEI-WEN Yu
University of Missouri—Rolla, Rolla, USA

ABSTRACT

During the past 14 years, several research projects related to cold formed
steel structures have been conducted at the University of Missouri—Rolla
(UMR) under the sponsorship of the American Iron and Steel Institute
(AISI). The purpose of these investigations has been to study the structural
strength of cold formed steel members and bolted connections. Some of the
research findings have been used in the AISI Specification for the Design of
Cold Formed Steel Structural Members.

1. INTRODUCTION

Since 1969, eight different research projects have been conducted at the
University of Missouri—Rolla (UMR) under the sponsorship of the Ameri-
can Iron and Steel Institute (AISI) for the purpose of studying the
structural strength of cold formed steel members and connections. The
subjects of these investigations were the compressive strength of perforated
elements, performance of thick sheet steels and steel plates, web strength of
flexural members, effect of bending on web crippling of beams, bolted
connections, design of stainless steel structural members, development of
tentative recommendations on load and resistance factor design of cold
formed steel, and automotive structural components composed of high
strength sheet steels. Some of the research findings have been used to
111
112 Wei-Wen Yu

improve the AISI Specification for the Design of Cold Formed Steel
Structural Members.' The objectives and descriptions of each of these
projects are given below along with a discussion of the research findings and
their application to the design of structural members cold formed to shape
from steel sheet, strip or plate. Details of the research are contained in the
publications referred to in the text.

2. RESEARCH PROJECTS

2.1. Cold Formed Steel Members with Perforated Elements

In cold formed steel structural members, holes are sometimes provided in
the webs and/or flanges of beams and columns for duct work, piping, ease
of handling and other purposes. The presence of such holes may bring
about a reduction in strength of individual component elements and the
overall strength of a member. For this case, the load carrying capacity of a
cold formed steel member with perforated elements depends mainly on the
configuration and arrangement of the holes, the material properties, and
the cross-section of the member.
The analysis and design of cold formed steel sections with perforated
elements are rather complex, especially when the shapes of the holes and
their arrangement are unusual. Even though the local buckling strength of
perforated plates has been studied by numerous investigators and a limited
amount of information on the design of relatively heavy steel sections
with perforated elements can be found in several design guides and
specifications,”~'® no specific provisions have been included in the AISI
Specification for this particular type of cold formed steel section.
Therefore, whenever perforated sections are used, special tests must be
conducted and evaluated in accordance with the AISI test procedures to
provide design data for engineers. Of course this requirement for testing of
perforated sections increases the cost of developing new products and may
limit the use of perforated sections.
In an attempt to develop design information for perforated cold formed
steel sections, a preliminary investigation was conducted at the University
of Missouri—Rolla in 1970 for the purpose of studying the structural
behavior of cold formed steel members with perforated elements.!*?°
Based on the results of the study, it was found the effective design width
equations previously developed by Winter?!:*? can be modified for
determining the effective design width for perforated, stiffened, compress-
ion elements having a single circular or square hole. It was also found that
 UMR Research on Cold Formed Steel Structures 113

reduction factors can be used to account for the effects of holes on local
buckling of unstiffened compression elements, shear buckling and web
crippling strength of webs when perforated elements are used.

2.2. Structural Behavior of Thick Cold Formed Steel Members

The AISI Specification’ was originally prepared for the design of cold
formed steel members less than one-fourth of an inch in thickness. During
recent years, cold formed members fabricated from steel sheets and plates
up to one-half of an inch thick have been successfully used in buildings,
bridges, transmission poles and other structures. In some cases, cold
formed steel sections up to three-fourths of an inch in thickness have been
used for steel transmission poles.
In 1971, a study was conducted at the University of Missouri—Rolla to
determine the validity of the 1968 Edition of the AISI Specification for the
design of thick cold formed steel members.”*:?4 This investigation included
studies of the effective design width of stiffened compression elements,
allowable stress for unstiffened compression elements, and the effect of cold
work on material properties. Consideration was given to the manufactur-
ing process, limitation on bend radius, practical width-to-thickness ratios,
material properties and initial imperfections. It was found that the AISI
design formulas for determining the effective design width of stiffened
elements and the allowable stress for unstiffened elements can be
conservatively used for sections cold formed from thick sheets and plates. It
was also found that the AISI design provision for the calculation of corner
yield point provides good results for thick steel sheets and plates having a
sharp yielding stress-strain curve with a small plateau and for cold rolled
sheets and strip. For steel sheets, strip and plates having a sharp yielding
type of stress-strain curve with a large plateau, the AISI design formula can
also provide reasonable results; however, modified formulas may be used
to improve the accuracy of prediction of tensile yield points for corner
sections. Consequently, the applicability of the AISI Specification was
extended in 1977 to the design of members cold formed from steel up to one
inch in thickness.?°
2.3. Web Strength of Flexural Members
In the United States, the design of cold formed steel beam webs is based on
Sections 3.4 and 3.5 of the AISI Specification. In recent years, new types of
cross-sections and materials have been introduced for use in buildings and
other applications. The use of unusual geometric shapes in conjunction
with different types of steel sheet and strip often complicate the design
problems.
114 Wei-Wen Yu

Since 1973, research concerning the webs for cold formed steel flexural
members has been conducted at the University of Missouri—Rolla. This
study dealt with the structural behavior of cold formed steel beam webs
subjected to bending stress, shear stress, combined bending and shear, web
crippling, and the effect of bending on web crippling load. In addition, this
project has also included an investigation of beam webs reinforced by either
transverse or longitudinal stiffeners. The details of the investigations and
the research findings are presented in refs. 26 through 32. This information
was used extensively in the revision of Sections 2.3.4, 3.4 and 3.5 of the 1980
Edition of the AISI Specification.
In the study of bending strengths of beam webs, the postbuckling
strength of web elements was determined to be a function of the web
slenderness ratio, the bending stress ratio of the web, the flat width-to-
thickness ratio of the compression flange, and the yield point of steel. A
formula for calculating the postbuckling strength factor was developed
from this study. As an alternative method, two formulas were derived from
the test data for reducing the bending capacity of beams with large web
slenderness ratios. In addition, for the range of cross-sectional parameters
covered in this study, ‘effective web depth’ equations were developed for
determining the ultimate bending capacity of cold formed steel
beams.*°:*7:3! These formulas differ from those developed by Bergfelt,**
Hoglund,** and Thomasson.?°
For web crippling and a combination of web crippling and bending,
additional tests of channels and I sections were conducted for end one-
flange loading, end two-flange loading, interior one-flange loading and
interior two-flange loading to supplement the Cornell tests conducted in
the 1940s and 1950s.°°3’ In addition, the publications of Yamaki,*®
White and Cottingham,*? Rockey, El-gaaly and Bagchi,*°~*° Baehre,*’
Ratliff*® and others and the current design criteria used in various
countries have been reviewed in the initial phase of the investigation. Based
on the available test data, modified design formulas were developed for use
in Section 3.5 of the 1980 Edition of the AISI Specification.

2.4. Web Crippling and Combined Web Crippling and Bending of Steel
Decks
The modified design formulas discussed above are based on the test data of
channels, I beams and hat sections having vertical webs with relatively
small R/tand N/t ratios (R being the inside bend radius, ¢ the thickness and
N the length of bearing plate). It has been realized that steel decks often
have inclined webs with large R/t and/or large N/t ratios. In addition, for
 UMR Research on Cold Formed Steel Structures ibis

the steel decks used for composite slabs, various types of embossments and
indentations are usually formed in the webs. These deformations may affect
the web crippling strength of steel decks. For this reason, additional studies
were conducted in 1979 at the University of Missouri—Rolla for the purpose
of studying the web crippling strength of steel decks under the
cosponsorship of Steel Deck Institute, American Iron and Steel Institute,
and H.H. Robertson Company. Based on the test data on one-flange
loading, it was found that the modified formulas for web crippling are
adequate for steel decks having flat webs, provided that the decks are
properly connected to support beams. However, the results of continuous
beam tests indicated that the interaction formula for combined bending
and web crippling developed from simple span beam tests of channels and
hat sections is conservative for multi-web steel decks used as continuous
beams if the localized deformation over the interior support is permissible.
This can be attributed to the different postbuckling behaviour of the steel
decks over the interior support and the moment redistribution that occurs
in the continuous beams. Consequently, additional studies have been
recommended for future research in this area.

2.5. Bolted Connections in Cold Formed Steel Structures

For the design of bolted connections, Section 4.5 of the AISI Specification
was updated in 1980 to reflect the results of additional research and to
provide a better coordination with the specifications of the Research
Council on Structural Connections and the American Institute of Steel
Construction. The major task of revising this portion of the Specification
was carried out by ajoint effort of the Task Group on Bolted Connections
of the AISI Sheet Committees, Committee 27 of the Research Council on
Structural Connections and Subcommittee 3 of the AISI Advisory Group
on the Specification for the Design of Cold Formed Steel Structural
Members. Reference 49 summarizes the changes of the design criteria and
documents the research data used for the revision.°°~°°
At the University of Missouri-Rolla, additional tests of bolted
connections were conducted in 1976 and 1977 for the purpose of studying
the bolt tension induced by an applied torque and the effects of washers and
torque on the bearing strength of connections. This information and the
reevaluation of the available test data have been used to revise the 1980
Edition of the AISI Specification.

2.6. Stainless Steel Cold Formed Structural Design Manual

In 1973, a project was conducted at the University of Missouri—Rolla to
116 Wei-Wen Yu

prepare a design manual for stainless steel, cold formed structural members
under the guidance of the AISI Advisory Group on the Specification for the
Design of Cold Formed Stainless Steel Structural Members (D. S.
Wolford, Chairman). This document was prepared on the basis of the 1968
Edition of the AISI stainless steel specification on annealed grades°’ and
on the research work conducted by Johnson, Wang, Errera, Tang and
Popowich at Cornell University under the direction of Dr George
Winter.°’~°3
The AISI Stainless Steel Design Manual was published in 1974.°* It
contains the following four parts: specification for the design of annealed
and cold-rolled grades of sheet and strip stainless steels, commentary,
illustrative examples and design tables and charts.

2.7. Load and Resistance Factor Design of Cold Formed Steel

Since 1976, a joint research project on ‘Load and Resistance Factor Design
of Cold Formed Steel’ has been conducted at the University of Missouri—
Rolla and Washington University. The purpose of this project was to
develop recommendations for the load and resistance factor design of cold
formed steel structural members and connections. It is based on the
concept of limit states, which describe a limit of structural usefulness. In
this method, load and resistance factors reflect the uncertainties of analysis,
design, loading, material properties and fabrication. They were derived on
the basis of the first-order probabilistic design principles. In this approach,
only the mean values and coefficients of variation of the loads and
resistances were used.
Following the statistical analyses of material properties and material
thicknesses and the calibration of various sections of the AISI design
provisions on allowable stress design, tentative recommendations on the
design of cold formed steel were developed.°*~ °’ A detailed review ofthese
tentative recommendations is being made at the present time.

2.8. Automotive Structural Components Using High Strength Sheet Steels

In recent years, automobile manufacturers have introduced new lines of
lighter vehicles for the purpose of achieving fuel economy. To construct
such cars, formable and weldable high strength sheet steels with yield
strengths of up to 140 ksi have been used for auto parts and structural
components. Because various types of new high strength sheet steels are now
available for the reduction of car weight and because these steels permit
the use of existing production equipment with virtually no change in
techniques or production rates, the design criteria for efficient and
 UMR Research on Cold Formed Steel Structures 117

economical use of these high strength steels in car bodies are needed by
structural designers and engineers.
In February 1981, the ‘Guide for Preliminary Design of Sheet Steel
Automotive Structural Components’ was issued by the American Iron and
Steel Institute for assisting automotive structural designers to achieve
weight reductions through the efficient utilization of carbon and high
strength steels.°® These design recommendations were based primarily on
the 1968 Edition of the AISI ‘Specification for the Design of Cold Formed
Steel Structural Members’ except for the following major differences with
respect to the AISI Specification written for the design of buildings:°?

(a) The design expressions presented in the Guide are based on an

ultimate strength basis.
(b) Because the design expressions are sometimes simplified in the
Guide, their range of applicability is restricted in some instances.
(c) The design expressions are extended to materials with yield
strengths up to 80ksi.

In view of the fact that many types of high strength steels with yield
strengths from 80 to 140ksi can be economically used for automotive
structural components, a comprehensive design guide for the use of a
broader range of high strength sheet steels is highly desirable.
Since early 1982, a new research project entitled “Structural Design of
Automotive Structural Components Using High Strength Sheet Steels’ has
been conducted at the University of Missouri—Rolla. The main purpose of
the project is to determine the material characteristics of typical high
strength sheet steels having yield strengths in the range of 80-140 ksi and to
develop the design criteria for the cold formed steel structural members
composed of such high strength steels.
The preliminary study has been completed at the present time. The
structural research will be initiated by a study of the strength of beam webs
and the structural components consisting of curved and flat elements.

3. CONTINUING EDUCATION PROGRAMS

The above discussion is related to the UMR research work on cold formed
steel structures. In view of the fact that many research projects on thin-
walled metal construction have been conducted by a large number of
researchers at various institutions and organizations throughout the world,
118 Wei-Wen Yu

six international specialty conferences on cold formed steel structures have

been conducted during the past 12 years for the purpose of reviewing the
research findings and the design methods developed in this field. The
Seventh Conference has been scheduled for 13-14 November, 1984.
Historically, the scope of the specialty conferences has included basic
and applied research, structural design, development of new products,
development of new design criteria, manufacturing techniques, con-
struction methods, economic use of cold formed steel and engineering
education. Previous conferences were cosponsored by the American Iron
and Steel Institute, National Science Foundation, Metal Building
Manufacturers Association, Steel Deck Institute and the University of
Missouri—Rolla in cooperation with the American Society of Civil
Engineers, Canadian Sheet Steel Building Institute, Rack Manufacturers’
Institute, Structural Stability Research Council and the University of
Strathclyde, Scotland. The conference proceedings have been widely used
by researchers and engineers as major references in the field of thin-walled
structures.
In addition, in order to provide engineers and other interested
individuals with a better understanding of the basic principles used in the
current AISI design provisions, eight short courses have been offered at the
University of Missouri—Rolla under the sponsorship of the American Iron
and Steel Institute and the University. The ninth short course has been
scheduled for 15-18 November, 1983.

4. CONCLUSIONS

The research projects, which have been conducted at the University of

Missouri-Rolla for cold formed steel structural members and connections,
have been briefly reviewed and discussed. Details of the investigations can
be found in the publications referred to in the text. The UMR continuing
education programs related to cold formed steel structures have also been
briefly described.

Acknowledgement
All of the research projects discussed in this paper have been sponsored by
the American Iron and Steel Institute. The technical guidance provided by
various AISI task groups and the AISI Staff is gratefully acknowledged.
Appreciation is also extended to Steel Deck Institute and H. H. Robertson
Company for cosponsoring the project on steel decks.
 UMR Research on Cold Formed Steel Structures 119

REFERENCES

. ‘Specification for the design of cold-formed steel structural members’,

Washington, D.C., American Iron and Steel Institute, 3 September 1980.
. HOWLAND, R. C. J. and Stevenson, A. C., Philosophical Transactions of the
Royal Society of London, Series A, 232 (1933-34) 155-222.
. BISKIN, L., Strengthening ofcircular holes in plates under edge forces, Journal
of Applied Mechanics, Transactions, American Society of Mechanical
Engineers (1944) A-140.
. WANG,C. K., Theoretical analysis of perforated shear webs, Journal of Applied
Mechanics, Transactions, American Society of Mechanical Engineers, B. No. 2
(June 1946) A77-A-34.
. Levy, S., Woo.tey, R. M. and KROLL, W. D., Instability of supported square
plates with reinforced circular holes in edge compression, Journal of Research,
National Bureau of Standards, 139 (December 1947).
. KRo.i, W. D., Instability of simply supported square plates with reinforced
and unreinforced holes subjected to shear, Journal of Research, National
Bureau of Standards, 43 (November 1949).
. Kumal, T., Elasticity stability of the square plate with a central circular hole
under edge thrust, Reports of Research Institute for Applied Mechanics, 1(2)
(April 1952).
. ScHLACK, A. L., Elastic stability of pierced square plates, Experimental
Mechanics (June 1964).
. YOSHIKI, M., et al., Instability of plates with holes (first report), Proceedings of
the Society of Naval Architects of Japan, No. 122 (December 1967).
. Kawal, T. and Ontsuso, H., ‘A method of solution for the complicated
buckling problems of elastic plates with combined use of Rayleigh—Ritz’s
procedure in the finite element method, Proceedings of the 2nd Air Force
Conference on Matrix Methods in Structural Mechanics, October 1968.
. Rockey, K.C., ANDERSON, R. G. and CHEUNG, Y. K., ‘The behavior of square
shear webs having a circular hole, Thin-walled steel structures, K. C. Rockey
and H. V. Hill (eds.), London, Gordon & Breach, 1969.
. BuLson, P.S., The stability of flat plates, New York, American Elsevier, 1969.
. Hutt, R. E., ‘Buckling of plates with unknown inplane stress using the finite
element method, Thesis, University of Washington at Seattle, Washington,
1970.
. JOHNSTON, B. G. (ed.), Guide to design criteriafor metal compression members,
2nd edn, New York, John Wiley, 1966.
. ‘Specification for steel railway bridges’, American Railway Engineering
Association, 1969.
. ‘Specification for the design, fabrication, and erection of structural steel for
buildings’, American Institute of Steel Construction, 1969.
. ‘Standard specifications for highway bridges’, American Association of State
Highway Officials, 10th edn.
. ‘Minimum engineering standards for industrial steel storage racks’, Rack
Manufacturers’ Institute, September 1964.
. Yu, W. W. and Davis, C. S., ‘Buckling behavior and post-buckling strength of
perforated stiffened compression elements’, Proceedings of the First Specialty
120 Wei-Wen Yu

Conference on Cold-Formed Steel Structures, University of Missouri—Rolla,

August 1971, pp. 58-64.
20. Yu, W. W. and Davis, C. S., Cold-formed steel members with perforated
elements, Journal of the Structural Division, Proceedings of the American
Society of Civil Engineers, 99, No. ST10 (October 1973) 2061-77.
Ake WINTER, G., Strength of thin steel compression flanges, Transactions, ASCE,
112 (1974) 527-54.
pd, WINTER, G., ‘Commentary on the 1968 edition of the specification for the
design of cold-formed steel structural members, American Iron and Steel
Institute, 1970.
mph. Yu, W. W., Liu, A. S.and McKINnNgEY, W. M., ‘Structural behavior and design
of thick, cold-formed steel members’, Proceedings of the Second Specialty
Conference on Cold-Formed Steel Structures, University of Missouri—Rolla,
October 1973, pp. 25-52.
24. Yu, W. W., Liu, A. S. and McKInney, W. M., Structural behavior of thick
cold-formed steel members, Journal of the Structural Division, Proceedings of
the American Society of Civil Engineers, 100, No. ST11 (November 1974)
2191-204.
25) ‘Specification for the design of cold-formed steel structural members—
Addendum No. 2 to the 1968 Edition’, American Iron and Steel Institute, 4
February 1977.
26. LaBouse, R. A. and Yu, W. W., ‘Study of cold-formed steel beam webs
subjected to bending stress’, Proceedings of the Third International Specialty
Conference on Cold-Formed Steel Structures, University of Missouri—Rolla,
November 1975, pp. 315-35.
Die LaBouse, R. A. and Yu, W. W., ‘Effective web depth of cold-formed steel
beams’, Proceedings of the International Coloquium on Stability of Structures
Under Static and Dynamic Loads, Washington, D.C., 17-19 May 1977,
pp. 573-84.
28. LaBousBeE, R. A. and Yu, W. W., ‘Cold-formed steel web elements under
combined bending and shear, Proceedings of the Fourth International
Specialty Conference on Cold-Formed Structures, University of Missouri-
Rolla, June 1978, pp. 219-51.
MY). HETRAKUL, N. and Yu, W. W., Cold-formed steel I-beams subjected to
combined bending and web crippling, Thin-walled structures, J. Rhodes and
A.C. Walker (eds.), International Conference at the University of Strathclyde,
Glasgow, 3-6 April 1979, London, Granada, 1980, pp. 413-26.
30. NoGuyen, P. and Yu, W. W., ‘Bending strength of cold-formed steel
longitudinally reinforced beam webs’, Proceedings of the Fifth International
Specialty Conference on Cold-Formed Steel Structures, University of
Missouri-Rolla, November 1980, pp. 201-24.
Sill, LaBousg, R. A. and Yu, W. W., Bending strength of webs of cold-formed steel
beams. Journal of Structural Division, Proceedings of the American Society of
Civil Engineers, 108, No. ST7 (July 1982) 1589-604.
328 NGuyeEN, R. P. and Yu, W. W., Longitudinally reinforced cold-formed steel
beam webs, Journal of the Structural Division, Proceedings of the American
Society of Civil Engineers, 108, No. ST11 (November 1982) 2423-42.
 UMR Research on Cold Formed Steel Structures 121

33: BERGFELT, A., Profile’s minces formes a froid, Bulletin Technique De La Suisse
Romande, 99, No. 17 (August 1973) 363-8.
34, HOGLUND, T., ‘Design ofthin plate I girders in shear and bending with special
reference to web buckling’, Bulletin No. 94, Division of Building Statics and
Structural Engineering, Royal Institute of Technology, Sweden, September
1973.
Bo: TuHomasson, P., ‘Livbuckling ens Inverkan pa Burformagan Has Trapets
Profilerad Stalplat’, Nordiske Forskningsdager for Stalkonstruksjoner, Oslo,
Norway, August 1973.
36. WINTER, G. and Pian, R. H. J., Crushing strength of thin steel webs, Cornell
Bulletin, No. 35, Part I, 1946.
37. ZETLIN, Lev, Elastic instability of flat plates subjected to partial edge loads,
Journal of the Structural Division, Proceedings of the American Society ofCivil
Engineers, 81 (September 1955).
38. YAMAKI, N., ‘Buckling of a rectangular plate under locally distributed force
applied on the two opposite edges’, Ist and 2nd report, The Institute of High
Speed Mechanics, Tohoku University of Japan, Vol. 3, 1953.
39. Wuite, R. N. and CottinGHaM, W. S., Stability of plates under partial edge
loadings, Journal of Engineering Mechanics Division, ASCE Proceedings, 88,
No. EMS (October 1962).
40. BaGcul, D. K. and Rockey, K.C.,‘A note on the buckling ofa plate girder web
due to partial edge loadings’, International Association for Bridge and
Structural Engineering, Final Report, September 1968.
41. Rockey, K.C. and BaGcui, D. K., Buckling of plate girder webs under partial
edge loadings, International Journal of Mechanical Science, 12 (1970).
42. Rockey, K. C. and Et-Gaaty, M. A., ‘Ultimate strength of plates when
subjected to in-plane patch loading’, International Association for Bridge and
Structural Engineering, Proceedings of the Seminar on Design of Plate and
Box Girders for Ultimate Strength, 1971.
43. Rockey, K. C., EL-GAALy, M. A. and BAGCHI, D. K., Failure of thin walled
members Gader patch loading, Journal of eral Division, ASCE
Proceedings, 98, No. ST12 (December 1972).
. Rockey, K. C. and Er-Gaaty, M. A., Stability of load bearing trapezoidal
diaphragms, International Association for Bridge and Structural Engineering,
Publications, 32-Il (1972).
45. E.t-GaaLy, M. A. and Rockey, K. C., ‘Ultimate strength of thin walled
members under patch loading and bending’, Proceedings of the Second
Specialty Conference on Cold-Formed Steel Structures, University of
Missouri—Rolla, October 1973.
46. Rockey, K. C. and Et-Gaaty, M. A., ‘Ultimate strength of plates when
subjected to in-plane patch loading’, International Association for Bridge and
Structural Engineering, Colloquium on Design of Plate and Box Girders for
Ultimate Strength, London, England, March 1971; published July 1972.
47. BAEHRE, R., ‘Sheet metal panels for use in building construction—recent
research projects in Sweden’, Proceedings of the Third International Speciality
Conference on Cold-Formed Steel Structures, University of Missouri—Rolla,
24-25 November 1975.
122 Wei-Wen Yu

48. Rat irr, G. D., ‘Interaction of concentrated loads and bending in C-shaped
beams’, Proceedings of the Third Specialty Conference on Cold-Formed Steel
Structures, University of Missouri-Rolla, 24-25 November 1975.
49, Yu, W. W., ‘AISI design criteria for bolted connections’, Proceedings of the
Sixth International Specialty Conference on Cold-Formed Steel Structures,
University of Missouri—Rolla, November 1982, pp. 675-98.
50. WINTER, G., Tests on bolted connections in light gage steel, Journal of the
Structural Division, ASCE Proceedings, 82, No. ST2 (March 1956).
51. WINTER, G., ‘Light gage steel connections with high-strength, high-torqued
bolts’, Publication of the International Association for Bridge and Structural
Engineering, Vol. 16, 1956, p. 513.
52 Porowicn, D. W., ‘Tension capacity of bolted connections in light gage cold-
formed steel’, Thesis, Cornell University, 1969.
Ssh DHALLA, A. K., ERRERA, S. J. and WINTER, G., Connections in thin low-
ductility steels, Journal of the Structural Division, ASCE Proceedings, 97,
No. ST10 (October 1971).
54. CHONG, K. P. and Mattock, R. B., Light gage steel bolted connections
without washers, Journal of the Structural Division, ASCE Proceedings, 101,
No. ST7 (July 1975).
5). GILCHRIST, R. T. and CHONG, K. P., Thin light-gage bolted connections
without washers, Journal of the Structural Division, ASCE Proceedings, 105,
No. ST1 (January 1979).
56. Yu, W. W. and Mossy, R. L., ‘Bolted connections in cold-formed steel
structures’, University of Missouri—Rolla, January 1981.
7, ‘Specification for the design of light gage cold-formed stainless steel structural
members’, American Iron and Steel Institute, 1968 edn.
58. JOHNSON, A. L., ‘The structural performance of austenitic stainless steel
members’, Ph.D. Dissertation, Cornell University, February 1967;
Department of Structural Engineering, Rept. No. 327, Cornell University,
November 1966.
oo: JOHNSON, A. L. and WINTER, G., Behavior of stainless steel columns and
beams, Proceedings, ASCE, 92, No. STS (October 1966) 97-118.
60. Jounson, A. L. and KELSEN, G. A., Stainless steel in structural applications,
Stainless Steel for Architecture, ASTM STP 454, American Society for Testing
and Materials, August 1969.
61. Wana, S. T., ‘Cold-rolled austenitic stainless steel: material properties and
structural performance’, Department of Structural Engineering, Rept. No. 334,
Cornell University, July 1969.
62. WANG, S. T. and Errera, S. J., ‘Behavior of cold-rolled stainless steel
members’, Proceedings of the First Specialty Conference on Cold-Formed Steel
Structures, University of Missouri-Rolla, August 1971.
63. ERRERA, S. J., TANG, B. M. and Popowicn, D. W., ‘Strength of bolted and
welded connections in stainless steel’, Department of Structural Engineering,
Rept. No. 335, Cornell University, August 1970.
64. ‘Stainless Steel Cold-Formed Structural Design Manual’, American Iron and
Steel Institute, 1974.
65. RANG, T. N., GALAMBOs, T. V., YU, W. W. and RAvINDRA, M. K., ‘Load and
resistance factor design of cold-formed steel structural members’, Proceedings
 UMR Research on Cold Formed Steel Structures 123

of the Fourth International Specialty Conference on Cold-Formed Steel

Structures, University of Missouri-Rolla, June 1978, pp. 11-39.
66. GALAMBOS, T. V., RANG, T. N., YU, W. W. and RAVINDRA, M. K.., ‘Structural
reliability analysis of cold-formed steel members’, Proceedings ofthe Specialty
Conference on Probabilistic Mechanics and Structural Reliability, American
Society of Civil Engineers, January 1979, pp. 146—S0.
67. RANG, T. N., Yu, W. W., GALAMBos, T. V. and RAVINDRA, M. K., ‘Load and
resistance factor design of bolted connections’, Thin-walled structures, J.
Rhodes and A. C. Walker (eds.), International Conference at the University of
Strathclyde, Glasgow, 3-6 April 1979, London, Granada, 1980, pp. 752-9.
68. ‘Guide for Preliminary Design of Sheet Steel Automotive Structural
Components’, American Iron and Steel Institute, 1981 edn.
69. ErrERA, S. J., ‘Automotive structural design using the AISI Guide’, SAE
Technical Paper Series 820021, 22—26 February 1982.
 Some Thoughts on Future Cold Formed Steel
Design Rules

J. RHODES
University of Strathclyde, Glasgow, UK

ABSTRACT

Some of the problems associated with the buckling behaviour of cold formed
steel structural sections are examined. The methods of approach to these
problems used in various design codes are outlined and compared. Effective
widths of compression elements, edge and intermediate stiffener require-
ments for these elements, and some aspects of column and beam behaviour
are considered. Suggestions are made for appropriate design methods to deal
with the problems examined. .

1. INTRODUCTION

In recent years there has been a great deal of effort devoted to the
compilation of new design codes for cold formed steel sections. New AISI!
and Swedish? design codes were introduced in 1980, and substantial work
has been carried out since then on drafting European Recommendations?
for cold formed steel design. At the same time, new national design codes
are in the process of being developed in various countries including the UK.
It is perhaps, therefore, an opportune time to give some thought to the
problems encountered in cold formed steel design and on how these can
best be dealt with. In this paper some impressions of typical aspects of cold
formed section design requirements are presented and some suggestions are
made as to how these aspects can be dealt with in a design code.
125
126 J. Rhodes

No attempt is made to be comprehensive, and only a few aspects of

design requirements are considered, dealing with buckling behaviour of
plates and members. The topics selected are mainly those which are not
adequately covered by the current UK design specification.*
It is mentioned here that in thinking of future design codes it would be
realised that, in the UK at least, the limit state philosophy will be used as
the basis.

2. GENERAL REQUIREMENTS OF A DESIGN CODE

In the drafting of a new design code there are factors other than purely
technical details which must be considered. In the case of cold formed steel
design, in which local buckling adds complexity to the various design
problems, one of the most important requirements of a design code is
simplicity of use. There is no merit in providing a designer with
sophisticated analytical methods if he is not able to use them.
Generality of application is also a requirement. Lack of generality is at
present the major failing of the UK code, which users find does not deal
with a number of commonly occurring situations.
If a design code must be simple and general then it cannot always be
accurate in dealing with every situation. In such cases, if simple rules are
inadequate to accurately cover all situations then the range of application
must be known and it must be ensured that the simple rules at least err on
the safe side. Conservatism is therefore a further requirement.
These three requirements are not always easy to satisfy simultaneously,
while at the same time ensuring efficient design. This is especially so when it
is considered that during the lifespan of a code there may be significant
changes in the use of cold formed steel. It is therefore necessary also to
make some attempt at keeping the provisions of a code suitable for dealing
with possible future needs.
With these thoughts in mind, some ideas as to the way 1n which the design
of some elements and members may be dealt with are given in the remainder
of this paper.

3. BEHAVIOUR OF COMPRESSION ELEMENTS

3.1. Effective Widths

Typical compression plate elements of cold formed sections are shown in
Fig. 1. These elements may be supported on both edges (stiffened) as in
 Future Cold Formed Steel Design Rules 127

is +
' es i} ll
ul II lI
i i Nl
Lee | cov i]

ca) T
Stiffened Element Hu

(ob)
Unstiffened Element

cc) (é
Edoe Stiffened Element aterediote. SiufFenead
Element

Fig. 1. Compression elements of cold formed sections.

Fig. l(a) or on one edge only (unstiffened) as in Fig. 1(b). The elements may
also have lips, or edge stiffeners (c) or intermediate stiffeners (d).
The major factor affecting the behaviour of compression elements is
local buckling. This is taken into account in most design codes using the
concept of ‘effective width’ which admirably describes the effects of local
buckling. Many design codes use the effective width formulation of
Winter° or slight modifications of this, although some codes use other
approaches. The UK specification uses, instead of effective widths, stress
reduction factors (CL factors) based on rigorous analysis by Walker® of a
simply supported square plate with a carefully devised imperfection
parameter which describes imperfections commonly found in cold formed
steel sections.
While the writer feels that an effective width formulation is more suitable
in design than a stress reduction factor approach, the rigorous basis of the
UK specification provides a consistently conservative prediction of plate
strength.
With this in mind, an effective width formulation which gives results in
broad agreement with the UK code predictions, or a little less conservative,
128 J. Rhodes

could be considered desirable. A suitable effective width expression is given

below:

am for 2% < 0,59

cr

4-022
ral +14 ([Pau 0,35) | for 2™*
> 0.59 (1)
b Gor Cor

where 8, is the effective width, 5 is the full width and o,,,, is the maximum
compressive stress.
This expression permits evaluation of effective width for any stress level.
The buckling stress o,, is given by
kn’°E py
Per ~ 19(1 — v2) (*) 2)
where the buckling coefficient, k, may be taken as 4 for a stiffened element
and 0.425 for an unstiffened element.
Equation (1) is a little more complex than the effective width expression
of the AISI code, but is a closer fit to the existing UK CL factor curve and
also gives results more consistent with curves derived for heavier structures
in the low b/t range. Figure 2 shows comparison of the effective widths
given by eqn (1) with those of the AISI specification and with the CL factors
of the UK specification. Also shown are the curves of Chatterjee and
Dowling’ and Scheer and Nolke® for bridge structures.
For simplicity in design, eqn (1) may be used for both stiffened and
unstiffened elements using the appropriate & factors. It is known (see ref. 9)
that the behaviour of unstiffened elements can be conservatively predicted
by a formulation set up for stiffened elements (with the appropriate k
factors) so long as the unstiffened elements can be prevented from bending.
However, as these elements have very low bending rigidity in the presence
of local buckling, safeguards should always be built in to take account of
this.

3.2. Edge Stiffeners

In the various design codes in existence a number of different formulae for
the requirements of edge stiffeners are given. These in most cases define
minimum stiffener flexural rigidity with regard to bending about its own
neutral axis parallel to the plate which is to be stiffened.
The simplest rules available at present are those of the UK specification.
 Future Cold Formed Steel Design Rules 129

These state simply that to ensure that the stiffened plate behaves as a
stiffened element (rather than as an unstiffened element) the lip width must
be equal to one-fifth of that of the element to be stiffened.
An examination of this problem shows that, while all the specifications
give rules from which the rigidity requirements of edge stiffeners can be
evaluated by consideration of the element in isolation, this is not the case in
reality. Actually the requirements of edge stiffeners are highly dependent

at (<->)

be ‘\ Eq? a1
b
os 0-3

be
b
oe 0-6 PRenax
fe) os J 1-5'Ccr

x AISI Specification
XN

Ok : Eq? ao
we ON
Specification

Oz
_ =

Fig. 2. Effective width curves.

on the complete geometry of a section. A further complication which is

encountered is that although an edge stiffener may adequately support the
edge of the element at the point of buckling this does not ensure that the
edge support is continued until the plate/stiffener arrangement fails as a
stiffened element. This problem is examined in ref. 10 and in paper | ofthis
volume.
A comprehensive approach to the determination of edge stiffener
requirements would therefore be very complex, and probably restrictive as
all possible situations cannot be dealt with. In view of this it can be argued
that the simplest approach which has been found to be, in general, safe is
preferable.
130 J. Rhodes

Therefore a suitable requirement would be, for an edge stiffener

iid a 3
fey Sea 7 3
where 6 is the width of the element to be stiffened and /,,;, is the second
moment of area of the stiffener taken about the element middle surface.
This requirement is automatically satisfied by a simple lip with width b/5
and inside radius 1.5t, and is therefore similar to the UK code requirements.
As stipulated by that code, and by various other codes, the maximum
width-to-thickness ratio of an element which is to be stiffened by a simple
lip should be 60. For this width of element the lip flat width required is 12t.
If the lip flat width is greater than this then the possibility of lip buckling
and its attendant complications arises.
As in the AISI code and others with a similar basis, it is suggested that
wider elements, up to 90t, may be stiffened by an edge stiffener satisfying
eqn (3) and having arbitrary shape other than that of a simple lip.
Figure 3 shows comparison of stiffener dimensions obtained using
eqn (3) with those of various design specifications. In the range b/t < 60 a
simple lip is used as the basis for comparison and in the range 60 < b/t < 90
a compound lip, as illustrated in the figure, is used. It will be noted from the
figure that for b/t greater than 60 the dimensions of stiffener required by
eqn (3) are greater than those required by the various codes. Although this
arises purely from the extrapolation of the current UK requirements given

European =
12 Recs. Dn
/

UK
3 ; .

2 Eq“ cad

fe) 20 LO CO 80 x
Fig. 3. Required minimum edge stiffener dimensions.
 Future Cold Formed Steel Design Rules 131

by eqn (3) it is not felt desirable to make any reduction to come into line
with the other specifications. It is quite possible that in the future the AISI
code and similar codes will require larger edge stiffeners in this range. The
European Recommendations give larger stiffener requirements than those
of the AISI in this range. These are based on recent research in the USA at
Cornell University’? in which it was shown, on the basis of experiments,
that the AISI requirements in this range are not sufficient. Furthermore, the
tests of ref. 11 were carried out on beam elements in which efforts were
made to ensure that the supported edges of the edge stiffened elements were
kept as close to the fixed condition as possible. There are many situations in
which edge stiffened elements will suffer more severe conditions so that in
the writer’s opinion even the increased rigidities required by this research
will be inadequate in some circumstances.

3.3. Intermediate Stiffeners

As with edge stiffeners the function of intermediate stiffeners must be
carefully considered when setting minimum stiffener requirements. The
majority of design codes at present have requirements formulated on the
basis that intermediate stiffeners restrain the buckling of sub-elements
between stiffeners so that these should buckle as stiffened plates with k = 4.
The draft European Recommendations, again based on experimental work
at Cornell,!? on this problem are formulated on the basis that this restraint
should be continued until the sub-element fails as a stiffened plate. If design
of the sub-elements is to use a stiffened element basis then the continuance
of post-buckling restraint is imperative and the requirements of the draft
European Recommendations are more realistic than those of other codes.
Figure 4 shows the required minimum / values suggested in ref. 12. There
are three different w/t ranges suggested by this research, and used in the
draft European Recommendations. To avoid the necessity of considering
three different w/t ranges, a single curve can be used to represent the
minimum stiffener requirements. This is
w\? o
Iegqminy = 9.045 (*) 580 te (4)

The requirements given by eqn (4) are shown in Fig. 4, and these are
substantially the same as those of ref.3, but provide a little more
conservatism for sub-elements of high w/t ratios beyond the limit of the test
results.
Although design codes tend to’ give minimum stiffener rigidity
requirements in terms of the Jc, of the stiffener, in actuality stiffeners bend
 132 J. Rhodes

log
TA 300

600 Ref C12) and

Evvopean Recs.

300

° 25 50 76 100 125 {50 alg

Fig. 4. Minimum / values for intermediate stiffeners.

about an axis closer to the plate middle surface than to their own neutral
axis. In order to ascertain more accurately the relative merits of different
stiffener shapes it is therefore more realistic to specify the rigidity
requirements in terms of the second moment of area about the plate middle
surface, say J,,;,. Since the ratio of J,,;, to [¢g varies from about 4 to 5 for
common stiffener shapes, a suitable value for required rigidity based on
eqn (4) is
| bk Wie 6
r= 0.2(—] — 5
t* (*) 280 ©)
For comparison purposes Fig. 5 shows the variation of required height, h
of a stiffener of equal height and width / obtained from eqn (5), ref. 3 and
the AISI and CSA’? specifications, with variation in sub-element w/t ratio.
td
12
JO

|=

20 +O GO BO {OO . 120 440

Fig. 5. Required minimum intermediate stiffener dimensions.
 Future Cold Formed Steel Design Rules 133

4. COLUMNS

4.1. Overall Buckling

A variety of different approaches are used in cold formed steel codes to deal
with the overall buckling of columns which are also subject to local
buckling. The UK specification uses a mean stress reduction factor
approach, which is equivalent to a uniformly reduced area approach. The
AISI specification uses a reduced area approach combined with stress
reduction factors when unstiffened elements are used. The Swedish
specification uses an effective section approach. Each method of approach
has strong and weak points. A long pin-ended column will fail primarily by
bending and bending behaviour can be most accurately assessed using an
effective section approach. A short column will fail primarily by
compression (although if the section is not doubly symmetric bending will
also be present) and an effective area or stress reduction factor approach
will be realistic and simpler to use than an effective section approach.
For a section predominantly composed of stiffened elements, the use of
an effective area or mean stress factor approach generally gives slightly
lower estimates of failure load than would be obtained using an effective
section approach. This, and the relative simplicity of analysis, suggest that
design on an effective area basis, using the CLM approach of the UK code
to find the effective area, is suitable for most cases. It should be mentioned
however, that there are situations in which this may be non-conservative.
The UK code uses a Perry—Robertson interaction formula with an
imperfection magnitude of 0.3 x 10~4(//r)*, where J//r is the column
slenderness ratio. This imperfection magnitude is very severe on long
columns and a more suitable value would be 0.002(//r). This value is
broadly in line with values suggested for hot rolled sections.
One area of potentially unsafe design in the current UK code occurs
when columns which have unstiffened elements perpendicular to the minor
axis of bending are being considered. In such a case, for example with plain
channels, the elastic buckling of the unstiffened elements leads to very
significant loss in flexural rigidity of the column and even relatively short
columns can fail elastically. This must be taken into account, especially if a
design code is to deal with unstiffened elements of width-to-thickness ratios
up to 50 or 60.
This can be taken into account either by using an effective section
approach, in which the reduction in flexural stiffness is modelled quite well
by the fact that regions of the unstiffened elements remote from the
supported edge become ineffective, or by incorporating a stress factor
approach similar to that of the AISI code.
 134 J. Rhodes

{20
Ah) Ref Ce) =3%
alo

ee) U.K. code

Te) ap proach = ob

Failouve ay oN ALSI1
Stress s ™~ fas pe
CN 2,00 > oe ,
/vamn) Cc) R
Suggested ce : oe,
approach ie mage

30

1@) 4O 80 4520 {GO 200 240

Fig. 6. Failure loads for plain channel columns.

In the AISI approach a limiting stress value is calculated on the basis of

‘the most flexible unstiffened elements. The effective area of the section is
then determined using this stress and assuming that only the stiffened
elements are partially ineffective. The results of this approach are shown,
with safety factors taken out, in Fig. 6 for a specific geometry of plain
channel section.
Also shown in the figure are (a) theoretical predictions of elastic failure
(ref. 14), and results of tests included in this paper, (b) failure predictions of
the approach of the current UK code, (c) predictions of the suggested
approach incorporating a stress factor approach similar to the AISI code.
The limiting stress was taken as (228 — 3.23b/t) N/mm? where b/t is the
flange width-to-thickness ratio. The non-conservatism of the current UK
code in this area is shown clearly, and the incorporation of a combined
stress and area factor for sections with unstiffened elements of the type
discussed is suggested.

4.2. Torsional Flexural Buckling

The treatment of torsional flexural buckling in the UK code is vastly
different from that of other specifications, which in general regard the
problem much more seriously. For example, in the case of channel sections
the UK specification indicates that if the web/flange ratio is greater than
1.25 then no possibility of torsional flexural buckling arises. In ref. 15
 Future Cold Formed Steel Design Rules 135

sections listed as subject to this type of behaviour according to the

Canadian Standard include channels with web/flange ratios of almost 3.
The approach of this standard is similar to that of the AISI and other
design codes.
The reasons for the discrepancy are to be found in the effects of warping
restraint considered in the different codes. In the UK specification it is
assumed that full warping restraint is always present at the column ends,
whereas most other codes assume no warping restraint. In reality the
conditions which will be encountered will be somewhere between these
extremes and depend very much on the connection methods used. It must
be said that the relatively unsafe design which may have resulted from the
non-conservative conditions used in the UK specification does not seem to
have caused any great trouble in the past. Nevertheless, this specification is
so far out of line with other codes on this issue that some changes would
appear to be warranted.
The UK code approach to the problem of torsional—flexural buckling is
to specify effective length multiplication factors (« factors) to be used in the

1.6
,No warping restrawmt

Partial Warping
restraint
J+

Full warping
/ restraint

Sa
oO 25 5 Us 10 12°5

Fig. 7. « Factors for channel section.

136 J. Rhodes

Perry—Robertson interaction curve discussed previously. This approach is

simple and easy to use and the only alteration required is the postulation of
alternative values for « depending on the type of connection used.
Three sets of « values are suggested, corresponding to fully welded
connections, partially effective connections and ineffective connections
with regard to warping restraint. The variation of factors which result for a
channel section of 1.25—1 web/flange ratio is illustrated in Fig. 7 for these
three cases. In setting up this figure effective column length factors with
regard to warping for the three cases were 0.5, tf./2 and 1. The factors
given by the lowest curve are the same as those of the UK code, whereas
those of the highest curve are loosely descriptive of the requirements of
other codes.

5. BEAMS

In the past few years there has been a number of developments in the design
analysis of cold formed beams which should be taken into account in any
new code. A few of these developments are discussed here.

5.1. Utilisation of Plastic Capacity

The practice of allowing some measure of inelastic capacity has been
introduced in the 1980 AISI Specification subject to a number of provisos
regarding maximum web depth and angle from the vertical, maximum
shear force, etc. The limitations prescribed are very well thought out and
should be included in any design code before allowing plastic analysis.
The AISI approach requires determination of moments to be based on a
partially elastic, partially plastic stress distribution around a beam. For
simplicity it would be useful to base analysis solely on M,, the first yield
moment, and M,,, the fully plastic moment ofthe section. On this basis, the
following is suggested.
For compact sections, with stiffened compression elements of b/t ratio
less than 25,/280/o,, plastic capacity of 0.95M, can be assumed. For
sections with 4/t greater than 40,/280/o, no plastic capacity can be
permitted, and elastic analysis using a reduced effective section should be
used. For intermediate b/t ratios the moment capacity may be determined
using linear interpolation between these values as indicated by Fig. 8.
For sections with unstiffened compression elements the corresponding
limiting b/t values should be 8,/280/a, and 13,/280/o,.
 Future Cold Formed Steel Design Rules 137

20 25 30 4O

Fig. 8. Variation of moment capacity with b/t ratio.

In the case of sections bent about an axis of symmetry and having low
shape factors, the plastic capacity so evaluated will not be a great deal more
than the elastic capacity, but in other cases significant increases in
calculated capacity can be obtained. An indication of the increased
capacity is illustrated in Fig. 9 which plots the ratio of plastic and elastic
moment capacities given by the above suggestions for a hat section beam
with compression flange b/t less than 25,/280/a,.
It is suggested that if a section is compact, as described above, plastic
limit analysis may be used for continuous beams. If the section is not fully
compact then the only advantage which is intended to be obtained using the
plastic capacity given above is the increase in moment capacity at a point,
and limit analysis should not be used for continuous beams.

a P Saal compression

tension

fo) 0-2 O- 0-@ 0:3 4

Fig. 9. Moment capacity of a hat section.

138 J. Rhodes

5.2. Bending Elements

There has been a trend in recent years to use sheeting and sections which
have webs of such high d/t ratios that local buckling effects in the webs are
significant. As there is an obvious desire to use slender webs then the effects
of local buckling must be taken into account in design.
There has been a great deal of research in this area recently, much of it
devoted to the prescription of an effective width formulation to deal
with bending elements. Such an approach is used in the Swedish
Specification and in the draft European Recommendations. While it
appears that this approach can be made to adequately describe the effects
of local buckling, it is to some extent cumbersome and involved. The AISI
approach is to specify a maximum value of bending stress in the webs which
is dependent on the web b/t ratio. This stress is then used as the basis for
determination of the compression flange effective width and the maximum
allowable moment.
An approach suggested by the writer‘® is to formulate expressions for
effective flexural rigidity, /.,,, effective area, A.,, and effective neutral axis
position of the web elements, y.,, and to use these in the evaluation of the
properties of the section. Suitable values for the effective quantities are

A gse/A = Iog¢/1 = 0.3 + 0.70,/b (6)

Vere/Y ae 0.5 2% b,/b) (7)

where b,/b is obtained from eqn (1) using the web 5/t ratio and the k factor
corresponding to the compressive stress at buckling of the web element.
This factor for a stiffened element can be accurately and conservatively
calculated using the expression

k =4(1 +0.9f?-4) (8)

where f is a factor defining the stress system such that ¢,,,, (tension) = o,,4,
(compression) x (f — 1), i.e. 6 = 2 defines pure bending, and B = 0 defines
pure compression.
Comparison of the failure predictions using these expressions with the
experimental results of various authors is shown in Fig. 10. Also shown are
a number of curves which illustrate the variation of failure predictions
obtained on the basis of other approximations. These curves, 2—5, were
drawn on the basis of comparisons of the failure loads predicted by the
different approximations for box sections of depth twice their width and of
various thicknesses with those obtained using the approach of curve 1 for
the same sections. Curve 2 neglects local buckling effects completely.
 Future Cold Formed Steel Design Rules 139

Curve 3 takes compression flange buckling into account but neglects web
buckling, and thus illustrates the significant effects of web buckling.
Curve 4 shows the predictions of failure using the AISI maximum bending
stress (with safety factors omitted) together with the compression flange
effective width determined on the basis of eqn (1). The predictions of this
approach are very good, although non-conservative, and the sheer

25
2
CNo buckling considered) 3
CWel buckling
els regarded)

Mexe

Meare. G Eqn? G-@)

{ =
He

5
CSuggested
o— Experimental results from approach)
[eXe) Various Seurces given mM
Ref CiG)

wax
oO 2 Ly @ 3 JO Ger

Fig. 10. Failure predictions for beams with slender webs.

simplicity of the approach is so attractive that in a bid to obtain a closer

approximation using this method curve 5 was set up using a similar
approach.
For curve 5 the maximum compressive stress in the web was taken as
[1.07 — 0.0001(b/t),/, ]o,, and this was taken as the controlling stress
governing flange effective width and the failure moment. The results are in
good agreement with experiments and this would therefore appear to be a
most attractive and simple approach to the design analysis.
140 J. Rhodes

5.3. Unstiffened Bending Elements

In the case of unstiffened elements bent in such a way that the free edges are
in compression, theoretical analysis!” shows that the tangent flexural
stiffness reduces on buckling to a little less than 10% of its pre-buckling
value. This is of about the same sort of reduction which will be obtained
using an effective width approach with the effective width situated adjacent
to the supported edge. For example, the tangent effective width immediately
after buckling is about 4/9 of the initial width and the J value calculated on
the basis of a plate of width 4/96 is about 64/729, i.e. 0.088 of that ofa plate
of width 5. Thus determining stiffness on the basis of effective width is in
this case quite accurate and simple. Furthermore, recent tests carried out at
Strathclyde on plain channels for the Cold Rolled Sections Association
indicate that if the maximum compressive stress is determined on the basis
of the effective section (quite a presumptious approximation in the
circumstances) then fairly conservative predictions of failure moment
result. Thus the concept of basing design of unstiffened bending elements
on an effective width approach is sound, agrees with theory and
experiment, and is simple to use. In general for unstiffened elements bent in
this way, the tension on the free edge produces k factors significantly larger
than 0.425. For the particular case of flanges of plain channels the & factor
does not deviate by very much from 1.2 for a wide range of geometries.
The general conservatism of the effective width formulation of eqn (1) for
unstiffened elements induces perhaps a little more safety than is required
for very thin elements of b/t, say, 50 or so, and suggests that there are
perhaps some grounds for introducing a different effective width expression
for unstiffened elements.

5.4. Lateral Buckling of Beams

This is possibly the area in which deficiencies in the UK specification are
most noticeable, as this code only considers I beams with respect to lateral
buckling. The AISI and CSA codes consider I, Z and C sections in this
respect, and provide simple conservative rules for these sections. However,
attention seems to be confined to these sections and other important
sections, such as T beams, cannot be analysed by the simple methods
given in these codes.
Much more refined and detailed analysis procedures have been proposed
for the design of hot rolled sections in the recent past. These are
undoubtedly more accurate and general than the simple approaches, but
tend to be complex.
 Future Cold Formed Steel Design Rules 141

On the basis of combining simplicity, as far as this is possible, with a

reasonable degree of accuracy and generality, perhaps the most suitable
approach is similar to that of the AISI and related codes together with
multiplication factors for stabilising and destabilising loads, in simple form
as given in the Australian code.'’ The treatment of selected additional
sections on an individual basis should also be given, together with
references to more detailed analytical procedures. Interaction between
yield and lateral buckling in such an approach would be taken account of
using a Perry—Robertson formulation.
This particular topic is worthy of a greater coverage in this paper.
However, as a detailed discussion would of necessity be lengthy, space
restrictions dictate only the rather cursory treatment given.

6. CONCLUDING COMMENTS

This paper has only considered a few of the topics which must be taken care
of in future design codes. Many other topics such as beam deflections,
stiffeners for bending elements, the use of enhanced k factors when these
are available, are of importance and should be part of a future design code.
Various aspects of cold formed sections behaviour other than buckling
are also important, for example connection behaviour, and these must also
be included, as well as guidance on the use of testing as a basis for design.
Examination of the various design codes indicates that the current UK
code is deficient in the range of design problems it can deal with. The AISI
specification is impressive and deals with a wide range of aspects of cold
formed section behaviour with simple, accurate analysis methods. This
code forms the basis of many other National Specifications and must be
considered carefully in the setting up of new design rules.
Finally, it may be remarked that both tabular and graphical presentation
of design information are very useful, and should be included where
possible in design codes. However, these should not be used as the sole basis
for design analysis, as has been the case, to some extent, in the past. The
availability and widespread use of cheap computational facilities makes it
imperative that design codes become primarily formula orientated for ease
in computerisation. Indeed, considering the advances in microcomputers
in recent years, and the likely continuation of these advances, one may well
ask the question: should the next generation of design codes, say in the
early 1990s, be written as microcomputer packages?
142 J. Rhodes

REFERENCES

. American Iron and Steel Institute, ‘Specification for the design of cold formed
steel structural members’, 1980.
. National Swedish Committee on Regulations for Steel Structures, ‘Swedish
code for light-gauge metal structures’, English translation, 1982.
. ECCS, ‘European recommendations for the design of profiled sheeting and
sections Part 2—Sections’, 1982.
. British Standards Institution, ‘Specification for the use of cold formed steel
section in building: addendum No. | to BS 449: Part 2: 1969’, 1975.
. WINTER, G., ‘Performance of thin steel compression flanges’, Cornell
University Engineering Experiment Station, Reprint No. 33, 1950.
. WALKER, A. C., The post-buckling behaviour of simply supported square
plates, Aero Quarterly, XX (1969).
. CHATTERJEE, S. and DOWLING, P. J., The design of box girder compression
flanges, Steel plated structures, P.J. Dowling, J.E. Harding and P. A. Frieze
(eds), London, Crosby Lockwood Staples, 1977.
. SCHEER, J. and Novke, H., The background to the future German plate
buckling design rules, Stee/ plated structures, P. J. Dowling, J. E. Harding and
P.A. Frieze (eds), London, Crosby Lockwood Staples, 1977.
. Ruopes, J., Effective widths in plate buckling, Developments in thin-walled
structures—1, J. Rhodes and A.C. Walker (eds), London, Applied Science
Publishers, 1981.
. Ruopes, J., ‘Buckling and failure of edge stiffened plates’, IUTAM Symposium
on Collapse, 1982.
. DESMOND, T. P., PEKOz, T. and WINTER, G., Edge stiffeners for thin-walled
members, J. Structural Div. ASCE, 107, No. ST2 (1981).
. DESMOND, T. P., PEKoz, T. and WINTER, A., Intermediate stiffeners for thin-
walled members, J. Structural Div. ASCE, 107, No. ST4 (1981).
. Canadian Standards Association, ‘CSA Standard $136—1974, cold formed
steel structural members’, 1974.
. Ruopes, J. and Harvey, J. M., ‘Interaction behaviour of plain channel columns
under concentric or eccentric loading’, Preliminary Report, Second
International Colloquium on Stability of Steel Structures, Liege, 1977.
. SCHUSTER, R. M., Cold formed steel design manual, Solid Mechanics Division,
University of Waterloo Press, 1975.
. Ruopes, J., ‘The post buckling behaviour of bending elements’, Proc. 6th
International Specialty Conference on Cold Formed Steel Structures, St Louis,
1982.
. Standards Association of Australia, ‘Rules for the use of cold formed steel in
structures AS1538— 1974’, 1974.
 Design of Profiled Steel Sheeting and Decking

E. R. BRYAN and J. M. DAvIEs

University ofSalford, UK

ABSTRACT

Figures are given for the increase in production of coated steel strip and the
economic importance of cold rolled steel products in industrial buildings. It
is shown how design methods for profiled steel sheeting have evolved and the
present position of British, European and International Codes is
summarised. Factors influencing the use of profiled sheeting in roofing,
decking and wall cladding are considered. Calculation procedures are given
for the strength and deflection of unstiffened and stiffened profiles, according
to the European Recommendations. It is advantageous for some of the
design assumptions to be based on test data—a procedure only applicable to
mass produced structures—and detailed information is given on the testing
of profiled sheeting. The results of many such tests are correlated with
theory. Finally, some factors influencing future trends in the design of
sheeting, decking and sandwich panels are discussed.

1. INTRODUCTION

Over the last four years, UK steel production in many areas has fallen by
more than 30°. The only area in which there has been increase in
production has been in coated steel strip.
In Europe, production of coated material increased from 284 million
square metres in 1980 to 323.4 million square metres in 1981, and there is no
143
144 E. R. Bryan and J. M. Davies

reason to believe that this trend will not continue. For the years quoted the
increase was 16 °% for coated steel and 9.3 °% for coated aluminium.’
Much of the output of coated strip is used in profiled sheeting and
decking, and there has been a dramatic increase in such usage in recent
years. The trend is most evident in Sweden, where the annual consumption
is approaching 3 square metres per capita” but other European countries
are now experiencing the same phenomenon.
It is of interest too to note the results of a survey of costs of four
industrial buildings made by two of the country’s largest steelwork
contractors.* In three of the buildings, the cost of the cold rolled steel
members and sheeting in the walls and roof exceeded the cost of the hot
rolled steelwork. Yet it is safe to assume that the latter accounted for nearly
all the design effort.
It follows that overall efficient structural design can only be achieved
when full consideration is given to the design of the cladding and its
structural implications. The design of the frame may be more important
from the point of view of safety but less important from the points of view
of serviceability, economics and overall aesthetics.
The design approach to mass produced components such as sheeting and
decking may be very different from that used in the design of individual
structures.* With the former it may well be possible to base the design
assumptions on observed behaviour, but such a procedure would rarely be
economic for an individually designed structure. Moreover, the brevity of
the design process for sheeting is not usually of prime importance since it is
worth spending a considerable time in evolving the most economic shape.
Nevertheless, it is of great help to those who check the design if the method
can be readily followed by the non-specialist.
It is against this background that the design philosophy and procedures
for profiled sheeting and decking are developed in this paper.

2. PRESENT POSITION

For many years after the war, sheeting and decking were designed either on
a semi-experimental basis or according to the principles set out in the AISI
specification for the design of cold formed steel structural members.* In
Britain, the only design guide which has been available is that published by
the Metal Roof Deck Association.® In addition, there are two relevant
Codes of Practice,”’*? but these are orientated towards installation and
performance rather than actual design. Nevertheless the upsurge of interest
 Design of Profiled Steel Sheeting and Decking 145

in profiled sheeting has resulted in some useful publications on sheeting in

Britain,? Germany!° and other countries.
During the past 12 years or so, this rather simplistic approach has been
overtaken by a much more exhaustive study of sheeting design carried out
in Sweden, which has resulted in the recent publication of a new Code.!?!
This work has had important repercussions worldwide and is influencing
projected National Codes, European Recommendations and ISO
Standards.
In this country, the new BS 5950 on ‘The structural use of steelwork in
building’ which is in the course of preparation. recognises the importance
of light gauge steel. It is divided into nine parts of which the following five
deal with thin gauge material:

Part 4 Code of practice for design of floors with profiled steel sheeting;
Part 5 Code of practice for design of cold formed sections;
Part 6 Code of practice for design of light gauge sheeting, decking and
cladding;
Part 7 Specification for materials and workmanship: cold formed
sections and sheeting;
Part 9 Code of practice for stressed skin design.

Part 4 has just been published‘? and includes a design method for profiled
steel sheeting which is more advanced than any previous method published
in Britain, and which owes a considerable amount to the Swedish work. In
due course the method will be further revised to bring it in line with the
latest procedures proposed for Part 6.
In Europe, Preliminary European Recommendations for the design of
profiled sheeting have been published by ECCS‘? and the revised version
will shortly be available. The final document will incorporate the European
Recommendations for the testing of profiled metal sheets, published by
ECCS in 1977.'* In addition, the European Recommendations for good
practice in steel cladding and roofing'? specify some structural perfor-
mance criteria. It is expected that these documents, which have benefited
from CIB and US participation will be largely accepted as the ISO
Standard in this field.

3. ROOF SHEETING

Profiled steel sheeting used as the outer waterproof skin of a roof, with the
insulation placed inside, is known as ‘cold roof’ construction. In this
146 E. R. Bryan and J. M. Davies

STON GENURAOURUURYLOTT ere

liner panel
perforated
spacer section
purlin
4
—————

Fig. 1. Insulation and liner panel.

situation weather-tightness, durability and insulation are all-important.

For maximum weather-tightness, the end laps should be adequate (sealed if
necessary) and the sidelaps should be under the crests of the sheeting. It is
no longer necessary for fasteners to be in the crests, since modern self
drilling/tapping screws with neoprene washers can effect a perfect seal.
The durability and appearance of the sheeting depends largely on the
choice of coating. In this country, the British Steel Corporation offers a
range of Colourcoat and Stelvetite finishes.'® Insulation is usually effected
by means of a glass fibre quilt held in place by means of a thin gauge steel
liner panel fitted over the purlins (Fig. 1).
Because roof sheeting is supported on purlins, longer spans mean not
only deeper sheeting but also heavier purlins. It has been found that the
most economical roof spans in practice are of the order of 2—3 m, so scope
for innovation in design is limited.
For many years, the simple trapezoidal profile (Fig. 2(a)) was used for
sheeting, but this profile is now supplemented by profiles with rolled-in
stiffeners (Fig. 2(b)) which are more efficient in bending. Another addition
to the range is a profile (Fig. 2(c)) with a very wide lower flange stiffened by
shallow intermediate stiffeners. Such a profile is very efficient in shedding
water in long span buildings with low roof pitches.
The steel quality of many sheeting profiles is still about 250 N/mm7?, but
an increasing number of manufacturers are using sheet witha yield stress of
320-350 N/mm2 and one manufacturer uses material of 550 N/mm? yield.
Under these conditions, the minimum acceptable thickness of sheeting is
no longer regarded to be 0.7 mm, and thicknesses of 0.5 mm are used quite
widely for particular applications.
To allow for loads incidental to maintenance, roof sheeting at a slope of
less than 45° should be able to sustain a concentrated load of 0.9kN on an
area 125mm x 125mm.° This requirement is often the design criterion for
thin or short span sheeting and stiffened profiles of the type shown in
 Design of Profiled Steel Sheeting and Decking 147

Fig. 2. Roof sheeting profiles.

Fig. 2(b) or (c) are often better able to meet this condition. Work is
presently in progress to define the ‘walkability’ of sheeting more objectively,
since the current position depends far too much on subjective judgements.
Because roof sheeting is usually continuous over several spans, the
maximum span is usually determined by strength rather than deflection.
Wind suction, particularly in local areas such as the eaves, gable or apex, is
likely to be a more important design case than dead plus imposed load, and
due attention must be given to the design of fixings.
If long span sheets are required, decking profiles used in the inverted
position (to improve water shedding) may always be used.

4. ROOF DECKING

If the insulation and waterproof membrane is placed on top of the sheeting,

then the sheeting is termed ‘decking’ and the method is known as ‘warm
roof’ construction. The method is usually reserved for flat or nearly flat
roofs, and demands considerable technical ‘know-how’ and experience. A
comprehensive guide to good practice has recently been published.'”
Because the decking spans directly between the main or secondary beams
without the need for purlins, it is usually advantageous for long spans to
148 E. R. Bryan and J. M. Davies

Fig. 3. Roof decking profiles.

be used. The profiles shown in Fig. 3, of depth about 100mm and yield
stress 350 N/mm7?, can be used for spans between 6 and 8 m for UK loading.
Since the permissible deflection of decking is limited to span/250,° the
design criterion is usually that of stiffness rather than strength.
Simple trapezoidal profiles have usually been designed according to the
MRDA code? but neither this document nor alternatives readily available
in the UK envisaged the complicated shapes shown in Fig. 3. Such profiles
can only be calculated according to the Swedish code'! or the European
Recommendations. '? In addition, the maximum trough width requirement
given in the MRDA code was related to the thickness of insulation used at
that time and so it may not be appropriate to the much greater thicknesses
used today.
The most recent development in long span decking profiles is the 206 mm
deep cassette profile, stiffened in both directions, as shown in Fig. 4. It can

Fig. 4. Deep cassette decking profile.

 Design of Profiled Steel Sheeting and Decking 149

span up to 12m and can thus span directly between main frames without
the need for secondary members. An assessment of the design has been
made at the Royal Institute of Technology, Stockholm and at the University
of Salford.'® It was first used in Britain in a large building for the
Warrington Development Corporation. In Sweden it is often used with an
additional metal skin to replace the traditional felt and bitumen finish.

Fig. 5. Architectural use of wall cladding.

150 E. R. Bryan and J. M. Davies

5. WALL CLADDING

For many applications, the profiles used for roof sheeting are also suitable
for vertical cladding. In many respects the loading is similar except that the
requirement for the profile to sustain a concentrated point load no longer
applies.
Since the walls are often the most prominent feature of a building, the
choice of wall cladding is often determined more by aesthetic con-
siderations than structural performance (Fig. 5). In fact, a number of
profiles have been specially developed to give different impressions of scale
and light and shade (Fig. 6). In addition, the sheeting may span vertically or
horizontally or even diagonally (Fig. 7). Curved sheeting around the apex
or eaves of a building may also be used to good effect (Fig. 8). The choice of
colour and texture of wall cladding is probably even more important than
for roof sheeting.
A review of the whole field of sheeting and cladding is given in ref. 19.

Fig. 6. Typical wall cladding profiles.

Design of Profiled Steel Sheeting and Decking 151

Fig. 7. Industrial building in Salford.

Fig. 8. Salford postal delivery office.

 ‘JSIY SPA osu UOTsud] UdYM UONNQI}SIP ssois daNeUIOIe (9) {ISI SploIA oduey UOTssoIduIO0S
udYM UONNQLISIP ssas (q) fUOTdeS-ssoIN ATO (e) “Burpuoq ut sopyosd [eprozodey soj jopow jeondjeuy °6 “SY
E. R. Bryan and J. M. Davies
152
 Design of Profiled Steel Sheeting and Decking 153

6. CALCULATION PROCEDURES

This section describes briefly the analytical model that forms the basis of
the Swedish code’? and the European Recommendations.!? A full account
of the derivation of the design expressions has been given by Hoglund.”°

6.1. Trapezoidal Profiles

For trapezoidal profiles in bending, the basic analytical model is shown in
Fig. 9(a) and (b). The effective width b,, of the compression flange is
calculated using the Winter expression with a modification for bends of
appreciable radius. With the compression flange reduced in this way, the
depth e, to the approximate position of the neutral axis is calculated. The
effective portions of the compression zone ofthe web are then positioned as
shown in Fig. 9(a) and have lengths given by

Sep, =0.76t./ E/o, (1)

Sep2 = 1-35
e61

where @, is the stress at the level of the compression flange in the effective
cross-section.
Using the doubly reduced cross-section, the new section properties are
calculated so that, for first yield in the compression flange as shown in
Fig. 9(b), the ultimate moment of resistance M@, may be found.
In the above calculation it can be shown that, for reasonably symmetrical
profiles, the web is fully effective unless the ratio of profile height to
thickness is above about 70. For shallow trapezoidal profiles, therefore,
with bends of small radius, the calculated bending strength is almost
identical to that given in the MRDA code.
The European Recommendations also make provision for an increase of
bending capacity to be allowed when first yield occurs in the tension flange
as shown in Fig. 9(c). The tension flange is allowed to become plastic and
the ultimate moment of resistance M, is calculated when the compression
flange also reaches yield. If advantage is taken of this concession, a, is
always equal to the yield stress of the material, o,.
For continuous sheeting, where the bending moment changes sign over
the support, the above calculation has to be repeated in the case of
unsymmetrical sheeting. The ultimate moment of resistance over the
support will not generally be the same as that near midspan.
Unlike the MRDA code, the European Recommendations also require
that the support reaction capacity R, is calculated and that account is taken
154 E. R. Bryan and J. M. Davies

of the interaction between moment and reaction when sheeting is

continuous over an intermediate support.
The value of R, at which web crippling takes place at a support or under a
concentrated load is

Ry =0.151?,/Eo,(1 — 0.1,/r/1)(0.5 + ,/0.02/s/1)[2.4 + (0/90)?] (2)

where t=sheet thickness (mm), E=modulus of elasticity (N/mm7),
o, =yield strength (N/mm?’), r=inside radius of bends (r<10f),
/, =length of bearing (10mm </, < 200mm), and 6 =web slope (50° <
= 90%).
The relationship between the moment M and reaction R at the design
load and the above moment and reaction capacities, M, and R,, are given
in the Preliminary European Recommendations by the interaction diagram
and formulae of Fig. 10. As a result of considerable further experimental
evidence (possible only with mass produced structures) it is expected that
these interaction formulae will be changed somewhat in the final version.

6.2. Stiffened Profiles

The analytical model for stiffened profiles in bending is shown in Fig. 11(a)
and (b). An important point to note is that both web and flange stiffeners,
including their associated effective widths, are designed to a reduced stress.
The calculation of this reduced stress is quite complex and is based on the
philosophy of treating the stiffener as a strut restrained by an elastic
medium representing the remainder of the profile. The design procedure

0:25

M_ when Ra <0:25,

Mg
M
then Mg < 1:0

ies —0:52
2 en R_ >0:25 <10,
d

M R
then Ma ae Oe eee

6)
10) OFS Yip
1-0 aabe

Fig. 10. Interaction diagram for permissible values of ultimate moment M and
support reaction R at an internal support.
 Design of Profiled Steel Sheeting and Decking 155

is given in the form of explicit expressions in the European

Recommendations and will not be reproduced here as there are a number of
alternative cases to consider depending on the number of stiffeners and the
position of the neutral axis.
A modification factor for eqn (2) is also given so that the influence of web
stiffeners on the support reaction capacity R, may be considered. The
interaction diagram (Fig. 10) is then the same for both stiffened and
unstiffened profiles.

6.3. Deflection Calculations

It is not always appreciated that the effective width for deflection
calculations is not the same as the effective width for strength calculations,
although conservative results will always be obtained if the latter is used. As
the design of decking is usually limited by deflection, this point is important
and so the European Recommendations include separate expressions for
the effective width (in the flange and web) to be taken for deflection
calculations. This necessitates a separate calculation for the relevant
second moment of area J, which is associated with a particular bending
moment. It may be noted that, for deflection calculations, stiffeners may be
assumed to be fully stressed so that the stress reductions shown in Fig. 11(b)
do not apply.
The rigorous approach to deflection calculation requires that J, be
calculated for several values of the bending moment so that a curve of J,
versus moment can be drawn. An accurate value of the deflection can then

la ' Le I Sefi=0-76t VE/Oy

a Set2= Ser 1(1+0:'5hs/ec)

eo Sipe Seti (1+0-5h3/ec)

Y [\Sen=t 5 Pa
NA

Fig. 11. Analytical model for stiffened profiles in bending.

156 E. R. Bryan and J. M. Davies

be obtained by dividing the member into slices and using either computer
analysis or another appropriate numerical procedure. However, this is
rather impractical for general use and a suitable value of /, for practical
purposes can be obtained by calculating a single value at a bending moment
averaged over the entire length.

7. TESTING OF PROFILED SHEETING

Although the design of profiled sheeting and decking is carried out by

calculation, it is usually most desirable to test as well in order to observe the
actual behaviour and to obtain data on which to base the design
expressions. The definitive document regarding testing is the European
Recommendations’* which sets out procedures for the following tests.

7.1. Mid-Span Bending under Distributed Load

This test is to determine the bending moment capacity at sections where the
shear force is negligible. The test load on a simply supported span may be
applied by means of an air bag (Fig. 12), vacuum chamber or by two or four
line loads applied by jacks or dead load.

Fig. 12. Airbag test on deep cassette profile.

 Design of Profiled Steel Sheeting and Decking 157

uniformly distributed load Side 1

k
1 x
' Side 2 A

| : ! :
oe Side 2
Conditions at central support above are
equivalent to conditions under point load: — - Side14
O4L

Fig. 13. Equivalent support test for double span sheeting.

7.2. Bending Capacity at an Internal Support

This test is to determine the capacity of the sheeting under a combination of
bending moment and support reaction and may be carried out in one of two
ways. The arrangement that is most representative of practical conditions is
to test two spans of sheeting loaded by any of the methods given in 7.1
above (Fig. 13(a)). However, this is an expensive test so a simplified
alternative which simulates the condition at an internal support is often
used. In this alternative test, the section is inverted and subjected to a
downward line load at the centre of asimply supported span of0.4 times the
actual span to be considered (Fig. 13(b)). This arrangement gives the same
combination of bending moment and reaction as the two span test but
removes any benefit that may be obtained from continuity. A photograph
of this equivalent test is shown in Fig. 14.

Fig. 14. Test for combined bending and reaction at support.

158 E. R. Bryan and J. M. Davies

Fig. 15. Test for end support capacity.

7.3. Capacity at End Support

The purpose of this test is to determine the capacity of the sheeting under a
concentrated reaction (or other load) in the absence of bending. In fact it is
particularly severe as it specifies that the bearing surface at the end support
should be inclined at a slope of 1 in 20 so that it imposes a knife edge load on
the sheeting. A photograph of this test is shown in Fig. 15. Further tests on
an internal support have confirmed the rule given in the European
Recommendations that the permissible reaction at the extreme end of a
sheet should be taken as half of the value when the reaction is some distance
from the end.

7.4. Capacity to Resist Concentrated Loading

Although BS 5427® specifies that sheeting shall be able to withstand a
concentrated load of 0.9kKN on an area of 125mm x 125mm, the
European Recommendations are even more severe and specify a load of
1kN on an area of 100mm x 100mm. The positions at which load should
be applied are specified, as is the recovery required after the removal of the
load. Although the test is so simple (Fig. 16), for a number of profiles it is
quite critical.
The European Recommendations on testing of sheeting specify not only
testing procedures but also methods of interpreting and evaluating the test
 Design of Profiled Steel Sheeting and Decking 159

Fig. 16. Concentrated load test.

results. In particular, methods of determining the characteristic strength

are given and so are the corrections necessary for variations in sheet
thickness.

8. CORRELATION OF THEORY WITH TEST RESULTS

The tests illustrated in Figs 12-16 were carried out at the University of
Salford during a comprehensive assessment of the decking profiles
concerned. The results confirmed the design principles of the Swedish
160 E. R. Bryan and J. M. Davies

Fig. 17. Interaction between bending moment and support reaction for 1500
trapezoidal profiles.
 Design of Profiled Steel Sheeting and Decking 161

Code'! and helped gain acceptability of the relevant design manual in

Britain.
A much more exhaustive test programme was carried out by Héglund?°
in the course of his fundamental investigation into the design of sheeting.
For 98 simply supported sheets in bending, the mean value of experimental
ultimate moment M divided by theoretical ultimate moment M, was 1.05
with a standard deviation of 0.06. This provides adequate justification for
the method of calculation of the moment of resistance. Héglund also
carried out 134 tests on roll-formed steel sheet under combined moment
and reaction (R) and obtained a mean value of R/R, of 1.10 with a standard
deviation of 0.09.
In Germany, a study of the effects of combined moment and reaction
have been carried even further by Baehte and Fick”! with some 1500 tests
on 27 unstiffened and stiffened profiles. The tests were of the type shown in
Fig. 14 and refer to the interaction relationship given in Fig. 10:
M R R
M, Ma = .16(for OS <2 R, < 1.0)

The results of Baehre and Fick’s tests as well as the Swedish tests are plotted
in Fig. 17 together with several interaction formulae suggested as lower
bounds. Of these A is the expression given above (and as presently included
in the Preliminary European Recommendations), B is the more
conservative linear expression:

ccheey 1:25 (to0.25 ene 10)

M, Ry Ry

and Cisa parabolic interaction curve favoured by the German DIN 18807.
It is probable that the final European Recommendations will adopt
expression B. In fact, because in practice the values of R/R, are rather
small, the important part of the interaction curves is the upper part where
the differences are quite small.
The present state of the art thus requires elastic or quasi-elastic design of
continuous sheeting based on the calculated moment of resistance at the
internal supports. Figure 17 shows that this can in itself be very
conservative because many of the test points lie well above the lines
representing the alternative design expressions. When the results of tests
on two-span sheeting are compared with the above theory a further degree
of conservatism is often revealed. Most sheeting and decking profiles are
capable of considerable redistribution of bending moment so that the
attainment of the calculated moment of resistance at an internal support is
162 E. R. Bryan and J. M. Davies

not immediately followed by failure. Instead, a degree of ‘plastic hinge’

action takes place, possibly accompanied by a reduction in the moment of
resistance, until failure takes place when the full moment of resistance is
achieved within the span. Because the sagging mid-span moment of
resistance is often greater than the (reduced) hogging moment of resistance
at a support, the increase in the load carried as a result of this redistribution
of bending moment can be considerable.
At present, the actual carrying capacity of sheeting and decking profiles
continuous over several spans can only be determined by testing. However,
a method to allow for partial plasticity over internal supports and
consequent redistribution of bending moment has been proposed by
Lawson and Crisinel?? with particular reference to decking profiles used for
composite floors. In this type of construction, the ability of the profile to
carry the weight of the wet concrete, without resort to intermediate props, is
often the determining feature of the permissible span, and is therefore of
importance. An investigation of this method will shortly be carried out at
the University of Salford.

9. FUTURE TRENDS

It is not many years ago that insulation to sheeting or decking consisted of

12mm plasterboard or fibreboard. Today, thicknesses of insulation of
50mm, 100 mm, or even 200 mm in Sweden, are commonplace. This trend
has affected and will continue to affect the design of profiled sheeting.
Whereas, in Britain, it has been usual to use decking with troughs at
150 mm centres in order for the insulation to bridge the gap, this restriction
will not in future be nearly so severe. Also, in roof sheeting, the trend
towards longer span buildings and lower roof pitches will necessitate wider
troughs in the sheeting to get rid of the rainwater.
Other factors which will influence the design of sheeting are the local
loads at fasteners under wind suction and their effect on local deformation,
the requirement of some sheeting and decking profiles to act as stressed skin
diaphragms, and the continuing demand for improved aesthetics of
buildings.
Another important development which is gathering momentum in all
European countries is the use of insulated sandwich panels using skins of
sheet steel or aluminium. Although good text books exist on the theory of
sandwich panels, there is very little literature on the practical design of
such panels, especially when they are continuous over two or more
 Design of Profiled Steel Sheeting and Decking 163

spans. To rectify this deficiency, ECCS Committee TC7 has set up a Study
Group to make proposals for eventual European Recommendations on
Sandwich Panels. It is hoped that this document will serve as a base for all
national codes and will stimulate further the use of sheet steel in building.

REFERENCES

. Makins, N., Coil coating comes into its own, Financial Times, 20 May 1982.
No. BAEHRE, R., ‘Sheet metal panels for use in building construction’, Proceedings
third international specialty conference on cold formed steel structures,
University of Missouri-Rolla, November 1975.
. BRYAN, E. R., ‘European Recommendations for cold formed sheet steel in
building’, Proceedings fifth international specialty conference on cold formed
steel structures, University of Missouri-Rolla, November 1980.
. University of Salford, Conference on light gauge steel, 20 May 1981, p. 34.
. American Iron and Steel Institute, ‘Specification for the design of cold formed
steel structural members’, first edition April 1946, latest edition 3 September
1980.
. Metal Roof Deck Association, ‘Code of design and technical requirements for
light gauge metal roof decks’, first published 1965, third edition 1970.
. British Standards Institution, ‘Code of practice for sheet roof and wall
coverings’, CP 143: Part 10: 1973.
. British Standards Institution, ‘Code of practice for performance and loading
criteria for profiled sheeting in building’, BS 5427: 1976.
. ‘Profiled steel cladding and decking for commercial and industrial buildings’,
London, Constrado, 1980.
. ‘Stahltrapezprofil im Hochbau’, IFBS, Stuttgart, Karl Kramer Verlag, 1980.
. Swedish Institute of Steel Construction, ‘Swedish code for light-gauge metal
structures’, Publication 76, March 1982 (in English).
. British Standards Institution, ‘Structural use of steelwork in building: code of
practice for design of floors with profiled steel sheeting’, BS 5950: Part 4: 1982.
. European Convention for Constructional Steelwork, ‘Preliminary European
Recommendations for the design of profiled sheeting and sections: Part 1—
profiled sheeting’, ECCS-T7-1981, March 1981.
. European Convention for Constructional Steelwork, ‘European Recommenda-
tions for the testing of profiled metal sheets’, ECCS-XVII-77-2E, April
1977.
. European Convention for Constructional Steelwork, ‘Preliminary European
Recommendations for good practice in steel cladding and roofing’, ECCS-T7-
1981, December 1981.
. ANDREW, T., Cladding for life, Building with steel, December 1979. See also
other articles in this issue and in that of July 1979.
. Tarmac Building Products Ltd, Flat roofing—a guide to good practice, RIBA,
1982.
. BRYAN, E. R., Long-span steel decking, The Structural Engineer, 57A, No. 4
(April 1979).
164 E. R. Bryan and J. M. Davies

Sy, Davies, J. M., Profiled steel cladding, The Consulting Engineer (February
1980).
20. HOGLUND, T., ‘Design of trapezoidal sheeting provided with stiffeners in the
flanges and webs’, Swedish Council for Building Research, Document D28:
1980 (in English).
DAN. BAEHRE, R. and Fick, K., ‘Berechnung und Bemessung von Trapezprofilen—
mit Erlatrungen zur DIN 18807’, Berichte der Versuchanstalt fiir Stahl, Holz
und Steine der Universitat Fridericiana, Karlsruhe, 1982.
22) Lawson, R. M. and CRIsINEL, M., ‘Evaluation of the properties of profiled
sheeting’, private communication, 1982.
 10

The Instability of Composite Structures with Initial

Imperfections

I. H. MARSHALL
Paisley College of Technology, UK

and

W. M. BANKS
University of Strathclyde, Glasgow, UK

ABSTRACT

In general imperfections in shape significantly complicate the stability

analysis of structural elements such as plates and shells, more so in the case
of such elements manufactured using composite materials. An attempt is
made in this presentation to put such imperfections into perspective by
highlighting their influence on the buckling and post-buckling character-
istics of typical structural elements. Special reference to the additional
complications arising through the orthotropic material properties, common
in composite materials, is made.

NOTATION

Ags Amplitudes of plate initial and final deflections.

a, Plate dimensions in x and y directions respectively.
Dy, D1,
D,6,D22 Plate flexural rigidities.
Seren en Elastic modulii, for x and y directions respectively.
e Ratio of buckle half-wavelength to plate width.
F Airy Stress Function.
G5 Elastic shear modulus.
H Definedias (UiGis —2)i5/ Bia):
165
166 I. H. Marshall and W. M. Banks

kaaks Plate curvatures in x and y directions respectively.

N,, N,,, N, In-plane stress resultants.
Pike Non-dimensional lateral loading.
Uu/U..s Non-dimensional plate edge compression.
u Non-dimensional lateral pressure.
Wo, W Plate initial and final deflection respectively.
w,/h Non-dimensional plate lateral deflection at centre.
Mane Plate coordinate axes.
y Deflections across buckled plate.

1. INTRODUCTION AND BACKGROUND

The buckling of thin, isotropic, homogeneous, flat plates has been the
subject of a vast amount of previously published work. Papers on this
subject have appeared in the technical press for approximately a century
with in excess of two thousand publications presently available. With so
many references available for consultation the researcher could be forgiven
for being not only overawed by the quantity of previous studies but also
slightly bewildered by the diversity of topics considered and the variety of
analytical approaches utilised. Indeed the diversity of past analytical
approaches used in the solution of seemingly identical problems appears to
leave the investigator more than sufficiently armed to solve the particular
problem in hand. Unfortunately such thoughts are often misleading.
From the classical paper by Bryan! almost a century ago, through
notable landmarks? ~* to the present date many analytical techniques have
been successfully employed. Apart from classical closed-form solutions for
specific problems (relatively few in number) approximate techniques
utilising energy, semi-energy, finite difference, finite element, etc. have
found favour.
Although ‘approximate’, such solutions can converge rapidly to yield
realistically accurate buckling and post-buckling characteristics up to two
or three times the critical load. In general the accuracy of these solutions
depends on the number of terms utilised in mathematically modelling the
problem, or the mesh size employed in the case of finite element techniques.
In the case of semi-energy solutions the earliest recorded example used to
determine plate post-buckling characteristics was in 1937 in a paper by
Marguerre.® Essentially the technique involved postulating an expression
to specify the plate lateral deflections (w) thereby allowing an exact solution
to the Von Karman compatibility equation to be obtained. Having
determined an appropriate relationship between plate middle surface
 Instability of Composite Structures 167

stresses and deformations the total energy of the system is minimised to

yield the required relationship between end displacements and lateral
deflections. Many variations on this theme have been published since this
valuable addition to platework analysis some forty years ago.

1.1. Initially Imperfect Plates

The Galerkin technique (English translation in ref. 7) as an alternative to
energy principles has been successfully employed in the solution of initially
imperfect plate problems. In a worthy trilogy of papers on the stability of
plates with initial imperfections subject to edge expression,®~ '° Yamaki
sought an approximate solution to the fourth order differential equation of
equilibrium using the Galerkin approach. This was combined with an exact
solution of the compatibility equation in the same manner employed by
Marguerre. A latter paper by Rhodes and Harvey’? reconsidered the
imperfect plate problems solved by Yamaki using the method of Marguerre
and produced identical results. The effect of load eccentricity has also been
successfully studied by the same authors using this latter approach.
Other alternative approaches utilising numerical methods of analysis on
initially imperfect plates subject to a variety of loading!*:'* have similarly
contributed to an understanding of the significance of initial imperfections
on plate stability characteristics. The aforementioned references are by no
means a complete catalogue of events to date—rather they serve to
illustrate some examples of previous work on the buckling and post-
buckling of isotropic plates with initial imperfections. In view of the large
number of publications in this field a paper by Leissa'* summarising
current advances in this subject is especially welcome.

1.2. Composite Plates

The stability of composite plates has also received considerable attention in
the past. However, the number of publications in this area is significantly
less than that available for equivalent isotropic problems. Since practical
interest in the potential of composites as viable materials of construction is
confined to the last two or three decades this is not unexpected.
Unquestionably the complexities of solution attributable to non-isotropic
materials are significant. In the extreme, solutions to the buckling and post-
buckling of anisotropic plates are few and far between. Notable exceptions
to this rule are refs. 16 and 17. The vast proportion of work on composite
plate buckling can be found where orthotropic, sometimes termed
orthogonally anisotropic, material properties are assumed. Recent books
by Ashton and Whitney,'® Jones,’? Vinson and Chou,”° Chia”! and Tsai
and Hahn”? are worthy of mention.
168 I. H. Marshall and W. M. Banks

Fig. 1. Coordinate system and convention for imperfect plate.

As the present paper is primarily concerned with highlighting the

influence of initial imperfections, or lack of flatness of the middle surface,
on the stability characteristics of composite structures, the significance of
laminated systems, stacking sequence, angle of fibre orientation etc. will be
excluded. Thus the term ‘composite plate’ will be deemed to specify a
relatively thin, initially unstressed, rectangular planform, homogeneous
plate possessing orthotropic material properties (Fig. 1) which behaves
according to the Kirchhoff hypothesis. The initial imperfection or ‘lack of
flatness’ considered herein will be of the form shown in Fig. 1.
It should be noted that the term ‘specially orthotropic’ occasionally
appears in the literature. This infers that the axes of reinforcement coincide
with the plate geometrical axes. However, as this notation is not exclusive a
certain cautiousness must be employed when surveying literature in this
subject.

2. ORTHOTROPIC PLATES SUBJECT TO IN-PLANE

STRESSES OR LATERAL LOADING

The classical equilibrium equation for thin, rectangular, orthotropic plates

can be written
6?w O*w 6*w
Diy 5g + 2D 12 + 2D56) wa
5a Oy ae
d?w 6?w 6?w
(1)
 Instability of Composite Structures 169

Providing the in-plane forces N, and N, are constant and the in-plane
shear loading is zero an exact solution to (1) can be found. Typically this
solution is in the form of an eigenvalue (defining the buckling load) and an
eigenvector (defining the mode of buckling).
The Von Karman-type compatibility equation for thin, flat, orthotropic
plates in the presence of large deformations can be written in the form
Ue Ee O*F - 1 O¢F (0?w\? 0?w d?w
Eos Ox oy? ES Gy Shdx oy dx? dy? (2)
Simultaneous solution of eqns (1) and (2) would yield an exact, closed form
solution for the buckling and post-buckling characteristics of the problem
under consideration. However, such solutions are rare and generally
limited to the most straightforward of plate boundary conditions.
Commonly, solutions represent the plate deformations and stresses as
trigonometric Fourier series, thereby producing a set of simultaneous non-
linear algebraic equations equal in number to the chosen number of terms
in the respective series. Mathematically such solutions are ‘exact in the
limit’, 1.e. an ‘exact’ solution only results if an infinite number of terms in the
Fourier series are used. However, since an iterative technique is usually
involved in the solution, numerical approximations are unavoidable.
As an alternative to solving eqns (1) and (2) simultaneously minimum
energy principles can be employed in lieu of equilibrium considerations.
Thus, after satisfaction of compatibility and constitutive requirements is
obtained by solution of eqn (2) the total energy of the system is minimised
according to the semi-energy approach to yield on upper bound solution to
the problem. Providing realistic functions are chosen when defining the
plate lateral deflections a satisfactory solution with acceptable accuracy
will result. The Galerkin technique has also been successfully employed to
obtain an approximate solution to eqn (1) along with the aforementioned
minimum energy principles or occasionally a ‘double’ Galerkin solution to
eqns (1) and (2). In this latter case it is not possible to bound the resulting
solution.

2.1. Orthotropic Plates with Initial Imperfections

In the case of plates with initial geometrical imperfections or ‘lack of
flatness’ additional terms require to be added to the right hand side of the
compatibility equation (2). At this stage it may be opportune to consider
the meaning of ‘initial curvature’ as it is often cited in analytical studies.
Essentially two forms ofinitial curvature have predominated in previously
published studies, viz. constant initial curvature (Fig. 2) and non-constant
170 I. H. Marshall and W. M. Banks

Fig. 2. Coordinate system and convention for constant curvature plate.

initial curvature specified using a trigonometric series (Fig. 1). Other cases
of hyperbolic, parabolic etc. curvature have received scant attention.

2.1.1. Non-constant initial curvature

In 1952 Yussuff?? published a paper based on his doctoral thesis extending
the isotropic work. of Levy,’ Hu et al.?* and Coan?> to include
orthotropic material properties. As in previous isotropic investigations the
‘lack of flatness’ defined in Fig. 1 was specified using trigonometric Fourier
series. A finite difference approach was used to analyse the behaviour of
imperfect orthotropic plates under symmetrical or antisymmetrical
systems of transverse and in-plane loading. The computed strain values
were compared with earlier experimental studies on glass reinforced plastic
plates.*° This was followed in 1969 with a paper by Aalami and Chapman?’
on the large deflection behaviour of initially imperfect rectangular
orthotropic plates under transverse and in-plane loading. Essentially this
work was an extension of the previous work on transverse loading of Basu
and Chapman.’ As in ref. 27 the effect of membrane actions was shown to
be significant at deflections greater than the plate thickness. This was again
highlighted in papers by Balaz*? and Bauld et a/.*° in which failure
considerations were investigated.
In recent years a number of generalised solutions to imperfect
orthotropic structure problems have been published.*!~ 34
Banks et al. considered the effects of imperfections in orthotropic plates
 Instability of Composite Structures 171

both theoretically** and experimentally.*> The method of approach was

to consider the initial imperfection as similar in form to the deflections due
to local buckling. This both facilitated the analysis and meant that the
results were conservative. The plates considered are shown in Fig. 1 with
initial imperfections w, and total deflection w. The equation then linking
the stress function and the deflection function is
1 gue. O*F ec? ee
E,, 0x" ax? dy? * E,, dy"
_[{ ew 0?w 0?w 0? wy \* i 07Wy 07Wo 3)
~ \éx dy Ox? dy? \dxdy ax? dy?
Since the effect of imperfections is most marked around the critical
buckling load it was considered sufficient to analyse for a one term solution.
The deflection function was taken in the form

w = AY cos ae (4)
eb
and wy

Wy = Ao¥ cos -; (5)

The ‘semi-energy’ method was then used to determine the deflections and
stresses in the plates.
The effect of initial imperfections on load-end Aimiacement behaviour is
shown in Fig. 3. As can be seen, the effect of the imperfections is most

Fig. 3. Load-end displacement curves for square SS GRP plates for various initial
imperfections.
172 I. H. Marshall and W. M. Banks

O25 I) es) CAS) Cos

Fig. 4. Relationship between post-buckling load and central deflection for square
SS GRP plate with various imperfections.

marked around the buckling load of the plate. A similar set of results to this
figure is obtained by plotting the post-buckling load against the maximum
membrane stress in the plate. It is thus evident that while the magnitude of
the stresses is significantly affected around the buckling load, they are
virtually unaffected at the higher loads.
The same relative effect is observed for the out-of-plane deflections
presented in Fig. 4.
Banks and Harvey**> found that for load out-of-plane deflection
behaviour for orthotropic plates the deviation of experimental results from
those for a perfect plate increased as the aspect ratio of the plates increased.
This was considered to be due mainly to the difficulty in fabricating the
plates and the resulting imperfections obtained. This initial imperfection
will of course occur in practical applications for composite structures. A
typical load out-of-plane deflection curve for a GRP plate is shown in
Fig. 5. The experimental points are seen to tie in with a theoretical
prediction of an imperfection of half the plate thickness.
Stress distributions for this same plate are seen in Fig. 6. Again it can be
seen that significant variation from the perfect plate obtains for the
logitudinal membrane and bending stresses around the buckling load.
Imperfections are obviously an important governing factor on stresses in
this region of loading.
The isotropic problems of Rushton'*:!* have been extended to include
orthotropic material properties in recent publications,*°3° viz. the
snap-buckling of imperfect plates loaded on the convex surface.
 Instability of Composite Structures 173

Fer

Fig. 5. Load-out-of-plane deflection curves for imperfect plate.

2.1.2. Constant initial curvature

The compatibility equation for rectangular plates with constant uni or bi-
directional curvature can be written in the form
ligiesk O*F Ios
Ph a st eee
ED, OX Oxsay- PERCY 4
3 0-w \? 6?w d2w 0*w 6?w
~ \éx dy On eOy Ase ig OX
(6)

The latter two terms in eqn (6) are those necessary to define the initially
deflected shape, i.e. Fig. 2. Thus, a solution of eqn (6) along with any of the
aforementioned analytical techniques would -yield the relevant load/
deformation characteristics. The case of constant curvature plates subject
to in-plane compression appears to have received scant attention in the

XO. EXPERIMENTAL
Og8 EXPERIMENTAL

10 fe) -10 f20, <8) ao

STRESS IN MN/me

Fig. 6. Variations of longitudinal membrane and bending stresses at centre of

imperfect plate.
174

Fig. 7. Load-out-of-plane deflection curves for constant curvature plate

(kx
= ky = 1/50m*).

past. In an earlier study Marshall*? showed that the problem solved by

Yamaki extended to include orthotropic material properties and constant
initial curvature exhibited similar load/deflection characteristics to the
previous isotropic study—Fig.7. Although fundamentally different
problems, it should be noted that Yamaki considered a maximum initial
imperfection of 0.1 of the plate thickness and ref. 39 investigated large radii
of curvature. It should also be noted that the term ‘shallow initial curvature’
has been applied to the problems described by both Figs. 1 and 2 in many
earlier investigations. In a number of cases this looseness of terminology
has led to erroneous comparison of results. Recently the more complex
problem of generally layered composite plates with constant initial
curvature subject to a variety of in-plane loading was considered in papers
by Zhang and Matthews.*°~4?
In 1955 Burmistrov*? presented a paper to Inzhenernyi Sbornic on the
subject of double curved rectangular planform orthotropic panels subject
to uniform lateral pressure loading. Two boundary conditions were
considered, namely simply supported and pinned edges. The method of
solution adopted required each term of the compatibility equation to be
represented in Fourier series form, including the initial curvatures. Since
constant initial curvature was considered this would seem to be an
unnecessary innovation. The phenomenon of snap-buckling when such
plates are loaded on the convex surface was indicated for a range of
geometrical parameters although no experimental evidence is cited. A
year later two highly mathematical papers by Ambartusyan**** also
considered the snap-buckling of shallow shells. Although essentially the
same problem as Burmistrov studied, albeit in greater depth, the looseness
 Instability of Composite Structures gs

CLAMPED
SUPPORTS

@ LOADING
0 UNLOADING
cl :
SIMPLE SUPPORTS
nz\i1300mm

weh
Fig. 8. Lateral pressure-out of plane deflection relationship for constant
curvature plate.

of terminology is again apparent. Some twenty years later the snap-

buckling of constant initial curvature orthotropic plates subject to uniform
pressure and central point loading was theoretically and experimentally
analysed.*? Clamped and simply supported edges were considered.
Typically the results of this investigation were of the form of Fig. 8.
Although material properties are of considerable importance, the stability
of such plates was shown to be predominantly a function of initial geometry
with small deviations producing marked changes in structural response.
Further investigations on this general theme and in particular the influence
of unsymmetrical buckling were subsequently published.*°~**® These
publications showed that initially imperfect orthotropic plates can undergo
the phenomenon of snap-buckling when loaded on the convex face. As in
earlier investigations the influence of plate initial geometry was shown to be
of paramount importance.

3. CONCLUSIONS

From the typical examples quoted herein it is evident that initial

geometrical imperfections significantly complicate the analysis of com-
posite structures. Depending on the type of structural element under
consideration significant reductions in load carrying capacity can result.
176 I. H. Marshall and W. M. Banks

REFERENCES

. BRYAN, G. H., On the stability of a plane plate under thrusts in its own plane,
with applications to the ‘buckling’ of the sides of a ship, Proc. London Math.
Soc., 22 (1891) 54-67.
. Cox, H. L., ‘The buckling of thin plates in compression’, Tech. Report of the
Aero Committee, 1933-1934, pp. 443-63.
. Levy, S., ‘Bending of rectangular plates with large deflections’, NACA Report
No. 737, 1942, pp. 139-57.
. Korter, W. T., ‘The effective width of infinitely long, flat rectangular plates
under various conditions of edge restraint’, National Luchtvaartlaboratorium,
Rep. S287, December 1943.
. STEIN, M., ‘Loads and deformations of buckled rectangular plates’, NASA TR
R-40, 1959.
. MARGUERRE, K., ‘The apparent width of the plate in compression’", NACA TM
No. 833, July 1937.
. DuNCcAN, W. J., ‘The principles of the Galerkin method’, ARC Tech. Report,
1938.
. YAMAKI, N., Postbuckling behavior of rectangular plates with small initial
curvature loaded in edge compression, J. App/. Mech., Trans. Am. Soc. Mech.
Engrs., 26 (1959) 407-14.
. YAMAKI, N., Postbuckling behavior of rectangular plates with small initial
curvature loaded in edge compression—(continued), J. Appl. Mech., Trans.
Am. Soc. Mech. Engrs., 27 (1960) 335-42.
. YAMAKI, N., Experiments on the postbuckling behavior of square plates loaded
in edge compression, J. Appl. Mech., Trans. Am. Soc. Mech. Engrs., 28 (1961)
238-44.
. Ruopes, J. and Harvey, J. M., Examination of plate post-buckling behavior,
J. Mech. Eng. Sci., ASCE, 103, No. EM3 (1977) 461-78.
. RHopEs, J., HARvEy, J. M. and Fox, W. C., The load-carrying capacity of
initially imperfect eccentrically loaded plates, Int. J. of Mech. Sci., 17 (1975)
161-75.
. RUSHTON, K., Large deflexion of plates with initial curvature, Int. J. Mech.
Sci., 12 (1970) 1037-51.
. RusHTOoN, K., Buckling of laterally loaded plates having initial curvature, Jnt.
J. Mech. Sci., 14 (1972) 667-80.
. Letssa, A. W., ‘Advances and trends in plate buckling research’, Proc. Symp.
on Advances and Trends in Struct. and Solid Mech., Washington, USA,
October 1982.
. LEKHNITSKI, S. G., Anisotropic plates, English translation by S. W. Tsai and T.
Cheron, J. E. Ashton (ed.), Technomic Pub. Co., USA, 1970, pp. 248.
. AMBARTSUMYAN, S. A., Theory of anisotropic plates, English translation by T.
Cheron, J. E. Ashton (ed.), Technomic Pub. Co., USA, 1970.
. ASHTON, J. E. and Wuitney, J. M., Theory of laminated plates, Technomic
Pub. Co., USA, 1970.
19. Jones, R. M., Mechanics of composite materials, Scripta Book Co., USA, 1975.
20. VINSON, J. R. and CHou, T. W., Composite materials and their use in structures,
London, Applied Science Publishers, 1975.
 Instability of Composite Structures 177

Ze Cua, C. Y., Non-linear analysis of plates, New York, McGraw-Hill, 1980.

De Tsai, S. W. and Haun, H. T., /ntroduction to composite materials, Technomic
Pub. Co., USA, 1980.
Ws, YussurF, S., Large deflection theory for orthotropic rectangular plates
subjected to edge compression, J. Appl. Mech., Trans. ASME, 19 (1962)
446-50.
24. Hu, P. C., LUNpquist, E. E. and BAtporr, S. B., ‘Effect of small deviation
from flatness on effective width and buckling of plates in compression’, NACA
TN No. 1124, September 1946.
25) Coan, J. M., Large deflection theory for plates with small initial curvature
loaded in edge compression, J. Appl. Mech., Trans. ASME, 73 (1951)
143-S1.
26. Horr, N. J., BOLeEy, B. A. and Coan, J. M., The development of a technique for
testing stiff panels in edgewise compression, Proc. Soc. for Exp. Stress Analysis,
5 (1948) 14-24.
ay AALAMI, B. and CHAPMAN, J. C., Large deflection behavior of rectangular
orthotropic plates under transverse and in-plane loads, Proc. Inst. Civ. Engrs.,
42 (1969) 347-82.
28. Basu, A. K. and CHAPMAN, J. C., Large deflection behavior of transversely
loaded rectangular orthotropic plates, Proc. Inst. Civ. Engrs., 39 (1966)
79-110.
29. BALAZ, I., DyuBEK, J., MAQuol, R. and MASSONNET, CH., Ultimate load of
orthotropic plates in compression, Constr. Met. V. (June 1979) 15-26 (in
French).
30. BAuLD, N. R. and SATYAMURTHY, K., ‘Collapse load analysis for plates and
panels’, Air Force Flight Dynamics Lab., Wright Patterson AFB, Ohio, USA,
AFFDL-TR-79-3038, May 1979.
Sil. Murtl, A. A. and Harris, P. J., Further developments in the use of linear
displacement functions in the finite element analysis of orthotropic plates and
shells, Eng. J. (Montreal), 54 (Jan.—Feb. 1971) 1-10.
B25 WHANG, B., Laminated orthotropic plates and shallow shells, Comput. Struct.,
1 (October 1971) 465-93.
333 Voir, A. S., PONOMAREV, A. T. and PopyTALov, S. A., Dynamic properties
of wing panels made of composite materials, Mekh. Polim (USSR), 10, No. 4
(July-Aug. 1974) 662-9.
34. Banks, W. M., Harvey, J. M. and Ruopes, J. ‘The non-linear behavior of
composite panels with alternative membrane boundary conditions on the
unloaded edges’, Proc. ICCM/2, Toronto, Canada, 1978, pp. 316-36.
3). BANKS, W. M. and Harvey, J. M., Experimental study of stability problems in
composite materials, Stability problems in engineering structure and com-
ponents, T.M. Richards and P. Stanley (eds.), London, Applied Science
Publishers, 1978, pp. 1-22.
36. MARSHALL, I. H., RHOopEs, J. and BANKs, W. M., The non-linear behavior of
thin, orthotropic, panels under lateral loading, J. Mech. Eng. Sci., 19, No. |
(February 1977) 30-7.
SM MARSHALL, I. H., RHODES, J. and BANks, W. M., Experimental snap-buckling
behavior of thin GRP curved panels under lateral loading, Composites (April
1977) 81-6.
 I. H. Marshall and W. M. Banks

. MARSHALL, I. H., RHopgs, J. and BANKs, W. M., General investigation of

shallow, orthotropic shells, Acta Technica Academiae Scientarum Hangaricae,
B7 (1978) 69-86.
. MARSHALL, I. H., ‘The non-linear behavior of thin, initially curved orthotropic
plates’, Ph.D. Thesis, University of Strathclyde, Scotland, 1976.
. ZHANG, Y.and MATTHEWS, F. L., Initial buckling of curved panels of generally
layered composite materials, J. Composite Structures, 1, No. | (1983).
. ZHANG, Y. and MATTHEWS, F. L., Post buckling behavior of curved panels of
generally layered composite materials, J. Composite Structures, 1, No.2
(1983).
. ZHANG, Y. and MATTHEWs, F. L., Post buckling behavior of cylindrically
curved panels of generally layered composite materials with small initial
imperfections of geometry, Composite Structures, 1, No. 2 (1983).
. Burmistroy, E. F., Calculation of flat orthotropic shells with allowance for
final strain, Inzhenernyi sbornie, 22 (1955) 83-97 (in Russian).
. AMBARTUSYAN, S. A., ‘On the theory of anisotropic shallow shells’, NACA TM
No. 1424, 1956.
. AMBARTUSYAN, S. A., ‘On the calculation of shallow shells’, NACA TM
No. 1425, 1956.
. Ruopes, J. and MARSHALL, I. H., ‘Unsymmetrical buckling of laterally loaded
reinforced plastics shells’, Proc. ICCM/2, Toronto, Canada, 1978, pp. 303-15.
47. MARSHALL, I. H. and RHopDEs, J., "Membrane analysis of discrete reinforced
plastic shells subject to uniform pressure loading’, Proc. ICCM/3, Paris,
France, 1980, pp. 759-69.
48. MARSHALL, I. H., The effects of shape imperfections on the mechanical
characteristics of GRP plates, Mechanics of structured media, A. P. S.
Selvadura (ed.), Amsterdam, Elsevier, 1981, pp. 437—S2.
 1]

The Background to Forthcoming Design

Proposals for Two Shell Buckling Problems

G. D. GALLETLY
University of Liverpool, UK

ABSTRACT
The background to design proposals currently being considered by the BSI
for two shell buckling problems is discussed in the paper. The problems are
(a) the buckling of torispherical shells subjected to internal pressure and (b)
the buckling of cylindrical shells under the combined action of external
pressure and axial compression. The shells considered are those occurring in
practice rather than theoretically perfect ones, i.e. they are likely to have
initial shape imperfections and contain residual stresses. Some comments on
further work which is needed are also given.

NOTATION
Amplitude of wo.
Perpendicular distance from the axis of revolution to the centre of
curvature of the torus.
Constants—see eqns (7) and (8).
Diameter of the cylinder to which the torispherical shell is
attached.
Modulus of elasticity.
Constants—see eqns (9)-(12).
Length of cylindrical shell (combined loading case).
Circumferential direct stress resultant.
(i) Internal pressure in a torispherical head, or (ii) external pressure
at which buckling occurs under the combined loading.
179
180 G. D. Galletly

P- Plastic collapse pressure of a torispherical shell (taken as the

internal pressure to cause twice the yield point deflection at the
crown).
Per (i) The theoretical internal buckling pressure of a torispherical
shell (zero strain-hardening assumed in the analysis), or (ii)
theoretical buckling pressure of a perfect elastic cylindrical shell
subjected to external pressure alone.
v Mal Buckling pressure of an ‘as-fabricated’ cylindrical shell subjected
to external pressure alone (the p, values are usually predicted
lower-bound values).
Radius of cylindrical shell.
p/p., Stress ratios in perfect cylinders under combined loading.
Radius of spherical portion of torisphere.
0,,/Oo,,x Stress ratios in perfect cylinders under combined loading.
DAADADAD
ee
rot
SON
Ores
Radius of curvature in the normal plane perpendicular to the
meridian.
e) Ss Meridional radius of curvature of a shell of revolution.
Radius of toroidal portion (knuckle) of torisphere.
p/p, Stress ratios in welded cylinders under combined loading.
o,,/0,,, Stress ratios in welded cylinders under combined loading.
x Safety factor.
SEM (i) Thickness of toroidal portion of torisphere (usually assumed
constant throughout the torisphere in theoretical analyses), or (11)
thickness of cylindrical shell.
Initial geometric imperfection in the radial direction in a cylinder.
(1 — v’)(L/R)(R/t).
Theoretical axial buckling stress of a perfect elastic cylindrical
shell under axial compressive load alone.
Axial buckling stress of an ‘as-fabricated’ cylindrical shell under
axial compressive load alone (the o, ,.values are usually predicted
lower-bound values).
Axial compressive stress at which buckling occurs under combined
loading.
yp
Yield point of the shell wall material.

1. INTRODUCTION

The pressure vessel ‘Design Methods’ subcommittee of the BSI (i.e. PVE-/5)
deals with many design problems relating to shells. Two of the current
 Design Proposals for Shell Buckling Problems 181

ones on shell buckling are internally pressurised torispherical shells and

cylindrical shells under biaxial compression. The author has been involved
with both of these problems and, in this paper, an outline of the
background to some proposed rules for them is given. These proposals are
still under consideration by the BSI sub-committees and the final versions
of the rules may not be exactly the same as those presented herein.
However, the availability of the following material should enable shell
designers and others to comprehend how the forthcoming rules were
formulated.

2. BUCKLING OF INTERNALLY PRESSURISED

TORISPHERICAL SHELLS

Very thin torispherical shells made from stainless steel and having diameter
thickness (D/t) ratios in the range 700-1300 are frequently used in the food
processing and brewing industries; the shells are also used, but to a much
smaller extent, in the nuclear industry, e.g. in the primary tanks of some
fast reactors and in the roof structures of containment buildings in PWRs.
Despite the fact that many of these thin torispherical shells have been
manufactured (and have given satisfactory service), there are no rules
covering their design in BS 5500. This is because there is, at present, a
restriction in Section 3.5 of the Code that D/t should not exceed 500. The
reason for this limit is not entirely clear but, in part, its purpose must be to
minimise the occurrence ofinternal pressure buckling in the knuckle region
of the heads.
During the last ten years there have been many experiments conducted
on internally pressurised torispheres. As-manufactured stainless steel
models up to 3.6m diameter were tested by Kemper! and by Stanley and
Campbell.* Small (0.13 m dia.) machined models of aluminium were tested
by Gill and his co-workers** and by Galletly>® (the latter also tested
some carbon steel models). The last ten years has also seen the emergence of
a powerful elastic-plastic shell of revolution buckling program known as
Bosor 5.’ With this program it has been possible to analyse the internal
pressure buckling problem under discussion, at least for nominally perfect
shells.2~1° Large-deflections need to be taken into account in the
theoretical treatment and the relevant failure mode, for steel shells, is
frequently asymmetric plastic buckling, with buckles occurring around the
circumference (see Fig. 1).
Using Bosor 5 to conduct parametric surveys, on ellipsoidal and
182 G. D. Galletly

Fig. 1. Buckling due to internal pressure in a 3 m dia. stainless steel torispherical

shell. Outward buckles (left), inward buckle (right).

torispherical perfect shells, the Liverpool group produced curves of plastic

buckling pressures versus various geometric parameters.'!~ '* They were
also able to reduce their results to simple formulae, which is important for
design purposes. Using these simple formulae, and making some
reasonable assumptions, Galletly'* was able to obtain good correlation
between theory and the tests on the ‘as-manufactured’ torispheres tested by
Kemper and Stanley and Campbell. As a result of this work, a design
procedure for preventing buckling in torispherical shells under internal
pressure'*:'© has been suggested to the BSI and is currently under
consideration. The background to this design procedure is given below.

3. BEHAVIOUR OF INTERNALLY PRESSURISED

TORISPHERICAL SHELLS

Just over 25 years ago, a 15m diameter fluid coker with a torispherical
bottom head failed while undergoing its hydrostatic proof test. A
subsequent elastic analysis of the head by Galletly!’:'® showed that the
direct hoop stress in the knuckle was compressive and exceeded the yield
point of the material at a number of locations. As the design seemed
questionable, it was decided to have a limit analysis of the head carried out.
This was done by Drucker and Shield and the results of the study provided
the stimulus for their subsequent work on the subject.!®:?° In essence, their
results confirmed the predictions of the elastic analysis with regard to stress
levels.
 Design Proposals for Shell Buckling Problems 183

ee adel Spherical cap

a
Ta
Toroidal shell

Fig. 2. Geometry of the toroidal part of a torispherical shell.

It may seem strange that the knuckle (or toroidal portion) of the
torisphere should have compressive stresses in it when the loading is
internal pressure. If the pressure vessel is divided into its component parts
(cylinder, spherical cap and toroidal segment) then the first two will move
outwards when subjected to internal pressure whereas the toroidal part will
move inwards. This latter can be shown from the membrane theory of
shells,*? from which one finds for the hoop direct stress resultant, No:

No =PR,(1 — Ro/2R,) (1)

The second term in the bracket in eqn (1) involves the ratio of the principal
radius of curvature in the meridional plane (R,,) to that in the normal plane
perpendicular to the meridian (R,). As the value of R, decreases (i.e. the
meridional curve becomes tighter), so the ratio R,/R, increases and will
equal 2 for some value of R,. For smaller values of R,, N, becomes
compressive and the toroidal segment moves inwards.
For a toroidal shell segment of the type under discussion (see Fig. 2) the
values of R,, R, and R,/R, are given by
R,=r (2)
R,=r+a/sing (3)
R,/R, =1 +a/r sin ~ (4)

Since a/r is usually greater than 1.5, the ratio R,/R, will be greater than
two for the toroidal segments occurring in practice.
Since the hoop stress resultants are compressive in the toroidal region, it
means that buckling of the knuckle in the circumferential direction is a
possibility.* Whether it will actually occur depends on several factors, e.g.

* See Fig. 8 in ref. 13 for a simple model for demonstrating internal pressure
buckling.
184 G. D. Galletly

the geometric characteristics of the head, the material of construction and

its properties. In analysing the problem, bending effects have also to be
considered and, for the D/t range of most practical interest, it is necessary
to utilise the large-deflection theory of shells.
The compressive hoop stress resultants in the knuckle region also affect
the limit pressures of torispherical shells. It was this problem which was
studied by Drucker and Shield some twenty years ago. They used the small-
deflection theory of shells and determined the limit pressure at which three
plastic hinge circles formed in the head.'*:*° Subsequent investigations

= Asymmetric
One 60000 Plastic Buckling

\bf/in \ 7-777 = Axisymmetric

150r \ Plastic Collapse
\ ‘

\
200r . x
Fe Asymmetric
& . Elastic Buckling
Fer 50+ <Soe ~
( bf /in* ) = SS
yp= 30000.
0 Ibf/in
500 1000 1500 2000

Fig. 3. Curves ofp,, and p, for a steel torispherical shell and two values of ¢,,.

have shown that the use of the large-deflection theory of shells, and
considering an elastic-plastic analysis, often results in higher collapse
pressures than those given by the limit analysis.?”
Both circumferential buckling (with waxes or wrinkles in the hoop
direction) and the large deformation axisymmetric mode (obtained from a
limit analysis or otherwise) have to be considered in the design of a head.
Which one occurs in practice depends on the geometric characteristics and
the material properties of the head. This point is illustrated in Fig. 3. While
buckling is the main concern in this paper, it should be realised that it is not
the controlling failure mode for the thicker heads. Also, while elastic
buckling may not be of concern for steel heads made from the normal
steels, it is of interest for heads made from plastic materials (PVC, etc.) and
sometimes aluminium.
 Design Proposals for Shell Buckling Problems 185

4. TYPES OF TORISPHERICAL HEAD

In practice, torispherical heads are manufactured in three main ways, viz.

pressing and spinning, the crown and segment technique and deep drawing.
The writer has not been involved with deep drawn shells as yet, so they will
not be considered any further herein.
With spun shells, the thickness in the circumferential direction is usually
very uniform. The thickness is normally a maximum at the crown and the
amount of thinning, caused by the forming operation, varies along the
meridian (with the maximum reduction in thickness usually occurring in
the knuckle). The larger spun heads also have meridional welds in them.
With heads made from strain-hardening materials (e.g. austenitic stainless
steel) the reduction in thickness causes a change in the mechanical
properties of the head and the change can be substantial. Even when the
amount of thinning can be estimated, the problem of assessing the
mechanical properties in the knuckle region of an ‘as-manufactured’
torispherical shell still remains.
With crown and segment heads, a spherical cap is welded on to a toroidal
portion which is itself made up of several portions welded together. With
this type of head, there are both circumferential and meridional welds and
the circularity of the heads is usually not as good as for the spun heads.
However, as there is not much reduction in thickness, the change in
mechanical properties is also less than with the spun heads.

5. DEVELOPMENT OF DESIGN RULES FOR THE BUCKLING

OF INTERNALLY PRESSURISED TORISPHERES

The sequence of events leading to the proposed design rules was more or
less as follows.

(1) The tests of Galletly and Lagae on small machined models were
analysed using the Bosor 5 shell buckling computer program.*’®’?°
The actual stress-strain properties of the shell wall material were
used in the analyses, as well as the meridional variations in
thickness. The wall thickness of the models in the hoop direction
was not constant but, for the D/t = 500 models, the variation was
not large and it was ignored in the analyses.
In general, the correlation between theory and experiment was
reasonably good. However, the definition of when buckling occurs
186 G. D. Galletly

is not fully resolved. In the Liverpool tests, it was taken as the

pressure at which the first buckle was visible to the naked eye. In the
UMIST tests, a sensitive rotating probe was used to determine the
location of the incipient buckles.
(2) Parametric studies to determine the buckling pressures (p,,) of
internally pressurised torispheres were then carried out using the
BOSOR 5 program. Various curves of p,, versus geometric
parameters were drawn and these are given in refs. 12 and 14. In
addition, an effort was made to derive simple formulae for the p,,
values, as it was felt that these would be useful to designers. The
discrepancies between the p,, values calculated from the simple
formulae and the computer results may be found in refs. 13 and 14.
The above parametric studies were carried out for shells made
from elastic, perfectly plastic materials (i.e. with no strain-
hardening) and the shell thickness of the three components
(cylinder, torus, spherical cap) was constant. Both steel and
aluminium shells were studied but only steel shells will be
considered herein. Some limited investigations of strain-hardening
were conducted but were only of an exploratory nature—see refs 11
and 12.
(3) The next step was to use the Bosor 5 program to analyse some ‘as-
fabricated’ torispheres for which experimental buckling results
were available. With one or two exceptions, the test results
published in the literature were for austenitic stainless steel
vessels!:? and the heads selected for analysis were four 54 in. dia.
spun torispheres tested at Nottingham.” These shells had been
made from single sheets so they had no meridional welds in them,
i.e. they were axisymmetric shells of revolution. A disadvantage of
using these shells was that strain-hardening had to be considered in
the analysis as they had been formed by spinning.
For the Nottingham tests, stress-strain curves for the toroidal
portion of the as-manufactured shells were not available. However,
test strips of the as-received plate material had been rolled down to
various thicknesses and information was available on certain
mechanical properties (e.g. the 0.2% proof stress) versus the
reduction in thickness of the as-received plate. A stress-strain curve
had also been obtained on one of the rolled test strips whose
thickness was known (curve A in Fig. 4). The thickness of the four
Nottingham spun heads had been measured after manufacture and
the average reductions in thickness in the toroidal portions were
 Design Proposals for Shell Buckling Problems 187

150
Stress
(Ibf /in? )
+10?
100

50

ed
0. le2igea5 10 5 20
Strain (in/,_x103)

Fig. 4. Stress—strain curves for the Nottingham spun heads.

known. It turned out that these latter values were approximately

the same and an average value for the four heads was, therefore,
taken. The 0.2% proof stress corresponding to this average
reduction in thickness was obtained from the experimental curve
relating these variables. This latter value, divided by the 0.2%
proof stress for curve A in Fig. 4, was used as the factor to obtain
curve B in Fig. 4. To determine the internal buckling pressures of
the four 54in. dia. spun heads with sBosor 5 the following
assumptions were made:

(a) The stress-strain curve B in Fig. 4 was used throughout, for all
the heads. Actually, it should only be employed for the toroidal
region of the heads; however, as this region is where buckling
occurs, there should be no great error in using it for the whole
head.
(b) The meridional variations of thickness were taken into
account.
(c) Residual stresses were ignored.
(d) The nominal radii of curvature were used in the analyses.

With the above assumptions, the Bosor 5 predictions of p,, agreed

to within 20% of the test values; this degree of agreement can be
considered to be very satisfactory.
188 G. D. Galletly

In addition to the above buckling analyses using Bosor 5, it was

decided to carry out buckling analyses in which elastic, perfectly
plastic stress-strain curves, with different values of o,,, were used.
The intention was to see, for the four Nottingham spun heads,
which materials property could be used (instead of o,,) for strain-
hardening materials. It turned out that the 0.2 % proof stress was a
reasonable compomise, although there is nothing immutable about
this choice.
(4) Rather than having a design method for these heads which requires
access to a large computer, it was decided to investigate the simple
approximate formulae mentioned earlier. For steel torispherica!
shells, the internal buckling pressure may be found from

__ 285[1 — 1250,,/E](r/D)°-*4
Per! Oyp = (Dif *2(RID Ys (5)

As mentioned before, this equation was obtained for perfect

torispheres (no residual stresses, no shape imperfections) made
from elastic, perfectly plastic steel (no strain-hardening) in which
the thickness was constant throughout (for cylinder, torus and
spherical cap). However, ‘as-manufactured’ torispheres are by no
means perfect. They contain residual stresses, they have shape
imperfections, their thicknesses vary along the meridian and, if
made from a strain-hardening material, their mechanical proper-
ties may change due to the forming operation. The problem is, of
course, to relate the perfect torisphere to the actual torisphere.
There is not much information available on residual stresses and
shape imperfections in thin torispherical shells. Thus, these will be
ignored and will have to be taken into account via a composite
‘safety’ factor. With regard to the thickness to be used in eqn (5), it
would seem that the average thickness in the knuckle region is a
reasonable choice (although for design purposes, the minimum
value in the knuckle might be better). For o,,, the calculations
discussed above indicate that the 0.2 % proof stress is an acceptable
substitute. The remaining problem is to account for the change of
mechanical properties in cold spun heads made from, say,
austenitic stainless steel. As a first approximation to this, it was
assumed that the o,, for these would be the 0.2% proof stress
corresponding to the average reduction in thickness in the toroidal
region of the ‘as-manufactured’ head.
 Design Proposals for Shell Buckling Problems 189

The foregoing procedure was applied to eleven stainless steel

spun heads (whose diameters ranged from 54 in. to 108 in.) tested
by Stanley and Campbell.? The value of o,, used in these
calculations varied from 1.46 to 2.19 times the o,,, of the as-received
plate material. All the experimental buckling pressures were higher
than the theoretical ones for these heads (the ratios ranging from
1.03 to 1.72). The procedure was also tried on the six crown and
segment torispheres tested by the Nottingham group.” For these
heads, all the residual stresses were again neglected and a,, and ¢
were taken as the values of the as-received plate. The ratios of the
experimental p,, to the theoretical p,, ranged from 0.82 to 1.27 in
these cases.
From the above comparison of the theoretical and experimental
results, it may be seen that use of eqn(5) for predicting the
experimental internal buckling pressures of as-fabricated toris-
pherical shells is very promising. However, in relation to a design
procedure, appropriate safety factors for the two head types need
to be introduced.
(5) In the written discussion of the paper in which the above procedure
was proposed,'* a suggestion regarding its implementation in
practice was made by Kendrick. This was (i) to multiply the 0.2 %
proof stress by a constant factor of 1.5 in order to account for the
strain-hardening in cold spun heads made from austenitic stainless
steel, and (ii) to use a constant factor of 1.75 to obtain the design
pressure from the buckling pressure predicted by eqn (5).
Apropos the 1.5 factor for strain-hardening, the values applied
to the 0.2 % proof stress of the as-received plate in the Nottingham
tests were, as stated above, in the range 1.46—2.19. The value of 1.5
is thus at the lower end of the range but it does not seem an
unreasonable choice. It could, of course, be changed to some other,
equally suitable, value.
The second factor mentioned above (i.e. 1.75) is suitable for spun
heads but not crown and segment ones. This becomes apparent
when as-fabricated torispherical shells other than the Nottingham
ones are considered, e.g. those tested by Kemper.' As is shown in
refs. 15 and 16, an appropriate factor for the crown and segment
heads, by which the theoretical p,, values should be divided, is
about 2.25.
Considering the Kemper tests and those of Stanley and
Campbell (and a few others) the use of the procedure outlined
190 G. D. Galletly
above gave a minimum margin of safety of 1.5 (see refs. 15 and 16).
This state of affairs can be considered to be quite satisfactory.
It should be noted that all of the tests in ref. 2, and most of those
in ref. 1, were conducted on torispheres made from 304 High Proof
stainless steel. It may be that some of the constants in the above
procedure will have to be modified for stainless steels which have
lower 0.2 % proof stresses (e.g. 316 S12). This may also be the case
with other materials. However, for carbon steel torispheres, the
design method seems satisfactory if o,, is taken as the normal yield
point stress—see ref. 16.

6. THE DESIGN PROCEDURE BEING CONSIDERED FOR

PREVENTING BUCKLING IN INTERNALLY PRESSURISED
TORISPHERES

When the steels normally used in dished end construction are considered,
an average value of the square-bracketed term in eqn (5) is approximately
0.8. Substituting this value in eqn (5) leads to
230)
(6)
Paley ea
(Dj)! 53(R/D)"
As mentioned earlier, the value of o,,, for austenitic stainless steel is taken as
the 0.2% proof stress of the ‘as-received’ plate material for crown and
segment ends. For cold spun ends, which experience thinning during the
forming operation, a factor of 1.5 is applied to the foregoing 0.2 % proof
stress.
As manufactured torispherical shells contain residual stresses (due to
welding and forming) and shape imperfections. To account for these and
other items, the theoretical p,, values given by eqn (6) have to be divided by
certain factors, C, to bring them into agreement with the experimental
buckling results. The value of C depends on the method used to
manufacture the ends and is as follows:
ee e for cold spun ends (7)
2.3 for crown and segment ends
The design pressure, p, for a cold formed torispherical shell is then found
from
P=P./C (8)
and this equation gives a safety factor of about 1.5.
 Design Proposals for Shell Buckling Problems 191

Amalgamating the various points above, the design pressure p is given by

the following equation:

P Ginyes
0.2% proof stress ae (Ditech) 0)
where
oe 100 for crown and segment ends (10)
203 for cold spun ends
and the 0.2 % proof stress is that of the as-received plate. Equation (10) is
useful for determining the design pressure of a torispherical head whose
dimensions are known. Inversion of eqn (10) gives
0.65 R/D Unne4

t/D = K, 0.2%0 proof

2 stress tg , 55
(r/D)”: (11)

where
0.050 for crown and segment ends
Kee 0.032 for cold spun ends
(12)
The thickness of the knuckle region can be calculated directly from eqn (11)
if the pressure, etc. are known. The form of eqn (11) is very similar to that
used in the 1982 edition of the French Pressure Vessel Code CODAP??
for this problem. However, CODAP does not appear to cite the source of
its buckling equation nor does it differentiate between spun ends and crown
and segment ends. The Appendix in ref. 16 discusses both of these points.
An equation which is slightly simpler than eqn(6) has also been
suggested by the author:

PerlOyp= 260(r/R,)(D/t)** (13)

As is shown in ref. 16, eqn (13) can be used to predict the test pressures of
the stainless steel heads in refs. 1 and 2 and its accuracy is nearly as good as
that of eqn (6).

7. FINAL COMMENT

The design procedure outlined above seems satisfactory insofar as cir-

cumferential buckling of internally pressurised stainless steel torispherical
shells is concerned. It also seems promising for torispherical shells made
from other materials.
192 G. D. Galletly

One of the problems remaining is how the new buckling rules outlined
above should be merged with the curves which are given in Section 3.5 of the
present BS 5500. These curves apply only to ends with t/D > 0.002 and their
origin is a little uncertain. The curves do not seem to have had much
experimental verification but it is believed that they refer to the
axisymmetric large displacement failure mode. It is also believed that the
present curves are safe. Indeed, for cold spun heads made from austenitic
stainless steel, they are probably too safe near t/D = 0.002. It is this facet of
the problem which is being studied currently. It is hoped to arrive at a
satisfactory solution for the transition region around t/D = 0.002 in the
near future.*
This may involve the use of the simple approximate equation for the
axisymmetric plastic collapse of steel torispheres. This is, from ref. 24,

_ 12.6[1 + 2400,,/E](r/D)!-°4
Peo (Di iyk2? (RD yas (14)

8. BUCKLING OF CYLINDRICAL SHELLS SUBJECTED TO

BIAXIAL COM PRESSION

Cylindrical shells subjected to combined external pressure and axial

compressive load are of interest to the offshore and nuclear industries.
However, this interactive buckling problem is by no means a new one and it
has been studied before (in the elastic region, in the main) by researchers in
the aerospace industry. A theoretical solution to the problem for perfect
cylinders was given by Fliigge’° in the 1930s using linear shell theory. On
the experimental side, a number of tests on 200mm diameter Mylar
cylinders were carried out in the USA about twenty years ago.”°
Interactive problems are often discussed in terms of stress ratios. For
perfect cylindrical shells, these will be denoted herein by R, and R,, where

Ry, =PIPex
R= ee >
and p=external pressure at which buckling occurs under combined
loading, o, = axial compressive stress at which buckling occurs under
combined loading, p,, = theoretical buckling pressure of a perfect elastic

* For crown and segment heads, the proposed buckling rules merge quite well with
the present BS 5500 rules near t/D = 0.002.
 Design Proposals
for Shell Buckling Problems 193

cylindrical shell subjected to external pressure alone, and o,, ..= theoretical
axial buckling stress of a perfect elastic cylindrical shell under axial
compression load alone.
In many cases, there is very little difference between the theoretical
buckling pressures for hydrostatic and lateral pressure and the distinction
will be ignored herein. In this paper, the pR/2t term due to the external
pressure will also be added to o,.
The theoretical interactive buckling curves given by the analysis in ref. 25
reduce to a straight line for reasonably long, simply-supported, perfect
elastic cylinders, 1.e.
R,+R,=1 (16)
For very short perfect elastic cylinders, ref. 27 queries the adequacy of
eqn (16). However, as is shown in ref. 28, the controlling buckling mode for
perfect cylindrical shells made from normal structural steels is the elastic-
plastic one and not the elastic one. The interaction curves for elastic-plastic
buckling are (see ref. 28) curved away from the origin and eqn (16) is a
conservative estimate of them.
When the elastic buckling test results of ref. 26 are plotted in terms of R,
and R,, the plot shown in Fig. 5 ensues. It is quite clear that many of the
points fall below the line R, + R, =1. In the main, this is due to the initial
geometric imperfections in the models which can reduce the theoretical
buckling load considerably, especially for axially compressed cylinders.
Some theoretical results on the effect of axisymmetric imperfections for this

[Ne
Model/LHe R ee
eo} 1 3 |400
[4 2 | 2 [400/526]
e| 3 | 2 |800/3056
[e[ 4 [15«@ li800 [1715|
L
SHS, © | 1 | 800 764
Riy@enl2 viseslzose
[=[9.10] 2 [So0|i906
eg: 4& 4
—1 , 19 e-entnac—|—__|_BR gy
005025040608 10 12,14 16,00
P Ocre

Fig. 5. Tests of Weingarten et al.*° plotted in terms of buckling stresses for

perfect cylinders (i.e. o,,., and o,,9).
194 G. D. Galletly

1:0

0:8

re) 06
Tah
0:4

0-2

02.04 06. 0:8. 10. - “s

Fig. 6. Effect of axisymmetric initial imperfections (w,) on the buckling stress of
axially compressed cylinders.

latter problem are shown in Fig. 6. They were obtained by Koiter (see
ref. 29), assuming the imperfections were axisymmetric and had the shape
shown in Fig. 7. As may be seen from Fig. 6, considerable decreases in
buckling resistance are predicted by this analysis for relatively small
amplitudes of the initial imperfections.
Simple linear, or non-linear, interaction equations have obvious
attractions to designers. Sometimes their form can be derived theoretically
but often they have to be determined empirically. An account of the
philosophy behind the derivation of the latter may be found in ref. 30.

=
wo=A cos q,(R)

q,2182/Rit

Fig. 7. Shape of axisymmetric initial imperfection considered in refs. 29, 31 and

32 (the above shape is for R/t = 100, Z = 200).
 Design Proposals for Shell Buckling Problems 195

Insofar as the present problem is concerned, a fairly obvious modification

to eqn (16) is to use buckling quantities which are appropriate to the
imperfect cylinders found in practice rather than those for perfect
cylinders. Introducing the quantities p, and ox,,where

Px = buckling pressure of an ‘as-fabricated’ cylindrical shell subjected

to external pressure alone
Ox,, = axial buckling stress of an ‘as-fabricated’ cylindrical shell under
axial compression load alone

the stress ratios S, and S, are then defined by

Ss. es 6,/Ox,x

By analogy with eqn (16), a linear interactive buckling equation in terms of

the stress ratios S, and S, is then
S,+S,=1 (18)
The adequacy of this equation for design purposes will be discussed later
when the various buckling test results are summarised.
Another way of assessing the effects of initial geometric imperfections 1s
to assume that a certain shape of imperfection is present in the cylinder and
then solve the buckling problem theoretically. Using the axisymmetric
shape shown in Fig. 7 for the initial radial imperfection, approximate?’
and more accurate*? solutions for the elastic buckling of imperfect
cylinders have been obtained. The two solutions, for a cylindrical shell with
the geometric characteristics R/t=100, L/R=1.45 and A/t=0.1, are
compared in Fig. 8. It can be seen that there is reasonable overall
agreement between the two solutions for the imperfect shells. With regard
to the effects of non-axisymmetric initial imperfections, the authors of
ref. 33 assumed that they had the same shape as one ofthe buckling modes
and Koiter’s analysis** was used to analyse the initial post-buckling
behaviour.
Most of the theoretical investigations to date on the effects of initial
geometric imperfections on the elastic-plastic buckling behaviour of
cylinders have assumed axisymmetric initial imperfections and used the
computer in their numerical studies.78*°° No general results have
emerged from these studies as yet, other than that the line S,+5,=1
appears to be a lower bound to the interaction curves.
196 G. D. Galletly

02 04 06 08 10“,
x x

Fig. 8. Theoretical elastic buckling interaction curves for perfect and imperfect
cylinders subjected to combined axial load and external radial pressure (R/t = 100,
Z = 200). (a) Simply supported ends, from ref. 31; (b) Clamped ends, from ref. 32.
, Perfect; --—~—, axisymmetric imperfection (A = 0.12).

9. THE EXPERIMENTAL DATA BASE

Rather surprisingly, there have not been many tests conducted on this
interactive buckling problem, particularly in the elastic-plastic region. A
summary of all the tests?°°*’~*! known to the writer, and published
before 1983, is given in Table 1. As may be seen, only a few of the tests have
been conducted on welded steel cylinders and only those tested by Miller
have been on large diameter cylinders.
More tests on this interactive buckling problem are currently under way
in the UK and the USA. However, it will probably be some time before these
results are available. When they are, they will enable any interaction
equations which have been proposed in the meantime to be validated (or
rejected).
If attention is now turned to the test data available on the buckling of
cylinders subjected either to external pressure alone or to axial compression
alone, then the situation is very different. For both of these cases, the
number of tests which have been conducted runs into many hundreds. The
test results for external pressure buckling (from ref. 42) are given in Fig. 9
while those for axial compression (from ref. 43) are shown in Fig. 10. There
is considerable scatter on both these plots, which may be due to initial
imperfections, boundary conditions, residual stresses, non-uniformity of
axial load, etc. However, for design purposes, most national Codes utilise
 9qey
T
Burpyong
sysoy,uo [eolpUT[AD
s|fays poyoafqng
07 pauIquioD
[eIxW UOIssoIdWIOD
PUB [BUIOXY SANSSoIg
fay ssoyinp /2pOW dupss [apow ‘ON y/T 1/4 -xoaddy ‘xosddy ainioy dappunog
"9 00x
joraaiou
— Sujuaffiis‘wip fo "q en fo -_ suopuox2 2
(ut) sjapou— ¥) /Aj (uu
Ny) (,wud/
yong Burj rp ‘
97 ‘udIeBUIaA\
TeAW ON WANED car =¢-l —-00r €8'r CL onse[q UseMjog —100'0 —100
uesiojy
pue ayeordnp 008 puess 000 €0'0
9Pl9g dO
SOsEe CUNTUTUIN ON COTO 8~y I =e 69 cs -onselq ‘xoiddy—-v0 -LIl
Aojye (yos) LY onseydSS 61 007
8¢ uesun| Sse[SIXxa]q ON esp'0 8 C798"
=6 00'S — onseg ‘xoiddy
— rs
whe We HO
6€ TONAL “[291§ BUlI—soX
SE9 ¢ =3'O OOS Vie DLE onseyq sury “Oo KEW
popjom sIouayys LOI siouayNs
9¢'0
Op sA0I)pue WITTY -BulI—soX
OTT: = b GaenCGi0) OL 69 OIP onseyq Bury =e OO
udsyLIpIG Aojye ysry) sIOUdNS bro siouayNs
p70
(yy3uans
Ip “‘IOMeM *[291$ ON 0 ce 9 0 sIE S061 c0c ele -onselg “xoiddy -€€0 O1¥
[esagpur poprom
pue 18°0 onseyd dO 880
ITROOW PoAdT]aI-sso3ys

, » porejngyeo
Wo GSQOq sUONepUSUIUIODEI
%) = ‘UONONpel
JO ‘UMOP-YOUY
10]9RJ
JOJ [VIX ‘(WorssaIdUIOD
198 G. D. Galletly

Lower bound of
test results

External Pre ssure |

rer/y,

Fig. 9. Test results on the buckling of cylindrical shells subjected to external

pressure (from ref. 42).

Lower bound
of
test results

1000 2000 3000 RA

Fig. 10. Test results on the buckling of cylindrical shells subjected to axial
compression (from ref. 43).
 Design Proposals for Shell Buckling Problems 199

lower-bound curves to the test data and give equations whereby the lower-
bound values may be calculated. These latter are the quantities py, and ox,
which appear in eqn (17) and they will be discussed further in the next
section.

10. THE MINIMUM PREDICTED BUCKLING QUANTITIES p,

AND ox,

Despite the fact that the various national Code-making bodies must start
with more or less the same test data, their predicted values of py and ox,.are
not the same. The reasons for the differences are discussed in ref. 44 but the
main point is the fact that differences do occur. The magnitudes of the
differences, for axial compression alone and external pressure alone, are
shown in Figs 11 and 12 respectively. As may be seen from Fig. 12 (external
pressure case), there is a large difference between the ASME III N-284
values and those of ECCS/BS 5500 in the elastic buckling region (i.e. the
portion near the origin).
Since the various Codes often predict different values of py and o, ..fora
cylinder with a given geometry, it is clear that the appearance of the same

— ECCS (with Y= “/3)

Oleh ssi ASME III Code Case N-284
Oy == DASt 013
A
10+ x x x x DnV

. PS
~SA ~ SS
0:2F (a) Oyp= 248 MN/m?
[

le
Sak
Sy 2F (b) Oyp= 344MN/m2 Sse,
0 i | al al a a
100 300 600” 100 S00 a econ
R/t R/t
Fig. 11. Predicted minimum buckling stresses according to several codes: axial
compressive load.
200 G. D. Galletly

10+ — DASt 013 pte Pieri

a :
Ox9 3
Oyp =0-6F ASME Ill Code Case N-284
BS 5500/ECCS

Fig. 12. Predicted minimum buckling stresses according to several codes:

hydrostatic external pressure.

test results, when plotted in terms of S, and S,, will differ and will depend on
which Code is used. Examples of this point will be given shortly.
The equations used by various Codes to determine py, and ox , are given
in ref. 44 and will not be repeated here.

11. TOWARDS A POSSIBLE DESIGN PROCEDURE

In essence, eqn (18) has been used by the US aerospace industry for the
elastic buckling of cylindrical shells. It is natural, therefore, to examine it
on the tests listed in Table 1. This will be done in this section and the Codes
used to evaluate py and ox ,will be ASME III Code Case N-284,*° DASt
O134°Dav* andsECGS.**
It is, of course, not necessary to utilise a linear interaction relation in a
buckling design procedure. A part-linear, part-quadratic equation which
includes the effective stress and the yield point of the material has recently
been suggested by Odland.*? The equation is not difficult to use and its
predictions for the tests of Walker er al. were conservative—see ref. 28.
This was also the case when his equation was used for Miller’s tests.
However, in this paper the main interest will be on the linear interaction
relation S,, + S, = 1. To examine its adequacy, all the tests in Table 1 were
plotted in terms of S,, and S, and the latter were calculated according to the
four Codes mentioned earlier. As examples of the results obtained, Figs. 13
and 14 show the plots for the two series of models made from welded
 Design Proposals for Shell Buckling Problems 201

Om0 1808 12816.20 5 0 04 0-8 12 16 20 &

0 1 ae 1 rt > L a 1 i 1

(a) ECCS (b) DASt 013

Model

S =f1_[07]
500]551 S
Ye [al 2 [o72{soof2u|
ro [#| 3 [os6fsoo]61] 5)
ef 4 [0-18] 500153]
eS
(paz 12

° 0-8
‘4 @ 5
a Spt Sy! Ce
L 1 1 | 4 4 »> 0 wie n ‘nas Fee
TS TER aa 0 04 08 1) 18 20 if
x
(c) ASME Ill N-284 (d) Dnv
Fig. 13. The elastic buckling tests of Miller? on 6.35 m dia. welded steel shells.
Results plotted in terms of S, and S, and according to four different codes.

steel.°?’*! As may be seen for these cases, the straight line S, + S, = 1 was
safe for all four Codes. It was safer for the smaller models of ref. 41 than for
those of ref. 39. Whether this is due to the fact that the smaller models were
stress-relieved or because their ends were clamped is not clear at the
moment. As mentioned before, it is also clear that the same test results may
look safer according to one Code than another (it should be noted that no
safety factors have been introduced into the discussion so far, although
they will need to be considered in a design procedure).
 G. D. Galletly

0 0
Oreo eOs 2 tb 7-20 OM O4e Oa 2 16820
(a) ECCS (b) DASt 013 _

[sa [072 1909] Sc

No.

[Sa [ost[=| 120)

(8ESSN t= 1120)
eps6 [os] 18,
0:39 [se]18|
se oat]8

0) 0
0) OA OMe Ie Ht AO S 0 04 08 12 16 20.
x
(c) ASME III N-284 (d) Drv %
Fig. 14. The elastic-plastic buckling tests of Walker er a/.*! on 0.32 mdia. welded
and stress-relieved steel shells. Results plotted in tems of S, and S, and according to
four different codes.
 Design Proposals for Shell Buckling Problems 203

28 rN

rte °& fo}

2o[t Stee 6
e@aa
Ole: ae ’ o 4
OS a. & a 8a
e, Ba ee a
1:2 eo a 9 4
io) a8

08 a
-
0:4
0 6

o—»>
DR04 20:8) T2eet he 087 eas OO OL OB VV25 7-682: oO “4 Se
(a) ECCS (b) DASt 013

lo] 1 | 3 [400[3434|
e[ 2} 2 400]1525]
e[ 3) 2 [200|2056,
rel_4 [+5 [800]1715,
Pa ffs,6| 1 |600]764] Sp
Fi) >(nze\|e2e)ges]2035) Opal
20 j=}910]
<2||500]1906] |
a
hor Onn
121° ca aa 4
%9. 4 4
0-8b oq, £
e 4
04 .
0 @®

O04 O85 12°16 920s.

(c) ASME III N-284 (d) Dnv
Fig. 15. The elastic buckling tests of Weingarten et al.?° on 0.20m dia. Mylar
cylindrical shells. Results plotted in terms of S, and S, and according to four
different codes.
204 G. D. Galletly

0
0 04 08 12 16 Sx
(b) DASt 013

reant[276[ra|73]
fa] IV [265] 117 [777 |
ralYe[225]128613 |
rolvs[r86[128[15|
re[wr [265]"9
ra vit2-06[735]25|
re[vit]2-66735]232]
rel[267]200]
tx 1351
WZ

0:8

04 StS yal
> ——s
OQ Oxs Oe WZ. ANS Sy 0. “O08 U8 12" 16 2-0 Sy
(c) ASME III N-284 (d) Dav

Fig. 16. The elastic buckling tests of Mungan*® on 0.45m dia. Plexiglass
cylindrical shells. Results plotted in terms of S, and S, and according to four
different codes.
 Design Proposals for Shell Buckling Problems 205

The elastic buckling results of the plastic models (Mylar and Plexiglass)
are also plotted in terms of the S,, S, coordinates in Figs. 15 and 16.
Considering all six series of tests listed in Table 1, the only Code (of the
four listed above) for which all the test points were above the line
S, +S, = 1 was DASt 013.°°°! The recommendations of ASME III Code
Case N-284 were the least conservative of the four Codes. However, this
lack of conservatism will be partly offset by the higher factors of safety
(1.67—2.0) used by the ASME III Code.
The ECCS recommendations give very similar results to BS 5500 for
external pressure alone and it is of interest to know how the ECCS
predictions fare for the combined loading cases. For most of these, the line
S,, + S,,=1 was below the test points. However, as may be seen from
Fig. 16, two of Mungan’s*® tests (under axial compression alone) were
below the line. Since Mungan used a small dynamic disturbance to initiate
buckling in these cases, some researchers believe that his results for axial
compression alone are too low.
It is also instructive to use the 1982 edition of BS 5500 for this interactive
buckling problem. For this purpose, section 3.6 was used with Enquiry
Case 5500/34 (although one is not supposed to do this). The end result
(allowing for the fact that BS 5500 deals with allowable values and not
lower-bound test values) was that all the test points were above the line
S,+S,=1. However, using BS 5500 in this matter gave the most
conservative predictions of the Codes investigated.
Insofar as the choice of a simple linear interactive design procedure is
concerned, one has several options. However, assuming one wishes to
utilise the BSI method for calculating the effects of external pressure, then
the choice reduces to:

(i) |Use of recommendations similar to those of the ECCS; however,

the reduction factors, «, would have to be modified slightly if one
wished to take account of the Mungan tests mentioned earlier.
(ii) Use of section 3.6 of BS 5500, in conjunction with a new section
dealing with the effects of axial compressive load. Enquiry Case
5500/34 was written with tall columns in mind and might be too
conservative for the present application.

If either of the above two options is chosen as a design procedure, one will
have to be careful about the use of safety factors. The ECCS, DASt, etc.
methods are concerned with predicted minimum test values and, therefore,
their predictions have to be divided by 1.5 (the safety factor) to obtain
206 G. D. Galletly

allowable values. The BS 5500 method deals with allowable values directly
and no additional safety factor (SF) is required. Thus one has
S, +S, = 1/SF
Le,
pity Se ae 1 (19)
Px/SF o,,/SF
The quantity p,/SF is the allowable external buckling pressure, which is
determined from Figure 3.6.3 in BS 5500. The quantity o, ,/SF is the
analogous quantity for axial compression—see the curve in Enquiry Case
5500/34.

12. FINAL COMMENTS

It is clear from Table 1 that the number of interactive buckling tests which
have been conducted to date on welded steel or aluminium shells of a
reasonable size is very small. Some more results will become available in the
near future when the series being conducted at present, under the auspices
of the American Bureau of Shipping and Conoco, is completed. Other
buckling results, obtained at Liverpool University on smaller models, will
be published later this year.°? This additional information on interactive
buckling will be very useful in evaluating any interaction equations which
may be proposed (the linear eqn(8) or any appropriate non-linear
equations).
Due to the lack of experimental data on welded models of a reasonable
size, any design equations proposed in the interim will have to err on the
conservative side. The simple linear interaction equation, used in
conjunction with the BS 5500 rules, seems to meet this requirement. This is
also the case when the linear equation is used with some other Codes.

REFERENCES

1. Kemper, M. J., ‘Buckling of thin dished ends under internal pressure’, Proc. of
the Symposium on Vessels under Buckling Conditions, I. Mech. E., London,
1972, pp. 23-32.
2. (a) STANLEY, P. and CAMPBELL, T. D., Very thin torispherical pressure vessel
ends under internal pressure: test procedure and typical results, Journal of
Strain Analysis, 16 (1981) 171-86.
 Design Proposals for Shell Buckling Problems 207

(b) Jbid., Strains, deformation and buckling behaviour, Journal of Strain

Analysis, 16 (1981) 187-203.
. Kirk, A. and GILL, S. S., The failure of torispherical ends of pressure vessels
due to instability and plastic deformation—an experimental investigation,
International Journal of Mechanical Sciences, 17 (1975) 525-44.
. PaTEL, P. R. and GILL, S. S., Experiments on the buckling under internal
pressure of thin torispherical ends of cylindrical pressure vessels, International
Journal of Mechanical Sciences, 20 (1978) 159-75.
. GALLETLY, G. D., ‘Internal pressure buckling of very thin torispherical shells—
a comparison of experiment and theory’, Proc. 3rd International Conference
on Structural Mechanics in Reactor Technology (SMIRT), London, 1975,
Paper G2/3.
. GALLETLY, G. D., ‘Some experimental results on the elastic-plastic buckling of
thin torispherical and ellipsoidal shells subjected to internal pressure’, Prelim.
Report of the 2nd International Colloquium on the Stability of Steel
Structures, Liége, 1977, pp. 619-26.
. BUSHNELL, D., BOSOR 5—program for buckling of elastic-plastic shells of
revolution including large deflections and creep, Computers and Structures, 6
(1976) 221-39.
. BUSHNELL, D. and GALLETLY, G. D., Stress and buckling of internally-
pressurised elastic-plastic torispherical vessel heads—comparisons of test and
theory, Journal of Pressure Vessel Technology, Trans. ASME, 99 (1977) 39-53.
. BUSHNELL, D., Non-symmetric buckling of internally-pressurised ellipsoidal
and torispherical elastic-plastic pressure vessel heads, Journal of Pressure
Vessel Technology, Trans. ASME, 99 (1977) 54-63.
. LAGAE, G. and BUSHNELL, D., Elastic-plastic buckling of internally-pressurized
torispherical vessel heads, Nuclear Engineering and Design, 48 (1978) 405-14.
. GALLETLY, G. D., Elastic and elastic-plastic buckling of internally-pressurized
2:1 ellipsoidal shells, Journal of Pressure Vessel Technology, Trans. ASME,
100 (1978) 335-43.
. GALLETLY, G. D. and RADHAMOHAN, S. K., Elastic-plastic buckling of
internally-pressurised thin torispherical shells, Journal of Pressure Vessel
Technology, Trans. ASME, 101 (1979) 216-25.
. GALLETLY, G. D., Buckling and collapse of thin internally-pressurised dished
ends, Proc. of the Institution of Civil Engineers, Part 2 (1979) 607-26.
. GALLETLY, G. D., Plastic buckling of torispherical and ellipsoidal shells
subjected to internal pressure, Proc. of the Institute of Mechanical Engineers,
195 (1981) 329-45.
. GALLETLY, G. D., ‘The buckling of fabricated torispherical shells under
internal pressure’, Buckling of Shells—Proc. of a State-of-the-Art Colloquium,
Stuttgart, E. Ramm (ed.), Berlin, Springer-Verlag, 1982, pp. 429-66.
. GALLETLY, G. D., A design procedure for preventing buckling in internally-
pressurised thin fabricated torispheres, Journal of Constructional Steel
Research, 2, 3 (September 1982) 11-21.
. GALLETLY, G. D., ‘Stress failure of large pressure vessels—recommendations
resulting from studies of the collapse of a 68ft. high x 45ft. dia. pressure
vessel’, Technical Report No. 45-57 Shell Development Corporation,
Emeryville, California, March 1957.
208 G. D. Galletly

18. GALLETLY, G. D., Torispherical shells—a caution to designers, Journal of

Engineering for Industry, Trans. ASME, 81 (1959) 51-66. Also published in
Pressure Vessel and Piping Design—Collected Papers 1927-1959, New York,
ASME, 1960.
118), Drucker, D. C. and SHIELD, R. T., Limit analysis of symmetrically loaded
shells of revolution, Journal of Applied Mechanics, Trans. ASME, 81 (1959)
61-8.
20. SHIELD, R. T. and Drucker, D. C., Design of thin-walled torispherical and
toriconical pressure-vessel heads, Journal of Applied Mechanics, Trans.
ASME, 83 (1961) 292-7.
Dy TIMOSHENKO, S., Theory of plates and shells, New York, McGraw-Hill,
1940.
2 GALLETLY, G. D., ‘Elastic-plastic buckling of dished ends subjected to internal
pressure’, Proc. 6th International Conference on Structural Mechanics in
Reactor Technology (SMIRT), Paris, 1981, Paper G4/6.
23) CODAP 80 (Code Frangais de Construction des Appareils a Pression), Revised
February 1982, Section C3.1.5., SNCT, AFIAP (10, Avenue Hoche, Paris).
24. RADHAMOHAN, S. K. and GALLETLY, G. D., Plastic collapse and the controlling
failure modes of internally-pressurised thin torispherical shells, Journal of
Pressure Vessel Technology, Trans. ASME, 101 (1979) 311-20.
25) FLUGGE, W., Stresses in shells, 2nd edn, New York, Springer Verlag, 1973,
pp. 449 et seq.
26. WEINGARTEN, V., MORGAN, E. and SEIDE, P., ‘Final report on development of
design criteria for elastic stability of thin shell structures’, Space Technology
Laboratories, Inc., STL/TR-60-0000-19425, 1960.
2k TENNYSON, R. C., BooTon, M. and CHAN, K. H., Buckling of short cylinders
under combined loadings, Journal of Applied Mechanics, Trans. ASME, 45
(September 1978) 574-8.
28. GALLETLY, G. D. and PEMsING, K., ‘Buckling of cylinders under combined
external pressure and axial compression’, Proc. International Union of
Theoretical and Applied Mechanics (IUTAM) Symposium on Collapse,
University College, London, 1982.
2). KoiTer, W., The effect of axisymmetric imperfections on the buckling of
cylindrical shells under axial compression, Proc. Koninklijke Nederlandse
Akademie van Wetenschappen, 66B (1963) 265-79.
30. SHANLEY, F.R., Strength of materials, New York, McGraw-Hill, 1957, pp. 633
et seq.
31. Hutcuinson, J., Buckling of imperfect cylindrical shells under axial
compression and external pressure, American Institute of Aeronautics and
Astronautics Journal, 3 (October 1965) 1968-70.
SPR, BootTon, M., ‘Buckling of imperfect anisotropic cylinders under combined
loading’, University of Toronto Institute for Aerospace Studies, Report 203,
1976.
33. PEDERSEN, P. T. and JENSEN, J. J., ‘On design criteria for elastic buckling of
shells’, Proc. International Symposium on Practical Design in Shipbuilding,
Tokyo, October 1977, pp. 87-93.
34. KoiTer, W., ‘On the stability of elastic equilibrium’, Ph.D. Thesis (in Dutch),
Technical University of Delft, Netherlands. English translation issued as
NASA TT-F-10, 1967.
 Design Proposals for Shell Buckling Problems 209

35). HARDING, J. E., Ring-stiffened cylinders under axial and external pressure
loading, Proc. Institution of Civil Engineers, London, Part 2, 71 (September
1981) 863-78.
36. Reis, A. J., WALKER, A. C. and Virtuoso, F. E., ‘Externally pressurized
cylindrical shells: buckling and collapse under combined loading’, Proc.
International Union of Theoretical and Applied Mechanics (IUTAM)
Symposium on Collapse, University College, London, 1982.
37k Lee, L. H.N., ‘Inelastic buckling of cylindrical shells under axial compression
and external pressure’, Proc. 4th US National Congress on Applied Mechanics,
ASME, 1962, pp. 989-98.
38. MunGan, I., Buckling stress states of cylindrical shells, Proc. American
Society of Civil Engineers, Journal of Structural Division, 100, No. ST11
(November 1974) 2289-306.
39) MILLER, C. D., Summary of buckling tests on fabricated steel cylindrical shells
in USA’, Paper No. 16, Buckling of shells in offshore structures, J. E.
Harding, P. J. Dowling and N. Agelidis (eds.), London, Granada, 1982.
40. Grove, T. and DipRIkSsEN, T., ‘Buckling experiments on 4 large ring-stiffened
cylindrical shells subjected to axial compression and lateral pressure’, DnV
Report No. 77-431, Oslo, Norway, 1977.
41. WALKER, A. C., SEGAL, Y. and MCCALL, S., ‘The buckling of thin-walled ring-
stiffened steel shells’, Proc. Colloquium on Buckling of Shells, University of
Stuttgart, May 1982.
42. KENDRICK, S. B., ‘Collapse of stiffened cylinders under external pressure’, Proc.
Conference on Vessels Under Buckling Conditions, I. Mech. E., London,
December 1972, Paper C190/72.
43. BrusH, D. O. and ALMROTH, B. O., Buckling of bars, plates and shells, New
York, McGraw-Hill, 1975.
44. GALLETLY, G. D., ‘Buckling of fabricated cylinders subjected to compressive
axial loads and/or external pressure—a comparison of several codes’,
University of Liverpool, Department of Mechanical Engineering, Report
A/059/82, March 1982. Also (minus the Appendix) in PVP-57 Pressure vessel
design, G. E. O. Widera (ed.), Bk. No. H00214, June 1982, New York, ASME.
45. ASME (American Society of Mechanical Engineers) Boiler and Pressure Vessel
Code Case N-284, ‘Metal containment shell buckling design methods’, Section
III, Div. 1, Class MC. Supplement No. 2 to Nuclear Code Case Book, 1980.
46. DASt (Deutscher Ausschuss fiir Stahlbau) Richtlinie 013, ‘Beulsicher-
heitsnachweise fiir Schalen’, Cologne, Germany, July 1980.
47. DnvV (Det norske Veritas), ‘Rules for the design, construction and inspection of
offshore structures’, Appendix C, Steel Structures, Oslo, Norway, 1977.
Reprinted 1979.
48. ECCS (European Convention for Constructional Steelwork), “European
recommendations for steel construction, Section 4.6, Buckling of Shells’,
London, The Construction Press, 1981.
49. ODLAND, J., Buckling resistance of unstiffened and stiffened circular cylindrical
shell structures, Norwegian Maritime Research, 6, No. 3 (1978) 2-22.
50. GALLETLY, G. D., ‘Towards a design procedure for the buckling of cylindrical
shells subjected to external pressure and axial compression’, presented to
British Standards Institution, Sub-Committee PVE-/5 (Design Methods),
October 1982 (Document 82/79229).
210 G. D. Galletly

51. GALLETLY, G. D., ‘Discussion of Paper 8558 by J.G. A. Croll (Proc.

Institution of Civil Engineers, September 1982). Proc. Institution of Civil
Engineers, Part 2, June 1983.
By GALLETLY, G. D. and PEMSING, K., ‘On design procedures for the buckling of
cylinders under combined axial compression and external pressure’, to be
presented at the 4th US National Congress of Pressure Vessels and Piping
Technology, Portland, Oregon, June 1983.
 12

Axisymmetric Collapse of Cylinders Including

Deformations of Ring Stiffening

J. G. A. CROLL
University College, London, UK

ABSTRACT

Axisymmetric elastic-plastic collapse estimates for cylinders, under

arbitrary combinations of axial and radial pressure loadings, are based
upon predictions offirst surface yield and first full section plasticity. Making
use of a perfect ‘base model’, the non-linear elastic behaviour is modelled in
the form of a simple column bifurcation. In this way arbitrary radial
pressures as well as Poisson bulging under axial load may be combined with
any prescribed geometric initial imperfections. Elastic-plastic collapse may
be summarised in terms of two imperfection parameters and the ratio of the
minimum elastic critical stress to material yield stress. Approximations of
the analytic solutions permit the axisymmetric elastic-plastic buckling to be
presented in sufficiently compact forms to recommend them for con-
sideration as the basis of design guidance. At the very least, the explicit
parametric delineation which emerges could permit numerically based
design guidance to be presented in more compact and universally applicable
forms.

NOTATION

Gy, 4,,... Coefficients used for Agy in eqn (13).

bo, 5;,... Coefficients used for Agp in eqn (15).
Ci, > Coefficients defined in eqns (18) and (22).
211
212 J. G. A. Croll

E Modulus of elasticity.
ve Function of Z used in eqn (19).
g Function of Z used in eqn (22).
a| Non-dimensional axial half-wave number (= (j/j,,)).
i Axial half-wave number.
k Coefficient used in eqn (19).
Kwe,kwe, Normalised equivalent load imperfections.
l Cylinder length or ring spacing.
M Moments for fundamental equilibrium.
m Incremental moments due to buckling displacements.
N Stress resultants for fundamental equilibrium.
n Incremental stress resultants due to buckling displacement.
Dp External pressure loading.
r Radius of cylinder.
t Thickness of cylinder wall.
W,, Wy, Total equivalent bending and membrane imperfections.
w Radial displacement—positive inwards.
x Axial coordinate.
Z Cylinder (Batdorf) parameter (=/*(1 — p?)!/?/rt).

B Ratio of nominal hoop stress to axial stress.

n Reduction factor due to ring displacement.
mM Poisson’s ratio.
6 Circumferential coordinate.
A Non-dimensional stress (= a/o,).
o Stress.

Superscripts
Ie Belonging to fundamental equilibrium.
e Relates to ‘equivalent loading imperfection’.
o Relates to initial geometric imperfection.
ae) Location for maximum stress or displacement.

Subscripts
cl Indicates classical critical stress or mode.
C; Critical stress associated with jth mode.
FY First surface yield state.
FP First full section plasticity state.
imp Indicates geometrically imperfect cylinder.
J Relates to jth mode.
 Axisymmetric Collapse of Cylinders 213

perf Indicates geometrically perfect cylinder.

ring Indicates geometrically imperfect cylinder having flexible ring
stiffeners.
x Relates to axial direction.
y Yield state.
0 Relates to circumferential direction.

1. INTRODUCTION

Tests show that cylinders under various combinations of axial and pressure
loading may collapse into one of two distinct deformation modes. For
short, thick cylinders the high bending stresses induced near boundary
constraints combine with the membrane stresses to produce a collapse
characterised by axisymmetric deformation modes. This form of collapse
has been the focus for much recent theoretical analysis.'~ 7 In contrast,
long thin shells are observed to collapse into modes having circum-
ferentially period, asymmetric, modes. Elastic-plastic collapse into both
these modes, shown in Fig. 1, displays an at times considerable degree of
sensitivity to the levels of their respective initial geometric imperfections.
Between these two limiting classes of stocky and slender cylinder there
exists a class of shell for which collapse could take the form of either an
axisymmetric or an asymmetric mode. Which will control buckling collapse
depends critically upon the precise shapes and relative amplitudes of the
axisymmetric and asymmetric initial geometric imperfections.

(a) (b)
Fig. 1. (a) Axisymmetric and (b) asymmetric elastic-plastic collapse modes for
cylinders.
214 JSGZASGroll

Existing design practice does not distinguish between these two quite
distinct forms of collapse. Consequently it fails to emphasise the importance
of controlling both axisymmetric and asymmetric imperfection forms. If
future design of the large cylindrical shell components being incorporated
into many forms of marine structure is to be put on a more rational footing,
it will be necessary to provide separate guidance on strengths against each
of the axisymmetric and asymmetric buckling collapse modes, and to relate
these to clearly delineated tolerance limits on imperfections in each of the
respective buckling modes. To do so will require analytical procedures for
estimating collapse that are reliable, but also sufficiently explicit and simple
that they can be effectively incorporated into design practice.
For periodic buckling such design guidance could conveniently be based
upon the recently described reduced stiffness lower-bound procedures.*~ !°
Analysis of axisymmetric collapse has been concentrated on sophisticated
finite element and finite difference computer codes which, for the numbers
of apparently independent parameters involved in cylinder collapse, are
not immediately suited to the presentation of compact and accessible
design guidance procedures. The following summarises an analytic
alternative which could provide a basis for the presentation of design
guidance against axisymmetric collapse of cylinders. At the very least it
could provide a parametric delineation which would permit numerically
based design guidance to be presented in more compact and universal
forms.

2. ELASTIC-PLASTIC COLLAPSE MODEL

The method adopted has recently been described in full,'! so only an

outline is provided herein.

2.1. Elastic Non-Linearities

As a ‘base model’ for the elastic axisymmetric non-linear behaviour, a
cylinder of radius r, thickness ¢, length /, and under axial compressive stress
o, is considered to be subjected to an external pressure p = pot/r. This
corresponds with a fundamental stress state

¢,=-9¢
(oy = —He (1)
which, because it involves no radial expansion, is independent of the nature
of end support conditions.
 Axisymmetric Collapse of Cylinders 215

Axisymmetric elastic buckling about this perfectly cylindrical base

model then takes the form of a classical bifurcation problem. For acylinder
having simply supported end conditions, the prospective critical modes
take the form
W=W, sin jax/1 (2)

Although the following concentrates on end boundaries with no moment

constraint, an extension to other classes of boundary constraint is
straightforward. As an example it will be shown how a ring stiffener, which
allows radial expansion at the end of the cylinder, may be handled through
a simple modification to the following analysis.
For a model having an initial geometric imperfection, w;, and an
associated loading imperfection, wj, in the deformation mode, w,, the total
incremental, that is stress producing, displacement could be represented by
the column equation

w—w® o wr+ws\.
( , ls » eae )sin (3)
izcilig Onna.

where the critical axial stress o,, associated with the jth mode is

a, [(t\? Cin)? i? 1
E -|(;) ay + (7) od oe
This spectrum ofcritical stresses reaches a minimum at the classical critical
stress

== BUS? (“) (5)

in an axial half-wave number

120. — pyr? |1/4

~ 0.592Z1/ (6)
i ec
where the Batdorf parameter Z =/?(1 — yv?)'/?/rt. It will assist subsequent
analysis if the critical stresses of eqn (4) are reformulated as
Cena
a3 (s+5) (7
cl

where J = (j/j,,)*. In the displacement expansion of eqn (3), wj would in the

present context result from an arbitrary external pressurep = Bot/r (B # y),
producing a radial expansion which is constrained by the end boundary
216 J. G. A. Croll

Fig. 2. Axisymmetric column buckling including mode coupling of eqn (3).

conditions. For the simply supported cylinder these pressure induced radial
expansions give rise to ‘equivalent load imperfection’ of the form"!
ws 2(B — p)
t (yJxBd—p*))"
The total response of eqn (3) is made up from the amplified model
responses as depicted in Fig. 2. Associated with this non-linear buckling
behaviour the elastic membrane and bending stress resultants would be
N,=NF +n,=-—ot
ss et ee be e
Ng= No +g = — pot — 7 (w; — w;) sinjrx/l
Siewaee

Et® jt 2 ei cctael
MS er eae » (=) (w;- w;) sin jax/l (9)

jaoney

Me Ma hee OS os(amy
pet »
jn\* oe:
(=) (w; — wj) sin jrx/l
psoas
This model of non-linear elastic behaviour has been shown to reproduce
that obtained from a full non-linear finite difference solution.’
 Axisymmetric Collapse of Cylinders 217

2.2. First Surface Yield

A first index of incipient collapse is the axial load at which surface yielding
commences (measured as o,y). Based upon the stress resultants of eqn (9)
the surface stresses depicted in Fig. 3(a) may be written

mek Weeeee Slee tps

Ag A)

(10)
No —A Ast
rang Wy mt
+ Wh,UW)

where all stresses are normalised with respect to the uniaxial yield stress,
o,, so that A=a/o,, A..=o4/o,, and the total equivalent bending and
membrane imperfections are defined as

W,=
“
» (Ag A) (weJ ;+e)
3 ah (v=
- A) 4 sin j7x/l

(11)
Wu = » EXC Vee eine (“2““1)sin jrx/l
Gay pe A)

All imperfections and displacements are considered positive when inward,

and in eqn (10) the upper sign refers to the outside face.
On the basis of a von Mises yield criterion
A? 2 ASA, PAP =1 (12)
these stresses will produce first surface yield at the lowest positive root of
the non-linear equation
k=4

> ie zA0 (13)

k=0

where the coefficients a,,k =0,1,... are

ag = — Ad
a, = +2A,

a,=+A2{( —pwt+y’) + (2- pW, t+ 2Qu—-1)(WytuW,)]

+ [Ws - Wy(Wy + uWy) + (Wy + UW)?]} - 1

a,=—A, [20 —p+p*)

+ 2—p)W, t+ 2u- 1)(Wy
+ HW)
a,=+(1—p+p’)
218 J. G. A. Croll

€ Va
a

Ox;
(a)

ie SS

y ee
Va ~

y / = =
Vie
£ me = /t-tn
fo) e fo) /,
Xe a 8i in 7
/ 7

Ox;

(b)
Fig. 3. Stress states used for (a) first surface yield, and (b) first full section
plasticity.
 Axisymmetric Collapse of Cylinders 219

It may be observed that apart from the two equivalent imperfection

parameters the only variable that affects the solution for Apy is A,,. With
eqn (13) representing the ‘exact’ solution for A,, it remains only to find the
location on the shells’ surfaces (that is, whether inside or outside and at
what value of x) corresponding with first yield. It has been found through
extensive experimentation that the critical axial location is at

5 ZESD8A9 AG a)
[
2= T2849 Ue)
(14)
21

and the critical surface depends upon the relative amplitudes and senses of
initial imperfection and external pressure loading.

2.3. First Full Plasticity

Between the shell reaching a first surface yield state and actually collapsing
a considerable redistribution of stress will occur, both through the shell
thickness and over its surface. However, when axial load is significant a
collapse mechanism could be anticipated as soon as one of the shells’ cross-
sections becomes fully plastic. In this situation, of a plastic hinge extending
around the circumference, any increase in displacement would from simple
static considerations inevitably increase the moment which must be carried
at the hinge; in turn, this will require a decrease in the average axial stress
possible at the plastic hinge. The result will be a drop-off in load carrying
capacity for deformations beyond that corresponding with first full
plasticity.
As an approximation of this first fully plastic section, the statically
admissible elastic stress resultants, eqn (9), are employed. On the basis of
the fully plastic stress block illustrated in Fig. 3(b), it has been shown’? that
the axial stress, o,;p, corresponding with first full plasticity is given by the
lowest root of the non-linear equation
by + Wa(b7 — 4b56,)'/?
=0 (15)
where
bp =(1— p+ pw)Agp — [201 — w+ 7) + Qu - Wy lAgAge
+ {AQIC — wt w7) + (2 - I)Wy + Wal — VAFe
+ 2A App — AG
—2(1—ptp?)AgAgp + (20 — wt 7) + (2u — 1) Wy JAZ App
SS
Ny—
ll
II +(L—ptp7)AZAfp
220 J. G. A. Croll

Again, solution effectively depends upon just the three parameters A,,, W,
and Wy.
It may be noticed that the equivalent imperfections (W,, W,,) depend
upon A,p. This means that the coefficients a, and b, in eqns (13) and (15)
respectively are themselves implicit functions of A,p. Solution is easily
undertaken as an iterative process, with just a few cycles required to
determine A, and A,;p. The whole process can be easily adapted to
microcomputer applications.
To illustrate the nature of the solutions from eqns (13) and (15), Fig. 4
summarises the behaviour for a cylinder having A,, =¢,./a, =4.0 in the
form of a two-dimensional imperfection sensitivity plot. In all cases shown
the surface or section corresponding with the lowest solution for eqns (13)
and (15) is given at the location indicated in eqn (14). Accompanied by
suitable charts or figures for specifying (W,, W,,), imperfection sensitivity

first full plasticity, App

—— — first surface yield, Ary

-1:0 -0°5 0 +05 +10

Wm
Fig. 4. Typical two-dimensional imperfection sensitivity plots for shells having
6.,/0, = 4.0.
 Axisymmetric Collapse of Cylinders 221

plots like those shown in Fig. 4 could be used for design purposes. Before
this can be achieved consideration needs to be given to the appropriate
forms of initial geometric imperfection.

3. INFLUENCE OF FABRICATION IMPERFECTION

Making use of the general forms given for the total bending and membrane
imperfection parameters (W,, W,,), eqn (11), it is possible to consider any
arbitrary form of initial geometric imperfection. Imperfections in the form
of a single axial harmonic j have often been considered. Of all the
wavelengths for imperfections it is that associated with the minimum
critical mode that produces the most adverse reductions in collapse
load.** This form of imperfection is also particularly convenient in the
present analytic formulation,’* and it would through appropriate choice of
imperfection amplitudes be possible to build-up a code specification based
upon the ‘worst shape’.
But of considerably greater direct practical relevance is an imperfection
induced by typical fabrication processes. As illustrated in Fig. 5 these
would arise from any lack-of-fit or shrinkage associated with the welding at

——S SSS,
CEE EN Ee
Ig =2l ring = lack
icone stiffener of fit

ring a
weld P,
l l
Sores | = = —) oo 2
| | ee SESH in
edge fg Wimax edge" W max
shear shear “
|
(a) (b)
Fig. 5. Localised distortions of shell wall caused by (a) weld shrinkage and
(b) the lack of fit of shell plating and ring stiffener, and modelled using fabrication
imperfection given by eqns (16)-(20).
222 J. G. A. Croll

circumferential seams or at connections between shell plating and a ring

stiffener or diaphragm. These would generally have a shape closely related
to that caused by a pressure loading—or equivalently by an end bending
shear. This means that modal contributions should have the relative
distribution
Ww;° Cy>
ee Te
(16)

It is a simple matter to show that for such a fabrication related distribution

of initial imperfection the maximum amplitude occurs at x =/, where

; ZiBS 6988 IE RLD)

pes}? (17)
ly hE tamedWA
Wet

Notice that for Z > 5.698, and on account of eqn (6), the gauge length
1, = 21 =~ 3.46(rt)'/? over which axial imperfections are likely to have their
maximum deviation is close to the 4(rt)/* recommended in many current
codes.'*~'* Choosing the maximum amplitude over the gauge length
Lael to be the prescribed geometric tolerance w,/t, allows c, to be
calibrated as
oO
WwW
t
= (18)
sin jri/l
w.. i 3 ~

To reflect some consistent fabrication process it has been shown that

° 1/2

_ ae =kf(Z) (") (19)

where a reasonable approximation of the weighting function f(Z) (shown in
Fig. 6) to allow for length effects is

iL) oT 2 (20)
ee Z
while the value of k would depend upon the precise nature of the fabrication
process. Design codes usually employ this form of tolerance specification,
 Axisymmetric Collapse of Cylinders 223

1:0
f,g 1
f(Z)=——— (Gn, (19))
e Fewoa
Sis ereeg)
7 1
06 giz) = ; cats (eqn (22))

Vi Neljseok ed

0-4

0:2

10 20 30 40
Zz
Fig. 6. Geometric parameters used in eqns (19) and (22) to define fabrication
imperfections.

with DnV adopting k =1/30'* and DASt Ri 13’? and ECCS'* using
k = 1/25. Weighting function f(Z) is taken to be unity in DnV,'? while
ECCS!* recommend f=0.25Z'/? and DASt!?
f= 500Z"/?/1.
Based upon the fabrication imperfection of eqns (16)—(20), and k = 1/30,
Fig. 7 shows typical interaction collapse loads based upon the solution of
eqns (13) and (15).

4. APPROXIMATE FABRICATION IMPERFECTIONS

A particularly convenient choice for the shape of fabrication imperfections

is to assume that the correct distribution of eqn (16) is approximated as
° >
Wipe
oath 2)
For slender cylinders, such a distribution puts a greater weight on the long
axial wavelengths ofj= 1, 3 than does eqn (16). However, its advantage, as
will be shown subsequently, is that it allows a direct combination with the
 Qa
224

470m wou ube (9,)

———
f +/3a woy uba (LZ) te
J/) L-0= wee
XV
n=g
0-=Xpo
3/40
= 5100-0 K
4/0= O€/L (4/4)
3, (2) Z-4 = jy-= og-
(e) J

wor
J. G. A. Croll

Ott Once

oi
kd_ ko
‘SIq°z uosueduiog
jo joexo pur ajewixoidde uoneoliqey ‘suonooysodurt
Burmoys
isry jjny Ayonseyd
se e ainsvour
Jo
OLJOWIUWIASIXe
asdeyjoo
Joy jerxe pue ainssaidpopeoy Ajduuts poysoddns ‘ssapurAo
stays Sutaey(e) 4/7= [0 pue (Q) 4/)= ‘S70
 Axisymmetric Collapse of Cylinders 225

‘pluor—L
*31q
diy

(Z)d

(94)
(12) ube wouy 47),

0E/L= (4/4)
ey

uba
wosy
SL00-0= 3/40
$20 =4/1

4/3m
4/0™

[
226 J. G. A. Croll

equivalent load imperfections of eqn (8). By choosing c, so that the total

imperfection again matches the required tolerance of w°/t at x =/, that is
w°?

ae w°
Cc, = sinjni/[]=—— t 9(7) (22)

Er
Hwee gece J

the imperfection profiles based upon eqn (21) differ very little from those
given by eqn (16) for the range of stocky cylinders (those for which Z < 50)
for which the present axisymmetric mode is the most likely form of
collapse. Not surprisingly, little difference can be observed in the first yield,
Ary, or first full plasticity, App, loads based upon these two alternative
fabrication imperfections for the range of shells shown in Fig. 7. If this
form of approximate imperfection was to be used for more slender shells it
would be necessary to recalibrate the tolerance specifications to ensure that
the results are representative of those obtained from the more realistic
fabrication imperfections of eqn (16).
The real advantage of the approximate fabrication imperfection of
eqn (21) over either that of eqn (16) or any of the alternative imperfection
forms is that it may be directly combined with the equivalent load
imperfections. This means that the total equivalent bending and membrane
imperfections of eqn (17) may be rewritten

W, =(B—B)kwe
(23)
Nia B)kwe,

where the normalised equivalent load portion of these imperfections is

given as
Ws = (B- Wwe
Ww = (B- Wkws, (24)
while from eqns (8) and (11)

hye y of a Cia eae sin jrx/I

RS) CN)
Sapa
(25)
k ao DIAS cl Ay ole .

ce JS
> t ea i
la owase
 43a4uad}J8ys XV x5
vt i
JIajsedwi
}9yS
asde}]}03
sn30)

ajewixoudde
ubisap ‘sn30)ube (€€) iy i
J/) S2-0=
4/4= O0L
3/Ao
= S$L00-0
/oM
4= (4/J)OE/L
2 (Z)3
Axisymmetric Collapse of Cylinders

31°§ Suiddeyy
jo yoajiod UONORIOIUTWIRISeIp
0} Als JOoJJOdur! WOHOeIN}U! ‘sueISeIp
BuIsnoy) uonewoj
jo ubasuen
(LZ)
227
228 J. G. A. Croll

In eqn (23) the value B, given by

eles (26)
where f’ = c,x[3(1 — n”)]'/?/2 represents for the geometrically imperfect
system the pressure loading required for the geometric imperfections to be
exactly nullified by the equivalent load imperfections—in much the same
way that the load eccentricity may nullify the out-of-straightness in an
axially loaded column. The constant c, required for the specification of f’ is
obtained from eqn (22) or directly from Fig. 6.
As a result of the above mapping the load interaction diagrams for an
imperfect cylinder (f’ 4 0) may be obtained directly from that of the perfect
cylinder (f’ = 0). All that this entails, as Fig. 8 illustrates, is shearing of the
perfect interaction diagram such that at the appropriate value, f;,,, for the
loaded but imperfect cylinder the collapse load would be the same as that
for the perfect cylinder with £,..,, where

Boert = Bim ay B (27)

Using this simple transformation of the perfect system allows considerable

computational economies, and means that once the perfect behaviour is
known it is possible to construct exact interactive curves for any desired
initial imperfection levels of either outward or inward geometric
imperfections.

5. EFFECT OF BOUNDARY CONSTRAINT

For more practical boundary constraints it is often reasoned that the

simple support condition provides a good approximation of the expected
collapse behaviour. Early flexural yielding at a fully clamped support
would result in the end condition at collapse being more closely related to
the simple support than the clamp. An exception, however, is when a ring
support is sufficiently flexible that it allows significant radial deformation to
occur at the end of the cylindrical subshell. This situation has been recently
considered using finite difference and finite element modelling.* Another
advantage of the tolerance specification in the form of eqn (21) is the ease
with which such radial deformations at the ends of subshells can be
incorporated.
Figure 9 indicates how for a given pressure the deformation of the
 Axisymmetric Collapse of Cylinders 229

p
aeSo aod oakVicia |
is Ws ee =)

ie ‘lr W=Wr

Pa a4 ie —_ ate

f, :
half ring force te
(qs 2/2 kp Wr) at} = aeons

shear
of
ring — Qs = Ks We = 1/2 kp We

Ws =TWs Ws = Ws
Fig. 9. Interaction between subshell and ring, allowing shell displacements for
flexible rings to be expressed in terms of those for infinitely stiff rings.

cylinder relative to a flexible ring, w¢, is related to the shell displacement for
a radially rigid ring, w,, through
w= nw, (28)
where yn = k,/(k, + 2k,). The ring and shell stiffness coefficients (k,, k,) may
be easily computed in terms of the ring parameter (A,//t) and the Batdorf
parameter Z, where A, is the cross-sectional area of the ring. Having done
so the total equivalent imperfections now become

W, = (n(B— 2) — B’lkws
(29)
Wy = [n(B — 1) — B’Kws,
and the interaction diagram for the imperfect ring supported cylinder once
again may be obtained from a straightforward transformation of the
230 J. G. A. Croll

perfect simply supported case. In the present context the collapse load of
the perfect system at .,,.,, will relate to the imperfect flexible ring, loaded in
such a way that the actual load ratio is f,,,,,ring? when

Does oe NB ing e B ne ud ss n) (30)

Otherwise the treatment of the flexible ring follows exactly the same pattern
as the approximate fabrication imperfection of the previous section.
For subshells that occur near end diaphragms a closely related technique
to the above could be developed. However, for these end bays, design based
upon rings having infinite radial stiffness should be conservative.

6. A DESIGN APPROACH

Even though the approximate fabrication imperfection shapes of eqn (21)

allow considerable simplifications in the construction of axial/pressure
load interaction curves for axisymmetric collapse, the procedure is
probably still too complicated for incorporation into design practice. A
further simplification may be achieved when it is noted that each of the
branches for the interaction curves shown in Figs. 7 and 8 have only mild
convexity. To approximate each of these curves using a suitable lower-
bound straight-line would accordingly be only moderately conservative
for design. Combined with the procedures outlined in the previous
two sections, such an approximation provides an attractive practical
proposition.
For a geometrically perfect cylinder the first yield and first full plasticity
states at 6 = yw are given by the plain strain squash condition

On = Opylpap = Trelp=, = O91 — p+) 7 (31)

which for 1 =0.3 means that o, = 1.1250,. Point A in Fig. 8 is therefore
determined by the Poisson’s ratio of the shell.
For a cylinder under external radial pressure, so that 8 = + 00, the hoop
compression, og, to cause first yield and that, dy,» to cause first full
plasticity are independent of initial geometric imperfection. Furthermore,
by taking Boo in eqns (13) and (15) it can be shown that (p_,> %,,)
depend only on the composite shell parameter Z. Figure 10 shows the
nominal hoop stresses to cause first yield and first full section plasticity for
shells of varying Z. Even though first full plasticity may sometimes
underestimate the actual collapse of cylinders under radial pressure, it
would seem prudent to take such a severe state of material failure as a
 Axisymmetric Collapse of Cylinders 231

“boiler code:
1-0 a a a es el Eicien ee
Pp/py
/0g
Oy
=

first full plasticity, Agp

09 — —- first surface yield, Ary
(oy/E =0-0015, =03)

10 20 30 40 50 60
Zi
Fig. 10. Radial pressure loading (f — 00) to induce first surface yield and first full
plasticity. Used to locate points B and C in Fig. 8.

limiting design state. Notice that for internal pressure the equivalent hoop
stresses will be equal but opposite to those under external pressure. With
the help of Fig. 10 the points B and C may be located in Fig. 8.
For an arbitrary inward, w;, and outward, w?, imperfection range,
appropriate values of f’ to locate the points A; and A, may be calculated
using the function g (shown in Fig. 6) to give

Bo=EU+¢3 -= g(Z)
(32)
R= U>cs— 8(Z)
where c, =2[3(1 —w)]'/7/2 ~2.592. Having determined the failure
pressure, Py = — Pc, from Fig. 10, design could then be based upon the
linear interaction formulas
A+c,P =c; (33)
7upyo J. G. A. Croll

where A =o/o, and P=p/p,, and

pera for A,B and A.B

1== (Ps zy. BA,)

pa eee for A,C and A,C

Ue ie pha)
peal ober for A,B and A,B
ee (Pp, - BA,)

pane fone cant e

5 = —

(= Fe Pig)
In the linear interactions of eqn (33) the values of 6’ appropriate for inward
and outward initial imperfections would be chosen from eqn (32) and used
to produce the two sets of interaction curves shown in Fig. 8. For design it
would be the lesser of these two solutions that should apply.

7. CONCLUSIONS

A new and explicit treatment of the axisymmetric elastic-plastic collapse of

cylinders is shown to provide a convenient means for summarising design
information. By choosing an approximation of the likely fabrication
imperfection shapes it is shown how they may be directly superimposed
upon those arising from boundary induced ‘equivalent load imperfections’.
A simplification of the resulting buckling loads allows the axisymmetric
collapse of any cylinder geometry, material, loading, boundary support
condition, and initial imperfection to be represented by a linear interaction
equation, which in turn is uniquely determined in terms of just one
composite shell geometry (Batdorf) parameter.
When future investigations of fabrication distortions allow more
complete calibration of likely initial imperfection shapes and sizes, the
present approach could provide a convenient, safe and reliable basis for
design.

REFERENCES

1. HARDING, J., The elasto-plastic analysis of imperfect cylinders, Proc. Instn.

Civ. Engrs., Part 2, 65 (1978) 875-92.
2. DowLING, P. J. and HARDING, J. E., ‘Current research into the strength of
cylindrical shells used in steel jacket construction’, Proc. 2nd Int. BOSS
Conference, 1979, Vol. 2, pp. 327-40.
 Axisymmetric Collapse of Cylinders 233

. SRIDHARAN, S., WALKER, A. C. and ANDRONICOU, A., Local plastic collapse of

ring stiffened cylinders, Proc. Instn. Civ. Engrs., Part 2, 71 (1981), 341-67.
. HARDING, J. E., Ring-stiffened cylinders under axial and external pressure
loading, Proc. Instn. Civ. Engrs., Part 2, 71 (1981) 863-78.
. HARDING, J. E. and EsTeFEn, S. F., Ring stiffened behaviour and its interaction
with cylindrical panel buckling, submitted for publication.
. ANDRONICOU, A. and WALKER, A. C., Buckling of ring stiffened shells subject
to pressure and axial loading, to be submitted for publication.
. Exxinas, C. P. and CRro_t, J. G. A., ‘Design loads for elastic-plastic buckling of
cylinders under combined axial and pressure loading’, Proc. BOSS ’82
Conference, MIT, Boston, August 1982.
. CROLL, J. G. A., Lower bound elasto-plastic buckling of cylinders, Proc. Instn.
Civ. Engrs., Part 2, 71 (1981) 235-61.
. CROLL, J. G. A., Elasto-plastic buckling of pressure and axial loaded cylinders,
Proc. Instn. Civ. Engrs., Part 2, 73 (1982) 633-52.
. Exvxinas, C. P. and Cro it, J. G. A., Elastic-plastic general buckling of ring
stiffened cylinders, IUTAM Symposium on Collapse: The Buckling of
Structures in Theory and Practice, September 1982, University College,
London (to be published by Cambridge University Press, 1983).
. CROLL, J. G. A., Axisymmetric elastic-plastic buckling of axial and pressure
loaded cylinders, submitted for publication.
. Det Norske Veritas (DnV), ‘Rules for the design, construction and inspection
of offshore structures—A ppendix C—Steel structures, Norway, 1979.
. Deutscher Ausschuss fur Stahlbau (DASt), ‘Beulsicherheitsnachweise fur
Schalen’, DASt Richlinie 013, Cologne, Germany, 1980.
. European Convention for Constructional Steelwork (ECCS), European
Recommendations for steel construction—Section 4.6—Buckling of shells,
London, The Construction Press, 1981.
 aos oe
1Pal-
- = eeemeiae
7 7 = eee
scan ed et ove

one a
oun &
 13

Local Loads on Cylindrical Vessels: A Fourier

Series Solution

G. DUTHIE
British Gas, Midlands Research Station, Solihull, UK

and

A. §. TOOTH
University of Strathclyde, Glasgow, UK

ABSTRACT

The theoretical background is presented and a Fourier expansion series

solution detailed to enable displacement and stresses to be obtained in the
region of local loads on cylindrical vessels. To facilitate the analysis of a
general load system, in the radial, tangential and longitudinal directions, a
compendium of ‘loading term solutions is presented. Four comparisons are
made with existing solutions available in the literature to provide a degree of
confidence to the user.

NOTATION

Radius of vessel.
Distance from end of vessel to centre of loading patch.
Half-width of the loaded area in x direction.
Extensional rigidity = Et/(1 — v?).
lave) Differentials with respect to x/a and ¢ respectively.
Modulus of elasticity.
Flexural rigidity = Er°/12(1 — v’).
TrmRHOTSTS
Non-dimensional factor = ¢?7/12a’.
235
236 G. Duthie and A. S. Tooth

l Length of cylindrical vessel.

M 4x Myo Stress resultants (see Fig. 1 and ref. 20).
m,n Harmonic numbers associated with x and ¢ directions.
Aewe i. Stress resultants (see Fig. 1 and ref. 20).

Dike Loading terms employed in Fourier analysis.

Do Das Dy External applied loading in x, ¢ and radial directions.
ONS Oy: Stress resultants (see Fig. 1 and ref. 20).
q Surcharge of internal pressure.
t Thickness of vessel wall.
u,v, W Mid-surface vessel displacements in x, @ and radial
directions—Fig. 1.
x, @ and z Coordinates in the axial, circumferential and radial
direction in Fig. 1.
Ziel Z coefficients employed in Fourier analysis.

of Angle from lowest generator to surface of fluid.

Bb, Half-angle of loading patch.
A mna/I.
v Poisson’s ratio.
p Specific weight of contained fluid.
oe Specific weight of the vessel material.
Oy, Ox Circumferential and longitudinal stresses.
dQ; Angle from lowest generator to centre of loading patch.

1. INTRODUCTION

Fourier expansion methods have been available for many years for the
derivation of displacements and stresses which occur in shells and pressure
vessels. They have been particularly useful for solving local loading
problems and a variety of solutions exist for vessels subjected to point, line
and distributed loads. Those relevant to cylindrical vessels are presented in
refs. 1-12. In general the solutions for the cylindrical vessels are based on
simplified shell equations of the form proposed by Donnell!? and,
because of the limited computer facilities available when the original work
was undertaken, have often only been evaluated to a limited degree of
accuracy. Nevertheless, a variety of results generated by these solutions
have been incorporated into National Pressure Vessel codes, for example in
 Local Loads on Cylindrical Vessels 237),

BS 5500,'* and when employed along with the appropriate design code
stress limitations have proved extremely useful and reliable.
{t has, therefore, become normal practice for designers to employ the
most appropriate results presented in the codes and the exception rather
than the rule for designers to attempt to develop rigorous analyses for
problems under consideration. This is not surprising when one considers
the complexity of Fourier expansion methods of solution, the timescales
and computer resources required, with the success of current code design
methods.
However, advances in micro-, mini- and mainframe computer facilities
have made direct application of Fourier expansion local loading computer
models attractive. Indeed such models need no longer be considered merely
as a research approach suitable for use by only a few. Although these
Fourier expansion solutions involve a significant amount of computational
effort they do not consume a great deal of computer core store and could
therefore be mounted and run on relatively small computer facilities.
The authors have used the method extensively in the derivation of the
stresses and displacements which arise when a cylindrical vessel is
supported on saddles.!°~ ‘8 In the process of developing the above analysis
a wide range of particular local load cases were examined to test the validity
of the approach. These cases are considered to have a general interest in
design and for that reason they are presented in this paper. Attention is
confined to the results which arise from one particular series solution for
circular cylindrical vessels subjected to mechanical loads in the radial,
tangential and longitudinal directions. These series are particularly useful
when the loads are applied remote from the vessel ends in such a manner
that the flexibility of the vessel ends plays no significant role in the
behaviour under consideration. A similar approach may be used for other
vessel end conditions using different series. Thermal stress problems may
also be tackled using a similar technique.’?

2. THEORETICAL ANALYSIS

2.1. Governing Differential Equations

Attention is confined to thin-walled circular cylindrical vessels operating in
the linear elastic regime. The behaviour of such vessels can be accurately
described using the differential equations proposed by Fliigge.”°
Rearranging these equations and representing the differential operators
238 G. Duthie and A. S. Tooth

0/0(x/a) and 0/0¢ by d and 6 respectively gives

(1 —v) 3 = YW
frat OS) aor — hah +{o- k ——
5 aro}

+ {l+k+k(d* + 2d76* + 64 +26*)}w =p,a?/D

eae a5} w= ~pya*/D (1)

The positive directions of the mid-surface displacements uw, v and w and the
loading components Px Po and p, indirections x, and zare given in Fig. 1.
An intermediate step in the derivation of the above equations generates
the following expressions for the stress resultants in terms of mid-surface
displacements:

N,= > [50 +w + vd] + [w + 6?w]

N, == [du + vd0 + vw] —5 [dw ]

ad gat eeegroa
coe =e aE vie a v) ;

K
M, ma [w+ 6?w + vd?w]

M,. =~LS (|d°w

osre + vow2 — du — vov]
a
K
Me. ae: (1 — v)[ddw + 46u —4dv]
 Local Loads on Cylindrical Vessels 239

Height of fill

Fig. 1. Positive directions of mid-surface displacements, stress resultants and

loading components.

K
OS) (1 — v)[dow
— dv]

OssK [(1
— v)(d* z ow
— d*v)
2
+ dw + 6°w
3
+ vd’ Disksdw]

O,.=—K [Cl — v)(dd°w

2y 4452 1
+ 36°u — sddv) + 5 52
d°w + vd*dw 2
— d*u °
— vddv]
a

(2)
Because of approximations made during their derivation, eqns (1) are
linear and can therefore be manipulated and simplified to give the following
partial differential equations relating mid-surface displacements to loading
components:
4
V
{DEXP}w =~ p,a?/D

{DEXP}v=— [2kd*65 + 2kd?53 — (2 + v)d?6 — 53] p,a?/D

|)
1—
{DEXP}u = a — vd?— 3k? \~_—
5 ) d?5? +dd* — ka p,a’/D
>|
1—
{DEXP}w= -7 jas vd? — 3k? |—_—
5 uz d*6? +.d65?— ka p,a’/D
240 G. Duthie and A. S. Tooth

my! D pay Bosal’ 5

wexPyo=;(7=]]e( 5 ja5+k}( 5 )vaif avs

(1 +) (1 —v)
2 doe REL Wid Oe Ta do

143
can = 2 as?map

{DEXP}u = eel] a
7 [ (i) d® + k(2 = v)d*d? +k(1 + 5?)?6?
1 ==
Gye epmy
ep omme oem
: ”)a4254 a | p,a?/D
1
(DEXP}w = —> [2kd*6 — (2 + v)d?6 + 2kd?6° — 5°] p4a7/D

{DEXP}v = = (=) Seo ee

Boy» d*6* + (1 — v*)d?

+k(2 —v)d?254 +k(2 —v + v2)d25? + 2vkd4

Sm _ ks + (19k |pya?/D
2
I 2 I
(DEXP}u=7 (— y. SS
AL eee ence:
rig Saeed, 3 + crv) do
5 do
Z

ba db tk oS” alr a?/D (3)

where:

{DEXP} = iv"a Ovid? 22(2 =v )d-o" 4224 =3)d70*

ie 2
cea) a + 609? + 3¢425°|
v8 a (d? ae Bly

VS Fe)?
 Local Loads on Cylindrical Vessels 241

2.2. Fourier Expansion Solution

A particular solution of eqn (3) can be obtained by expressing the loading
components and mid-surface displacements in the double Fourier series
form shown below:

Py, = Do Pn.m€OS Ng Cos (Ax/a)

n=0 m=0

Py= » » PnmSin ng sin (Ax/a)

n=1m=1

P= »,» Pn.mCOS N@ sin (Ax/a)

n=0 m=1

“= > » Un.mCOS N@ cos (Ax/a)

n=0 m=0

d= » . Vn.m Sin ng sin (Ax/a)

n=1 m=1

wie »:» Wa.m COS NO sin (Ax/a) (4)

n=0 m=1

The choice of the above expansion implies that the loading system is
symmetric with respect to the generator passing through ¢ = 0. They also
imply that certain boundary conditions, discussed in detail in 3.3, must
exist at the ends of the vessel. Other expansions can be used to represent
loading conditions which are not of the symmetric form described above.
Substituting these expressions into eqns (2) and (3) leads to the following
solution for the mid-surface displacements and stress resultants:
5 ie0) io6)

w a5 » > Z 4mPnimCOSN® sin (Ax/a)

n=0 m=1

zy
u= > » » Z 21m PnmCOS NG COS (AX/a)
n=0 m=0
242 G. Duthie and A. S. Tooth

2 = =

nimPn.m Sin ng sin (Ax/a)

aza » \ Z3
n=1 m=1

a?))Z4,, m Pn mCOSN® sin (Ax/a)

n=0 m=1

a :
N, =F Z5
1mPn,m COS Ng Sin (Ax/a)
n=0 m=1

ie ae |
a my » » Z6,, mPn,m Sin ng cos (Ax/a)
n=1 m=0

— =
rm)syZ6A 4 mPn.m Sinn cos (Ax/a)
n=1 m=0

M, =a Z7nmPnmCOSN@ sin (Ax/a)

n=O m=1

M, =a’ DsZ8.,,, mPnmCOSn¢ sin (Ax/a)

n= 08m =I

M,2=a@7(1—) »)»:Z9
1mPnm Sinnd cos(Ax/a)
n=1 m=0

M,,=4 2(1 —v) s Z 10), m Pnm Sin ng cos (Ax/a)

n=1 m=0

= «@

O,=a » »;Z 11), mPn.m Sin n@ sin (Ax/a)

n=1 m=1

= y » Z 12), mPnim COS NG COs (Ax/a) (5)

n=0 m=0
 Local Loads on Cylindrical Vessels 243
The Z coefficients used in the above expressions depend on the
particular
loading component considered. They are defined as follows.

For radial loading

ZA, = (4 4 97)7/DEN

L2n m= aa + vA} — 3k? a Aen? — An? ~ kin)[DEN

Z3,,0m = [— 2kA4n — 2k1?n3 — (2 + v)A2n — n>]/DEN

For tangential loading

Z nm = [—2kA%n — 2k1?n? — (2 + v)d?n — n3]/DEN

LOSER -_— E a (A°n + 2A°n3 + An’) —4(2 + 3v— v?)A3n
—y

—4(1 + 3v)An3 + — in|/DEN

= ee Oo in? +2) 32 4 2 —nt

= ee)

==:

—(2—v+v7)A?n? — 2yvA4 + Cay

2k

wee n° —(1 -»ynt| DEN

For longitudinal loading

Za =|RaSh ie A3n? —An? kin* |DEN

Tig gs NOI ecaag yee yy ty GR Ye mar
Dare (PET) |ErS
(5 —v) (1 — v)
Clg
SA tene
Py t MAP
tt eesgeceras ee DEN

2k [(l+v),, (3 —v) be Civ

eee
Vie = eal 5 An ae + 1) Aen +—_—
5 _ Jn

|— 1+3
egg anyiicnes = ale ; y in|DEN
244 G. Duthie and A. S. Tooth

In each of the equations

DEN = (A? +. n?)* — 2vd® + 2(2 — v)A?n? — 2(4 — v)A?n*

+
Cis f= 64° Fa one
k
For each loading component considered the Z4,,,, to Z12,,,, coefficients
can be calculated from the corresponding Z1,,,, Z2,,, and Z3,
coefficients using the relationships

Z4im = UL +k —217)|Z 14m — VAZ29m +L3nm

1, mn— AZ 2pm + YNZ
Z5nm = (V+ KA7]Z 3,
ZGjgp = RANZA aL 2) ALS,
LGA, = KANLA i RL eae lO,

Zee (lane A721,

TA oe Ohsane SVAN Wa VS
n A
LY nm = —NhZ
n l=
nym D) Lda
nym p) LS n,m

70 an Ze
ZAl nm = —N( =n? —2?)Z 1am + (1 —v)A7Z39on
N Ss ] =A? +n) Zl E =;a | Zo. ni a ;” Zam

Referring back to eqn (5) it is noted that the only unknowns are the loading
terms p,,,,.The method of deriving these is considered at a later stage of this
path.

2.3. Boundary Conditions

As indicated above the Fourier expansion series was selected to facilitate
the representation of a cylindrical vessel loaded by forces which are
symmetrical about the vertical diameter of the vessel. These forms, given in
eqns (4), imply certain boundary conditions. Since the origin of the
coordinate system is taken at one end of the cylinder all the Fourier
expansions, or their derivatives containing the term sin (Ax/a), vanish at the
ends of the cylinder. This implies that:

(a) The ends cannot deform in the plane of their profile.

 Local Loads on Cylindrical Vessels 245

(b) No rigid body displacement or rigid body rotation of the ends can
occur.
(c) The ends cannot carry applied axial loading.
(d) Generators are free to rotate in a plane normal to the profile.

If the vessel ends conform to the above boundary conditions then the
Fourier expansion of eqns (5) are a complete solution to the problem. No
complementary solution need be added to the particular solution since
eqns (5) satisfy both the governing differential eqns (1) and the boundary
conditions of the problem. However, in practice some deviation from these
conditions is likely to occur; for example, in the case of a vessel with a
torispherical end closure condition (a) is not ideally satisfied. In such cases
it is still possible to use the results of eqns (5) with confidence if the local
loading is applied some distance from the vessel ends.

2.4. Fourier Series Representation of the Applied Loads

The only unknowns in eqns (5) are the loading terms p,,,,.These terms are
found by expressing the loading system in double Fourier series form. This
is achieved by multiplying both sides of eqns (4) by suitable (orthogonal)
expressions such that integration over the surface of the cylinder eliminates
all but one of the terms in each Fourier expansion.
To illustrate the method consider a vessel subjected to a radial pressure
of p, over all, or part, of its surface. From eqns (4) this loading is expressed
in the form

p= »;» Pn m©OS n@ sin (Ax/a)

n=0 m=1

=) (Pom+ ) Pumcosnd)sinmnsil = mail

m=1 n=)

To illustrate the procedure consider the case n = 0, that is:

fo.6)

p= » Po.msin mnx/I
m=1

Multiplying both sides of this equation by sin (m’1x/1) dx d@ and, since the
loading is symmetric about the vertical diameter, integrating over half of
246 G. Duthie and A. S. Tooth

the vessel surface (0 to 2) we obtain

1 / /
MTX
{| asin dxd@ = intYmain ee stl dxd@
0 10)

Noting that

4 dx =U when m 4m’
l
== 1/2 when m =m’
we obtain
2 1

|Ly iy a ae |p, sin = dxdd

It No 0

Repeating this procedure for all values of n, m and all loading components
the following set of expressions for p,,,, are obtained.

For radial loading

pPaniaG ie e
Pan
i
||pis dxd@
wala lon ta l
(n=0)
4 l ua
as x ||p, sin cosnodxdd (n > 0)
co ped 2a bates l

sf
For tangential loading

4 |
= Po as sinn@dxd®@
0 (n>0Q)

For longitudinal loading

1 1

Pan
’ =F1 |
a \p,.Aaxd@ C0)

2 1

ran =z |IE
p.cos —— dx dg (n=0,m>0)
In Jo
9) i

Pan | ie
p,cosngdxd@d (n>0,m=0)
BG Mis
4 1
Ian=F | |p.cos
: lie ite
~~ cosnp dxd¢ (n>0,m>0) (6)
 Local Loads on Cylindrical Vessels 247

The method of solution consists of substituting the equation describing the

particular loading component (p,, py, p,) under consideration into the
appropriate equation above, solving for p,,,,and finally evaluating eqns (5)
to give the mid-surface displacements and stress resultants. When the stress
resultants are determined the stress values in the vessel shell at distance z
from the mid-surface may be obtained from the usual linear distribution
forms; for example:

N, 12M,z Ng 12My6Z
Pas Byaane 7s tab t ie
ete. (see ref. 20):
The above approach is of particular value in dealing with loads
distributed over discrete areas (local loads) and can deal with both
uniformly and non-uniformly distributed loading. The solution is not
intended to apply to point and line loading problems which can be dealt
with more efficiently using other methods.”'8~ '! An acceptable solution for
a point or line loading problem can, however, be obtained by considering
the loading to act over a small but finite area which is, in fact, a reasonable
approximation to what occurs in practice.

3. A PARTICULAR EXAMPLE OF A FOURIER SERIES

REPRESENTATION OF A LOCAL LOADING SYSTEM

To illustrate the derivation of the Fourier series forms consider an end

supported cylindrical vessel of length / subjected to a rectangular patch of
uniform radial loading of intensity p located symmetrically about the

Fig. 2. Rectangular patch of loading.

248 G. Duthie and A. S. Tooth

lowest generator (f = 0°) and at a distance b from one end of the vessel as
shown in Fig. 2.
This loading system is defined by the expression
P,=p inthe region —8,<@<f, and (b-—c)<x<(b+c)
P,=9 otherwise
Substituting this expression into eqns (6) gives
2 le Se manx
Poni | |Pp, Sim ] dxd@
0 JO

= > sin sin (W=0.

m= 1723: 3. 2y-"Ga)

4 1 us
Pram = 7 ||p, sin = cosn@dxd@d
Tt (0)
4 b+c (B,
= — psin oe cosnddxd¢
In l
Spe tres IRD +, (INR ©
= > sin sin i sinnp,

(12. Sacto I 2B)

For this loading system the value of p,,,,,1s given by the two equations (7a),
(7b) taken over the respective ranges of mand m, and may be used in egns (5)
to determine the displacements and stress resultants.
The series form of the applied radial load for this case may also be written
from the po, and p,,, expressions by substituting into

P= » (Pmfe>,Prim COS nd)sin mnx/1

m=1 n=1
co co

4 ] Bik : es b .
P, = 5 ae E + »ysin np, cos na |Sita ogee sie (8)
m= n=1

Expressions for p,,,, are presented in Appendix A for a range of radial,

tangential and longitudinal loading systems. These expressions can be used
 Local Loads on Cylindrical Vessels 249

in eqns (5) with the corresponding values of the Z coefficients to determine

the displacements and stress resultants for a variety of common pressure
vessel problems. Judicious use of superposition can greatly extend the
range of problems which can be dealt with using these solutions.

4. CONVERGENCE OF THE SOLUTION

When considering loading systems which involve a step or rapid change in

the loading distribution it is found that the Fourier series expansions for p,,
P,» and p, given in eqns (4) converge slowly. To illustrate this phenomenon
the case of the end supported cylinder subjected to a discrete patch of
uniform loading shown in Fig. 2 is considered. For this problem the length /
is 1219mm, half-patch width c is 51mm, half-angle B, is 5° and the
distance from the end b is 305mm. The same number of terms were taken
in each direction x and ¢ (i.e. m =n), namely 25, 50 and 100. The Fourier
series representation of the loading is given in eqn (8). Figure 3 shows the
extent of agreement between the actual loading and the Fourier series
representation. The results are plotted:

(1) along part of the datum generator (¢ =0°) in Fig. 3(a),

(2) around part of the profile at x = b in Fig. 3(b).

It is noted that there is a significant lack of agreement between the actual

loading and the Fourier series representation even when 100 terms are
taken in both the n and m series. At first sight this would appear to
invalidate the Fourier expansion approach; however, it must be
appreciated that we do not solve the problem for the approximate loading
system obtained by evaluating eqn (8) for a finite number of terms. Rather
we must consider the solution of eqns (5) which do not depend on the
convergence characteristics of eqn(8) but on the vessel dimensional
parameters, the form of loading considered and the rate of change of the
displacement or stress resultant modelled. In general, displacement
expansions converge faster than stress resultant expansions and increased
vessel length, reduced thickness and reduction in size of the loaded region
all have an adverse effect on convergence. The transverse shear forces 0,
and Q, expansions invariably converge more slowly than those for other
stress resultants but are seldom of significance. Failure to take sufficient
terms in the solution will have a greater effect on peak stress predictions
than on stresses at locations where low stress gradients exist.
 S|!
250

ras
Hele
ve
:
ne
EDs wue-o5=9“wUGLjZI=
L i | “5='4 wusoe=4 ;S
9 a=
= Ss is jo
vaquinn Ul
SWJa4 2
E4 S
Salas W=U zy
4 al; G2 eo bey “
ee
ow 3
”

( ,ww/N ) d aunssaud
El —s
|

JeIpey
[ joexg a
ae
0

L-

q <+—X

(e) Buojy Yoyesauah

= 90 (q) punosy ayljoud
$e q=xX
“BI“¢ IBINOsolos UONRJUOSoIdol
JOeB svjnsurjooI
Yyoued
Jo “SuIpeo]
 afseq yyed yews ydyed })ews yayed
ydIYL J8SS8A YUL JASS8A uly] Jassaa
00L 001 001
06 06 06
08 08 08
OL
ae w OL Vie ji
—
ie: Cc
LE STANEES
ot a eral7 dal
- 9
= 04 Oe Ras fa
5S = Mm
06 Oca” ocr
eye 4/2=hGL= r Og 4/e Sb= a Of =4/P=k
OOE
° 0 ot
3 =¢9io SCOS fe e/) S0-0=
Ol KS 4 9= e/a S0-0=
OL
Local Loads on Cylindrical Vessels

0 2.01 102 0G :07 OS 0 OF O2 O€ 07 0S 0= e0l” “OC O7S08OS

yaquny
Jo SWJa} (W=U) vaquiny
jo SWJ34 (W=U) JaqunN
JO SWJa} (W=U)
Sly‘p IBAUOD,
suds B—sonsiiojdeIeYO
[BolIpulfAo
yassoa payoalqns
0} osenbs yoied
Jo [erpes Surpeoy pozed0]
ye ye.yuao 2[youd
2/1)
= (8
2aS|
252 G. Duthie and A. S. Tooth

To illustrate the above comments the end supported vessel subjected to a

rectangular patch of uniform radial loading, shown in Fig. 2, is again
considered. In this case a square patch loading is located at the vessel centre
(b =1/2) and centred about ¢ =0°. Two steel vessels and two sizes of
square patch loads are considered. The vessels have a/t ratios of 15 and 300
and patch size ratios c/a of 0.05 and 0.25. Figure 4 shows how the radial
displacement w and the more important stress resultants converge as the
number of terms in the series is increased (note n =m), for the various
cases. The ‘exact’ value was taken as the solution obtained when 500 terms
were used in both the m and n series. The values presented in Fig. 4 are for
the centre of the patch location.
As indicated earlier this work is particularly relevant to those cases where
the load is located some distance from the ends. In these cases the m series,
that is the value associated with the axial direction, tends to converge more
slowly than the n series and improved accuracy could be obtained by, say,
putting m = 2n. An alternative means of improving the convergence is to
use an acceleration procedure such as that proposed by Aitken.*! However,
in general the authors have found that 200 terms in each series will give a
very satisfactory result for most problems unless the loading patch is
extremely small.

5. COMPARISON WITH OTHER SOLUTIONS

To illustrate the application and investigate the validity of the Fourier

expansion solution—eqns (5)—a series of cases are examined where well
established solutions are already available.

5.1. Circumferential Band of Radial Loading

Consider first an end supported circular cylindrical vessel subjected to a
rotationally symmetric band of radial loading of width 2c and intensity p
(force per unit area, N/mm7), located at a distance b from one end of the
vessel. Using Appendix A, Table 3, example No. 2:
?P,=p (0O<@<2n and b—c<x<b+c)

=) otherwise

Pam = F2=—— sin —— sin (n=0,m=1, 2,3...)

=i() otherwise
 Local Loads on Cylindrical Vessels 253

Table 1. Comparison of Classical Infinite Cylinder and Fourier Expansion.’

x/I wx 10° (mm) M,(Nmm/mm) M,(Nmm/mm) Ng (N/mm)

0 — 1.8793 + 2.6320 + 0.7896 — 5.4362

(— 1.879 7) (+2.531 7) (+0.749 5) (—5.4306)
0.025 — 0.7202 — 0.458 3 —0.1375 — 2.0832
(—0.720 4) (—0.459 6) (—0.141 6) (—2.0852)
0.050 +0.0159 —0.3289 — 0.0988 + 0.0460
(+0.016 4) (—0.329 5) (—0.098 8) (+0.0449)
0.075 +0.067 1 —0.028 1 — 0.008 4 +0.194 1
(+ 0.067 3) (—0.028 1) (—0.007 9) (+0.1947)
0.100 +0.0126 + 0.0232 + 0.0069 +0.0365
(+0.012 6) (+0.023 5) (+0.007 1) (+0.036 5)
0.125 — 0.003 3 + 0.007 4 +0.002 3 — 0.0096
(—0.003 3) (+0.007 4) (+0.002 3) (—0.009 7)

* Fourier expansion data in parentheses.

It is noted in the above that because of rotational symmetry only one term
in the n series (n = 0) exists and in eqns (5) the terms v, N,, Ny,, Nig, Mgx;
M,,, and Q, all vanish. The remaining terms can be obtained using eqns (5).
The best known classical solution for this problem is that of the infinite
cylinder carrying a uniform line load round a cylindrical profile.?°?* A
small steel cylinder, length /= 508 mm, mean diameter 2a = 229 mm and
thickness t = 1.6mm, was used for the comparisons. In the case of the
Fourier expansion solution a band-width of 0.8 mm was selected. In both
solutions the loading was taken as 1 N/mm of circumferential length.
Table 1 shows the comparison for w, M,, M, and N, with increasing
distance x from the band of loading. The only significant variation occurs
in the M, and M, values at the x = 0 position, no doubt due to the fact that
the classical analysis assumes a line load of zero width. In point of fact if the
modification to include the band-width given in ref. 22 is used the
agreement is within 0.1%. A series of different band-widths (3.2mm,
12.7 mm, 50.8 mm) were examined using both the Fourier expansion and
254 G. Duthie and A. S. Tooth

Table 2. Comparison of Classical and Fourier Expansion.“

x/I w M,. M, Nz QO.

(mm) (Nmm/mm) (Nmm/mm) (N/mm) (N/mm)

0 0.000 0 0.0000 0.0000 0.000 0 —5.108

(0.000) (0.000 0) (0.000 0) (0.0000) (—5.091)
0.01 +0.0079 — 24.265 —7.277 +113.83 +0.219
(+ 0.007 9) (— 24.269) (—7.175) (+ 113.80) (—0.221)
0.02 +0.1125 — 13.491 — 4.043 +161.64 +1.041
(40.1125) (— 13.460) (— 3.888) (+ 161.64) (+ 1.044)
0.03 +0.1166 — 3.105 —0.930 + 167.68 +0.536
(+0.1166) (— 3.074) (—0.770) (+ 167.77) (+0.540)
0.04 +0.1130 + 0.685 +0.205 + 162.52 +0.111
(+0.1130) (+ 0.747) (+0.374) (+ 162.52) (+0.112)

“ Fourier expansion data in parentheses.

the infinite cylinder case with the band-width modification. The agreement
is always better than 0.1%.

5.2. Pressurised Vessel with Constrained Ends

This case is exactly that handled by the Fourier series proposed in eqns (4).
The classical solution used is to combine the membrane solution with that
of the line load at the cylinder end—the so-called ‘edge-bending’ solution.
The magnitude of the end load was such that the resultant radial
displacement at the cylinder ends is zero and therefore consistent with
eqns (4).
Again a small steel cylinder, length /=1219mm, mean radius
a=229mm and thickness t= 1.6mm, was used for comparisons. The
cylinder was subjected to 6.9 bar (100 psig). This case is shown in Table 3,
No. |:
Pam = F1=4p/mn (= 0, iat 3)

It was found that, with the exception of the zones near the cylinder ends (i.e.
 Local Loads on Cylindrical Vessels 255

x/1<0.06), the stress and displacement values were identical to those

obtained using the membrane shell theory. Significant edge bending values
occurred in the boundary zones 0 < x// < 0.04 and 0.96 < x// < 1 and are
shown in Table 2 for-both the classical and Fourier series for various values
of x/l. It is considered that peak values of the displacement and stress
resultants are in reasonable agreement.

5.3. Partially Filled Cylindrical Vessel

Consider an end supported circular vessel partially filled with fluid of
specific weight p. The height of fluid is indicated by the angle « measured
from the bottom of the vessel (i.e. the datum generator ¢ =0°) to the
surface of the fluid (see Fig. 1). It is assumed that the radial pressure
remains constant along the length and is directly proportional to the depth
at the point of consideration. In this case Fliigge has obtained a solution for
a vessel half-filled with water (7 =90°) using a refined Barrel Vault
theory.”° The dimensions of the steel vessel used here are those used by
Flugge, viz. length / = 12.19m, mean diameter 2a = 2.44m and thickness
t= 12.7mm. This case is shown in Table 6, No. 1, and for the general
height of fill « gives values of p,,,, indicated below:

4pa_
Pam=—— — acos a)
(sina m=1, 3,5...)
(n=0,
mn

nm Se
mn? a in2 a)
sin ; (n=1,m=1,
A=1,mM=1,2,3,5...)
ehec!

8pa ; :
Pn (sin na COS & — NCOS na sin a) (ies ede
~ mren(n? — 1)
TaN)

Figure 5 presents circumferential bending stresses (6M,/t?) and longitu-

dinal direct stress (N,./t) predicted around the central profile (x = //2). The
measure of agreement is considered very satisfactory.

5.4. Patch of Radial Loading on the Cylinder

Reference has already been made to this case in Sections 3 and 4. Since a
classical solution does not exist for this case the intention here is to compare
the results obtained using the present approach with those of Bijlaard?
who also used a Fourier analysis method. A typical set of results for My, per
unit radial force P, is shown in Fig. 6 for a square patch of loading. In the
case of the Bijlaard results the small circles are the points obtained from the
analysis and the full line the interpolated or extrapolated values.
256 G. Duthie and A. S. Tooth

A Barrel vault solution (ref 20)

Fourier expansion solution.

diameter,

Za

(MN/m*)
Axial
N,/t
stress
membrane

0 30 60 90 120 150 180

@ (degrees)

of

6M,IP
(MN/
stress
bendi
Circumferential

¢ (degrees)
Fig. 5. Comparison with barrel vault solution.
 Local Loads on Cylindrical Vessels 257

Load
—o— Bijlaard (ref 3)
A Duthie and Tooth
eS euhy

Mg/P

/
Circumferential
Radial
Moment

0 0-05 0-10 0-15 0-20 0-25

B = Half patch c/radiusa
Fig. 6. Circumferential moment per radial load for square patch of radial loading
at centre of cylindrical vessel (//a = 8).

6. CONCLUDING COMMENTS

When dealing with a vessel subjected to local loading over small regions
of its surface it is common to assume that each loading component is
uniformly distributed, or in the case of an applied moment is linearly
distributed, over the area considered. This assumption implies that the
loading system, for example a bracket, is flexible enough to take up the
displaced shape of the vessel. Obviously this is not the case when the vessel
is loaded through a very rigid attachment or the loaded area is large like a
saddle type support. However, for cases in which the loaded area is small,
the maximum stresses predicted for the assumed flexible loading patch are
found to agree reasonably well with the stresses occurring in practice
although the maximum values occur at different locations.
Recognising this fact it is possible to use the compendium of p,,,,loading
terms, presented in Appendix A, to analyse the vast majority of loading
cases likely to occur in practice and obtain stress values which are of
genuine significance for design. The ease whereby these can be obtained
258 G. Duthie and A. S. Tooth

should encourage an ‘in house’ derivation thus providing solutions to a

wide range of different and perhaps unusual cases.
To estimate the stress levels likely to occur when the bracket or
attachment is very rigid the influence area method, employed to analyse the
saddle support,'>~ '® could be used. An alternative, though less precise,
approach would be to assume a non-uniform distribution given in
Appendix A, Table 4. This loading should be a form consistent with
the likely rigidity of the bracket.
When the vessel has a local geometric imperfection and departs slightly
from the true circular shape in the region of a local attachment—due, say,
to welding distortion—the predicted stress values are in general agreement
with the measured stresses when the vessel is only filled with liquid (i.e. a low
value of internal pressure). In the internal pressure case the stresses are
found to be very sensitive to small deviations from the ideal shape and for
this case a non-linear shell analysis is required.

Acknowledgements
The authors wish to record their thanks to the United Kingdom Science
Research Council and to the British Gas Corporation for the support given
to this work. This paper is published by permission of British Gas.

REFERENCES

1. BULAARD, P. P., Stresses from local loading in cylindrical pressure vessels,

Trans ASME, 77 (1955) 805-16.
2. ScHORER, H., Line load action on thin cylindrical shells, Proc. A. Soc. C.E., 61
No. 3 (1935) 281-316.
3. BIJLAARD, P. P., Stresses from radial loads in cylindrical pressure vessels,
Welding Journal, Res. Supp., 33 (1954) 615s—23s.
4. BIJLAARD, P. P., Stresses from radial loads and external moments in cylindrical
pressure vessels, Welding Journal, Res. Supp., 34 (1955) 608s—17s.
5. BIJLAARD, P. P., Additional data on stresses in cylindrical shells under local
loading, Welding Research Council Bulletin, No. 50 (May 1959) 10-50.
6. BIYLAARD, P. P. and Crancu, E. T., Interpretive commentary on the
application of theory to experimental results for stresses and deflections due to
local loads on cylindrical shells, Welding Research Council Bulletin, No. 60
(May 1960) 1-2.
7. BULAARD, P. P. and Crancu, E. T., Stresses and deflections due to local
loadings on cylindrical shells, Welding Journal, 39 (1960) 306s-8s.
8. Horr, N.J., KEMPNER, J. and POHLE, F. V., Line load applied along generators
of thin walled circular cylindrical shells of finite length, Quart. App. Maths,
11(4) (1954) 411-25.
 Local Loads on Cylindrical Vessels 259

. Mortey, L. S. D., The thin-walled circular cylinder subjected to concentrated

radial loads, Quart. J. Mech. & Appl. Maths (1960) 24-37.
. YUAN, S. W., Thin cylindrical shells subjected to concentrated loads, Quart.
App. Maths, 4 (1946) 13-26.
. Cooper, R. M., Cylindrical shells under line load, J. Appl. Mech., 24 (1957)
553-8.
. WICHMAN, K. R., Hopper, A. G. and MEeErRSHON, J. L., Local stresses in
spherical and cylindrical shells due to external loadings, Welding Research
Council Bulletin, No. 107 (August 1965) 1-69.
. DONNELL, L. H., ‘Stability of thin walled tubes under torsion’, NACA Tech.
Rep. No. 479, 1933.
. British Standards Institution, ‘Unfired fusion welded pressure vessels’,
BS 5500: 1982.
. Forbes, P. D. and Tootn, A. S., ‘An analysis for twin saddle supported—
unstiffened cylindrical vessels’, Conf. on Recent Advances on Stress Analysis,
Royal Aeronautical Society, London, 1968, pp. 47-58.
. WILSON, J. D. and Tooth, A. S., ‘The support of unstiffened and cylindrical
vessels’, 2nd Int. Conf. on Pressure Vessel Technology, San Antonio, 1973,
pp. 67-83.
. Dutuie, G. and Tootu, A. S., ‘The analysis of horizontal cylindrical vessels
supported by saddles welded to the vessel—a comparison of theory and
experiment’, 3rd Int. Conf. on Pressure Vessel Technology, Tokyo, 1977,
pp. 25-38.
. Durtuie, G., WHITE, G. C. and Tootn, A. S., An analysis for cylindrical vessels
under local loading—application to saddle supported vessel problems, J. of
Strain Analysis, 17, No. 3 (1982) 157-67.
. Wixson, J. D., ‘Thermal stresses in thin cylindrical shells’, 2nd Int. Conf. on
S.M.I.R.T., Berlin, September 1973, Vol. III, G1/10, pp. 1-12.
. FLUGGE, W., Stresses in shells, 2nd edn, Berlin, Springer-Verlag, 1973.
. Puitups, G. M. and Tayior, P. J., Theory and application of numerical
analysis, New York, Academic Press, 1973.
. Roark, R. J. and YouNG, W. C., Formulas for stress and strain, 5th edn,
Tokyo, McGraw-Hill Kogakusha Ltd.

APPENDIX A

Compendium of p,,,,, Solutions

The Fourier expansion solution presented in this paper contains only one
set of unknowns, the loading terms p,,,,. These loading terms are
determined by expressing the loading system in double Fourier series form
(eqns (4)) using eqns (6). This appendix presents p,,,,Solutions for a variety
of radial, tangential and longitudinal loading systems—Tables 3 to 8.
Figure 7 illustrates the parameters employed in these tables to define the
size, location and distribution of the loading. The symbolsp, A, B, po, No, p,
260 G. Duthie and A. S. Tooth

ariitt
TE
LE SS SN

Eas

Fig. 7. Parameters employed to describe loading systems.

p,, T and X are constants used to define the intensity or magnitude of

loading.
These p,,,, €xpressions or combinations thereof can be substituted into
eqns (5) to give solutions for the loading systems considered. When a
number of loading systems of the same type (ex. radial) act simultaneously
the calculations can be simplified by combining the p,,,,,expressions before
substituting into eqns (5). In other cases the values must be considered
separately and only combined at the final displacement or stress stage.
 aIqu],
“¢ BuIpeoTsway
10J WOsUA,[eIpeyY SuIpeOT ‘SUId]SKg
‘ON woi301
fo q SUuIpDOTwajsds ed
sulpvo] wajsds =) 09 (asimaayjo =) 0'Q s0f‘u ui sanjpa
jou (pajonb
|

I
: dad
(esos)
1>x30
aa)
“T= gE

ST Cease “d
UZ>PS0
P+)
SxS(9-
O=u
=u
Oe
“eT

|
Local Loads on Cylindrical Vessels

ps0)
zs
ee ete pau) (
u~Soso oy ef
ee” On
261

(panuijuo2)
 aiqeL ‘piuos—e
N
\o
ON wossvig
fo SUIpvOTwWasds | PG a
suUlpvo] wajsds =) 00 (asim1ay10 =) (0 40f‘u uw sanjpa
jou (pajonb

I C="
'g- os dS eae eee
7 a= boty)
fom: as
Il eG
6 = =u6 6 © eee
(gustaen)i (
| ee wy o
4
oans(I \SOSCJ—'P)
ee z hire
o=u ee 3
aM
S F 1>x>0 ty be 66 tee =
eels aS)
=
+'o-)So5('9
Cg
eee
—'0-)
-
(‘pu)
[7
urs
gu)
(
= x
soo
Ue uu , ; a Cre
Be
—
S

=S
td J Ei) (
ee sear
a
9 Cc Hy Secuve

'g-— isos
yee (9+q)5xS(9-q)
ig
— usp (Tf) eee, )
ee
 'g@So5'¢-
«2 (dea
1)) ia
(9+9)SxS(9-@) Can

Jena
'gsos'g v =u ‘TT‘€
— uIspy (‘gu)

(
a)ae

a
G7 =e Cag= )
”
‘p) (eine Q=uU

(9-9)
<5>
ec da “) re("
(ing)
—"O- \SoS('g (Gao

4
Sxs(0—¢9
OF)) ~ soop.7(pu)urs (‘u)‘)i
u
=

seyee)
lat
Z e
9) S(9—=x V+9)

ayiaIe i

Ol
are (eee
)a- 9)Sx(9+
5(9-q)
SoS (19+)
ja—19) ('g
(
8

soapy“
(9+9—))>x5(0—q=) (“ou)us
) (gu) =u os (
Local Loads on Cylindrical Vessels

—'o-)
('g SoS '-) ("9+ Sa SeeEe BOS

(Ogss ) ei (2a ie
)a-
263
 264
“p IIqeL “suIa}sAS BuUIPeO’T [eIpeY ofoqeieg JO} SWI] SUIPVOT

fo wossvig BUuIpDOTwajsds Rati gf

=) 00 (asim1ay10

‘ON
wajsds Suppo] 0'0=)
40f‘u uw sanjpa

jou
(pajonb
jig Q=u

Speen pea “ee

_ G+ q)sx>FC-q) € ‘7, =u
F Lar CH ee
en)

uz! eral
( Z Jiga (
(9+9)>xS(9—-@q)
heta Sian
7 asesy-) ( : )ea Cees
(9+¢-)S>xS(9-qG-))
tet Ce
G. Duthie and A. S. Tooth

-‘9)('9 S55 (9+) 0

17 ae
( ger i) (
a
ay oe Set
See Se
=) <')(9 0.5S Cote)
a AA oie( =u‘TT €
Ta
e cal
See cs Oa ep (omso
) oca
(

Teau\., eae
Ter
a (ome
—'f-) +'9
(*G¢-)So5('f
een ee(
 N19 e
Os ere
f= ( eae (
as(9-4) SxS9) (9+
eae ras
ores

9-*)V=17
+ 8d ((9-D—-*)¥=7T
+ 899d -b9=€T
d+ EI Sra recsurs X('gu)
uu uu

195 \
uls= pe gua) a sum) UISs=79qua}
soa!
\ Ee ay =F3 ae WAOUS Wel= 91aoa CAISE

d JZ é
i ly 'fd+ (‘gu)us7g=7x
+ soo gu) —('
= us (Igu)5+ uis (‘gu)
gq
Local Loads on Cylindrical Vessels

= Od)+ 29K— 1S((-2M)/1KT

+ TZ) TS((UM)/J9
265
 bile PEELE ee cite
eer ay Sd gsos'g— OL
Boe) tira Pe
(9+9)SxS(9—-@q) ar as
Na)? ( (
meee la
Gh scan SN ( aseee iee
G. Duthie and A. S. Tooth

Ce (+b) —"-)('f S05

Be A he
ean) (
ee ee
bine eis —) (9 55 4b)(J
Rene oe
SSE Ov eee HY Gso5'7—
Q=u Id
jou
(pajonb
sanjpa w‘u 40f 0 =) (asimsayj0 (0) =) wWajsds SUuIpDO]

‘ON
ead wajsds SUuIpvOT fo wos3vIG

“SUIdISAS
SuIpeoT
SUIPCOT
[PIPPY
IVIUIT
SUID,
sqey
JOJ
°s
266
 —"))(19 565!o) (f+

lI
4)\.,

“d
Uv(a
i sere Cee PO peu) (
eee
tH Cr Oe

ae eee Ce
( e else
(9+q-))SxS(9-q—-eo
)) gargs Se
p=) (ig o> l= = Cg
mae ee ee
(

TaD es) PBC Cie eres) Zea og urs Z('gu)

F A uu uuu

(‘ou)sooz7dxZ7=¢d
=rqd = 4a'g =¢q er urs —_AA("gu) (‘pu)soogg*z7=9q
UU : UU
Local Loads on Cylindrical Vessels

- quu uM \quu* osuu} 1°y \quu* suum}

— uu 97)uu
— (7 JUU fay amin! pny) —97Z)uu
(1
Lf= uls aS uIS | °y soo pea! cia uIS f +
gum}
uls zit soo a | ny edie
eee
ee ee a n
267
 268

sIquyL
‘9 BuipeoyTsway
JOJ [eIpeY BUIPLOT "SUI9]SAS

wopisvig
fo
BUIPDOT
‘ON
Wajsds CERO

sulppo] wajsds =) (0) (asim4ayjo = 0.0 40f‘u uw Sanjpa

jou (pajonb

ee
peau SSNS
|H="d (~SOS0-)

ea
PODS

ce hee
G. Duthie and A. S. Tooth

ee
=—d
Geers

len
 PD LF LP FL a a Ze NN RCE
®y,
u~Soso
A ew="a
)
say iieeiaes
PPLPLE LE A OE CTP
u~Soso
Q—¢g—S45
)) 1) Ot¢

|
SR,
ee (gay)
“d =
1>x>S0

LH = yd = $09)pd
— soo(x 1)°d=7y
+ (soo Psool'd= Ddp
1D EeH
Local Loads on Cylindrical Vessels

US)
» — S09”
(0
zulu

® voz )=(
07]
— urs [(0Z) ana?
uls] dp° quu} IUUL
= zulu
(nu) ns00
— soou (xu)uls[w = pO —_=uls} ——
,¥)u,um
— (1
(

wu ]
1

=¢D (um)/1*dp
269
 ORL
“L BuIpeoTSWAY
IOJ [eNUAsUL]ZUIPROT "Sua}SAG
270

‘ON woisvig
fo SUIPDOT wajsas GL
BUuIpDO] Wajsds =) 00 (asimsayjo =) 0.0 40f‘u ul sanjoa
jou (pajonb

DAAAAARAARALS
= u>o>0
Ad RA RRA eta —e\()
I Tay tote
A? ARAARBAAL lg
tee yeneetty >P>u
UT cers
ae
(
S
4
u>o>Q =
: Sx50-9)
Q oo :
oe x-=.
ieee 4
S
u>o>gQ
§
F ee)
ae S‘eT=u
Cee Soke
ie
XS Xe

ee (ee

lees (cad

ee) ae
u~>p>uu>o>g u~>p>u
 —'6) So5('
+o) (9g
Ae
io) xs Cheam
(a) UTS

"4 gs0s'y_)@use=Me ae
ea se
eet Sak
th= Ce
sa ie (cae
£ 7Y= [7 uls (aa
us (se)¢g = zg us us(‘ou)
(‘gu) ust—'d
(dz)
]w=pq
Local Loads on Cylindrical Vessels

89 = ag)
7) ['g(u— Djus igenchar ener Ga
3 eal
(u—1)Z [)Z (u+ zu / /
271
 °*g aIQUL “suId}skS SUIPeOT [PUIPNIIZUOT JOJ SUID, SUIPCOT

‘ON woisvig
fo 8UuIpDOTwmajsds De bali
Dads

suIpdo] masks =) (0 (asim41ayj0 =) (0 40f‘u ui sanjoa

(paionb
jou

P>0
UTS
i
tea! sera
: Coorer us
Se
I P>0 UZ Tau ‘eT
a ee © ee
oe

UZ
>P>0
ao poe See
x ) ere Ne O=u
c cv
waaPG PEG) uz~sos ai) ere SES
Z) ae eae
[
—9)CI SOS CI+'P)
t
G. Duthie and A. S. Tooth

saa, LON OPE(

(
—6—
(9

+P) SOS (C9 —b Cg

ee
ia, cancers i@yam =
+'6-)So5('f

(9 +'o-)SO5('9 —'¢-)
ae4(9+
("gu
b)
(“*)
is
ial
(1
——=|V
EV
TW
USS
*7=
uls
vv
'd
ey
usu)
cv
900
soo
=Cy
==
uu Z uu ] ] u
( q—-)SxS
 14

In-Plane Bending Behaviour of Flanged Pipe

Bends Under Steady State Creep

G. THOMSON
Ferranti PLC, Edinburgh, UK

ABSTRACT

An approximate solution is developed describing the creep behaviour of thin-

walled smooth pipe bends with rigid flange terminations under an in-plane
bending moment. The minimum total potential energy rate theorem is
employed together with suitable approximate velocity series. Integration
and minimisation of the energy function is performed numerically to give
flexibility rate factors. These are compared with the unconstrained bend
results of Spence and with elastic results for flanged bends with Poisson’s
ratio of 0.5. The effects of the various assumptions involved in the analysis
are discussed together with possible improvements.

1. INTRODUCTION

With current design trends in the nuclear and chemical industries towards
higher operating temperatures, the problem of creep in piping systems has
become increasingly important. Furthermore, the necessity for accurate
predictions of the safe working life of components in high capital cost plant
is essential. The fundamental problem in piping structural analyses is to
design a system with sufficient flexibility to contend with thermal loads and
stresses. Smooth pipe bends are one of the essential parts of any pipework
flexibility assessment; being more flexible than an equivalent length of
straight pipe they can provide a significant contribution to the overall pipe
system flexibility.
273
274 G. Thomson

The linear elastic behaviour of thin smooth pipe bends under various
loadings has been extensively examined by many authors in the last seventy
years.’ In 1957, Kachanov? investigated the effect of creep on pipe bends
under in-plane bending. Using a complementary energy method with a
creep power law he derived an upper bound approximate flexibility factor
using a two-term series. Spence later published a series of papers? * giving
more accurate creep flexibility factors employing upper and lower bound
energy analyses with up to five-term series. In his analyses he used a Norton
power law of the form

(é/ép) = (G/6,)" (1)

where é is strain rate and a is stress.

Spence’s results clearly demonstrated that creep flexibility factors can be
considerably higher, for practical values of n, than elastic flexibility factors.
In 1975, Boyle® published a dissertation on rational creep mechanics
including some work on creep in pipe bends. One important conclusion
from his work was the verification of the steady state results of Spence. In
ref. 7 Spence and Boyle presented results for out-of-plane bending of
curved pipes with finite bend angles. Various studies, both experimental
and analytical, have been published in recent years advancing knowledge
on the influence of creep on smooth pipe bends. Reviews of some of these
can be found in ref. 1.
Linear elastic examinations of the bends with end constraints, e.g.
tangent pipes and flanges, have revealed the significant influence of these on
the bend flexibility,':®° particularly for flanged ends. To date, no analytical
studies have been published on the effect of end constraints on the flexibility
creep rates of smooth bends. In the following work, a tentative solution is
presented to the problem of a bend terminated by rigid flanges at both ends,
subjected to an in-plane bending moment under steady state creep.

Zen LEbORY

2.1. Introduction
The analysis presented herein is a combination of the elastic solution of the
present author! together with the method used by Spence* to extend the
elastic theory to the creep regime. Compatible velocity fields are selected
which satisfy the kinematic constraints of the flanged bend problem,
employing trigonometric series. A suitable energy function based on the
 In-plane Bending Behaviour of Flanged Pipe Bends Dip

relevant constitutive relationships is derived from the velocities. This is

minimised to provide the required results.
Although this procedure is relatively straightforward for bends without
end constraints, it is many times more complex when these are included. In
the elastic analysis it is possible to make use of the quadratic nature of the
energy function to perform the minimisation in a reasonable time on a
computer. The complexity of the creep energy function requires the use of
numerical integration together with an iterative minimisation algorithm
which is several orders slower than the quadratic procedure. Furthermore,
numerical iterative minimisation in the 105 dimensions (unknown
constants) of the relatively complete elastic solution presented in ref. 8 is
prone to numerical error.
To enable a reasonable first tentative solution to be derived, a number of
assumptions will be made. Some justification for these can be obtained by
examination of the more complete elastic solution.®

2.2. Creep Law

The stationary creep law employed herein is a Norton power law of the
form given earlier in eqn (1). This is selected because its relatively simple
form can be shown to be typical of real material behaviour, particularly at
low stresses. Furthermore, the material constants in this equation are
available for a range of common materials.

2.3. Compatible Velocity Fields

Details of pipe bend geometry are presented in Fig. 1. Bends can be
modelled as part of a thin toroidal shell for which the governing equations
are widely available. To simplify the solution the following assumptions
will be applied:

(a) Mid-surface meridional strain rate (K,) is zero.

(b) [(r/R)cos ¢] <1.
(c) Direct shear strain rate (7,4) is zero.
(d) Circumferential and shear curvature rates (K, and K,,) are
negligible.
(e) Rigid section circumferential direct strain rate (&) is zero at
p= +7/2.
(pee:

Note that in assumption (b) radius ratio (r/R) itself is not being
neglected.
276 G. Thomson

Applying some of these assumptions to the more general strain

rate—velocity relationships gives

eC Utena
ig Rlgp~ 880 + 005 0)

) bil OL yee
YeoR\ G6 + ag
. 1fdv 6?w 5
fect) 2
For convenience in the current problem, velocities uv, 0 and w are divided
into two components, the first termed the ‘rigid section velocities’,
associated with the movement of the tube cross-sections relative to the bend
centreline with no change in circularity, and the second are termed
‘distortion velocities’ representing the distortion rates of the initially flat
and circular cross-sections.
Rigid section velocities tig, tg and Wg can be found from U, and V,, the
bend centreline (¢ = +7/2) velocities, i.e.
iif e
Up = U,+),rcos@ je RUF]

tg =V, sing

We =V,cosd (3)
y, is the rotation rate of the bend centreline. Substituting these into the
complete strain-displacement rate eqns(1l) together with the above
assumptions, particularly (e), gives
& = Yop = Ky =0

ees
Ree iaa: : Fae is eU, c
ae (’. 7) (4)

An appropriate series for », which satisfies the necessary boundary

conditions is
6 :|o = 27K.
jot Bin) {hs SNe (5)
 In-plane Bending Behaviour of Flanged Pipe Bends 20a

Fig. 1. Bend geometry.

Note that the rotation rate between the ends of the bend at 0= +«/2
153):
The distortion displacement rate component normal to the shell surface,
Wp, can be approximated by the following series which satisfies the relevant
flange and symmetry boundary conditions:

F 7
>=) )Cancos geos? (“2 m=1,3,5..., 9=2,4,6... (6)

m 4
278 G. Thomson

Vp and up can be determined by the use of assumptions (a) and (c)

respectively, i.e.

7
r [Cts
R } 00

2.4. Total Strain Rates

Using the velocities given in eqns (5), (6) and (7) together with the strain rate
eqns (2) and (4), the total strain can be determined as

Eq =5 [(®+), (=) cos (=*)), cos d

+)Dai qpcos cu
yale
— gg sin @) cos (mm
3:

DSye
ma (=)
R cos gp cos =)

k, iy >
: ]
nla? = Neos gdcos?(™22)
:
(8
mq

These equations are analogous to those required for a linear-elastic

analysis.

2.5. Total Potential Energy Rate

For a smooth pipe bend under an in-plane bending moment, M, with the
above assumptions, an approximate total potential energy rate function
can be derived as!°
do 4 (n+1)/2n 1/2 2n
. n
a t
V=
BTU ee NS pe ee
t? f (n+ 1)/2n

x (i+42 K) rRdg
dd — Mj},
where
n 2n/(n+1)
A=
(5;+ :) (9)
 In-plane Bending Behaviour of Flanged Pipe Bends 279

This can be minimised initially with respect to 7, to give

(4) )Biec) & oO

— (3\"* D2 (Ra\/Mel\" J

where ® is the minimum, with respect to all D, and Cosof

ae 20) JO eae a ow

1 m0
ne «((cosadeos 4 +1 sin gd sin
eos (“=)

ge
+—|— ] cosq@cos
R \ qa
72 i
ak a
mnO\ 12) "+ vi2n
+A by Y Cnt? — 1)cos qd cos? (=) dg dé

Rt & . a @ : a
A= =D, (=) Cut Oe, (=) (11)

Integration and minimisation of ® was performed numerically on a vAx

11/750 computer.

3. FLEXIBILITY FACTORS

The end rotation rate for a similar length of straight pipe (Ra), under the
same bending moment, M, is?
_ (Ra\( Me" 1
Ome PNT ig De

|2
0 =4 | (sin p)"*/"d@ (12)
0)

This can be used to define a flexibility factor K as

Kral) () 9
‘ (n+1)/n n

(0)
280 G. Thomson

SPENCE [4] (No end effects) n=1

az
PIPE FACTOR (A)
Fig. 2. Comparison of flexibility factors.

which is the end rotation rate of a pipe bend in creep divided by the end
rotation rate of a corresponding straight pipe in creep.
Flexibility factor convergence required about three terms in each velocity
series for A>0.1, «<< 180°, R/r < 10 and n<7. Acceptable convergence
could be obtained with fewer terms inside these ranges.
Figure 2 presents a comparison of the flexibility factors from the present
theory n = 1, with the elastic theory of ref. 8 with y =0.5. Although this
comparison is not strictly valid, since the present theory gives flexibility
rates, its use can be justified by the elastic analogy.'' It can be seen that the
present theory with all its assumptions shows a similar trend to the analysis
of ref. 8 which has more complete displacement series and virtually uses
only the assumptions of thin shell theory for the same problem. However,
the current results are appreciably lower in value, particularly for smaller
radius ratios, R/r, and bend angles, «. Attempts at improving the present
type of analysis for the elastic problem’ have shown that a significant
 In-plane Bending Behaviour of Flanged Pipe Bends 281

100
R/r=10
O¢=180"

20

{‘

FACTOR

FLEXIBILITY

1
04 02 05 1 04 02 05
PIPE FACTOR (A)
Fig. 3. Creep flexibility factors for flanged bends with « = 180°.

improvement in the results can be obtained by the removal of the zero direct
shear strain assumption. However, this requires the introduction of a
further independent displacement series for up. Usefully this can have the
same form as used in the current analysis but with independent unknown
coefficients. Introducing these constants in the present analysis is not
difficult but it requires a substantial increase in the solution computation
effort. After this, the biggest improvement can be obtained by use of more
complete displacement series, 1.e. functions which are more capable of
modelling the true displacements. Removal of the zero direct meridional
strain, é, = 0(as used by Karman in 1910) and (r/R) cos @ < | assumptions,
have been shown in the elastic analyses to have little significance. A fully
282 G. Thomson

FACTOR

FLEXIBILITY

—— — SPENCE AG]

01 02 05 1
PIPE FACTOR (A)
Fig. 4. Creep flexibility factors for flanged bends with a=90°.

convergent solution from an analysis of this type may require the use of
about 80 displacement rate coefficients. Numerically minimising a non-
linear problem of this size would be a difficult task, possibly prone to
numerical errors.
Also shown in Figs. 2 and 4 are the results from Spence’s* analysis which
neglected the influence of end constraints. It can be seen that flanges on the
ends of a bend will significantly reduce the potential bend flexibility.
Spence’s results will be more representative of bends terminated by long
tangent pipes.
Figures 3 and 4 present creep flexibility factors of bends with rigid flanges
under in-plane bending for bend angles of 180° and 90°, as determined
 In-plane Bending Behaviour of Flanged Pipe Bends 283
00

—— PRESENT THEORY
—-— BOYLE & SPENCE[12

Ss
FACTOR

FLEXIBILITY

i
01 02 05 - 1
PIPE FACTOR ())
Fig. 5. Comparison of present theory with approximate formula of ref. 12.

from the present analysis. From the figures it can be seen that flexibility
increases with reducing pipe factor, A, as was the case in Spence’s analysis,
and increases with increasing radius ratio, R/r, n, and bend angle, «.
Boyle and Spence!’ proposed the following formula:
creep = p" creep
end effects = no re effects

where
p= |Ge etna, Iap effects (l4)

which estimates the flexibility of a bend with end effects under steady state
creep using available data on creep in bends with no end effects and elastic
results for K withn = 1, with and without end constraints. This formula was
284 G. Thomson

derived from an argument based on reference stresses. Watanabe et al.'?

made use of this formula in their computer programs for inelastic analysis
of piping using beam theory with flexibility factors. Figure 5 presents a
comparison of results from this formula with the present theory where it
can be seen that the formula appears to underestimate flexibility. In view of
the assumptions inherent in the present theory this comparison is
inconclusive in a general sense, as the results may improve with the use of
more accurate flexibilities.

4. CONCLUSIONS

A method has been presented for the determination of flexibility factors of

smooth curved pipes with rigid flange termination under in-plane bending
in creep. When compared by analogy with elastic solutions, the present
theory demonstrates similar behaviour but can be significantly in error for
lower radius ratios and bend angles, i.e. for shorter length bends where
flange interaction is more significant. Possible improvements to the
analysis have been proposed.

Acknowledgements
The author gratefully acknowledges advice and assistance from Professor
J. Spence and Dr J. T. Boyle of the Department of Mechanics of Materials
at the University of Strathclyde.

REFERENCES

1. THomson, G., ‘The influence of end constraints on pipe bends’, Doctoral

Dissertation, University of Strathclyde, 1980.
2. KACHANOV, L. M., Theory of creep, Gostekhizdat, 1957.
3. SPENCE, J., ‘Creep behaviour of smooth curved pipes under bending’, Ist Int.
Conf. on Press. Vess. Tech. Delft, 1969, pp. 309-15.
4. SPENCE, J.,Creep analysis of smooth curved pipes under in-plane bending, Jrni.
of Mech. Eng. Sc., 15, No. 4 (1973) 252-65.
5. SPENCE, J., Stationary creep stresses for elliptical cross-section pipe bends
subject to in-plane bending, ASME Jrnil. of Press. Ves. Tech., 96(1974) 162-71.
6. Boy Le, J. and Spence, J., Flexibility of curved pipes in creep, ASME Jrnil. of
Press. Ves. Tech., 99 (1977) 844-S3.
7. Boy Le, J. and SPENCE, J., “An analysis of out-of-plane bending of curved pipes
in creep’, Proc. 1.W.G.F.R. Conf. Fast Breeder Reactors, Pittsburg, 1976.
 In-plane Bending Behaviour of Flanged Pipe Bends 285

. THOMSON, G. and SPENCE, J., The influence of flanged end constraints on

smooth curved tubes under in-plane bending, Jnt. Jrni. Press. Ves. and Piping,
13 (1983) 65-83.
. WHATHAM, J. F. and THomson, J. J., The bending and pressurizing of pipe
bends with flanged tangents, Nuc. Eng. and Des. (1979) 17-28.
. MACKENZIE, A. C., On the equations for steady state creep of thin shells, Jrni.
Mech. Eng. Sci., 7 (1965) 114.
. OvevisTt, F. K. G., Mathematical theory of creep and creep rupture, 2nd edn,
Oxford, Clarendon Press, 1974.
2, Boy Le, J. T. and Spence, J., Inelastic analysis methods of piping systems, Nuc.
Eng. and Des., 57 (1980) 369-90.
Bs WATANABE, O. ef al., Inelastic analysis of complex piping systems in an elevated
temperature environment by simplified methods, ASME Special Pub.
Benchmark problem studies and piping systems at elevated temperature, Press.
Ves. and Piping, 66 (1982).
ae eoseal.: bons ee PV
ets (oye .»
ball ators, ein
Los vere Uae?
 15

Plastic Design of Submarine Pipelines

J. W. B. STARK and P. E. DE WINTER

Institute TNO for Building Materials and Building Structures,
Delft, The Netherlands

ABSTRACT

In this paper a method is presented which enables the designer to determine

the collapse load of pipelines, which are subjected to combinations offorces
and deformations such as occur in deep water. This method makes use of a
simple kinematic model based on plastic design principles. It is
demonstrated for two loading cases. Special attention is directed to the
influence of loading sequence on the ultimate conditions. The theoretical
results are compared with test results.

NOTATION

Area [L’].
Average diameter D = 2R [L].
Modulus of elasticity [KL ~*].*
General notation of normal force [K].
Saha
Theoretical value of the yield force [K].
(xDto,.) of a pipe cross-section.
ss Rocker force (contact force between rocker element and bending
element) [KL *].

* For the sake of simplicity the derived dimension of force is here designated by the
symbol K (K = [MLT] ’).
287
 J. W. B. Stark and P. E. de Winter

Theoretical value of the yield rocker force (t?¢,/D) [KL~*].

Moment of inertia [L*].
Bending moment [KL].
Theoretical value of the full plastic bending moment (D?tc,) of a
pipe cross-section [KL].
Theoretical value of the full plastic bending moment (3t*c,) of a
rectangular cross-section [K].
Bending moment per unit length in circumferential direction [K].
Stress resultant in axial direction [KL ~'].
Theoretical value of the yield force of a rectangular cross-section
(n, =to,) [KL~*].
Stress resultant in circumferential direction [KL ~*].
Area of the tube cross-section [L?].
Theoretical value of the elastic buckling pressure (4E(r/R)*)
Kee |
External radial pressure [KL].
uo}
Theoretical value of the yield pressure (2tc,/D) [KL ?].
Radius (R =4D) [L].
Wall thickness [L].
Displacements [L].
7S)é weN
AS
es) Coordinates [L].
Coefficient (« = 1.17).
Ovalling parameter (out-of-roundness).
fo}
Initial out-of-roundness parameter.
SS
oo
RS oS Strain in the middle of a cross-section in axial direction.
Angle [rad].
Curvature [L~*].
a 4
Critical curvature [L~'].
Se)o
ty
GS
Ch Theoretical value of the curvature at which yielding occurs in the
extreme fibres (20,/ED) [L~'].
Axial stress [KL~].
Yield stress [KL].
Yield function.

1. INTRODUCTION

In deep water the loads and deformations that a pipeline can safely
withstand are reduced by external water pressure. The magnitude of such
reduction to a large extent depends upon the pipe parameters and the loads
 Plastic Design of Submarine Pipelines 289

and deformations to which the pipe is subjected. The pipe will collapse if the
load carrying capacity or the deformation capacity is exceeded.
This paper presents an approximate method of analysis enabling the
relation between pipe parameters, load combinations and the deformation
capacity to be determined (see Table 1).

Table 1. Parameters in the Mathematical Model.

Pipe parameters Load combinations Deformations

Diameter Bending Curvature

Wall thickness External liquid pressure Ovalisation
Initial out-of-roundness Axial force Axial strain
Yield stress

Protech International commenced experimental work on this subject in

1975. Later, the Netherlands Organisation for Applied Scientific Research
(IBBC-TNO), Protech International and the Universities of Delft and
Eindhoven carried out further investigations in order to determine the
mechanics of pipeline collapse. These investigations were constantly
aiming for easily applicable, limit state interaction formulae. This resulted
in the development of an approximate method describing collapse that may
occur in submarine pipelines.
This method can be explained with reference to a physical interpretation:
the pipe cross-section is replaced by two elements:

—a bending element for modelling the circumferential stiffness;

—a rocker element to simplify the geometric non-linear behaviour of a
tube.

The collapse load of such a model can be calculated in a relatively simple

manner. In frame analysis a similar method is being used for calculating the
buckling load of framed structures. In illustrating explanations, a
simplified yield function has been adopted to describe the physical non-
linear behaviour.
This method is demonstrated for two loading cases:

—collapse under combined external pressure and axial force, including

initial out-of-roundness;
—collapse curvature under combined bending and external pressure,
including effects of loading sequence.
290 J. W. B. Stark and P. E. de Winter

Tests have been carried out on model pipes. A brief description will be given
of the test specimen and the test set-up. The results of this test programme
and test results from other sources will be compared with the theoretical
outcome.

2. THEORETICAL MODEL

It should be possible to perform stability calculations in the plastic range

with the model to be formulated. In addition to this, the model should be
able to cover the influence of deviations from the ideal shape of the pipe. It
must be simple to understand, and the amount of arithmetical work should
be limited.
From experimental simulation it was observed that almost all
deformation in the collapsed cross-section of a pipe was concentrated in
four ‘plastic hinges’ (Fig. 1).
Therefore the following model for ovalisation has been adopted:

—four rigid quarters of a circle;

—mutually connected by pin joints.

This part of the model is called the rocker element. The geometrical non-
linear deformation in the circumferential direction can now be determined
very simply with only one parameter: the ovalisation angle f (see
appendix). By virtue of the four hinges the rocker element can be deformed
without any resistance. Therefore, a flexural element, representing the
circumferential stiffness, is also needed.
The stiffness of the pipe cross-section can be modelled either by
rotational springs at the hinges or by a bending element. Professor J.
Witteveen of Delft University proposed the use of the bending element
(Fig. 2) because it allows, in principle, for lower bound solutions.° It is
essential to know that the rocker element is only connected at the hinges to

(a) (b)
Fig. 1. (a) Cross-section after collapse. (b) Cross-section at the onset of collapse.
(c) Theoretical idealisation.
 31}DWeUIy
japow UJIM
D
Bulpueq juewe}e
J9HDOI JUBWI]e Buipueq 1UeWej}a
Plastic Design of Submarine Pipelines

“BY«°*7 OneUIOUTY
‘JEpoul
291
292 J. W. B. Stark and P. E. de Winter

the bending element. When ovalisation occurs, the rocker element

transmits point loads to the bending element, so that resistance to the
ovalising deformation is developed. To clarify the representation the two
rings illustrated in Fig. 2 show different diameters. For analysis purposes
they coincide.
Primarily, the load is applied to, and supported by, the rocker element.
The bending element only serves to resist ovalisation. However, its stiffness
is affected by the loading that the rocker element supports. As far as
possible the development has been restricted to ‘manual’ calculations, to
reduce reliance on computer analysis. This means that derived functions
must be linearised already at an early stage, which in turn means that the
model is valid only for small ovalisations. To describe the equilibrium of
the model the principle of virtual work will be used.
The fundamentals for the analysis are: plane sections remain plane, and
the cross-section remains perpendicular to the axis of the pipe. In the
appendix, the static, kinematic and constitutive (elastic and rigid plastic)
equations for the model are presented for the combined loading case of
external pressure, axial force and bending. In the appendix the basic
eqn (24) has been derived under the assumption that the plate bending
moment in axial direction m,=4m,. This means that the influence of
longitudinal plate bending is included. In the next sections various loading
cases are covered in order to demonstrate the calculation method.

3. TEST PROGRAMME

Tests have been carried out in order to verify the results of the theoretical
model. Test results from other sources have gratefully been used when
available and permitted.
In order to minimise the influence of geometrical imperfections some of
the test specimens used at IBBC-TNO have been machined carefully, first
internally and then externally, on a lathe, thus ensuring that the test
specimens had a truly circular cross-sectional shape with constant wall
thickness.
Variations in both wall thickness and diameter were less than +0.02 mm.
The mechanical properties and dimensions of the specimen which have
been tested are presented in Table 2.
All tests specimens were amply instrumented. Specially designed
callipers were attached to all test specimens to measure the change of
diameter during loading.
 Plastic Design of Submarine Pipelines 293

Table 2. Mechanical Properties and Dimensions.

Test Type CO. On & ER 1b), t

(Nimm*) (Nimm*) (%) (%)_— (mm) (mm)
] WwW 315 430 Zi DS 89.5 3.05
2,3 H 249 370 0.5 31 100.3. 4.06
A, D5, 6, 4;,8 S) 320 460 7 25 100.5 4.0
9, 10 S 320 460 Le? Ds) 997 3).
11 WwW 300 400 4.0 30 101.3 3.44

S: specially machined test pipe; H: seamless commercial pipe; W: welded pipe; o,:
yield stress; og: failure stress; ¢,: strain beyond which work hardening occurs; é.:
strain at failure.

Curvature measurements have been performed in two separate ways.

First, the curvature change is measured locally by several sets of strain
gauges. Second, with specially designed displacement transducers, it was
possible to measure the average curvature change over a certain length of
pipe. By applying two types of measurement, the test results are interpreted
much better. A fully instrumented pipe section is shown in Fig. 3.
The tests were conducted in a pressure vessel designed by Protech
International and made available for the current study (Fig. 4).
The test rig consisted of a pressure vessel of 14” OD. The test specimens
protrude from the pressure vessel. The loading equipment needed to induce
bending and axial force is situated outside the pressure vessel. This
configuration allows easy access to the loading equipment even during a

Fig. 3. Fully instrumented test specimen. (A) Callipers to measure change of

diameter during loading (6); (B) strain gauges to measure locally change of
curvature (8); (C) deflection transducers (2); (D) displacement transducers to
measure change of curvature over approximately 30cm of pipe length (4).
294 J. W. B. Stark and P. E. de Winter

dead weight
compensation

pressure vessel

outline of the test facility

Fig. 4. Test facility.

test. An additional advantage is that the need for a large pressure vessel is
greatly reduced.
With the test set-up as described above, the theoretical model can be
verified adequately. However, the distribution within the pipe does not
exactly correspond with the actual situation in the sea. In Fig. 5 the test
situation (no endcap) is compared with the actual situation in the sea
(endcap).
For this reason IBBC-TN O is considering constructing a new test facility
in which the specimen is installed inside the pressure vessel. The diameter
will be 1200mm (48”) and the operating pressure of the vessel will be
approximately 250 bar. An additional advantage is that larger specimens
can be tested to a larger curvature.
 Plastic Design of Submarine Pipelines 295

test facility

Fig. 5. Pipe laying situation.

R me

nm elastic
Uv ae)

ov

[P/Pp]
hydrostatic
pressure

Po ovalization e

Fig. 6. Pressure ovalisation diagram.

COLLAPSE
OP RESSURE [P/Pp]

00 02 04 06 08 40 roundness [35]
Fig. 7. Collapse pressure of out-of-round tubes.
296 J. W. B. Stark and P. E. de Winter

4. COLLAPSE BY EXTERNAL PRESSURE AND AXIAL FORCE

4.1. Theory
In this section the theoretical model, described in Section 2 and the
appendix, is used to derive an interaction formula for external pressure in
combination with axial tension. The effects of initial out-of-roundness will
be included. For this loading case the equilibrium equations for the rocker
element as given in the appendix can be reduced as follows:
M=0
F=|o,,dA
8pR?B =4fR (1)
The elastic pressure—ovalisation curve (Fig. 6) can be found by combining
the equations for the rocker and bending elements. The same applies for the

[o)
©

(=)
— Sa

S
wu
= ©
q

oO
S)

je)
@

Oo
©
W225 26
- F/Fp : -10%. - 90%
ss A : Ys = 450 N/mm2 —
: Ys = 700 N/mm2
Oo

2 : Ys = 900 N/mm2 —4

eo}

a) 20 40 60 80 400 {20 140 460

collapse pressure in test [N/mm2]

Fig. 8. Tension + external pressure: Nippon test data vs. theory.

 Plastic Design of Submarine Pipelines 297

rigid-plastic pressure—-ovalisation curve. A detailed elaboration for the

‘ring’ situation has been presented in ref. 2.
The upper boundary for p is reached at the intersecting point of the
elastic and the plastic curve. At this point the following equation applies:

(=) Uetancals)eve+Ce P(E)1-8 ©

A graphical example of the upper limit interaction curve is presented in
Fig. 7.

4.2. Experimental Verification

In order to verify the above-mentioned formula a number of tests have been
carried out by Protech International and in the current research
programme. The results are presented in Fig. 7.
A large number of tests on commercial casing have been carried out by
Nippon Steel Corporation. The above-mentioned formula is agreeable
with these results, which are plotted in Fig. 8.

5. COLLAPSE BY EXTERNAL PRESSURE AND BENDING

5.1. Theory
One of the loading cases that occurs in connection with the installation of
subsea pipelines is the load due to external pressure and bending. The pipe
emerges from the pipelaying vessel at a slope (Fig. 9) and has to curve to a
horizontal position in order to settle down on the sea bed. Deformation
associated with this change of direction is imposed upon the pipe. In
connection with this imposed character it is of considerable importance to
know what maximum curvature the pipeline can safely absorb.

installation methods reduction of maximum

curvature by external pressure
Fig. 9. Bending under external pressure.
298 J. W. B. Stark and P. E. de Winter

The maximum curvature of such a subsea pipeline cannot be determined

with an elastic analysis. Long before the critical curvature is reached, large
plastic deformations have taken place. Therefore the elastic analysis will
not be elaborated. The elastic analysis is treated in ref. 2.
The plastic analysis will be limited to a ‘rigid plastic’ analysis. The main
advantage of this simplification is that it substantially reduces the amount
of mathematical effort. A disadvantage is that there is no direct means of
relating stress to strain. Calculation of the deformations will be based upon
applying the normality principle as conceived in the plastic theory.
The following fundamental points have been adopted in the rigid-plastic
analysis:

(1) the external pressure (P) is kept at a constant level, while the
curvature is increased from zero to the critical value;
(2) a three-dimensional yield surface w for bending moment (M),
external pressure (P) and rocker forces (f) is adopted;
(3) the normality principle is used for calculating the critical curvature;
(4) this analysis is performed on pipes with endcaps (hydrostatic
pressure).

The yield function w(M,P,f) represents all loading cases (bending

moment, external pressure and ovalisation force) which cause yielding of
the pipe. In ref. 2 a detailed derivation of the yield surface for a ring
(m, = 0) has been presented. In a similar way the yield surface (Fig. 10) fora
tube (m, =4m,) can be derived:
1 1
W=M-M,(-W,
+ =W,)=0
6 3
ex P 2 if 1/2

Cals
s Ghana
nfo)
wi -a(f)
The yield surface has a boundary for

Pp

In the analysis given here the pressure P is kept at a constant level and the
curvature is increased from zero until collapse of the pipe occurs. This
means that it will be sufficient to consider a straight section through the
 Plastic Design of Submarine Pipelines 299

y = M-Mp (2 Wy» Wz)=0

115 P/Pp

“Zk
a 78 aV3t =1-3(
Fig. 10. Yield surface.

4 M/M>

ee)

t/tp
ftp
Fig. 11. Loading path by constant pressure.

three-dimensional yield surface. In Fig. 11 the loading path is drawn

schematically in the yield surface.
The deformation parameter x is associated with the bending moment M;
the displacement parameter RAP is associated with the ovalisation force f.
Making use of the normality principle, the relation between increase in
deformation force is
OW __0M
canis
eae)
OM
300 J. W. B. Stark and P. E. de Winter

Table 3. Test Results.

Summary Collapse curvature Collapse pressure

*10°°mm K/Ke bar PIP

1 Pure bending 470 14 = =

2 Bending test under
constant external
over-pressure 110 4.5 100 0.50
3 Idem Zi itl 150 One
4 Idem 135 4.2 147 0.57
5 Idem 51 1.6 177 0.69
6 Pure bending 700 DS — —
7 External pressure at
constant curvature 170 5.8 168 0.66
8 Bending test under
constant external
over-pressure 140 4.7 135 0.53
9 Idem; test on thin-
walled pipe 100 3:3 90 0.45
10 Idem 75 2.4 103 0.51
11 External pressure on
initially not exactly
round pipe; out-of-
roundness 2.3% — — 63 OSI

Elimination of the scaling factor 4 and substituting uw=4Rf means

This differential equation represents the relationship between increase of

ovalisation and curvature.
On solving this differential equation and evaluation for point C the
quotient df/dx tends to oo. This means that ovalisation greatly increases
(implosion) while the curvature remains constant.
The analysis can be simplified substantially if the yield contour (the line
BC in the above figure) is straight. Then the ratio dM/dfis a constant, so
that the above differential equation is simply solved.

Oo Re K
Se s-3(FY (4)

Py
 Plastic Design of Submarine Pipelines 301

For point C, where the critical curvature is attained, the following equation
for plastic collapse curvature can be derived:
(ze) to

re"
jo et
t 4\P
Ker = 4/3 55 ‘

The above formula applies to pipes with pressure on the endcaps. For the
‘no endcap’ situation, an analytical solution is not available.

5.2. Experimental Verification

A number of tests on model pipes (D = 100 mm) have been carried out. A
summary of the material properties and the test results are presented in
Tables 2 and 3.
The above-mentioned formula (eqn (5)) applies to the situation with
endcaps. In the test facility used there are no endcaps. It is not possible to
derive an analytical formula for this situation. Therefore a computer
program has been developed, which is still based on the theoretical model.
In Fig. 12 theoretical values from the computer program as well as the test
results for the ‘no endcap’ situation are presented.

bending+external pressure (F=0)

f=)

09 4
0. 2s test results ee
| | r) D/t =25 pe A
07 o Ot =32

8
g
jo)

OoWw

Ooi)

external
CP/Pp]
pressure
——»
oO=

Oo RO 4 6 8 Toe WA SIS)
——+» critical curvature [C/ce]
rigid plastic interaction curve
Fig. 12. Rigid plastic interaction curve.
302 J. W. B. Stark and P. E. de Winter

2 4 6 8 10 12 [C/Ce]
test
Fig. 13. Bending + external pressure: Shell test data vs. theory (formula).

= 260 - 460 N/mnf

lovality = -.0 and Q005

aS ee

6 8 ! 12 [C/Ce]
Fig. 14. Bending + external pressure: Shell test data vs. theory (computer).
 Plastic Design of Submarine Pipelines 303

Shell Development completed an extensive research programme in 1977.

The test results generated in this programme are plotted partially in Fig. 13.
In this figure the interaction formula which is derived in section 5.1 is
compared with the tests.
In the above plot tests in pure bending (P = 0) have been omitted. The
reason for this is that the buckling mode clearly deviates from the collapse
mode that has been assumed for the derivation of the theoretical model.
In Fig. 14 the same test results are compared with the results of the
computer simulation. Contrary to the analytical formula (eqn. (5)), the
computer program allows for elastic deformations and initial out-of-
roundness, while the exact yield contour (curved) is also used.

6. INFLUENCE OF LOADING SEQUENCE

In Section 5 the analysis was based on a loading sequence where first

external pressure is applied and then bending. This idealised sequence
applies to the installation in deep water.
In the test programme, one particular test has been carried out with a
reversed loading sequence (increasing the radial pressure with constant
curvature). Comparing this test to others, a higher pressure curvature ratio
was found. Therefore it could be necessary to review test results and
calculation methods in regard to the loading sequence used.
In this section the theoretical model will be used to establish the influence
of loading sequence.
At zero pressure a pipe is bent until the curvature x, is reached. In Fig. 15

M/Mp

P/Pp

t/fp
Fig. 15. Loading path by constant curvature.
304 J. W. B. Stark and P. E. de Winter

pressure (P/Pp) meiN

‘y “ vi sP

loading sequence «-P

loading sequence P-x

curvature (*/x,)
O 510" 15° *20'" 25°77 30
Critical pressure-curvature combinations
Fig. 16. Critical pressure-curvature combinations.

the loading path is drawn schematically on the yield surface (route O-A-B).
As with other structures, the collapse load here is also path dependent.
The computation of the curvature up to point B is done in the same way
as described in section 5.1 for pressure P = 0 and f=f,. At this point the
curvature is kept at a constant level, while the pressure is increased from
zero until collapse. The path over the yield surface depends on the
equilibrium equation of the rocker element and the normality principle.
With the normality rule the deformation increments on each point of the
yield surface can be established:

The curvature is kept at a constant level, i.e. the curvature increment is

zero:

Since 0/0M #0, the scaling factor 4 has to be equal to zero. Proceeding
 Plastic Design of Submarine Pipelines 305

over the yield surface, while the curvature is kept at a constant level, the
other deformations are also maintained constant.
b= k= A= 0
On the boundary of the yield surface the ultimate pressure is reached
(point C).
The ultimate pressure in point C can be calculated using the
deformations f and x, attained in point B. In Fig. 16 the computed
instability curves for xP and P-« are plotted together. In this particular
case there is considerable influence (D/t = 25, ¢, = 320 N/mm’).

7. CONCLUSIONS

The results may be summarised as follows:

—Because of the simplicity of the calculation method, designers are able

to gain a quick insight into the influence of pipe characteristics on
collapse due to combined loading.
—A simple analytical relationship can be derived for the interaction
between external pressure and axial force, including the effects of
initial out-of-roundness.
—The curvature under external pressure and bending is adequately
predicted by the theory presented. The results of the computer
calculation (Fig. 14) are somewhat closer to the test results than the
analytical solution (Fig. 13). This is caused by the numerical solution
procedure and the inclusion of elastic bending. For relatively high
external pressure the influence of initial out-of-roundness on the
magnitude of the curvature at collapse is considerable.
—It has been demonstrated that the loading sequence influences the
deformation capacity. This is important for the evaluation of test
results and the use of computer calculations.

Acknowledgement
This contribution forms a part of the Marine Technological Research
Programme sponsored by the Dutch Government. The work is carried out
jointly by the Institute for Building Materials and Building Structures of
the Netherlands Organisation for Applied Scientific Research (IBBC-
TNO), Protech International and the Universities of Delft and Eindhoven.
The authors wish to thank Nippon Steel Corporation and Shell SIPM for
making their test data available to us.
306 J. W. B. Stark and P. E. de Winter

&q0:dx

SS,
axial deformation due to axial force

curvature due to bending moments

Fig. 17. Deformations.

Fig. 18. Deformation of the rocker element.

 Plastic Design of Submarine Pipelines 307

APPENDIX

For the kinematic model, three degrees of freedom will be considered (see
ig Wy:

—ovalisation (f)
—axial deformation (¢,0)
—curvature (k)

The external loads considered are:

—external pressure (P)

—axial force (F)
—bending moment (M)

Basic assumptions are: Bernoulli’s hypothesis is valid and the cross-section

remains perpendicular to the axis of the tube. For the sake of simplicity,
formulas are linearised at an early stage and the ovalisation angle PB is
considered to be small.

Rocker element
Geometry (see Fig. 18)

undeflected: z = R sin 0 (6)

deflected: z + u, = R(sin(0 + f) — sin B) (7)
linearised: z + u, = z(1 — B) (8)
Displacements along the axes:

u, = R(cos B — sin B — 1) (9)

u, = R(cos
B + sin B — 1) (10)
Linearised:
u,=—Kp 1° variation du, = — Rdp (11)
UTD 1° variation du, = Rop (12)
Area change of cross-section:
AA =4R? sin? B (13)
308 J. W. B. Stark and P. E. de Winter

Linearised:
AA =4R’B? 1° variation 6AA = 8R?BdB (14)

Kinematic equations
Strain in axial direction:

Eaa = E40 + K(Z + Uz) = Eo + KZ(1 — B) (15)

1° variation d¢,, = z(1 — B)dx — Kzd6B + €,0 (16)

Static equations
Equilibrium of the cross-section is expressed using virtual work theory (see
Figs 17):
Jo,.66,,4dA = MOx + Foe,)
+ POAA — 4fR6B (17)
After substitution of the variations derived above, an equation is found
which is valid for random variation of the three degrees of freedom.
Therefore, three equilibrium equations may be written:
M =\{o,,2(1 — B)dA (18)
F=\q.dA (19)
8PR?B = —|o,,kzdA +4fR (20)
The first two equations can be easily recognised. The external bending
moment (M) equals the resulting bending moment caused by axial stresses
in the deformed model. The same applies for the axial force (F). The third
equation shows the connection between both external pressure and bending
with the ovalisation of the tube.

Bending element
The bending element is loaded only by those (opposite) contact forces (/)
which act on the rocker element. Stiffness of the bending element is
nevertheless influenced by the load acting on the rocker element. The
starting point is a linear analysis in which the displacements on the
horizontal and vertical axes are of equal magnitude.

Static equation
Because of symmetry it is sufficient to consider the equilibrium of a quarter
of the cross-section (see Fig. 19):

(ae (21)
 Plastic Design of Submarine Pipelines 309

US
iA
Mr

Mr, Ve

ly,
Fig. 19. Equilibrium of the bending element.

Deformation of the bending element

For elastic deformation only, the following relationship between diameter
change and bending in the circumferential direction may be derived:*

R(B — Bo) =u, = u,=

oe IE] ( )

For rigid plastic material behaviour, a simplified yield function is used. The
yield function describes the interaction between bi-axial bending and
extension.
On the assumption of three zones in an elementary block of material,
each satisfying the Von Mises yield condition, the following equation can
be derived:

Cl rastal ealaarss te
2 2 2 2 2

m, mM, mM, ny Ny ny

where
m, = 1/4176, ne = 10; n,=—PR
From the tests observed and from the collapse mode in the theoretical
model, it is clear that the change of curvature in the pipe wall material in the
axial direction is negligibly small in comparison to the curvature change in
the circumferential direction.
K, <K,
In the further analysis it is assumed that
K, =0
310 J. W. B. Stark and P. E. de Winter

Fig. 20. Yield condition.

The yield condition (Fig. 20) can be substantially simplified now. x, =0

means that there is a fixed relationship between the plate bending moments
m, and m,.
In the yield point x, = 0 for any arbitrary value of c the following relation
holds:
re
m,= 7M,

vah-EJ-QPEII so
The bending moment in the circumferential direction now becomes

The normality rule is applied for deriving the deformation in the plastic
area (see interaction between bending and external pressure). Based on the
relationship presented above, the interaction between bending, external
pressure and axial force can be derived.

REFERENCES

1. pE Winter, P. E., ‘Deformation capacity of steel tubes in deep water’,

Proceedings 13th Annual Off-shore Technology Conference, Houston, 1981,
Paper 4035.
 Plastic Design of Submarine Pipelines 311

. DE WINTER, P. E., ‘A method of analysis for collapse of submarine pipelines’,

Proceedings of the 3rd International Conference on the Behaviour of Offshore
Structures (BOSS), Boston, 1982.
. Haacsma, S.C. and Scuaap, D., Collapse resistance of submarine lines studied,
Oil and Gas Journal, No. 2 (1981) 86-95.
. TIMOSHENKO, S. P. and Gere, J. M., Theory of elastic stability, Tokyo,
McGraw-Hill Kogakusha, 1961.
. VROUWENVELDER, A. and WITTEVEEN, J., Lower bound approximation for
elastic buckling loads, HERON, 20, No. 4, Delft, 1975.
. Nippon Steel Corporation, ‘Summary ofcollapse test results under axial stress’,
API Task Group on Performance Properties, Bal Harbour, Florida, 1981.
. Murpui, C. E., ‘Ultimate pipe strength under bending collapse and fatigue’,
Shell Development Company, Houston, 1975.
 ee sn 1Maeeie
2
pans ome’
_ bees 7 ape Fy "i
2 ae
. 8 ail oeevo ‘ yy eas
 16

The Inelastic Response of Pipes Under External

Pressure and Bending
CeIeTAY
National University of Singapore, Singapore

W. J. M. STEEL
Britoil PLC, Glasgow, UK

and

J. SPENCE
University of Strathclyde, Glasgow, UK

ABSTRACT

Subsea pipelines for oil or gas transportation are being installed in

progressively deeper water locations. Inherent in the normal pipe laying
procedures is a pipe configuration which induces bending type loading in the
transition region between the position of the pipe on the sea bed and the
suspended position of the line. In deep water the combination of bending with
the external pressure loading due to water depth may give rise to plastic
buckling of the pipe. Part of the present project developed a simplified elastic-
plastic analysis for the response of a pipe under combined loading conditions
with the assumption of linear work hardening material. The theoretical
analysis is described and results presented for various geometric material
and load configurations.

1. INTRODUCTION

In the early 1970s techniques had been developed for routinely laying
offshore pipelines in water depths up to about 122m. However, the
capability has since been extended as the search for crude oil and natural
gas has moved into deeper water. Examples include a line in 265 m of water
off the California Coast, the planned construction of a line across the
313
314 C. J. Tay, W. J. M. Steel and J. Spence

Norwegian Trench at a maximum depth of 305m and three lines from

Tunisia to Sicily in water depths of nearly 710m. The trend towards ever
deeper water is expected to continue. Indeed a study has indicated the
feasibility of aMediterranean crossing in water 2000 m deep while areas off
the east coast of the United States (water depths up to 2140 m) have recently
been leased for exploration.
Reports have, however, suggested that the technology does not exist for
repairing pipelines in such deep water. It has been suggested that it is not
feasible to repair the lines already installed in 710m of water. Thus
although the problems associated with the laying operations increase
significantly as the water depth increases, it becomes even more important
to avoid damage to the line.
During the construction of submarine pipelines either by the
conventional or new generation semi-submersible laybarge, stresses are
developed as a result of the pipe configuration (see Fig. 1). As the figure
shows, the pipe is bent into an elongated S-curve as it is being laid. The
overbend is supported by rollers on the barge and stinger while, beyond the
stinger, the pipe is freely suspended to the ocean floor in what is called
the sagbend.
In the overbend, stresses are induced as a result of the curvature, the
stinger and the pipe tensioner, although the curvature is the predominant
factor. In the sagbend, however, in addition to the bending stresses induced
by the curvature, external hydrostatic pressure due to the water depth also
produces circumferential stresses in the pipe wall. The stresses that are
developed may be above the elastic limit of the material and given sufficient
curvature the pipe will buckle plastically.

tensioner

barge“ y
\ stinger anchor cable

\
pipeline

Fig. 1. Schematic diagram of laybarge pipelining method.

 Inelastic Response of Pipes 315

The buckle, once initiated, may propagate along the pipe under the
influence of the pressure alone even into regions free of bending. The
financial and time penalties resulting from such a buckle are severe. It is
therefore of paramount importance to have a comprehensive understand-
ing of the bending/pressure buckling behaviour to ensure the integrity of
offshore pipelines during the laying process.
Previously a great many investigations had been carried out on the
buckling problems of cylinders under different loading conditions. The
contributions of Ades,! Seide and Weingarten,” Lee,? Wilhoit* and
Berdnikov and Nozdrin° reported work being carried out on pure bending
of cylindrical shells. Others such as Stokey er al.,° Gerdeen et al.’ and
Larson® reported work on pipes subjected to combined internal pressure,
bending, axial load and torsion. However, relatively little attention has
been focused on the plastic behaviour of pipes under combined bending
and external pressure.
In recent years experimental work on the buckling of pipes under
combined bending and external pressure has been reported by Johns et al.?
on 51mm diameter pipes of an unspecified grade of steel, and by
Broussard!° on 152mm pipes with a d/t ratio of 24. The pipeline material
was unfortunately not specified and also the detailed development of the
theoretical analysis was not given.
An experimental programme on the bending/pressure behaviour and
buckle propagation has been undertaken by the present authors under the
sponsorship of BP (Trading) Ltd. The test facility for this investigation
comprised a pressure vessel together with suitable instrumentation and the
supporting systems for externally pressurising and/or bending the pipes.

Fig. 2. Buckied pipe.

316 C. J. Tay, W. J. M. Steel and J. Spence

The pressure vessel, which was 1.07 m in diameter and 6.7 m long, allowed
the model pipes (approximately 1/5 scale) to be subjected to external
pressure while being simultaneously subjected to bending by means of two
external 200kN hydraulic rams. The rams were mounted on an external
test rig and acted through glanded nozzles on top of the vessel.
In addition, a number of small scale tests were conducted on mild steel
pipes using the test facility. This background stimulated the theoretical
work outlined herein. The detailed results of the test programme are not
reported here. Theoretical results of the work on some aspects of buckle
propagation and sleeve arrestors is reported separately.'? For interest a
typical buckled cross-section is shown in Fig. 2.

2. THEORETICAL ANALYSIS

As discussed above, the stresses developed during the pipelaying operation

are predominantly caused by the combined loading of bending and
external hydrostatic pressure. The open literature has provided little
information on the complete load—-deformation relationship for pipes
under such loading conditions. As a knowledge of this relationship is of
fundamental importance in designing against pipe collapse during
pipelaying, the present theoretical investigation seeks to obtain the
moment-strain response for constant external pressure, based on a
simplified plasticity theory.

Fig. 3. Pipe coordinate system and stresses.

 Inelastic Response of Pipes 317

€, €z

-€;5 mes

pure bending combined bending

and external pressure

Fig. 4. Typical axial strain distributions.

2.1. Assumptions and Coordinate System

The pipe geometry and loading considered in this analysis are shown in
Fig. 3 while Fig. 4 illustrates typical axial strain distributions for a pipe
under pipe bending or simultaneously subjected to bending and external
pressure.
The following basic assumptions are made:

(i) Plane cross-sections remain plane.

(ii) Stresses and strains are invariant in the axial direction.
(iii) Stresses are taken to be uniform across the thickness of the pipe.
For the radial stress, an average value equal to one half of the
external pressure is assumed (a common assumption in pipe
analysis).
(iv) The pipe material has a bi-linear stress-strain relationship.
Specifically, the material behaviour is assumed to be linear and
elastic up to the yield stress but subsequently strain hardening
occurs at a constant rate. The post-yield behaviour is then
characterised by the tangent modulus (see Fig. 5).

2.2. Equilibrium Considerations

For a thin cylinder we have

pr P
C= = Uta (1)
318 C. J. Tay, W. J. M. Steel and J. Spence

where o, and o, are the averaged circumferential and radial stress

respectively, p is the external pressure and r and ¢ are the radius and wall
thickness of the pipe, respectively.
Also, the external applied moment, M, may be written as
+ 2/2

M =2tr? | o, sin 0d@ (2)

=o

where go, is the axial stress.

2.3. Compatibility
From assumption (i), the axial strain may be expressed as

Blige pon is Deets IT

o = psing +2 >(") (F+5)| (3)

where R is the radius of curvature, E is the elastic modulus and vy is

Poisson’s ratio.

2.4. Elastic Considerations

In the elastic region, the stresses and strains are related by

c= glo,— vo, + 20) (4a)

i= glor— M0, +0.) (4b)
f= Zl, — Wo +2.) (4c)
Elastically the axial stress is simply
_Msin@_ pr
S> Merriam 2) (5)
where the terms denote the contribution to the stress from bending and
external pressure, respectively.

2.5. Plastic Considerations

For material that is plastically deformed, it is convenient to assume that the
total strain may be taken as the sum of the elastic and plastic strains, i.e.

e, =e tee (6)
 Inelastic Response of Pipes 319

Stress

:
bi-linear idealisation

tan” E,
\ actual

Strain

Fig. 5. Idealised stress-strain graph.

The plastic strain behaviour is assumed to be governed by the

Prandtl-Reuss relationship’* so that the plastic strain increment at any
instant is proportional to the instantaneous deviatoric stress,

de?
Zod,
oO Zz
(7)

where d/ is a non-negative constant which may vary throughout the

loading history. The deviatoric, or reduced, stress is defined as
1 On ar Oy a Oy
OO he
3
= 319, — 249+ 9,)] (8)
and hence eqn (7) becomes

de? =3[0, —H(a,+0,)] dA

Zz (9)

It can be shown that dJ may be written as

Oi (10)
320 C. J. Tay, W. J. M. Steel and J. Spence

where 4G, the effective stress, is defined as

6 =[0? +07 + 62 — 4,0, — 0,0, —60,]'!” (11)

and according to the Von Mises yield criterion yielding occurs when
=,

where g, is the yield stress of the material. In eqn (10) dé? is the effective
plastic strain increment and may be expressed, in terms of the strains, ina
similar form to eqn (11).
A relationship between effective stress and plastic strain may be obtained
from a simple uniaxial test:

o = g(e") (12)
As stated earlier, the material is assumed to have a bi-linear constitutive
relationship which, from Fig. 5, is

Eae er :
ars GSB,

relVS laoS
edOe
5 eye be
aes
(13)
=é* +e?
Then

ag — 109 — %)

‘ (14)
_ dé
a

Assuming that the radial and circumferential stresses remain constant

during bending (which is reasonably valid if the pressure is constant) the
axial elastic strain increment is, from eqn (4a),

d
dee =
E
ue (15)

On substitution of eqns (10) and (14) into eqn (9) we have

dep = 22 260
0
+ &)] do
Ey
(16)
 Inelastic Response of Pipes 321

and thus the total axial strain is

de! = det + deP (17)

or
o,—+(6,+0,)d6 1
d
gl |
—_ Zz 2
eae
6 r
ae 5 |
se.
(18)
From eqn (11)
dg «o,—4(o,4+4,)
7 ni aa ie
Substitution of this expression with eqn (11) into eqn (18), rearranged in a
suitable form for integration, gives
1 A 1
de, =| —| 1 —|d 20
ve Fal “2 —(a, ae a Ye oo)
where
A =3(4,05—3(0; + 98)]
B=07
+065 -6,64
and the superscript ‘T has-been omitted for convenience. On integrating
eqn (20) we obtain

1 1 A _,| 6, —(6, + 6,)/2

fs = alate) tae ila ne
| aman C Rat |+D Dil
Ce
where
C = [3(o? + 4,)* — 40,0,]1/”
The value of the constant D may be found using the condition that when
yielding first occurs (¢ = o,) the expressions for pre-yield and post-yield
strains must be continuous. Thus from eqns (4a) and (21)

ae [o, — v(o, + o)]

é iad os
There A _,[o,—(6,+4,)/2
=o alate
ee tans an |
bag |C + iD 2D
(22)

which identifies the constant as

eee
jon 4) ] -a|; at -|- és on |
[o, — v(o, +
by (Per eC Es
322 C. J. Tay, W. J. M. Steel and J. Spence

When a = 9, the value of o, is obtained from eqn (11) while o, and a, are
given by eqns (1).
In the case of pure bending, the external pressure is zero and thus
eqn (21) becomes simply

ana(ste)-%
=
1
—+—)]-—
1 oO
(23)
23

The form of the equations developed above permits a solution to be

obtained for specific pipe dimensions and material properties if desired. To
allow general results to be obtained from the theoretical analysis the
variables in the equations may be non-dimensionalised.
The bending moment acting alone to produce first yield in the pipe is

M,= nr? to, (24)

Hence eqn (3) may be re-written as

M 2
—,— =—JS,sin9dé (25)
REO SH
where

saeoO Ox

The axial strain in the elastic and plastic regions of the pipe, eqns (4a) and
(21) respectively, may also be non-dimensionalised as follows:

= [S,— WS, + 5,)] (26)

cay E (A/o;) S Sy —S.aS, 2
Pau (. a) * (Co, (Ey/ENE/o,) | (Clo) |D
(27)
with the obvious corresponding non-dimensionalisation of D and where

The solution procedure using these equations will be outlined in the

following section.
 Inelastic Response of Pipes 323

2.6. Computational Procedure

The method of determining the axial stress at any point in the pipe, and
hence the overall bending moment, is briefly described below.
For particular values of the external pressure (p/o,), pipe material and
geometric parameters (dy/E, r/t) are selected and for some given curvature
the axial strain at any position on the cross-section may be found from
eqn (3). The axial stress was then obtained by solving either using eqn (25)
or (27) depending on the strain level. In a tube which has only partially
yielded, expressions for the stresses in both the elastic and plastic ranges are
required. In the computer program used in the present analysis, an iterative
loop was provided to decide which expression to use.
Knowing the stresses at various points on the cross-section allowed the
bending moment to be calculated from eqn (25). Repeating the above
procedure for increasing values of curvature resulted in the moment-
deformation curve at the particular pressure.
The theoretical development here is essentially similar to that of Franzen
and Stokey.'* However, they approximated the stress-strain curve by a
Ramberg—Osgood power law. This leads to an equation which is more
complex than eqn (27) and requires numerical procedures for solution. By
adopting a bilinear idealisation to the material behaviour, it is possible to
integrate and obtain the solution directly. In practice the bilinear
idealisation has been found to provide a reasonable approximation for
pipeline grade tube material. Moreover it was reported that the assumption
of a power law led to difficulties in the numerical solution of the problem,
expecially for large values of the index.

3. RESULTS

Typical results obtained from the analysis are shown in Figs. 6-8. As
mentioned above, the material of interest for pipeline evaluation is typically
API 5L Grade X65 which has a nominal yield stress of 450 MN/m?
(65 000 Ibs/in?). The dimensionless parameters in Figs. 6-8 were therefore
chosen relative to this yield stress and a Young’s modulus of 210 MN/m?
(30 x 10° lbs/in?). If these basic material properties apply, Fig. 6 then
shows the effect of varying the tube size when the external pressure is
3.45 MN/m? (500lbs/in*) and the tangent modulus is 3.45 GN/m?
(0.5 x 10° Ibs/in?). This value of the tangent modulus was judged to be
reasonably representative of Grade X65 material. For one particular size
(r/t =15), Fig. 7 illustrates the variation in the moment versus strain
relationship resulting from two different external pressures. For the
324 C. J. Tay, W. J. M. Steel and J. Spence

M Oy
p = 130 Ee 60
9 == 3 & 461-5fia
, ==
O “

1:0

0-75

05

0:25

0 2 4 6 8 10 12
Extreme fibre strain x10"
Fig. 6. Moment vs. extreme fibre strain.

nominal yield stress quoted these would be equivalent to 3.45 MN/m?

(500 lbs/in?) and 13.8 MN/m? (2000 lbs/in”). The moduli are the same as in
Fig. 6.
The method of determining the tangent modulus from the stress-strain
curve is somewhat arbitrary and in some cases slightly different
approximations can be made to the same curve. Figure 8 was therefore
plotted to indicate the effect of changing the modulus by a factor of 4. Fora
Young’s modulus of 210MN/m? (30 x 10° Ibs/in), the ratios on the graph
correspond to tangent moduli of 6.9 MN/m? (1 x 10°lbs/in?) and
1.725 GN/m? (0.25 x 10° lbs/in) respectively.
All of the above curves are of a similar form to the stress—strain curve of a
simple tensile test and represent the typical shape for pipes subjected to
combined bending and external pressure. The collapse moment is obtained
when the curve approaches a horizontal tangent, i.e. when the increase in
moment is less than a specified value for a given increase in strain.
 Inelastic Response of Pipes 325

= 461-5

10

0-75

0-5

0-25

0 Z 4 6 8 10 12
Extreme fibre strain x10°
Fig. 7. Moment vs. extreme fibre strain.

4. CONCLUSIONS

A theoretical analysis has been developed for the collapse moment of a

subsea pipe which included the effect of material strain hardening.
Moment-curvature relationships were obtained from this analysis and
examples have been given to indicate the typical variation with each of the
main parameters. From various comparisons with experimental results not
detailed here it can be said that the analysis provides a reasonable
indication of the maximum moment to which a tube could be subjected.
However, it provides no means of determining the extreme fibre strain or
curvature at buckling. The buckling or collapse curvature would be more
useful for preventing failure during the pipelaying process as it is easier to
estimate and control. This limitation of the analysis is a consequence of the
assumption of small displacement behaviour of the cross-section during
326 C. J. Tay, W. J. M. Steel and J. Spence

v te
petri Dy = 130 lar
ago =Ors a
ul

0-75

05

025

0 2 4 6 8 10 12
Extreme fibre strain x10>
Fig. 8. Moment vs. extreme fibre strain.

bending which results in the bending moment continually increasing with

increasing strain. However, if large displacements in the cross-section were
allowed the moment versus strain relationship would indicate the buckling
load and curvature and give a good indication of failure of the tube.
Further development of the analysis in this direction would therefore be
useful.

REFERENCES

1. Apes, C. S., Bending strength of tubing in the plastic range, J. Aero. Scis., 24
(1957), 605-10.
2. SEIDE, P. and WEINGARTEN, V. I., On the buckling of circular cylindrical shells
under pure bending, Trans ASME, 83, Series E, J. of Applied Mechanics (1961)
112-16.
 Inelastic Response of Pipes 327

. Leg, I. H., ‘Buckling and ovaling of pipeline’, M.S. Thesis, Rice University,
Houston, Texas, May 1971.
. Wituoit, J. C. JR, MERwIN, J. E. and Jirsa, J. O., Critical plastic buckling
parameters for pipes in pure bending, ASME Publication paper, 72-pet-29,
1972.
. BERDNIKOV, Y. N. and Nozprin, A. S., Stability of cylindrical shell in the
elasto-plastic region with pure bending (yield theory), Soviet Aeronaut., 17,
No. 4 (1974) 11-17.
. STOKEY, W. F., PETERSON, D. B. and WunpkerR, R. A., Limit loads for tubes
under internal pressure, bending moment, axial force and torsion, Nuclear
Engineering and Design, 4 (1966) 193-201.
. GERDEEN, J. C., SIMORREN, F. A. and Hunter, D. T., Large deflection analysis
of elastic-plastic shells using numerical integration, AJAA Journal, 9, No. 6
(June 1971).
. Larson, L. D., ‘Inelastic response of pressurised tubes under dynamic bending
and torsional loads’, PhD thesis, Mech. Eng. Dept., Carnegie—Mellon
University, 1973, Microfilm order No. 73-22872.
. JOHNS, T. G., MESLOH, R. E., WINEGARDNER, R. and SORENSON, J. E., ‘Inelastic
buckling of pipelines under combined loads’, OTC paper No. 2209, 1975.
. BROUSSARD, D. E., Deepwater pipelaying research under way, Oil and Gas
Journal (June 1975).
. Tay, C. J., ‘Inelastic collapse of pipes under external pressure and bending’,
Doctoral dissertation, University of Strathclyde, June 1979.
. STEEL, W.J.M.and Spence, J., On propagating buckles and their arrest in sub-
sea pipelines, Proc. I. Mech. E. London, 197a (1983) 139-47.
. FRANZEN, W. E. and STokey, W. F. ‘The elastic-plastic behaviour of stainless
steel tubing subjected to bending, pressure and torsion’, 2nd Int. Conf. on
Pressure Vessel Technology, Part 1, Design and Analysis, 1973, pp. 457-67.
. JOHNSON, W. and MELLoR, P. B., Plasticity for mechanical engineers, New
York, Van Nostrand, 1962.
. Hopce, P. G. and ROMANO, F., Deformations of an elastic-plastic cylindrical
shell with linear strain hardening, J. Mech. and Phys. of Solids, 4(1956) 145-61.
 ae eee
F Af tay
eye #7 Sognkion
ee a
gee meteooi valeer as ‘ "
on aw a yet vet rea vein
wa i ore ery a¥ aa “9 W wwett..o vw eS
- s ins keg emai DOLE ,crea nies Sipe cath ty amt t 4
; Vets As ey dvinwel Gee cme wryaatSs, 9 a) -ereae oe Yi
4 4
413 maf — Teve6 ama) .7
ied gay: tri, (meri Hew waitbi ptr ethan Jb 73 eet
: la are mit onjbtint hy ees) ae) Vier
i reed Re orean?, A4RSA2st «4 As d,
ae : ik Ai) gO wee) YW iatkh A &
or chien’ a ined A Pig Liha ae pines 2 a ae
Bt, Sow! [A aeety gaaeS reTa ace
ah SOF! epygh re Nye).Lc eet 98)er ord lesen ted
ames tee, in” “tind i

my Ye ities ae ; H
ies ae
oprk Anu ne
; hbhaseaace a me ys che. AnLe aa
Rees
tadoliakg atin SD % gia :
Lites s wy aes ie

: - : =P
® , I el rr. =

She iy & _ om

o_o v

4% + lewes

. by rp
 17

The Strength of Plates Subjected to Biaxial Forces

A. F. Dier and P. J. DOWLING

Imperial College, London, UK

ABSTRACT

A numerical procedure to analyse the elasto-plastic large deflection

behaviour of plates is briefly described. The main results of a parametric
study on plates under biaxially applied in-plane compressive and tensile
forces are presented. The effects of aspect ratio, slenderness and initial
imperfections (consisting of geometric deformations and residual stresses)
are investigated. It is shown that the numerical results are inagreement with
experimental results. Inconsistencies,within the relevant rules of thenew UK
steel bridge code, BS 5400 Part 3, are identified and suggestions are made
which make the rules safer and are simple to apply.

NOTATION

a Length of plate.
b Breadth of plate.
t Thickness of plate.
E Young’s modulus.
K, Design coefficient for longitudinal direction.
K, Design coefficient for transverse direction.
4 Design coefficient.
B Non-dimensional plate slenderness (=6/t,/09/E).
69, Geometric imperfection amplitude of a single half-wave component.
329
330 A. F. Dier and P. J. Dowling

693 Geometric imperfection amplitude of a three half-wave component.

& Uniaxial yield strain.
E Average longitudinal strain.
g Average transverse strain.
7 Design coefficient.
o, Uniaxial yield stress.
Average longitudinal stress.
Average transverse stress.
Maximum average longitudinal stress when o, = 0.
Maximum average transverse stress when o, = 0.
Average compressive residual stress.

1. INTRODUCTION

Since the mid-nineteenth century plates have been used extensively in steel
structures as elements participating in the primary load carrying action.
The first solution to the linear plate buckling equation for a simply
supported plate under edge compression was due to Bryan’ in 1891. Since
then many solutions, accounting for different boundary conditions, types
of loading, etc., have been found and a large number of these are collected
together by Bulson.? For stocky plates the elastic critical stress lies above
the yield stress but sensible results can be obtained by using a reduced value
of Young’s modulus in the critical stress expressions. During the 1940s the
deformation (history independent) theory of plasticity was used by workers
such a Bijlaard,* Ilyushin* and Stowell® to evaluate the reduced tangent
modulus. The more correct incremental flow theory of plasticity was first
applied by Handelman and Prager.°®
Further knowledge of elastic plate behaviour was obtained by solving the
large deflection equations of von Karman which can account for the post-
buckling reserve of slender plates. Among the notable early workers in this
area are Levy,’ Coan,® Hu et al.? and, more recently, Yamaki.!° These
‘classical methods’ gave way to less exact and numerical methods such as
Rayleigh-Ritz, finite differences, finite elements, etc. The perturbation
approach'! has been used to find design coefficients for uniaxially and
biaxially compressed plates.'* By assuming a plate fails when the midside
edge stress reaches yield an estimate of the load carrying capacity of the
plate may be found. Much of the information obtained from the above
analyses and other works, including experimental, has been used to derive
effective width expressions for plates in compression. Faulkner!? has given
 The Strength of Plates Subjected to Biaxial Forces 331

a comprehensive review of these expressions and Rhodes!* has also

discussed their use in design.
The collapse of four box girder bridges in the late 1960s prompted a great
research effort into the behaviour of plated structures. One of the
developments of that effort is the rise of powerful analytical tools to
examine the large deflection elasto-plastic behaviour of plates—see refs.
15-19, for example. This was made possible by the parallel development of
large digital computers which are necessary as solutions can only be found
by the numerical methods, a closed analytical solution not being possible.
It has been shown’° that these tools predict similar ultimate loads,
although there are differences in the slopes of unloading curves (depending
on the plasticity model used). The computer programs have been used for
several parametric studies on the ultimate behaviour of plates,”!~ 7° the
results of which are being used to-rationalise aspects of design codes.
An area which has received relatively little attention is the behaviour of
plates under biaxial in-plane loads. The linear elastic solution,” by which
the critical stress may be calculated, has been known for some time. A
knowledge of the likely buckling modes can be inferred from this solution.
For example, a plate of aspect ratio a/b =3 will buckle into three half-
waves for longitudinal compression, a single half-wave for transverse
compression, and into two half-waves for a small intermediate region. The
post-buckling behaviour has been investigated by Williams and Walker,'*
although for only two ratios of the transverse-to-longitudinal stress. The
earliest reported work on biaxially loaded plates including the effects of
plasticity is due to Coombs.”’ He incorporated ‘initial deformations, but
not residual stresses, in his analysis which was based on the approach
explained in the next section. Valsgard?* has also studied biaxial
compressive strength but again residual stresses were not included
implicitly although some account is taken of them by a method proposed
by Crisfield.”®
This work is a summary of a parametric study on biaxially stressed
plates,?°?? although additional computer runs were undertaken to extend
the range of applicability of the results.

2. METHOD OF ANALYSIS

The analysis takes account of non-linear behaviour due to both large

deflections and yielding. It is based on a finite difference approach in which
the large deflection equilibrium equations of von Karman, modified by
332 A. F. Dier and P. J. Dowling

Marguerre®° to include initial geometric imperfections, are expressed in

their finite difference equivalents at each mesh point. The finite difference
equations are solved iteratively by dynamic relaxation which has a physical
analogy of following the dynamic response of a damped system set
oscillating by the application of load.
During the dynamic relaxation iterative loop, boundary conditions are
satisfied by the use of fictitious mesh points lying outside the plate
boundary. Convergence to the static solution is accelerated by using
fictitious plate densities and near critical values of damping factors. To
maintain stability over the maximum load, loading is effected by applying
edge displacements.
With the onset of yielding, the edge displacements have to be applied
incrementally to ensure non-violation of the incremental approach to
plasticity. At the end of each load increment the strains and stresses are
calculated at seven layers through the plate thickness for each finite
difference mesh point. Using Prandtl—Reuss flow rules, the layer elasto-
plastic rigidities are calculated when the layer stresses (substituted in the
von Mises yield criterion) show that yielding is taking place or where, if the
node was previously yielding, the stress increments are such that the stress
increment vector is directed outwards from the yield surface. Otherwise
elastic rigidities are used. When all the layer elasto-plastic rigidities have
been updated the plate rigidities are found by quadrature. The revised
plate rigidities are used in the next load increment. Numerical trials have
indicated little difference in results whether or not the stresses are corrected
back to the yield surface where penetration has occurred. Typically forty
load increments are used for one program run.
The mesh spacing used for square plates is a/9 in both directions. This
spacing has been shown*! to be fine enough to give results of sufficient
accuracy. For other aspect ratios the number of meshes in the longitudinal
direction is increased such that the spacing remains approximately the
same as that for square plates. Advantage is taken of symmetry to analyse
only one quarter of the plate.
In all cases the plates are assumed to be ee supported, i.e. there are
no rotational restraints at the plate edges and out-of-plane movement is
prevented. The plates are also assumed to have no tangential edge restraint,
thus there are no applied edge shear stresses. Two boundary conditions for
the normal in-plane direction have been used. First, and most commonly
used, the edge displacement is predetermined and the corresponding stress
distribution is found in the iterative loop. Second, the edge is again kept
straight but is allowed to move in or out such that the net average stress is
 The Strength of Plates Subjected to Biaxial Forces 333

zero. These boundary conditions will be referred to as restrained and

constrained respectively.
Further details of the analytical approach and the implementation of
boundary conditions etc. may be found in refs. 17 and 18.

3. NUMERICAL RESULTS

3.1. Parameters Examined

The material properties, slenderness and aspect ratio of a plate are the main
factors determining the biaxial strength. It is now well established that, for
similar shapes of the material stress—strain curve, the material properties
(c,, EZ) are suitably accounted for by using appropriate non-dimensional
factors. Thus, although the following numerical results were generated by
taking the yield stress, oj, and Young’s modulus, E, as 245 N/mm? and
205 000 N/mm? respectively, i.e. values appropriate for mild steel (m.s.),
the results can be applied to plates fabricated from different steels if they
have the same aspect ratio, slenderness (in terms of #) and imperfections
(Off O2/05):
Plates of slendernesses up to b/t=110 (m.s.) have been examined,
corresponding to £ = 3.80. Aspect ratios of 1 and 3 have been used for the
major part of the study, although others have been used to examine the
variation of transverse strength with aspect ratio.
All plates have initial imperfections consisting of both geometric
deformations and residual stresses. For biaxially loaded square plates a
half-wave sinusoidal-shaped geometric imperfection of magnitude
6,/t =0.1f? has been used, while for an aspect ratio of 3 an additional three
half-wave component of magnitude 6,3/t = 0.05? is considered. In the
transverse strength study only the single half-wave component is used for
all aspect ratios.
The geometric imperfection is seen to be dependent on the square of the
slenderness whereas in most parametric studies and Codes of Practice it is
linearly related. A quadratic relationship has been used here as
measurements have shown that this is more typical*? of ships for which
application this study was initially intended. For the purposes of examining
the effect of imperfections, plates with three times and one quarter of the
above stated magnitudes have been investigated.
A biaxial residual stress field is assumed to be present in all plates. In the
standard case the average longitudinal compressive stress is taken as 0.20,
and twice or a quarter of this is taken for the imperfection study.
 334

he
q/0 = €
SOIl}DJ
We. = ave CS)

A3:x3
yoy uOlyDUD|dxa
JO
SaAUND
Y PpuDG 22S 1X2]
A. F. Dier and P. J. Dowling

ON
oe

poy
JO UOONIWSUOD
JO UOTOVIIUT
"SAINI

‘|
3q
 The Strength of Plates Subjected to Biaxial Forces 335

Transversely, the width of the tensile residual stress block is taken as half
that of the longitudinal block for 3:1 plates and for plates in the transverse
strength study. However, in the investigation of square plate biaxial
strength, the widths of the tensile blocks are taken to be the same in both
directions. This makes the x and y directions indistinguishable for a square
plate and thus only approximately half the number of program runs are
needed to generate interaction curves leading to savings in computer costs.
Due to out-of-plane components of the residual stresses the plate is not
initially in equilibrium. Before the plate is loaded it is allowed to find its
equilibrium configuration using the constrained edge condition. It is
generally found that the out-of-plane components magnify the geometric
imperfection and so the initial magnitudes are factored down and a second
equilibrium configuration found. If the resulting magnitude is now close to
the desired value, loading can begin; otherwise the process is repeated until
it does become acceptable.

3.2. Generation of Interaction Curves

As mentioned in Section 2, loading is accomplished by applying edge
displacements. In any one program run the ratio of the transverse to the
longitudinal displacement is kept constant. However, the resulting ratio of
the average stresses in the transverse and longitudinal directions do not
generally stay constant. Indeed, it is often found that one of the stress
components (usually the transverse one) reaches a maximum and unloads
while the other component is still increasing. It may be thought that a
suitable failure criterion would be to take the stresses at the point where one
of the components reaches a maximum. This would be unnecessarily
conservative in the predominantly longitudinally loaded region as the
transverse stress reaches a maximum long before the longitudinal stress
peaks. Figure 1 shows the stress paths (each representing one program run)
in the compressive quadrant ofthe stress space for a 3:1 plate of slenderness
b/t = 60 (m.s.). It is typical of the results obtained. It can be seen that the
failure criterion adopted is the envelope which is tangential to the stress
paths. It should be emphasised that all points on the stress path are
physically attainable stress states.
To substantiate this approach two special runs were carried out for the
above plate where non-proportional displacement loading was used. The
first loading sequence [A] consists of longitudinal loading followed by
transverse loading to failure. When the plate is being loaded longitudinally
the long edges are constrained (and therefore the average transverse stress
is zero). Transverse loading is applied by imposing displacements from the
336 A. F. Dier and P. J. Dowling

positions the long edges take up at the end of the first stage, meanwhile
keeping the short edges fixed. The second loading sequence [B] is a reversal
of the above, i.e. transverse loading followed by longitudinal loading to
failure. The stress paths resulting from the two load sequences have been
plotted with those obtained from proportional displacement loading
(Fig. 1).
In case [A], longitudinal loading is applied up to ¢,/é) =0.5 (i.e. that
displacement which would cause a strain of half yield in a perfect plate). At
this stage the classical three half-wave buckling mode has developed and as
a consequence of this the longitudinal stress initially increases with the
application of the transverse load. However, as the plate is still
substantially in the elastic state, it quickly snaps through and takes up the
single half-wave mode of deformation. From this point the longitudinal
stress decreases with increasing transverse load. As can be seen the stress
path grazes the interaction curve produced from the proportional
displacement stress paths. At the point of contact the ¢,:¢, ratio of plate [A]
is approximately 1/4:1 and it is interesting to note that this ratio is the same
as that of the proportionally loaded plate which fails at almost the same
place. However, the absolute magnitudes of ¢, and ¢, are not the same for
the two cases, being 18° more for the non-proportionally loaded plate.
In case [B] transverse loading is applied up to ¢,/é) = 0.5 with the single
half-wave mode predominating. This mode is maintained while longitu-
dinal loading is applied, although there is a tendency for the three half-wave
mode to develop. The maximum longitudinal stress occurs outside the
drawn interaction curve and lies very close to the 1:1/2 (¢,:¢,) ratio stress
path failure point. The interaction curve has not been extended to this point
as the single and three half-wave buckling modes interact to anomalously
strengthen the plate in this region. Small deviations in the initial
imperfections could bring the failure point back towards the interaction
curve as drawn. Here again, the ¢,:e, ratio of plate [B] at collapse is
approximately the same as that of the associated proportionally loaded
plate (¢,:¢, = 1:1/2).
Further evidence for the soundness of the approach will be found in the
following section where, for the special case of loading along the stress axes
(i.e. proportional stress loading), approximately the same strengths are
predicted.

3.3. Basic Behaviour

The interaction curves for square plates of various slendernesses are shown
in Fig. 2 together with the von Mises yield ellipse. The corresponding curves
 The Strength of Plates Subjected to Biaxial Forces 337]

Fig. 2. Elasto-plastic interaction curves, a/b = 1.

for plates with an aspect ratio of 3 are shown in Fig. 3. For both aspect
ratios, increasing slenderness causes a steady decrease of the proportion of
squash that a platecancarry where at least one of the stress components is
compressive (i.e. positive stress values).
In the absence of transverse loading (a, = 0), approximately the same
load can be carried in longitudinal compression by plates of the same
slenderness but of different aspect ratios. Physically, this may be explained
by considering the 3:1 plate, which buckles into three half-waves, as three
338 A. F. Dier and P. J. Dowling

Fig. 3. Elasto-plastic interaction curves, a/b = 3.

square plates arranged end to end. When only transverse loading is present
the plates buckle into a single half-wave mode, the longer plates being
weaker due to column-like behaviour in the mid-field of the plate. For
combined loading on 3:1 plates, the transition of the single half-wave and
the three half-wave buckling modes is indicated by sharp changes of
curvature in the interaction curves (Fig. 3). It may be noted that these
transitions do not occur on the radial line predicted by the classical linear
solution.
The effect of aspect ratio on transverse strength is further illustrated in
Fig. 4. These results were generated by loading in the transverse direction,
keeping the short edges constrained. As can be seen, most of the fall-off in
 ‘ 4/4 (sw)
—~_ OZ

Ss
~
SS
SS
ASS
SS OE

SS
SS
=
~
SS
Ora

a
aS
Ss

ae
The Strength of Plates Subjected to Biaxial Forces

a<
=$ Coon
-<
OlL
wW— =
q/p
4D9dsSD Ol}Du

*BIY«“p OYA
Jo oodse ONIUO ssJoAsuRI] “YuINs
339
 340

a © Ssan|DA U2YD} WOs} du}

Ni E=q/D UO!WDDI2\U|
S2AUND
SS
N sijnsay wou} Buisn du}
pauipuysuod
~2Hpa2 UO!I}IPUOD

q/o
=)

bib
JO

L:€
(241d
U}bHu2u4s
|DUIPNz1BUC|

AD
b-s
A. F. Dier and P. J. Dowling

Lie Lia
L:G

OL 4/Q (SW)
SO OL SL Oz Guc O-€ Se ov d
“Bly*¢ OYA
JO SsaUIApUITS
UO ISIOASURI] ‘YIBuas
 The Strength of Plates Subjected to Biaxial Forces 341

strength occurs for the lower aspect ratios. Also shown in Fig. 4 are the
limiting strengths for infinitely long plates.
The results of Fig. 4 are replotted in Fig. 5 using the slenderness for the
abscissa. The strength values from the interaction curves for 3:1 plates are
also indicated on this figure. The good agreement between these values
(produced by proportional displacement loading) and the constrained edge
results (proportional stress loading) establishes confidence in the method of
interaction curve construction and in the use of constrained results for the
transverse strength study.
Although Figs. 4 and 5 indicate the sizes of interaction curves for
different aspect ratios, they do not give any indication as to the shape of the
curves. Figure 6 shows the interaction curves for three aspect ratios

Fig. 6. Effect of aspect ratio on interaction curves.

 342

° PduUIDIZSUOD
S3|NSDU

x WOU4 UO!}DD42\UI
S$2AUND
—
sbi4zg pud ‘€ $2u-pud
62)(
A. F. Dier and P. J. Dowling

—
Moy
|

ybiu
aBouaao

————
SUOI}I2jJyadw
ele| lor o3|

O
yi
4) bo.
_ | oat 3/q (sw)

suoljo2ajuadwi
SO OL L S O~ Ve Qe Ge Ov 9)

suol}a2syuadw1
BIgLZ poy
Jo suoNdajJoduI
uo ‘yiBuaNs
=q/b
'€
 The Strength of Plates Subjected to Biaxial Forces 343

(a/b = 1, 3, 5) for two slendernesses (b/t = 30, 80m.s.). For the longest
plates, a combined single and five half-wave imperfection mode has been
assumed. As can be seen there is a considerable change in the shape of the
interaction curves for aspect ratios between | and 3. However, the curves
for plates of aspect ratios 3 and 5 are similar in shape for any given
slenderness.

3.4. Effect of Imperfections

Generally, the effect of increasing imperfections is to reduce the ultimate
strength of a plate. Different combinations of the geometric and the
residual stress imperfections were investigated in ref. 29 for both square
and 3:1 plates. It was found that the interaction curves appertaining to
different imperfections were very similar in shape for any given plate. In
other words the interaction curves for different imperfections can be
inferred from a standard case by simple scaling of the stress axes. The
scaling factor for the longitudinal direction is that obtained from the ratio
of the longitudinal strengths for the different imperfections and similarly
for the transverse direction. Two other general conclusions are worth
mentioning. First, the effect of one type of imperfection (geometric or
residual stress) is reduced in the presence of the other. Second, the greater
part of the reduction in strength occurs for relatively small values of
imperfections.
Figure 7 shows the maximum stresses occurring on the stress axes for 3:1
plates with three levels of imperfections: the low levels of imperfections
acting in combination; the standard case as used previously; and the high
levels in combination. The levels of imperfections are given in Section 3.1.
For plates in the slenderness range b/t = 40-80 (m.s.), the results of the high
and low levels of imperfection were obtained from existing full interaction
curves;*° outside this range new results have been generated using
constrained plates. As can be seen the sensitivity to imperfections is reduced
for plates of increasing slenderness. A similar variation of of exists for
square plates.

4. COMPARISONS WITH EXPERIMENTS

Before any design procedure based on theoretical work can be used with
confidence, it must be shown that the theory is in accordance with
experimental results. Some carefully conducted tests carried out at
Cambridge by Bradfield?* have shown that elasto-plastic large deflection
344 A. F. Dier and P. J. Dowling

jDJUaW142dxKyA
ere yuzuoduos

|
( An
¥€) patjddp
¥SD|
c-&G
6-LE
$24 1/9 pel
a= -¢
Y
@
‘
s3jnsau
a (a) oy To
LZiGOm
4

‘syuouttiodxs
q/p= ‘¢
uosuedWOD
YIM
“BI°*g

Me 4 S

=_ 2
Ko
 The Strength of Plates Subjected to Biaxial Forces 345

programs, including some of those mentioned in Section 1, can give good

agreement with experimental results for plates under uniaxial loading. The
program used in the present work is most similar to that of Harding’s!’ and
gives almost identical results. However, although this may instil confidence
for uniaxial results, further verification is required for the more general
biaxial stress state.
Unfortunately, there is a paucity of tests for plates under biaxial loading.
The most systematic series of tests reported to date are those due to Becker
et al.?* Square cross-section tubes were fabricated from steel sheet of
nominal 0.03 inch thickness. The panels of the tubes were proportioned so
that a/b =3 and 4/1 ranged from 30 to 90. The test results are plotted in
Fig. 8, together with the interaction curves of Fig. 3. The b/t values shown
in Fig. 8 are the equivalent b/t values (calculated by using f) for the material
properties assumed here.
In general, the computed results give a lower bound to the test results and
the agreement is especially good for plates of b/t = 32 and 53. Where just
uniaxial loading (longitudinally or transversely) has been applied in the
tests, somewhat lower strengths are found compared to both an
extrapolation of the biaxial strength data, and indeed to other uniaxial
resuits;.e-¢. ref. 33:

5. DESIGN CONSIDERATIONS

Design rules should be:

(i) simple and easy to apply,

(ii) accurate but safe,
(iii) physically meaningful to the designer,
(iv) and, of increasing importance, amenable to computer coding.

It is usually found that not all four points can be satisfied simultaneously
and a compromise must be sought, especially with regard to the first point.

5.1. A Tentative Format

Previous work has concentrated on plates under compression only and, for
3:1 plates, various interaction formulae have been proposed including
parabolic,*> elliptical,*° and more general curve forms.** However, none
of these formulae is satisfactory when one stress component is tensile and
the other compressive. Although it should be possible to devise formulae
 346

sanjD~A
WOJ}
PaUIDIGO
UOl}DID4I4zUI!
S2AGND

——
S2UI|
+ Sag OOS SaN|IDA
JO} |:€ $2yDjd

——
P2}D|Oduazu;

Xy
j\D‘'
SoOlyDu
yoadsp
A. F. Dier and P. J. Dowling

oz, 1/9 (sw)

SO Ol om O~ Gz O€ Ge Ov d

°6 “SIA “SJUDIOYJIOS UZISO:

 The Strength of Plates Subjected to Biaxial Forces 347

that can represent the interaction curves (and which are also valid for
tensile components), they would undoubtedly be too complicated for use in
a design code (see point (i) above). However, an examination of Figs. 2, 3
and 5 suggests an expeditious method to account for tensile components;
that is to use horizontal or vertical lines as appropriate to approximate the
interaction curves in these regions. For square plates (Fig. 2) the ordinates
and abscissae of the lines would be safely given by the values indicated
where the interaction curves cross the stress axes. In some instances for 3:1
and 5:1 plates (Figs. 3 and 6), slightly reduced values would have to be used
to ensure conservatism in the combined tension—-compression regions.
It is widely recognised that for plates of the same slenderness,
approximately the same longitudinal strength is expected, no matter what
the aspect ratios may be (see Fig. 6 or compare Figs. 2 and 3). It is thus
common to use a single design curve cavering all aspect ratios for the
longitudinal strength. In the present case the curve would be that obtained
from a consideration of the longer plates, again to ensure conservatism.
Based on the present numerical results and the above proposed format, the
values of the design coefficients are shown in Fig.9. Some design
coefficients of BS 5400°’ are also indicated and these shall be referred to
later.
An interaction formula is needed for the compressive quadrant to
complete the design curve. Figure 10 shows the normalised interaction
curves (with respect to the K values of Fig. 9) in the quadrant for square and
3:1 plates. The curves do not pass through unity on the axes because the K
values, on which the normalisation is based, are lower than the
corresponding o™/o, values of Fig. 5. The general interaction formula
proposed in ref. 23, which covers other commonly used formulae, is of the
form

[eeference ra leg eee

Chali 6,/F9 F,/Fo Cio

For square plates a was taken to be equal to 2 and yn a function of the

plate slenderness. The advantage of the coefficient 7 is that any shape from
a straight line (y = 2) to the von Mises ellipse (y = — 1) may be represented.
It is true that slender plates have straighter interaction curves than stocky
plates in the compressive quadrant (Fig. 2) but, because of the manner in
which they have been normalised, it seems more appropriate to keep 7
constant and allow K, (=K,) to take on values greater than 1.0 for the
stocky plates (say b/t < 30, m.s.). (Conservatism is ensured by also applying
the von Mises formula.) This is the philosophy adopted in BS 5400
348 A. F. Dier and P. J. Dowling

Fig. 10. Normalised interaction curves.

although there is no cross product term (i.e. 7 =0). A safe value of 7 for
square plates calculated from the curves of Fig. 10 is 0.45.
Having decided on a value ofn the values of K for stocky plates may now
be found. As an example the K value for the b/t=20 plate may be
considered. The value of o,/a) when o, =, is 0.88 and thus K is given by

or
 The Strength of Plates Subjected to Biaxial Forces 349

The one major disadvantage in this approach is that the designer may be
confused as to how a plate can have K values greater than unity implying
strengths greater than squash (see point (iii) above). It is, of course, just a
design artifice for biaxially loaded plates and care in the wording of the
rules should ensure that it is not misapplied. The corresponding
normalised curves for 3:1 plates (Fig. 10) are more widely spaced than
those of the square plates and they are non-symmetric about the diagonal.
Parabolic*® and skew parabolic”? forms have been proposed for which
a = 1 in eqn (1). However, for the sake of simplicity, it is felt that keeping
a = 2 is still satisfactory. A function for 7, dependent on f, can be found to
take account of the different shapes of the interaction curves.
For longer plates, the same values of y as those of 3:1 plates can be used
as the shapes are very similar (Fig. 6), and using, of course, the appropriate
K values. It is suggested that for plates of aspect ratios between 1 and 3, a
linear interpolation of 7 should suffice.

5.2. Comments Relating to BS 5400

In the design of web panels according to the BS 5400 Code two clauses have
to be satisfied. The first (clause 9.11.3) is an application of von Mises and
the second (clause 9.11.4.4) is a check on interactive buckling which
considers shear and in-plane bending as well as biaxial stress effects.
It is informative to see the biaxial strengths given by the Code for mild
steel plates of 3:1 aspect ratio (Fig. 11). These curves contain the relevant
safety factors. One shortcoming is immediately obvious: where one stress
component is tensile and the other compressive, the curves enter non-
conservative regions—compare with Fig. 3. Even with the inclusion of the
safety factors, the Code may give an unsafe design for slender plates. This
shortcoming also exists for other aspect ratios. As discussed in the
preceding subsection, this may be easily amended by using horizontal and
vertical cut-off lines.
A second modification which seems to be necessary is the inclusion of the
cross product term of eqn (1). If the term is included some adjustment
would have to be made to those K values which are greater than 1.0 (i.e. for
very stocky plates).
Finally, it has been shown that if the ‘cut-off approach were adopted,
reduced values of o%"/¢, and o}'/o) would have to be used (i.e. K, and Ky). In
Fig. 9, the values of K, and K, based on the present numerical work were
presented along with BS 5400 design coefficients for square and 3:1 plates.
The BS 5400 points for square plates (or plates of other aspect ratios under
longitudinal loading) are satisfactory, whereas for slender 3:1 plates loaded
350 A. F. Dier and P. J. Dowling

Fig. 11. BS 5400 interaction curves, a/b = 3.

transversely they appear to be non-conservative. This will be only partly

due to the greater imperfections assumed in the present study for slender
plates (Fig. 7 shows a small imperfection sensitivity in this region). In
BS 5400 the transverse coefficients are implicitly calculated from the curve
for square plates by using an effective slenderness which is found by
multiplying the actual slenderness by the aspect ratio. Thus, the transverse
coefficient for a 3:1 plate of b/t = 40 is the same as that for a square plate of
b/t= 120 and the transverse coefficient for a 3:1 plate of b/t=120 is
obtained from that of a b/t= 360 square plate. In the latter case it is
suspected that the slope of the linear approximation design curve in this
 The Strength of Plates Subjected to Biaxial Forces 351

slender plate region is not steep enough and thus gives rise to high
coefficients for transversely loaded long slender plates. It may be noted
that, by using this approach, very long slender plates are out of the range of
the BS 5400 curves.

6. CONCLUSIONS

(1) Interaction curves for square plates and plates of 3:1 aspect ratio
have been presented for biaxial loading.
(2) Although longitudinal strength is not greatly affected by aspect
ratio, the transverse strength is. Curves have been presented
showing the transverse strength of plates of practical aspect ratios.
(3) Adesign method covering the range of possibilities from combined
compression—compression to compression-tension has been
outlined.
(4) Thenew UK bridge code has been shown to be deficient for cases of
combined compression and tension and also needs a re-
examination for transversely loaded plates of large aspect ratios.

REFERENCES

1. BRYAN, G. H., On the stability of a plane plate under thrust in its own plane
with application to the buckling of the side of a ship, Proc. London Math. Soc.,
22 (1891).
2. BuLson, P. S., The stability of flat plates, London, Chatto & Windus, 1970.
3. BIJLAARD, P. P., Some contributions to the theory of elastic and plastic
stability, International Association of Bridge and Structural Engineering, 8
(1947).
4. ILyusuin, A. A., ‘Stability of plates and shells beyond the proportional limit’,
National Advisory Committee for Aeronautics, T.M.116, 1947.
5. STOWELL, E. Z., ‘A unified theory of plastic buckling of columns and plates’,
National Advisory Committee for Aeronautics, T.N.1556, 1948.
6. HANDEELMAN, G. H. and PraGer, W., ‘Plastic buckling of rectangular plate
under edge thrust’, National Advisory Committee for Aeronautics, T.N.1530,
1948.
7. Levy, S., ‘Bending of rectangular plates with large deflections’, National
Advisory Committee for Aeronautics, T.N.846, 1942.
8. Coan, J. M., Large deflection theory for plates with small initial curvature
loaded in edge compression, Journal of Applied Mechanics, 18, No. 2 (1951)
143-51.
352 A. F. Dier and P. J. Dowling

>: Hu, P. C., LuNpquist, E. E. and BATporz?, S. B., ‘Effect of small deviations
from flatness on effective width and buckling of plates in compression’,
National Advisory Committee for Aeronautics, T.N.1124, 1946.
10. YAMAKI, N., Postbuckling behaviour of rectangular plates with small initial
curvature loadedin edge compression, Journal of Applied Mechanics, 26, No. 3
(1959) 407-14. Continued in 27 (1960) 335-42.
ll. THOMPSON, J. M. T. and WALKER, A. C., A non-linear perturbation analysis of
discrete structural systems, International Journal of Solids and Structures, 4
(1968) 757-68.
1 WILLIAMS, D. G. and WALKER, A. C., Explicit solutions for the design of
initially deformed plates subject to compression, Proceedings of the Institution
of Civil Engineers, 59, Part 2 (1975) 763-87.
iley. FAULKNER, D., A review of effective plating for use in the analysis of stiffened
plating in bending and compression, Journal of Ship Research, 19, No. 1 (1975)
1-17.
. Ruopes, J., Effective widths in plate buckling, Developments in thin-walled
structures 1, J. Rhodes and A.C. Walker (eds.), London, Applied Science
Publishers, 1982.
. Dow.inc, P. J., ‘Some approaches to the non-linear analysis of plated
structures’, Symp. on Non-linear Behaviour and Techniques in Structural
Analysis, Transport and Road Research Laboratory Supplementary Report
164 UC, 1974, pp. 86-93.
. CRISFIELD, M. A., ‘Large-deflection elasto-plastic buckling analysis of plates
using finite elements’, Transport and Road Research Laboratory Report LR
5931973.
. HarpinG, J. E., Hopss, R. E. and NEAL, B. G., The elasto-plastic analysis of
imperfect square plates under in-plane loading, Proceedings of the Institution
of Civil Engineers, 63, Part 2 (1977) 137-S8.
. Frieze, P. A., Hospps, R. E. and Dow Lina, P. J., Application of dynamic
relaxation to the large deflection elasto-plastic analysis of plates, Computers
and Structures, 8 (1978) 301-10.
. Littte, G. H., Rapid analysis of plate collapse by live energy minimisation,
International Journal of Mechanical Science, 19, No. 12 (1977) 725-44.
. BRADFIELD, C. D. and CHLADNY, E., ‘A review of the elasto-plastic analysis of
steel plates loaded in in-plane compression’, Cambridge University Report
CUED/D-Struct/TR 77, 1979.
. FRIEZE, P. A., DOWLING, P. J. and Hosss, R. E., Ultimate load behaviour of
plates in compression, Stee/ plated structures, P. J. Dowling, J. E. Harding and
P. A. Frieze (eds.), London, Crosby Lockwood Staples, 1977, pp. 24-50.
Pd, HARDING, J. E., Hosss, R. E. and NEAL, B. G., Ultimate load behaviour of
plates under combined direct and shear in-plane loading, Steel plated
structures, P.J. Dowling, J. E. Harding and P.A. Frieze (eds.), London,
Crosby Lockwood Staples, 1977, pp. 369-403.
2% VALSGARD, S., ‘Ultimate capacity of plates in biaxial in-plane compression’,
Det Norske Veritas Report 78-678, 1978.
24. LittLe, G. H., The collapse of rectangular steel plates under uniaxial
compression, The Structural Engineer, 58B, No. 3 (1980) 45-61.
 The Strength of Plates Subjected to Biaxial Forces 353

yS), DowL.ina, P. J., HarpinG, J. E. and SLatForp, J. E., ‘Strength of ships’

plating, plates in biaxial compression’, Imperial College, London, CESLIC
Report SP4, 1979.
26. Dier, A. F. and Dow ina, P. J., ‘Strength of ships’ plating, plates under
combined lateral loading and biaxial compression’, Imperial College, London,
CESLIC Report SP8, 1980.
TI. Coomss, M. L., ‘Aspects of the elasto-plastic behaviour of biaxially loaded
plates’, M.Sc. Thesis, University of London, 1975.
28. CRISFIELD, M. A., Full range analysis of steel plates and stiffened plating under
uniaxial compression, Proceedings of the Institution of Civil Engineers, 59, Part
2 (1975) 595-624.
US). Drier, A. F., ‘Collapse of metal plates’, Ph.D. Thesis, University of London,
1981.
30. MARGUERRE, K., ‘Zur Theorie der gekruemmten Platte grosser
Formaenderung’,, Proc. Sth. Int. Congr. App. Mech., Cambridge, 1938,
pp. 93-101.
31. Frieze, P. A., “Ultimate load behaviour of steel box girders and their
components’, Ph.D. Thesis, University of London, 1975.
3. FAULKNER, D., ‘Compression strength and chance’, Defence Fellowship
Thesis, MIT, published by Dept. DS15, Ministry of Defence, London, 1973.
33) BRADFIELD, C. D., “Tests on single plates under in-plane compression with
controlled residual stresses and initial out-of-flatness’, Cambridge University
Report CUED/D-Struct/TR 78, 1979.
34. BECKER, H., GOLDMAN, R. and PAZERYCKI, J., ‘Compressive strength of ship
hull girders. Part 1—Unstiffened plates’, Ship Structure Committee Report
SS@2I7-21970!
3S: FAULKNER, D., ADAMSCHAK, J., SNYDER, G. and VETTER, M., Synthesis of
welded grillages to withstand compression and normal loads, Computers and
Structures, 3, No. 2 (1973) 221-46.
36. HARDING, J. E.and DowLING, P. J., The basis of. the proposed new design rules
for the strength of web plates and other panels subject to complex edge loading,
Stability problems in engineering structures and components, T. Richards and
P. Stanley (eds.), London, Applied Science Publishers, 1979.
3c British Standards Institution, ‘Steel, concrete and composite bridges. Part 3.
Code of practice for design of steel bridges’, BS 5400, Part 3, 1982.
 na
¢ . J i. A
wre hep da sini — pes
we blaccertaii ee a :
s ous 7 he 7» ce
At Agiions haat Sali tee
ter
i ‘7. - a7 4? 6
peie:
vd ona iad oe ee Oe (cuaale >
oe, *:, ahd
: eal Vi et) of yb
(Op
:
om
ni

ow: Beret lie Peas i> Oe fast : mA, od ‘1 BO,

: a Le 5 | be; a S | cL —

; OMT Cur EM tyke sy

Ome METS. Hooee :
Abe rk
UM wer Rog

Jb fe saaeth
He aM plihe ’ 4 i’. : et aed | vx 1 itapayy vy ma ug we

——~ j
yo
; ; oneal ieee
=; lat
| 7 =
| 7
a
( =|
ra.
1 y a ‘YY pormhineg. Vi,
i iA) ‘ ane “Aga aca a ey Mkag y Soe Wy Ww wohilits ;
ut PD, ' Peete “Tie ; vat i i hier anf pie 4)
:
7 x
vent
) P R t -
4a} 7 .

: A od @
 18

Elastic Buckling of Stiffened Steel Plates of High

Aspect Ratio Under Uniaxial Compression

W. KATZER and N. W. MurRAY

Monash University, Clayton, Australia

ABSTRACT

It is shown that simply supported stiffened panels with high aspect ratios can
buckle either locally, in which case the cross-section undergoes considerable
distortion, or globally, in which case the panels undergo an overall dishing.
In shorter panels this dishing is often suppressed with the result that the
panels behave like wide columns because then the effect of the support along
each side on the buckling behaviour is small. By using a so-called buckling
plot obtained by using thefinite strip method it is shown that the change from
local to global buckling can be predicted accurately. Experiments carried
out to verify the application of the finite strip method to stiffened plates
demonstrate that this change can be brought about by changes in the number
of stiffeners or in the depth of the stiffeners and results confirm the
theoretical work.

NOTATION

Total cross-sectional area of panel.

Panel width.
Width of plate strip.
Stiffener depth.
Panel length.
AS
bh
we
log
Buckling length.
355
356 W. Katzer and N. W. Murray

L, Length of local buckle.

m Number of buckles in a specimen in longitudinal direction.
n Number of stiffeners.
P... Critical load for local buckling.
P, Experimentally measured frictional force acting on sides of
specimen.
P, Squash load (=o, x cross-sectional area).
5 Stiffener spacing.
Applied stress.
o, Buckling stress.
The smaller of o,, and o,.
Critical stress for local buckling.
Critical stress for global buckling.
Mean compressive residual stress in the plating.
Thickness (¢ = ¢, = 1,).
Thickness of plate.
Thickness of stiffener.

1. INTRODUCTION

This paper presents a study of the elastic buckling behaviour of rectangular

steel plate panels with a relatively high aspect ratio and which have a few flat
stiffeners attached to one side. The panels are assumed to be simply
supported along all four edges and are subjected to a uniformly distributed
uniaxial compressive force in the longitudinal direction (Fig. 1). Such

All edges
get
AA q Pe a simply - supported

Oy n stiffeners oe fe

Fig. 1. Stiffened plate panel.

 Elastic Buckling of Stiffened Steel Plates 357

Fig. 2. Local and global buckling modes.

panels are used in the compression flanges of long, narrow box girders, in
the legs of derrick cranes and in box columns. The panels can buckle in at
least two different modes, viz. a local buckling mode and a global buckling
mode (see Fig. 2). Local buckling involves a distortion of the cross-section
with the buckles having a relatively short half-wavelength approximately
equal to the stiffener spacing. For usual cross-sections local buckling
results in the buckling of the plate strips in the well-known chequerboard
pattern with sympathetic twisting of the stiffeners. Global buckling of a
stiffened panel is like that of an orthotropic plate and involves bending of
the stiffeners about their major axis. It is, however, not sufficient to analyse
a panel separately for local and global buckling as these buckling modes
can interact and this leads to a lower buckling stress than both the stresses
for local and global buckling would suggest. This phenomenon is well
known and examples are given by Timoshenko and Gere.’ It is important,
therefore, for a theoretical analysis to analyse the whole cross-section for
buckling in all possible modes and not simply elements of the cross-section,
as for example the plate strips between stiffeners.
The study carried out here is both theoretical and experimental. For the
theoretical analysis the finite strip computer program PLATE, which was
developed by Murray and Thierauf for the preparation of their book of
tables for the design of stiffened steel plates, was used. This program is
similar to the one developed by Stanley.? It is based on a small-deflection
elastic analysis and calculates critical buckling stresses of panels with given
dimensions and boundary conditions. Some results of an experimental
program carried out by Katzer* are used to check and interpret theoretical
results.
A systematic and comprehensive parametric study of stiffened plates is
impossible as there are too many independent parameters. This paper
investigates local and global buckling behaviour and the interaction of
these buckling modes. After briefly describing some of the background of
the theory used (in Section 2) and the experimental program (in Section 3),
358 W. Katzer and N. W. Murray

two studies of the transition from local to global buckling are presented. In
the first, the number of stiffeners n and in the second, the stiffener depth d,
are systematically varied. In Section 6 theoretically and experimentally
obtained buckling stresses are compared and in Section 7 a number of
conclusions which can be drawn from the present study are presented.

2. THEORETICAL ANALYSIS

The stiffened panels described below have been analysed for buckling using
Cheung’s® finite strip method. Details of its specific application to the
analysis of stiffened panels are given by Murray and Thierauf’ and by
Thierauf et al.’
Their method solves for the eigenvalues of a panel with known cross-
section as the length L, of the buckle is increased. The information
contained in their book is the stress at which local buckling occurs, the
corresponding length of the local buckles and the length of panel above
which the panel will buckle in a global manner.

3. EXPERIMENTAL PROGRAMME

Altogether 36 small-scale models fabricated from cold rolled mild steel

sheet metal of 1.4-2.0mm thickness were tested.* The specimens were
manufactured by using the plasma welding technique which is a non-
electrode consuming fusion welding technique. This technique had earlier
been used successfully at Monash University by de George* for the
manufacture of trough-stiffened plates. All panels had a total length of
950 mm and a total width of 350 mm. The specimens were tested in a special
test rig which is shown in Fig. 3. The rig provided simple supports for the
two longitudinal edges of the specimens. At the top and bottom edges of the
panels the plate strips were clamped into supporting blocks. Load
distribution blocks were used in order to apply an evenly distributed load.
All tests were carried out in the deflection-control mode of a Baldwin
testing machine which has a capacity of 50 tonnes.
The load increments were applied with a constant displacement rate of
1.7 um/s. Load-shortening curves were traced by the testing machine.
Additionally a Moiré Fringe technique was used to study the development
of the buckling deformations. The loading process was stopped at regular
intervals in order to take photographs of the Moiré Fringe patterns which
 Elastic Buckling of Stiffened Steel Plates 359

Head of testing machine

4

Fig. 3. Test rig.

were contour maps of the buckled specimen. In every case two nominally
identical specimens were tested in order to check on the reproducibility of
the results.
Experiments on the type of stiffened plates investigated here had earlier
been carried out by Stanley.? Stanley used specimens made from
acrylonitrile butadiene styrene (ABS), a material which has a large elastic
range. His specimens were fabricated by gluing plate strips together and
they were therefore relatively perfect and stress free. In the present testing
programme mild steel plate was used because it was desired to study not
only elastic buckling but also the collapse behaviour of the specimens. All
360 W. Katzer and N. W. Murray

S) compression Stresses in MPa

80
60
40
20

20
40
60
80
100
120
140
160
Oy = 171 MPa

+) tension

Stiffeners : SS +1587
+171:0

<a55 +168-°5
[345 +624

+103'5

Resultant compressive force : 15:-9KN.

Resultant tensile force -47-2KN.
Fig. 4. Residual stresses in a three-stiffener panel.*
 Elastic Buckling of Stiffened Steel Plates 361

specimens had initial imperfections and also carried residual stresses owing
to the jigging and welding process. Residual strains were evaluated in some
specimens by measuring the length of the as-welded specimen at certain
grid points of the cross-section, then cutting the specimens into strips in the
longitudinal direction and finally measuring the length of all strips again.
Figure 4 shows the result for a three-stiffener specimen.

4. STUDY NO. 1:2 VARIED BETWEEN 1 AND 3

In the first study the effect of varying the number of stiffeners on the elastic
buckling behaviour is investigated. One major problem in a theoretical
buckling analysis is that stiffened plates can buckle in many different ways.
From the designer’s point of view, of all possible buckling modes the one
which gives the lowest buckling stress is most important. This buckling
mode is not known beforehand so the theoretical analysis can only be
carried out for an assumed number of buckling half-wavelengths and one
must search systematically for the buckling mode corresponding to the
lowest critical buckling stress. One way of doing this is to assume the
buckling mode is a sine curve with half-wavelength in the longitudinal
direction of L, = L/m with m= 1, 2, 3,.... Another possibility is to vary
the buckling length L, froma very small value in finite steps up to the actual
panel length and calculate the stress o, which is required to maintain the
specimen in equilibrium in the configuration ofa half-sine wave of length
L,. If these values o, are plotted against values of L, a curve or so-called
‘buckling plot’ is obtained. The full line in Fig. 5 shows a typical buckling

ce} SL; 4Ly L Lp

Fig. 5. Buckling plot.
362 W. Katzer and N. W. Murray

Op
(MPa)
TEST C1-23
owe \ SR PER ada EE)
Vomecs ee Bee
' 175 175
400 \ t= 1-97
\

|
|
=950m
TH
5c |
7
al
ish
re)
-_

=
ca |
o |
af
OL, 1000 2000 3000 | (mm)
Fig. 6. Buckling plot for a one-stiffener panel.

500
TEST C2-24
Ween
77777
pen news
777
tes

gee e ‘= a
t= 1:95
300

Og o
200 \8

SS ee ee
panel
of
Length
° Lz fe) } re) 2000 3000 L(
mm)
Fig. 7. Buckling plot for a two-stiffener panel.
 Elastic Buckling of Stiffened Steel Plates 363

Ob
(MPa) |

50 |
| Apt
c H TEST C3-27

400

300
Oc. »
Ogata: oo ae
200 | §

| BIN
100 Ll ont
| | ™. EURER
| 2| i> ee
aij — oe

OL, 1000 2000 3000 1(mm)

Fig. 8. Buckling plot for a three-stiffener panel.

plot for all of the panels investigated here. It shows two distinct minima.
The first minimum at L, is due to local buckling. The corresponding
buckling stress o,, is referred to as the ‘critical stress for local buckling’. The
buckling stress corresponding to L 1s called the ‘critical stress for
global buckling’, o,. The smaller value of o,, and o, is referred to as 04m.
For the panels considered here the second minimum usually occurs at a
length which is greater than the actual panel length L. This buckling plot was
found to be the most satisfactory way of presenting the results of this study.
It is self-evident that ifa panel of length L, can buckle locally at o,, and with
a buckle length equal to L, panels of length 2L,, 3L, and so on will also
buckle at the same critical stress. The dotted lines in Fig. 5 show how this
effect can be included in a buckling plot but since it does not provide
additional information the presence of these additional lobes is tacitly
assumed in the discussion which follows.
Figures 6-8 show the buckling plots for three specimens each with a plate
thickness of approximately 2 mm, a stiffener depth of 40 mm, and one, two
and three stiffeners, respectively. Panels of these cross-sections and a length
of 950 mm have been tested in the laboratory. From the buckling plot of the
one-stiffener specimen one would predict that the panel buckles mainly by
364 W. Katzer and N. W. Murray

(c)
Fig. 9. Contour maps of buckled specimens (¢~ 2mm, d= 40 mm).
 Elastic Buckling of Stiffened Steel Plates 365

local buckling, as o,, is significantly smaller than o,. This behaviour was
indeed observed in the laboratory as can be seen from the contour map
shown in Fig. 9(a) and obtained by using the Moire technique. Figures 9(b)
and (c) show the buckling modes observed for the two- and three-stiffener
specimens. As the number of stiffeners is increased above one the b/t ratio
of the plate strips increases and therefore also the resistance against local
buckling. Consequently the first minimum of the buckling plot is shifted
significantly upwards. For the two-stiffener specimen (Fig. 7) o,, is still
slightly smaller than o,. The panel shows an interactive buckling with local
and global modes and this is reflected in the contour map shown in
Fig. 9(b). For the three-stiffener specimen (Fig. 8) o,, is greater than a,. The
contour map (Fig. 9(c)) for this panel shows that there is very little local
buckling and that the main shape indicates global buckling approximating
to that of an overall dishing.
For the sake of comparison the column buckling stress of the specimens,
the Euler stress o,, is also included in all three buckling plots. o, was
calculated using the simple formula o, = n*E/(L/r)* of a pin-ended Euler
strut, thus neglecting the influence of the two lateral supports. It can be seen
that the difference between the Euler stress o, at L and the critical stress for
global buckling is greatest for the one-stiffener specimen. For the three-
stiffener specimen the Euler curve is an asymptote for the part of the
buckling plot which corresponds to global buckling. However, with
increasing L, the difference between o, and a, grows. For large L, the Euler
stress approaches zero whereas for those specimens which had simply
supported sides the global buckling stress reaches a minimum and increases
again with increasing L,. This latter behaviour is similar to that of a very
long simply supported orthotropic plate for it is well known!’ that such a
plate develops long buckles at its critical stress.

5. STUDY 2: d IS VARIED BETWEEN 0 AND 100mm

In Figs. 10 to 12 the results of the analysis of a series of panels each with a

plate thickness of 1.5mm and a panel length of 950 mm are presented. The
critical buckling stress o,,,, is plotted as a function of the stiffener depth d for
the one-, two- and three-stiffener panel of the test programme. For d=0,
which corresponds to unstiffened panels, the buckling mode consists of
three roughly square buckles. As d is increased by small amounts the
buckling mode changes to a single long buckle with L, = L meaning that
there is then global panel buckling. With d= 15mm global buckling is
366 W. Katzer and N. W. Murray

NOTE : Logarithmic Scale and False Origin for Ordinate

Local ee a
10° LGlobal buckling 7

panel length
= 950 mm

O 10) 20) So) &@) SO Go no) sto) So jer

d(mm)
Fig. 10. o,,, aS a function of d for a one-stiffener panel.

predicted to occur first for all three panels, i.e. m = 1. The observed contour
maps from the Moiré technique of the relevant specimens after buckling are
shown in Fig. 13(a—c). The one-stiffener specimens showed severe local
buckling deformations although o, was about 25 % smaller than o,,. The
two- and three-stiffener specimens showed the global buckling behaviour
predicted by theory.
For the one-stiffener specimen and d>20mm the buckling mode
changes to local buckling (Fig. 10, m = 5, thus L, = 190 mm). Up to about
d= 60 mm the curve is fairly horizontal indicating that in this region mainly
local buckling of the plate strips occurs. With further increase in stiffener
depth o,,, decreases and thus the cross-section becomes weaker against
local buckling. The corresponding buckling mode involves both plate and
stiffener deformations. The two- and three-stiffener panels have a lower b/t
ratio and hence a greater resistance against local plate buckling. The graphs
in Figs. 11 and 12 show that the change from the global buckling mode to
the local buckling mode consequently takes place at a greater value of o,,,.
 Elastic Buckling of Stiffened Steel Plates 367

One
5x10° NOTE : Logarithmic Scale and False Origin for Ordinate

Global Local Cetra

buckling buckling ~~“ m=l Se
7 N
7 N
Oa SS

10°

5x107

icaGtgie panel length

107% WI S° GF = 950mm

12 !126 ‘112

O 10° S2ZONg30N £40 950 560i 70. 80 ~-90 400

d(mm)
Fig. 11. o,,, as a function of d for a two-stiffener specimen.

In the case of the three-stiffener panel the horizontal plateau, which

corresponds to a region of mainly local buckling of the plate strips,
disappeared. This indicates that in that case local buckling, which governs
for a stiffener depth greater than 32 mm, is always associated with stiffener
buckling. A further group of specimens with a stiffener depth of 30mm
was tested and they all showed the predicted buckling behaviour (see
Fig. 14(a-c).
The dashed lines in Figs. 10 to 12 are theoretical curves which show the
buckling stress for global panel buckling (m= 1). It is seen that the
resistance against global buckling increases rapidly with increasing d and
reaches a maximum at about d = 70 mm. The buckling stress then decreases
with further increase in d due to the weakness of the slender stiffeners
against buckling with a half-wavelength of L. However, this branch of the
stability analysis is fictitious because the critical buckling modes involve
twisting of the stiffeners with shorter wavelengths.
A series of panels of nominally identical geometry were also tested but
368 W. Katzer and N. W. Murray

5x10° NOTE: Logarithmic Scale and False Origin for Ordinate

Local Saati
Global eas ode Se
buckling ote aad nae pear

10°

5x10

m=5
+ t=1-5
dt AST sipanel length
104 7707 7777 = 950mm.
eae
80 95 95 80

O~ 10" 20530 “740.50; S607 SO, 660 790." Oo

d (mm)
Fig. 12. 0,,, aS a function of d for a three-stiffener specimen.

they had light transverse stiffeners at midspan. Their local buckling

behaviour was almost identical to that of the panels which had no
transverse stiffener. However, the global buckling mode was in every case
suppressed.

6. COMPARISON OF OBSERVED AND PREDICTED

BUCKLING STRESSES

All panels had initial imperfections and therefore did not buckle suddenly
at their critical load. Table 1 shows a comparison between the theoretically
obtained critical buckling loads and the load range at which local plate
buckles formed during testing. Measurements of residual strain had
shown‘ that the plate strips carry compressive residual stresses o, over most
of their width. o, was found to be approximately 30, 50 and 70 MPa for the
 Elastic Buckling of Stiffened Steel Plates 369

Fig. 13. Contour maps of buckled specimens (¢ ~ 1.5mm, d~ 15mm).

370 W. Katzer and N. W. Murray

(b)

Fig. 14. Contour maps of buckled specimens (tf ~ 1.5mm, d~ 30mm).

 Elastic Buckling of Stiffened Steel Plates 371

Table 1. Comparison of Observed and Predicted Buckling Loads for Local Plate
Buckling.

Specimen Theoretical P,,° Theoretical P’.,° Observed P.,

(KN) (kN) (kN)

a f ; =
AD 13 68 35 55-10
Blagps (a8 : a
GIL din ; ma
Head no : a
1233 3 22 55-15
r Pe = (6.,A, te P,).

4 Re = (Soy 7 0,)A, as Pe
© The first value in this column indicates the approximate load at which first local
plate buckles were observed anywhere in the panel and the second value indicates
the approximate load at which a regular pattern of local plate buckles had formed
over the whole panel.

one-, two- and three-stiffener specimens respectively. It could therefore be

expected that the specimens may buckle elastically at an average applied
stress of o,, — ¢,. The corresponding critical buckling loads are also listed in
Table 1. It can be concluded from Table 1 that local plate buckling
occurred at an average applied stress approximately equal to o,, and not at
a (lower) stress o,, — ¢,.
Pure global buckling was only observed in two specimens of the test
programme. It occurred very gradually and there was not a distinct
buckling phenomenon at a critical load.

7. CONCLUSIONS

(i) The test series described here is confined to stiffened panels with
high aspect ratio, with simple rectangular stiffeners and with simple
supports along each longitudinal edge.
372 W. Katzer and N. W. Murray

(il) One aim of the programme was to check that the elastic buckling
stresses and patterns predicted by Murray and Thierauf * who used
the finite strip method could be confirmed by experiments.
Although the testing programme was limited in its scope it can be
concluded that experimental and theoretical results so far obtained
agree quite well. No discrepancies have yet been observed.
(iii) Buckling plots, typically shown in Figs. 6, 7 and 8, show that as the
number of stiffeners increases, the buckling mode may change from
local to global with interactive buckling in the transition region.
Contour maps of the buckling patterns obtained by experiment
(Fig. 14) confirm this.
(iv) Graphs of the theoretical buckling stress against depth of stiffener,
typically shown in Figs. 10, 11 and 12, show that as d increases
from zero, where the panels behave as an isotropic flat plate, up to
an intermediate value the panels behave as orthotropic plates.
Beyond a certain intermediate value the buckling stress decreases
and the local buckling mode is largely governed by torsional effects
in the deep narrow stiffeners. Contour maps of the buckling
patterns obtained by experiment (Figs. 13(b) and 14(b)) confirm
this change from global to local buckling as d increases.
(v) The global buckling is relatively easily suppressed by use of
transverse stiffeners which do not significantly increase the local
buckling stress.

Acknowledgements
The financial help of the Deutsche Forchungsgemeinschaft, Deutscher
Akademischer Austauschdienst and Monash University which have
enabled the authors to pursue this research is gratefully acknowledged.

REFERENCES

1. TIMOSHENKO, P.S. and Gere, J. M., Theory of elastic stability, 2nd edn., Tokyo,
McGraw-Hill and Kogakusha, 1961.
2. Murray, N. W. and THIERAUF, G., Tables for the design and analysis ofstiffened
steel plates, Braunschweig, Wiesbaden, Vieweg, 1981.
3. STANLEY, C.R., ‘The critical buckling stress of plates stiffened with longitudinal
stiffeners’, M.Eng.Sc. Thesis, Dept. of Civil Engg., Adelaide University, 1973.
4. KaTzerR, W., ‘The collapse behaviour of stiffened plate panels of high aspect
ratio under uniaxial compression’, M.Eng.Sci. Thesis, Monash University,
August 1982.
 Elastic Buckling of Stiffened Steel Plates 373

5. DE GEORGE, D. M., ‘Collapse behaviour of trough stiffened steel plates’, Ph.D.

Thesis, Monash University, 1979.
6. CHEUNG, Y. K., Finite strip method in structural analysis, Oxford, Pergamon
Press, 1976.
7. THIERAUF, G., KATZER, W. and Murray, N. W., ‘Application of the finite strip
method to the design of stiffened plates for buckling’, International Conference
on Finite Element Methods, Shanghai, August 1982.
p14 & 6 a 6
, oa 7 —
rs

OPRF ’
 19

A Brief Review of Plate Buckling Research

A. C. WALKER
University of Surrey, Guildford, UK

ABSTRACT

The paper presents a brief and somewhat subjective review of the

developments in plate buckling research. It seeks to highlight the early
major achievements and to survey in a little more detail the recent work
which has been the basis of new design codes for box girders, bridges and
analysis of cold formed steel sections.

NOTATION

b Piate width.
Bs Effective width.
c Coefficient.
E Young’s modulus.
Hy, uy Coeihcients.
DP. Critical load.
t Plate thickness.
w Deflection.
Wo Initial deflection.
€ Plate end shortening.
Eo Yield strain.
v Poisson’s ratio.
Average applied stress.
375
376 A. C. Walker

6. Critical stress.
Om Maximum stress.
Fo Yield stress.
oe Ultimate value of average applied stress.

1. INTRODUCTION

A review of the whole field of the development of plate buckling would be a

formidable task. The subject has been active in a research sense for about
one hundred years and, as Fig. 1 shows, the number of papers published
which contain information relating to plates and buckling is over four
hundred in the last decade alone. These papers cover a wide range of
aspects and materials, including static, dynamic, thermal and impact
loading, concrete and composite materials and a variety of plate
geometries.
It is really the job of books to review the entire field and to present a
coherent picture. Therefore the present paper will be restricted to a very
narrow range of topics and will concentrate on developments in the
understanding of the behaviour of the buckling of rectangular plates
subjected to fairly simple forms of in-plane loading. Furthermore, the
treatment will attempt to highlight some of the important landmarks in the
history of the development. Naturally, this kind of treatment of a subject is
subjective and potentially contentious. I apologise in advance for seeming
to ignore the work of some people while mentioning other contributions
which, it may be felt, are of lesser importance.
In considering the papers published during the past decade, it appears
that less than 10% had any experimental content. It could be thought that

Number of ay Wa
papers 40 y
published Vy)

: We
30 y g

10 | Z g
AY AY
SERRE
REE ES RRS
Fig. 1. Papers published on plate buckling.
 Plate Buckling Research 377

this is a fairly recent phenomenon and reflects the tremendous growth in the
power of computers and numerical analysis. However, this is not the case;
the great majority of the papers on plate buckling during the past ninety
years has been concerned with theoretical developments. Notwithstanding
the value of these contributions, I feel that the really important
developments in understanding the phenomenon of plate buckling and
developing information for designers have been made when theory and
experiment are brought closely together. In this paper, therefore, the
experimental aspects of research are emphasised.
Indeed it has been one of the great merits of the work carried out in the
UK that the development of analysis and testing facilities has proceeded
hand-in-hand. Ona personal note, I have a feeling of gratitude to Professor
Harvey and his colleagues who impressed this attitude so strongly upon me
when I worked in their department at Strathclyde, or the Royal College of
Science and Technology as it then was.

2. 1850-1930

The earliest incidence of practical importance relating to plate buckling is

associated with the Britannia Bridge. In 1845 Robert Stephenson projected
the construction of a suspension bridge across the Menai Straits as part of

Fig. 2. Britannia Bridge during construction.

378 A. C. Walker

iy
wsq

Fig. 3. Cross-section of box beam.

his railway then under construction. He had decided to use wrought iron
which could be produced in small sheets and rivetted together to form
larger plate structures. He designed a novel form of structure in which the
train would pass through the centre of the box which itself would rest on
piers and be supported by chains. Figure 2 shows the general arrangement
and Fig. 3 shows the cross-section.
Realising that his structure was in fact quite unique in concept and in
advance of contemporary engineering practice, Stephenson arranged with
Fairbairn to carry out a number of tests to determine the strength of box
beams having various cross-sectional forms. It is important to remember
 Plate Buckling Research 379

fig

Fig. 4. Typical collapse modes.

that at that time the engineers in Britain were generally ignorant of

mathematical structural analysis, and indeed quite scornful of it. It was to
be another ten years before the value of the theories currently being
developed in France by Navier would be accepted in England.
The tests were carried out in the laboratories of University College
London with the cooperation of Hodgkinson who was in effect the first
professor of engineering science. Thus we can see that from the beginning
of the studies of engineering mechanics, plate buckling has been a factor to
be considered. Another interesting aspect of the tests is that they were
carried out using small scale models of the proposed box beams and
evidently Stephenson had no qualms about accepting the validity of the
results from these tests.
In the event, the tests on a number of cross-sectional forms showed that
local buckling, as we would now call it, proved to be a limiting factor which
governed strength. Figure 4 shows the type of failure mode occurring in
some of the test specimens; of course, no theoretical basis of this form of
collapse was possible then and it was not until a century later that the
analysis of this type of failure mechanism was reconsidered. Stephenson’s
bridge was built and proved to be so strong that the chain supports were
never installed. However, it was an expensive structure and no further
380 A. C. Walker

practical interest in box beams or plated structures occurred in the next

forty years.
The renewal in interest in buckling resulted from the use of relatively
high strength steel in ship construction. The thin plates used in the hulls
evidently were subject to inplane loading and there arose the requirement
for design information. A theory of plate buckling had been developed and
in 1891 Bryan’ published a paper which provided a solution of the initial
buckling load of a square simply supported plate. This was the precursor to
a large number of papers during the next twenty-five years in which the
critical loads of plates having a wide variety of boundary conditions were
calculated. Pre-eminent among the researchers was Timoshenko.
Although von Karman, in 1910,” had published a paper presenting the
mathematical theory describing the non-linear relationship between
stresses and deflections which exist when a plate is loaded into the post-
buckled regime, no solution was forthcoming until about 1932. Up to that
date, the design philosophy was to assume that plate and column behaviour
was similar and to dimension plates from a knowledge of their critical
buckling load together with the specification of a factor of safety. No
specific allowance was made for imperfections; that is, initial deviations
from flatness.

3. 1930-1950

Throughout this period extensive theoretical work was carried out to

satisfy the demands of aircraft designers who were eager to obtain
lightweight monocoque structures. In Britain, Germany and United States
of America a considerable body of knowledge concerned with the critical
loads of plates was amassed.° Since the interest from aircraft engineers lay
in light weight it became necessary to investigate the post-buckling load
carrying capacity of plates. To this end, a number of experimental
investigations were performed, notably in 1930 by Schuman and Back* and
in 1932 by Sechler.> Complementary to von Karman’s large deflection
theory, an energy approach had been formulated and a single term solution
was derived for the complete non-linear plate loading path. It was evident
from the theory that thin plates, adequately supported along edges parallel
to the direction of loading, should support loads well in excess of the critical
load. Moreover, the deflections which resulted at these loads gave rise to a
redistribution of membrane stresses in a non-linear manner with a
concentration at the longitudinal edges.
 Plate Buckling Research 381

The experiments which were carried out on a variety of materials

generally supported the theoretical predictions and in a brilliant intuitive
proposal von Karman® suggested a simple interpretation of the test results.
He considered that in the post-buckled regime the applied loading is carried
by the plate along two strips then constituting a simply supported plate of
width b, and that the critical stress for this plate would be
iS
30 NES? oe
He argued that the plate width,’ b,, would be such that at failure the critical
stress would equal the material yield stress. In other words the plate would
adjust the internal stress so as to optimise its failure stress. Thus
Tt E
b= Tama Va i
and the ultimate average stress, as a ratio of the yield stress, is
Om a t Eres
gery se =) Ga je-Vz oe
If eqn (3) is put in the form

it is seen that von Karman’s value for the coefficient c would be c= 1.9.
However, as Fig. 5 shows, the value of c is indeed variable and Sechler
0.067 Dural
0:012 Steel
0-050 Aluminum
0-016 Steel
0.052 Steel
ex 0:016
—-Q20 Brass

0:2 0-4 0-6 08 1-0

tye
b 0
Fig. 5. Experimental value of coefficient for effective width.
382 A. C. Walker

Loading bar

specimen

Deflection reading
apparatus
0
0000000
qajiooo0oo00e0

Testing machine
Fig. 6. Testing arrangement.

estimated from his test results that the constant should have a value of
almost 2 for very thick plates but reduce significantly for thin plates. Thus
the simple concept of effective width provided a good basis for the
incorporation of test results; however, it is worthwhile to remember the
basis of these tests. They were carried out on single plates loaded axially by
flat plattens. Sometimes the plate edges were rounded to ensure that the
load was introduced in the mid-plane. The longitudinal edges were
supported in vee grooves or between knife-edges and nominally were free to
pull-in as the plate deflected (see Fig. 6, ref. 3, p. 424). It was generally
assumed that these supports gave conditions which were similar to the
theoretical simple support. However, it is by no means certain that simple
support was indeed attained and the interpretation of the load—deflection
curves using Southwell’s method will not be reliable in assessing the efficacy
 Plate Buckling Research 383

of the experimental boundary conditions. This second factor is discussed

later in the present paper. Despite these reservations, the conjunction of
theory and experiment resulted in a very valuable step forward and
provided much-needed design guidance but, nevertheless, much work was
required subsequently to ensure that the actual relationship between the
ultimate, yield and critical stresses gave as accurate as possible guidance to
engineers.

4. 1950-1965

Subsequent to about 1950 the situation in aeronautics altered, so far as

plate buckling was concerned. Increased engine power resulted in higher
operating speeds for aircraft and it was no longer tolerable from an
aerodynamic viewpoint to have locally deflected surfaces. Thus the concern
within aircraft engineering was to ensure that the geometries of plate panels
was such that they operated below the critical stress range.
At about the same time, civil engineers concerned with the uses of cold
formed sections were developing their own approaches to the evaluation of
effective width. In beams with thin-walled cross-sections made from steel
sheet it was observed that collapse resulted when the flange in compression
failed by buckling. George Winter’ carried out a large number of tests on
such beams and by making measurements of the strains at the edges was
able to derive the following formula for effective width:

b,=1.9t
JicLt-048(5) JerI
/—]1-—0.475{—]
oe b
/—
Gin
(4)

in which g,, is the maximum stress on the plate element. In other words,
Winter showed that the effective width defined by eqn (4) is applicable
throughout the elastic post-buckled regime of the plate and not just at the
ultimate load. This formula acted as one of the bases for design for cold
formed sections in the USA until 1968 at which time it was considered that
sufficient experience had been accumulated to change it to

pai [E[i—oais(t)
Os b
[F
Oe.
(5)
or, in terms of the collapse stress,

#22 (1-022 [%) (6)

09 699 a)
384 A. C. Walker

—=-= €g.1 5): be =1-9tVE (1.0-0.415 +t E )

Om Gm

0 0:2 0-4 0-6 08 10 2 14 1-6

vb ify
b= =Om
Fig. 7. Correlation between the effective design width formula and test data.

This is shown in Fig. 7, where it is seen that there is fair agreement between
the test results of Winter and the design equation.
In a very interesting recent study Roorda® carried out a critical analysis
of a number of effective width formulae relevant to cold formed section
plate elements and concluded that Winter’s form of relationship is the best
of those proposed in the available literature, but commented that some
revision of the coefficients would result in even better agreement with the
published experimental results.
In Britain about this period there was also active interest in the buckling
behaviour of cold formed sections. Research into local buckling was being
actively pursued at Bristol University and in the Royal College of Science
and Technology in Glasgow. Chilver® proposed a formula for plates as
1/3

ow a6 @ (7)
a) a)

and Kenedi et a/.!° showed that in comparison with the result of tests on
some 200 mild steel sections of different shapes the limits of the scatter of
these tests are shown in Fig. 8; eqn (7) defines the average maximum stress
 Plate Buckling Research 385

Scatter of test results

Equation ( 7)

0 1:0 2-0 3-0

[09 /0¢)"?
Fig. 8. Collapse tests on mild steel sections of different shapes.

with reasonable accuracy, while a conservative form of this relation,

approximating the lower scatter boundary, can be written as
1/3

E85 0:66 (*:) (8)

a) Oo

Equation (8) was also shown to be adequate for mild steel single plates with
different types of longitudinal edge support, which suggested that this
expression could be used for a wide range of thin-walled mild steel sections.
Indeed, eqn (8) formed the basis of the calculations of buckling strength
of cold formed sections and was contained in Addendum No. | to BS 449.
The proposed approach was to calculate the strength of a section as the sum
of the individual strengths of the component plates, assuming these to be
simply supported or free on the longitudinal edges. In a later paper‘! it was
shown that the approach could be in error in the range 0.25 < (a,/a9) < 1.
However, this factor was recognised in a subsequent revision of the
Addendum; that design curve is shown in Fig. 9.
It can be seen that up to that time the mechanics of behaviour of single
plate behaviour was fairly well explored’? '* and that design formulations
were based on non-linear analysis and the empirical interpretations of test
results. There was also a growing interest in the development of more
precise solutions of von Karman’s non-linear equations. A particularly
significant paper was published in 1959 by Yamaki'® in which he used a
multi-term deflection series solution and evaluated the coefficient from the
386 A. C. Walker

+ Schuman and Back (ref 4)

0:8 © Sechler (ref. 5 )
Ou
Oo 06

0-4

0:2

1 2 3 4 5 6 i
VO9/ Oc
Fig. 9. Comparison of generalised collapse with test results for simply supported
plates in various metals.

non-linear algebraic equations by numerical means. An important aspect

of Yamaki’s work is that he considers a number of in-plane boundary
conditions, including the situation where the plate is free to pull-in along
the longitudinal edge. The theory therefore provided a valuable yardstick
against which to compare the available experimental results which typically
had attempted to simulate this mathematical boundary condition. Stein,’®
about this same time, had developed a solution to the non-linear plate
behaviour using a perturbation approach. This reduces the mathematical
formulation to a series of linear equations which can be solved sequentially.
Stein showed that the implicit series descriptions was strongly convergent if
the load was used as the perturbation parameter, a factor which was
important in later analyses.

Bearing holder

Loading beam

Edge restraint Make -up piece

( knife edge)
Fig. 10. Loaded-edge at the extremity of the plate.
 Plate Buckling Research 387

Interest was also being shown in improving experimental techniques. A

significant contribution was made by Boley'’ who carried out some very
careful tests on single plates in which the simple support on the loaded
edges was provided by an ingenious arrangement of split roller bearings.
This approach was used in tests at Strathclyde University’® on the effort of
eccentric loading on the strength of rectangular plates. The longitudinal
boundary supports were provided in these tests by knife edges and doubt
still remained with regard to the influence of friction etc. on the plate
behaviour (see Fig. 10).

J 1965-1983

This period has been one of tremendous activity in the research of plate
buckling, as evidenced by the large number of papers published. So far as
steel plates are concerned two factors have driven this work: one is the
phenomenal increase in the power of computers and their ready
availability to researchers; the other factor has been the unfortunate
accident to a number of box girder bridges which caused engineers to
review their state of knowledge with regard to the design of these complex
structures.
Taking the analytical development first, there can be no doubt that the
advent of high-speed digital computers has revolutionised our approach to
plate analysis. Although the mechanics of plate behaviour was well
established, the design approaches were essentially empirical. The
numerical analysis developed during the past fifteen years have permitted
the design of plates having a wide variety of boundary and loading
conditions to be put on to a firmer theoretical basis. Plasticity and non-
linear geometrical deformations can both be incorporated into the
numerical analyses and from a knowledge ofthe initial deflections and state
of stress it is possible now to predict, with acceptable accuracy, the
maximum load. The problems with the communication of this information
from researcher to designer is something I will return to later in the paper.
Two schools of research in Britain have been pre-eminent in the
development of computer methods although, ofcourse, individual workers
elsewhere have made significant contributions. The work at Cambridge
University directed by John Dwight led to the first publications'®:?° in
which the important factors of non-linear deformations and plasticity were
combined effectively in a single numerical analysis scheme. Later work at
Cambridge, particularly by Moxham and Little, developed these ideas to
388 A. C. Walker

produce very significant information concerning the behaviour of plates

with, for example, initial residual stress states. Tom Graves-Smith
confirmed theoretically the long held notion that for thin plates with edges
free to pull-in, the failure load will be reached when the maximum
membrane stress at the edge reaches yield. This had been used in semi-
empirical analyses, e.g. ref. 21, as a means of deriving design curves and it
was very valuable to have theoretical confirmation. Moxham was able to
show from his analysis that as plasticity develops there is a progressive
change in the length of the buckle wavelength. The maximum load for a
plate may coincide with a wavelength significantly shorter than the plate
width.
The other research school of note, with regard to numerical analysis, is
located at Imperial College where engineers, first under the direction of Dr
Chapman and later Professor Patrick Dowling, were active in the
utilisation of computers in plate analysis. Initially the method of approach
was to incorporate elastic non-linear deformations using a finite difference
description of the von Karman equation. The algebraic equations were
solved by a novel approach called ‘dynamic relaxation’. Later, plasticity
was incorporated and then the finite element method was used. Typical
results from this work are contained in refs. 22 and 23.
Valuable input with regard to the concepts of plasticity in plate analysis
were provided by Dr Crisfield** of the Transport and Road Research
Laboratories. During the decade 1967—77 when the various methods were
being developed, a number of international symposia were organised. It
was interesting to hear the various claims and counter claims for greater
accuracy, effectiveness, etc. by the researchers for their computer
programs. Undoubtedly these discussions were very valuable in the
assessment of the published work and reflected a research vigour which was
most refreshing.
It is a potential problem with computer based analyses that the results are
not always easily communicated to other researchers or design engineers.
There was seen to be a need, therefore, for simplified interpretation of the
computer output. One obvious way to do this is to incorporate the results
into acode of practice, and indeed the theoretical work described above has
been used as the basis for the revised Bridge Code BS 5400. The curve for
single plates with edges free to pull-in is shown in Fig. 11 in which it can be
seen that the computer has been used to obtain a good lower bound
description of presently available corresponding test results. An alternative
approach to obtaining simplification of theoretical results has been to
assume that the deformation, stresses etc. can be described by strongly
 Plate Buckling Research 389

8 e

Ou 49 Bx, * Ref 33 Ref 36

— ht ha © Ref 34 O Ref 37
0 2, °° + Ref 20 @ Ref 38
er 20° x Ref 35

Fig. 11. Correlation of effective width and test results.*!

1 (b/t) ¥ (Oo/E) = 1.037

(b/t)
¥ (Oo/E) = 2-074

Simply supported square plates

unloaded edges constrained to
remain straight
Wo/t = 0-188
Rhodes method
— —— Frieze et al. (ref 22)

E /Eo
Fig. 12. Load end shortening curves for constrained plates (wot = 0.188).
390 A. C. Walker

UMOYS

3a

‘asplruquiey
|
AjUO‘ssaBuUI) wo2q

NOILOSS
V-V¥
3oq 12>

dwoj> poDoO7uvoly>024y

+++4+44
42 4 ae
104
ysoddnsvawisads NW

Saket
U2dIWDId
$)0H
uawisads
I
"|

teeh]
a6pzZ

Bulpyong
3
=a!
i /

ie
—/ yy
uF /
/ / /

I. |
/
va saro0s6-A
/

++

++
ae
dbo

Joj ayeid
JUsWUIDsUBLIe
ye
x0qJD26 Ze Ul
ssajjou
buls02q disys
Se
¥4JO}OWM2J2S Le
ce Wier
ajpaau
6
ul} Buispaq
ysNJYy
1adD) uaqD
WOJ} DWIY ‘saping VE
% jmaoy Id
‘26pam
AIIP
ANU Ulz/y-|— VA
{UHL

7
ISO,
oe
BAIIQ ca pe
UIDUD
TWMMLUIIIE
—
“BIA “ET
tee

eae
ttt

Aj0\010qQ07

8
‘q poxig SUd|]OU
ysay DOIG

- asoo7
uld Pd}OIO}
Bulsnlpy ul yoisul Jo6ulj)
MdIIS
GudjD BxOKA SPUB}YXD
O}
YUOU
Y xD0Q
JO Jd6ulj)
ayo, payoddns
AQ DBIpzou
42}j}ou Buiy22foudY6nosy, 4s6uij
BulAOWwBxOK0} 442] saysnd
uid
dn adojs
0} diuh UdBWIDDIS
uo4 podwo)}>d
BOpD BSN OM) suid
PuDdD DAOW see UDI}OU
OS DxOA
Ssy>d0U j1yUN yyOQ Suid Giu6

so6ui4
Plate Buckling Research

yabuly pduuid
©}
Buljsoddns
woaq

B14 ‘pluoo—€]
391
392 A. C. Walker

convergent series.*> Computer methods, such as finite difference models,

are used to determine the coefficients. Thus the problems of publication are
reduced since it is necessary for engineers only to have knowledge of the
coefficients relevant to his particular plate geometry, loading and boundary
conditions to be able to reconstitute, using the series expression, the plate
non-linear behaviour.
This method, as presented, was restricted to elastic deformation,
although some limited surface plasticity would result in little loss of
accuracy of predictions of membrane stress. Recently, however, Rhodes”®
has shown that it is possible to obtain a simple description of the behaviour
of a uniformly compressed plate loaded well into the plastic regime. A
typical result is shown in Fig. 12 where it can be seen that the agreement
between his idea of extending the effective width into the plastic region and
the computer results are remarkable. He is currently developing this
approach in combination with the series method mentioned above. In this
way, therefore, there is a good combination of the power of computers and
the capability of simple methods to communicate a great deal of
information.
It was stated earlier in the paper that the most significant contributions in
plate buckling have resulted when theory and testing have proceeded in
close conjunction. This has certainly been the case in the work in Britain
during the past fifteen years. Both at Cambridge and Imperial College the
researchers were active in experimentation and assiduous in comparing
their theoretical results with the facts emerging from their laboratories. The
Cambridge team carried out a large number of tests on individual plates
and developed a very ingenious device to ensure that the edge conditions
were as close to simple support as possible. This comb-like attachment is
shown in Fig. 13. In addition, the effects of residual stresses were carefully
tested by the introduction of simulated weld passes along the plate edges.
As part of his work investigating the effects of edge restraint on test
plates, Spencer at University College London developed a method akin to
the Southwell Plot used for columns. Figure 14(a) shows the Southwell Plot
interpretation of results from a plate test.2” No accurate information can
be inferred from this plot, simply because plates do not behave in a similar
manner to struts. By a reinterpretation of plate theory, Spencer?® devised a
plot shown in Fig. 14(b); the coefficients H, etc. are parameters ofthe plate
geometry etc. It is apparent that the application of Southwell Method, as
advocated by Timoshenko, may have resulted in some misinterpretation of
early tests, e.g. those by Sechler.
The experimental work carried out at Imperia! College was primarily
 Plate Buckling Research 393

W
Se
a3
<
1 ; Southwell plot

0-10 045
Ah aan seen tye target
(a)

all
Dc = 2850lb ( 12-7 kN)

= 10
E
=
"o

PE
=

W,= 0-03" (0-75mm)

0 H2
Wes
0 0:5 1:0 MM 4.5
(b)
Fig. 14. Southwell Plot and modified Southwell Plot for plate test in ref. 27.

concerned with providing information to ensure the safety of design of box-

girder bridges. Obviously these boxes could not be tested at full scale and it
is one of the major achievements of that research programme that methods
were devised to manufacture at small scale and still incorporate the
important features such as material characteristics, initial geometric
imperfection and residual stresses. Figure 15 shows the arrangement for a
typical test in the Imperial College Laboratories. Concurrently, testing of
stiffened plates was under way at Manchester University under the
guidance of Professor Horne and at Monash University led by Professor
Murray. From this work and the significant contribution by researchers
under the leadership of Professor Massonnet in Liege, has emerged a good
394 A. C. Walker

Fig. 16. Small scale stiffened plate.

 Plate Buckling Research 395

understanding of the behaviour of plates and plate panels subjected to a

variety of loading conditions.
The effects of fabricating plates and panels at small scale were considered
theoretically?”°° and it was found that, provided the model material,
welding and initial deflection were representative of the full-scale
counterpart, the model test results would provide accurate representation
of the strength of practical structures. Small scale modelling of stiffened
plates was developed at University College London to provide a tool for the
wider experimental investigation of stiffener and plate buckling. These
models were carefully welded to provide practical levels of residual stress
and the results indeed proved to be comparable with those from
Manchester, etc.*! Figure 16 shows a typical small scale panel in which
stiffener buckling has occurred.

6. CONCLUSIONS

In this paper I have attempted to highlight a sequence of development in the

understanding of plate buckling. The treatment has concentrated on single
plates with uniform loading but of course the practical range of loading,
geometries etc. is much wider than this. However, investigation of the
behaviour of plates subject to, for example, shear loading or concentrated
loading has paralleled the work on simple compressive loading and
therefore many of my remarks are applicable there also. In this context I
would like to mention the work of Professor Rockey and his co-workers at
University College Cardiff. I regret that shortage of space here has
prevented me from doing justice to the breadth of Professor Rockey’s
contribution to plate buckling research or from presenting the ingenious
simple design methods he developed.
The present state of knowledge on plate buckling is very extensive. If an
engineer has access to a computer and the corresponding plate program he
can obtain information on the buckling and plastic failure characteristics of
almost any plate shape and loading configuration. Alternatively, if his
problem is fairly standard he will obtain adequate guidance from codes of
practice dealing with cold formed sections or box and plate girder bridges.
The information on the buckling of plates made from concrete or
composite materials has also improved considerably in the last decade.
It is evident from the decay in the number of recent papers that there is a
reduction in research interest in plate buckling. Perhaps this is natural since
research interest and funding follows closely the need of industry for
information.
396 A. C. Walker

Unfortunately, this waning in interest does present a problem of

communication and of ensuring that the fruits of the last few years are
gathered together so that engineers may benefit from them. Plate buckling
is still a specialist subject and little is taught in universities or polytechnics,
except in one or two postgraduate courses. Moreover, there are only a few
good text books on the subject. A recent addition* is very welcome but
even there, with 87 pages dealing with plate buckling in which perhaps 750
papers in total have been published, the treatment is still elementary. It may
be that it is too much of a specialist subject for publishers to be interested in
books on plate buckling, but I fear that much of the value of recent work
will be lost to practical engineers who never seem to have enough time to
find and peruse research papers. I hope that the research groups in
universities etc. will continue to find problems of academic and practical
interest to ensure a healthy supply of engineers trained in the knowledge
and application of plate buckling in practical situations.
Two factors seem to me of general importance, which emerge from the
development in plate buckling. The first concerns the balance of the
analytical approach; it is necessary, I think, to have simplified methods
proceeding in parallel with more complex numerical computer based
analyses. The former provide a channel for communication of the more
general results from the computer analyses. The work of Dr Rhodes and Dr
Williams are good examples of such a balanced approach. The second
feature is the collation of research results, and these remarks are
applicable more widely than to plate buckling. The various agencies which
funded the research resulting in the papers counted in Fig. 1 have invested
funds possibly in excess of £3 m. Yet none has to my knowledge suggested
or funded a collation exercise which ensures that all the work of the
previous decade has been put on a common base. Without this exercise
much ofthe potential benefit of the research will be lost and a proportion of
the funding unnecessariiy wasted. Surely a small fraction, say 2-3 %, of the
research funds could usefully be set aside to ensure that such exercises are
properly carried out and that future proposals for research are indeed
dealing with new problems and not repeating previous, uncollated work.

REFERENCES

1. BRYAN, G. H., Proc. London Math. Soc., 22 (1891) 54.

2. VON KaRMAN, T., Enzyklopadie der Mathematishen Wissenschaften, 4 (1910)
349.
 Plate Buckling Research 397

. TIMOSHENKO, S. P., Theory of elastic stability, New York, McGraw-Hill, 1961.

. SCHUMAN, L. and Back, G., N.A.C.A. Report No. 356, 1930.
WwW
&
an . SECHLER,
E. E., Publication No. 27, Guggenheim Aeronautics Lab., California
Institute of Technology, 1933.
. VON KARMAN, T., SECHLER, E. E. and DONNELL, L. H., Trans. ASME, 54 (1933)
Sih,
. WINTER, G., Trans. ASCE, 112 (1947) 527.
. Roorpba, J. and VENTKATARAMAIAH, K. R., Paper No. 147, Solid Mechanics
Division, University of Waterloo, 1978.
. CHILVER, A. H., Civil Engineering and Public Works Review, 48 (1953) 1143.
. Kenept, R. M. et al., Pub. IABSE, 20 (1960) 137.
. Reiss, M. and Cuitver, A. H., Proc. of 8th Congress IABSE, 1968.
. MARGUERRE, K., N.A.C.A. TM No. 833, 1937.
. Koiter, W. T., N.L.L. Report $287, Nat. Aero. Res. Inst., Amsterdam, 1943.
. Coan, J. M., Trans. ASME, 73 (1951) 143.
. YAMAKI, N., J. of Applied Mechanics, 26 (1959) 407; 27 (1960) 335.
. STEIN, M., N.A.S.A., TR R-40, 1959.
. Horr, N. J., Botey, B. A. and Coan, J. M., Society
for Exp. Stress Analysis, 5
(1948) 14.
. WALKER, A.C., in Thin walled structures, A. H. Chilvers (ed.), London, Chatto
& Windus, 1967, p. 208.
. GRAVES-SMITH, T. R., in Thin walled steel structures, C.K. Rockey and H. V.
Hill (eds.), London, Crosby Lockwood & Son, 1967.
. Dwicut, J. B. and RATCLIFFE, A. T., in Thin walled steel structures, C.K.
Rockey and H.V. Hill (eds.), London, Crosby Lockwood & Son, 1967.
. Dawson, R. G. and WALKER, A. C., Proc. ASCE J. of Struct. Div., 98 (1972)
7D.
. Frieze, P. A., DOWLING, P. J. and Hosss, R. E., Steel plated structures, P.J.
Dowling, J.E. Harding and P.E. Frieze (eds.), London, Crosby Lockwood
Staples, 1977.
. HARDING, J. E., Hopss, R. E. and NEAL, B. G., Proc. Inst. ofCiv. Eng., Part 2,
63 (1977) 137.
. CRISFIELD, M. A., Transport and Road Research Laboratory, Rep. LR919,
Crowthorne, 1979.
. WILLIAMS, D. G. and WALKER, A. C., Proc. Inst. Civ. Eng., Part 2, 59 (1975)
763.
. Ruopes, J., Proc. Inst. Civ. Eng., Part 2, 71 (1981) 165.
. SCHLACK, A. L., Proc. Soc. Exp. Strain Anal., 25 (1968) 69.
. SPENCER, H. H. and WALKER, A. C., Experimental Mechanics, 15 (1975) 303.
. CuiLver, A. H., Study No. 6, Stability, Solid Mechanics Division, University
of Waterloo, Canada, 1973.
. Davies, P. and WALKER, A. C., Proc. of Fifth Australasian Conf. on
Mechanics of Structures and Materials, Melbourne, 1975, p. 127.
. ELSHARKAWI, K. and WALKER, A. C., Proc. of ASCE J. of Struct. Div., ST8,
106 (1980) 1695.
32. ALLEN, H. G. and Butson, P. S., Background to buckling, New York,
McGraw-Hill, 1980.
313. MoxuaM, K. E., Report No. CUED/c—Struct/TR3, 1971.
398 A. C. Walker

34, VepA, Y. et al., J. Soc. of Naval Architects of Japan, 140 (1976) 20S.
35% KRAGERUP, J., Report No. 143, Series R, Technical University of Denmark,
1982.
36. BRADFIELD, C. D., J. Constr. Steel Research, 1 (1980) 27.
3). Becker, H. et al., Report No. SSC-217, Ship Structure Committee,
Washington, D.C., 1970.
38. BECKER, H. and Cotao, A., Report No. SSC-267, Ship Structure Committee,
Washington, D.C., 1977.
 20

Buckling of Aluminium Plates in Compression

D. S. MOoFFLIN and J. B. DwiGHT

Cambridge University, UK

ABSTRACT

The paper describes a study of the buckling strength of individual aluminium

plates loaded in uniaxial compression, having unloaded edges simply sup-
ported and free to pull in. The experimental programme comprised tests on
76 specimens in two alloys, conducted in a special purpose rig. All the
specimens were given a controlled imperfection before testing. About halfof
them were welded down the long edges to introduce residual stresses and HAZ
softening. A few were also welded transversely. Two theoretical models were
developed. One was afinite strip analysis based on large deflection elasto-
plastic behaviour, which allowed for the gradual spread of yield through the
thickness. The other was a simple analysis in which the plate was divided into
yielding edge zones of width 0.2b anda buckling central zone of width 0.6b.
This gave remarkably good agreement with the finite strip analysis at a
fraction of the computing cost.

NOTATION

a Length of plate.
b Width of plate.
b. Effective width.
Cy Om/Fo.2 = 5/5.
é Width of residual tension block at one edge.
399
400 D. S. Mofflin and J. B. Dwight

E Modulus of elasticity.
n Ramberg—Osgood formula index.
t Thickness of plate.
Wo Initial imperfection amplitude.
u, v, w, 8 Displacement parameters.
V Volume of plate.
eye Coordinate axes.
B ./ 69 ./Oc, = (b/t)/1.92 ./o9.,/E for aluminium (v = 0.33).
E Applied strain.
£2 Strain at the 0.2% proof stress.
o Stress or average applied stress.
Bo 0.2% proof stress.
Centre zone stress (simple model).
Elastic critical stress.
Modified elastic critical stress (simple model).
Edge zone stress (simple model).
Limiting value of average applied stress.
Residual compressive stress.
Yield stress of elastic-perfectly plastic materials.

1. INTRODUCTION

Previous code treatments of local buckling strength have been largely

based on empirical data. For example in the USA the famous Winter
effective width formula was the result of a large test programme. In Britain
the C, values adopted in the BS 449 light-gauge addendum mainly
stemmed from the work of Chilver, using an empirical aircraft-type
formula to relate o,, to o,,. These relatively crude treatments were employed
because a full solution of the plate strength problem was not then available.
In the last decade rigorous numerical analyses of plate strength have
been developed, taking proper account of large-deflection elasto-plastic
behaviour, and including the effects of out-of-flatness and residual stress.
These have enabled ‘stress-strain’ curves to be generated for individual
plate elements. Copious results have by now been produced, mostly for
internal (i.e. web) elements.!~ ° Outstands have been studied too.° These
computer studies, all for structural steel with a sharp yield, have been
supported by programmes of tests on individual plates.°:!*
 Buckling of Aluminium Plates in Compression 401

Simply Supported
Free to Pull-in

Fig. 1. Problem considered. Uniform end shortening is applied to the plate.

The present research extends this work to material having a curved

stress-strain curve, with particular reference to aluminium. The specific
problem studied is the ultimate buckling strength of a simply supported
aluminium web, treated in isolation, when loaded in uniaxial compression
and with its unloaded edges free to pull-in (Fig. 1). Important parameters
are (Fig. 2): shape of stress-strain curve, initial out-of-flatness and effects
of welding. Two alloys were studied:

6082-TF Fully-heat-treated material having a relatively abrupt knee on

its o—€ curve.
5083-M Non-heat-treatable material with a much more rounded o-¢
curve.

Welding is important because it introduces residual stress and, for the

heat-treated material, HAZ softening. Figure 2(b) shows the typical
pattern of longitudinal stress in a plate with edge welds, in which c defines
the width of each tension block.
The experimental programme’ comprised 76 tests on individual 6mm
plates in these two alloys. Two theoretical models were developed, known
as the finite strip’ and simple'° analyses. Comparable analyses have been
presented by Dier'! and by Little.‘
Current rules for the buckling of light-gauge steel can be criticised in that
they take no account of the form of the stress-strain curve. Often sections
are produced from strip having properties greatly enhanced by cold
reduction. Such strip has a clearly rounded o-¢« curve. Is it reasonable to use
the same design values for it as for conventional steel with a sharp yield?
The work described in this report, although primarily directed towards
aluminium, could throw light on this question.
402 D. S. Mofflin and J. B. Dwight

(d)

Fig. 2. Factors affecting plate strength. (a) Initial out-of-flatness; (b) residual
stress; (c) HAZ softening; (d) stress-strain curve.
 Buckling of Aluminium Plates in Compression 403

3. STRESS-STRAIN CURVE

To study materials having a gradual yield, it is essential to have a means for

describing the shape of the stress-strain curve. We have used the well-
known Ramberg—Osgood formula’? for this purpose:

oO @ \"
é=—, + 0.002 (=) (1)
90.2

which produces a curve which automatically has the correct strain at 0.2%
proof stress level. The knee-factor nis a measure of the shape of the curve, a
high n indicating a relatively abupt knee and a low na more rounded one
(Fig. 3).

0002 €
Fig. 3. Ramberg—Osgood stress-strain curves.

Figure 4 shows typical o—é curves obtained from compression coupons,

compared with Ramberg—Osgood approximations thereto. It is seen that
by careful selection of n excellent agreement is possible.
Table | lists n values for the materials used in the present study. The 6082
alloy had a remarkably consistent shape of o—¢ curve (n = 25-28). The 5083
was much more variable (n = 8-18), due to looseness in the definition ofthe
M condition in the material specification.
 L590 [61 4 «81 8 700°0 LE/9MSSZ
LLI‘0 961 = — 1 €00°0 Le/9NSsz
€LL'0 p77 97 61 1S00°0 c/9MES7
L780 Orz 7 91 b 700'0 €/9TESZ
780 pz 8Z 07 Z 100'0 ¢/9TSSZ
L980 SZ = —= 0500°0 €/9NASZ
5680 $80 097 — = 0z7000 82 167 AL-7809 Sz c/9NSSZ
$660 16 €p 0'Pr 9L00°0 6/VMdS7
1r6'0 98 = = ¢ 700'0 6/VNASTZ
£960 Sr'0 88 = =: 90000 Ol 16 O-€805 SZ 6/VASSZ
D. S. Mofflin and J. B. Dwight

8860 p71 MM Pa € Z00'0 L8/SMSS7Z

6760 al = = S L8/SASSz
1760 SII LE 67 0200'0 8/SMES7
1860 ETI LE 67 $ 500'0 8/STASz
9960 IZI = — ¢$00°0 8/SNasSz
1L6°0 6s'0 IZI — = 71000 ~=SsoWv S71 W-€80S SZ 8/SASSZ
0SL'0 SIZ a *T1 d 12/9702
$$8°0 8r7z M M 8 500°0 Z/9M E07
076'0 L9TZ = a 7$00'0 Z/9Nd07
8960 890 18Z = = C1000R PLC 067 AL-7Z809 0Z Z/9NSOZ
L960 10Z 2 ro S LL/SASOZ
8L8°0 €8I M M 0900°0 L/SMA07
000°I 807 = = 8 500°0 L/S N07
190'T 8S5°0 07 = = £0000 = LI 807 W-€80S 02 L/SASOZ
(,mu/N) (,wu/N) (,wul/ jn’) 1/q
eS g “oO 2 1/9 g/m u eo Ao]j]P —[DuIWON fu
“s][NSoI 4S9} pale1oq “1 eqeL
404
 9/SASOE O€ W-€805 COC 1000
1 SLI 998°0
9/SN0E 1900°0 8S 18Z°0

$80
9/STSOE 100°0
1 9S 0 Ll
9/ST0€ 0$00°0 9IL0
9/SMSOE 100°0
~

Stl
Sbl'0
9/SMA0E $00°0
7

Is]
Sbl'0

IS]
6/VOSOE O-€80S 16 100°0
0 L8 9°°0 0 7$6

O€
Ol
6/VNA0E 1$00°0 18 6880
6/VMA0E 09000 68 1€60
b/9NSOE AL-C809 COC 000'0
$ OVC vv30

0€
NE
b/9NA0E $000
1 VEC c08'0
c0'l

b/9TSOE 1000
~ 87 6LL'0
b/9TEOE 0900°0 LNG 0 cb
b/9MSOE 100°0
0 807 0 cIL
p/OMA0E 700°0
9 SOC 0 COL
1/SASOr W-€80S 000°0
v 0 0€L

OV
91
aor 1/SN

C81
700°0
8
cel

vOL0
£01

1/STd0r
8c

$00°0
$ 1790
1/SM0r 65000 S90
6/VASOr OV Ol 000°0
8 0 vl 0 tv8

16
ag Lil LL

6/VA0r 1$00°0 v6L0

EL

6/VW180r L00°0
¢ 0820
Buckling of Aluminium Plates in Compression

O-€806
0b 6/VM 0900°0 0€L'0
We 99

¢/9NSOr OV AL-C809 C67 L000°0 0

6£9
|I

SC
s/9Nd0r 05000 [p90
Gs

¢/9180r L000 LE9'0

S/9MA0P 900°0
€
L8I 981

€LS'0
Ls/9NS0r 0 c00
€ 9150
L8I a IST
405

(panuijuod)
 406 D. S. Mofflin and J. B. Dwight
aspo

ue) 6690ps0
brs 8eL'0
€19°0 ges'0 plr'0€1s‘0 p8r'0
€7S'0
6SS'0 8LE0oss'0
Srs'0 ple0
Lee‘0
ISI]
0
sooUaP

g 6c 760 99'| 691 rs'l Ir'Z

I JT
(,wu/N) “S00'0
WO ZI 86 66 L9 9 791£91€SISel p0l 86 O1l 69 89 8L OL
~
F/M
(,wu/) sajoUusp
to = 8Z tp = 9] = — 6 {Z += a a = — a M

g
°[00'0
] = ee LY = hes a = Su ce = 67 «1 = — = M
~
g/°M
g/on 710008 0900°0
€s000
ZS00°0 8 LS00°0
1100°0 0500°09500°0
€600°0 6 0S00'0
6900'°0
700°0 500°0 S 700'0
S sajOUAP

~3=
#91 OL 8 bl LI §
aiqeL
‘pluoo—{
y ce
“PlaM
(,wu/A’)
ISIOASUB
700 781 16 767 Z0Z STI 807

SO]OUDP
Aon
W-£805 O-€80¢AL-7809 W-€805 W-£805 W-€805

,
“UW
jDUIWO
1/q 0S 0S (Ns 09 OL i) Q
=
NN SSoUYIY
°0¢'()
“©

6/VM s/9N S/9M9/409

1/SM6/VNa0s L8/SASOL L/ISMAS8
Lr/9TS09
1/SASOS
1/STa0sS ¢/9TA0S 9/STA09
¢/9NSOS 8/SNGO0L L/SNS8 [PUTWION
ssors

‘fou
=

40S 40S 0s 40S

p=
=.
0S00°0
601
009°0
1/Sagos 79b'0 767
M409
9/S
0S00'0
Ce
LZ
€6 AL-7809
6€1
Chic
97
¢$00'0
¢/9NG09
09
SLr'0
=— O€
6700'0
91
7
¢/9M
0
6Ir
409 AL-7809
167
87
8700'0
611
8E°7
€/9NG0L
OL
OIr'0
=a=SLe/9NSOL
€0l
co
ps0 oa
601
8
AL-Z809
062
i
LS00°0
osc
Z/9NAS8
ple
=NM
M
001
L£00°0
Z/IOMAS8
ere sI YOo[G
:SMOT[O
EROG
UOISUd
Paseq
9Bpo
‘spjam
uolsu
sajous
Aavay
‘SP[IM
yoo[q
AA
=ssan
SB
2“°°
 Buckling of Aluminium Plates in Compression 407

Oo

300

ae
6082-TF

200

508 3~m

100

ee 5083-0

Sas Ramberg OS good

~— Experimental

7tar car al ae a ths ee re pores

0.005 €
Fig. 4. Comparison between experimental stress-strain curves and Ramberg—
Osgood approximation.

4. TEST PROGRAMME

A programme of tests on 76 individual plate specimens was carried out, as

reported in full in ref. 7.

4.1. Testing Arrangements

The special plate testing rig (see Fig. 5 and Chapter 19, Fig. 13a) had been
previously developed by Bradfield for similar tests on steel plates.’*
408 D. S. Mofflin and J. B. Dwight

Referring to Fig. 13a of Chapter 19, the special features are the motor-
driven wedge-straining device and the array of‘fingers’ along the unloaded
edges of the specimen. The role of the fingers, which owed their original
development to Moxham, was to provide a determinate edge condition,
namely simple support with free pull-in, without taking any appreciable
part of the longitudinal load.'°
In each test straining continued without pause, readings of load and
specimen shortening being fed into an X—Y plotter. The duration of a
typical test was about one hour. Load was measured using a conventional
load-cell, and shortening by means of two displacement transducers
located either side of the specimen.

4.2. Specimens
The following list summarises the various specimens tested (see also
Table 1):

Nominal thickness 6mm

b/t (B) 20-85 (0.58-2.85)
Length 4b
Unloaded edges s.s., free pull-in
Loaded edges clamped
Material (¢, 5,7) 6082-TF (290-292 N/mm7?, 25-28)
5083-M (125-208 N/mm2, 8-17)
5083-O (91 N/mm?, 10)
Nominal imperfection 0.0015 and 0.0056
Residual stress (¢,) 0-49 N/mm?

The 6082 was all in the fully-heat-treated TF condition. The bulk of the
5083 plates were tested in the as-received M condition (i.e. hot rolled). A
limited amount of the 5083, however, was annealed in the laboratory before
testing (O condition).
All the specimens were given a controlled ‘bump’ (out-of-flatness
imperfection w,) using a hydraulic jack. Springback made it difficult to
control this operation precisely, and Table | lists actual w,-values achieved.
A controlled fusion weld was made down the longitudinal edges of about
half the plates, using a TIG torch mounted on a tractor. The resulting
level of compressive residual stress o, was monitored by taking strain
readings in the middle of each plate with a Weldscan extensometer,!° the
corresponding value of c being then calculated by statics. The results
 Buckling of Aluminium Plates in Compression 409

Fig. 5. Buckled plate in test rig.

(Table 1) show that the heat-treated alloy 6082 tends to suffer less residual
stress than the 5083, as was predicted by Wong.'’ The extent of the HAZ
softening adjacent to the weld was determined by means of a hardness
survey on two of the 6082 specimens, and was found to correspond roughly
to the width c of the residual tension zone.
A transverse weld bead was laid at midlength on ten of the specimens by
the MIG process, the bead being ground off before testing.

4.3. Results
Table 1 lists failure loads for all the specimens in the form of C,(=a,,/do >).
Typical load-shortening curves appear in Fig. 9.

5. FINITE STRIP ANALYSIS

5.1. Capability of the Method

The following is a short description of a finite strip analysis developed to
model the buckling of plates in uniaxial compression.® The method takes
proper account of large-deflection elasto-plastic behaviour, and allows for
the gradual spread of yield through the thickness. It can cater for steel or
aluminium type materials, taking account of initial out-of-flatness and
welding effects (residual stress, HAZ softening). It can be applied to
410 D. S. Mofflin and J. B. Dwight

individual webs or outstands with various edge conditions. It has also been
successfully extended to cover assemblages of plates.
The method uses an energy solution procedure, which provides stable
convergence characteristics and allows unloading to be followed. Here we
describe its application to the specific case of a web simply supported on all
four edges, with the unloaded edges free to pull in and the loaded edges
constrained to stay straight. The plate is assumed to buckle in one half-
wave.
In this application the method has similarity to the earlier analyses of
Moxham! and Little.* The main difference is the splitting up of each
quadrant of the plate into strips, so providing greater flexibility in the
description of the deflected form.

5.2. General Description of the Programme

The symmetrical web is divided into longitudinal strips, which need not be
of equal width. In the present study six strips have been found sufficient
(Fig. 6), the stresses and strain energies being computed at 600
representative points, throughout the plate, as shown.

(a)

Fig. 6. Finite-strip model. (a) Typical strip division; (b) definition of parameters;
(c) monitoring points within each strip.

The Appendix gives the displacement functions used to describe the

deflected form, which satisfy the boundary conditions. Fourier series are
employed for the longitudinal variation in u, v and w, and polynomials for
their variation transversely. For this problem it was found that a single
Fourier term longitudinally (as shown) gave adequate accuracy.
Transversely the polynomial coefficients vary from strip to strip, but
continuity between strips is automatically satisfied.
 Buckling of Aluminium Plates in Compression 411

The program enables a curve of load vs. overall shortening to be

generated, proceeding as follows for each incremental step:

(1) Set a new value for the end shortening.

(2) Compute the deflected shape and hence strains throughout the
plate in terms of the displacement parameters (u,v, w, 0),>.
(3) Calculate stresses, using elastic theory with the Prandtl-Reuss
flow rule. Strain hardening is assumed to be isotropic and is
modelled by the Ramberg—Osgood formula.!
(4) Compute the incremental change in energy A¢( =A¢! + Ad®) and
also its gradient, with respect to the displacement parameters,
where:

Ad’ = change in internal energy

= |[{o} + 7{Ao}]
{Ae} dV
A@g* = change in potential energy of the loading system.
(5) By minimising the energy with respect to the displacement
parameters, using a quasi- Newton minimisation procedure, obtain
the equilibrium deflected shape.
(6) Compute the corresponding stresses and hence the applied load.
This provides one point on the required curve.

The procedure is repeated for further values of end-shortening, and the

complete load-shortening curve is obtained, from which can then be read
the plate strength o,,.

5.3. Imperfection and Welding Effects

At the start of the analysis the plate is given an initial deformation w,
representing the given geometrical imperfection (Fig. 2a). This is of the
same form as the function used for w.
Residual stresses due to edge welds are taken into account by assigning a
width for the two outer strips equal to the width c of the tension zone
(Fig. 2b). Longitudinal stresses in these strips are initially set to oo ,
(tensile), and in the other strips to —o,. When both initial out-of-flatness
and residual stress are incorporated, the plate is not initially in equilibrium.
In such cases a preliminary step is conducted, whereby the residual stresses
are held constant, while the plate deflects to attain equilibrium. The actual
412 D. S. Mofflin and J. B. Dwight

out-of-flatness for welded plates will therefore be very slightly greater than
the values quoted.
HAZ softening in welded plates is dealt with, when necessary, by
assigning to the two outer strips (representing the zones of residual tension)
values of ¢) , and appropriate to the weakened material. This implies that
the extent of the HAZ is effectively the same as that of the tension zones, a
convenient assumption to make until more is known about HAZs.

5.4. Computed Results

Before running the program in earnest it was necessary to decide on a
realistic value for the aspect ratio a/b, corresponding to minimum strength.
A preliminary study was conducted holding w,/a constant, and it was
found that taking a/b = 0.8 produced the lowest ultimate load over a range
of cases. This ratio was adopted for all subsequent runs. It was felt logical to
hold w,/a rather than w,/b constant, since the latter leads to an
unrealistically low a/b value.
The program was used to obtain load-shortening curves for a wide range
of cases, covering different b/t, shapes of stress—strain curve (n value), out-
of-flatness and welding effects. Some typical curves appear in Fig.9.
Further results are given in refs. 8 and 9.
A comparison was made with similar results published by Little,!? over a
wide range of cases. Good agreement was obtained. A direct comparison
with Dier’s computed results'! was unfortunately not obtainable at this
stage owing to the different cases which he ran.

6. SIMPLE ANALYSIS

6.1. The Basic Model

In this section we describe a simple method for determining the buckling
behaviour of simply supported plates in uniaxial compression.'°
Employing simple physical concepts, it enables realistic load-shortening
curves (and hence ultimate loads) to be predicted. Out-of-flatness and
welding effects are readily allowed for. The approach was conceived by
Calladine at Cambridge.
Referring to Fig. 7 the plate, which is taken as square (a = )), is divided
into edge zones of width 0.25 and a centre zone of width 0.65. As axial
straining (€) proceeds the stress in each zone is assumed to be uniform and
equal to o, or o,, as shown. o, in the edge zones follows the material
stress—strain curve with buckling effects ignored, while a, in the centre zone
is suitably adjusted to allow for buckling.
 Buckling of Aluminium Plates in Compression 413

SSS ee

fee ey
Fig. 7. Simple model. Assumed stress distribution.

Consider first mild steel plates (elastic-perfectly plastic) which are

perfectly flat (w) =0) and unwelded. Take the case of a non-compact
geometry (high 6/t) such that o,, < o,, where o,, is the conventional elastic
critical stress. Figure 8 shows how og, and a, are assumed to vary with the
applied strain ¢ for such a plate (curves | and 2, full lines, respectively). It
also shows the variation of the mean stress a,, (=0.40, + 0.60,) (curve 3).
Curve | is simply the material stress-strain curve. Curve 2 follows curve |

Fig. 8. Simple model. o—¢ curves for plate with high b/t. (1) Edge zone; (2) centre
zone; (3) average.
414 D. S. Mofflin and J. B. Dwight

up to ¢6,=6,,, at which point there is a horizontal cut-off followed by a

falling branch. The limit of the horizontal portion is calculated as the strain
at which localised surface yielding begins at the centre of the plate. The
falling branch is obtained by putting o, equal to a reduced elastic buckling
stress a/,, calculated for a plate with the yielded material removed. A simple
model has been devised for this latter calculation.
Curve 3 is the required load-shortening curve for the plate. It is seen to
have an initial post-buckled stiffness of 0.4£, the appropriate value for a
plate with free pull-in. Also, the peak value (¢,,) depends on the width
obtained for the horizontal portion of curve 2.
The dashed curves in Fig. 8 show the response of the same steel plate
having an initial out-of-flatness wy. Curve | is unaltered. In calculating
curve 2 it is assumed that the out-of-plane deflection varies sinusoidally
along the plate, its amplitude w, increasing with o, as follows:

W, =Wo <* (2)

This enables a relation to be obtained between o, and «. In applying eqn (2)

o., is taken as the conventional elastic value in the earlier stages, and as the
reduced value o%, after yielding has begun.
Plates of low 6/t, such that o,, > o,, are studied in a similar manner. In
this case curve 2 can never reach o,, and must of course lie below o,. This
presents no problem provided there is some initial imperfection, however
small.
Residual stresses (Fig. 2b) are readily allowed for by treating them as
preloads. Curve | is obtained by assuming that each edge zone consists of
two parts carrying initial stresses of o, (tension) and o, (compression)
respectively, which must be superposed on the stresses caused by the
applied shortening. Curve 2 simply receives a downward shift equal to o,.

6.2. Application to Aluminium

There is no difficulty in applying the method to aluminium-type materials,
using the Ramberg—Osgood equation (1) to represent the o-e relation.
Curve | simply follows the stress-strain curve as before. In calculating
curve 2 it is necessary, at stress levels above the proportional limit, to take
G,, as the plastic rather than the elastic buckling stress. For this purpose Fis
replaced by a reduced value E” given by

E'=3(E, + £,) (3)

 Buckling of Aluminium Plates in Compression 415

g 1.0 2.0 HO 2.0

Fig. 9. Comparison between theoretical and experimental stress—strain curves.

(a) 5083-M, wy/b = 0.005, non-welded; (b) 6082-TF, wo/b = 0.005, non-welded; (c)
5083-M, w,/b = 0.005, heavily welded; (d) 6082-TF, w,/b = 0.005, heavily welded.
, Experimental; ———-, finite strip; ...... , simple theory.
416 D. S. Mofflin and J. B. Dwight

where E£, is the tangent modulus and E, the conventional reduced modulus.
In order to determine the falling branch of the curve, involving the
calculation of the reduced critical stress a/,, it is assumed that material is
fully effective until the strain reaches oo.,/E and ineffective thereafter.
The effect of HAZ weakening adjacent to edge welds is readily allowed
for by splitting each edge zone into two parts. The outer part, representing
the HAZ material, is given a suitably reduced stress-strain curve.

6.3. Computed Results

A small program has been written, based on the above model. This is
suitable for use on a microcomputer and enables load-shortening curves to
be predicted. An incremental approach is used, whereby stresses are
calculated at some 50 points on the curve.
The program has been applied to a wide range of cases.1° Some typical
load-shortening curves obtained for aluminium are included in Fig. 9.

7. DISCUSSION

7.1. Load-shortening Curves

Figure 9 shows some typical load-shortening curves, experimental and
theoretical, covering unwelded and edge-welded plates. In order to model
the tests as closely as possible, the theoretical curves are based on the actual
values of ¢) , and n, and—for the welded plates—residual stress and HAZ
softening (6082 only). In every case they assume an imperfection of 0.0056
which was the target value for these particular plates. The following
comments are made:

(1) There is a clear difference between the curves for the two alloys,
with the 5082 tending to keep on rising. This is clearly a function of
the o—é curve (Fig. 4).
(ii) For the 6082 the experimental curves exhibit a much steeper
unloading path than the theoretical ones. This is not a defect of the
theory. It is merely that the test specimens were all of the length 4),
whereas the analyses considered a plate only one buckle long
(a =0.8b or b). For a long plate loaded in isolation only one buckle
is critical. The other buckles unload into it after peak load is passed,
producing a more steeply descending curve when plotted in terms
of average strain on the overall length.'?
 Buckling of Aluminium Plates in Compression 417

(ili) It is less easy to explain discrepancies in the rising parts of the

curves, with the theoretical curves tending to become non-linear
sooner than the experimental ones. This is noticeable at high b/t.
Two contributory factors are (a) the test specimens contained a
single bump at mid-length, which would receive some restraint
from the rest of the plate, whereas the theories assumed simply
supported ends; and (b) for the welded specimens, the residual
stresses die away at the ends.
(iv) Apart from these aspects the two analyses give realistic predictions.
The simple theory performs remarkably well, considering the
relative crudity of its model.

7.2. Definition of Plate Strength

A problem arises in defining the ultimate strength of plates in material
having a very rounded stress-strain curve, such as 5083-M. It is seen that in
some cases, especially for the welded specimens, the load-shortening curve
goes on rising without any well defined maximum stress. This tends to
happen when n falls below about 10.
In the present work we get over this difficulty by taking the plate strength
Om aS follows (Fig. 10):

(a) Ifno peak on the load-shortening curve is reached before ¢ = é 5,

take o,, equal to the stress at this ¢.
(b) Ifa peak is observed for ¢ < €, ,, take a,, equal to the peak value.

We believe this to be logical, seeing that a strain of €) ,is deemed to bejust

acceptable under normal (non-buckling) conditions.

7.3. Plate Strength Results

Experimental values of plate strength are given in Table 1. Comparison
with theory is made on a non-dimensional basis in Fig. 11. In the figure the
theoretical curves are based on representative values for the parameters
A, Wo and c, which tended to vary from specimen to specimen. A rather low
value of n(=5) was used in computing the 5083 curves, in order to facilitate
comparison with Little.
It is seen that the theoretical strengths agree reasonably with the test
data, but with a tendency for the finite strip analysis to underestimate the
strength of the 6082, especially if welded.
It can be seen from Fig. 11 that the unwelded plates are fairly insensitive
to the degree of geometrical imperfection. The effects of welding appear to
418 D. S. Mofflin and J. B. Dwight

Fig. 10. Definition of ultimate strength. (1) o—-e curve of material; (2) high-n
material; (3) low-n material.

be more important, the welded strengths being significantly lower than the
non-welded.
The absolute strengths of the 5083 are of course well below those for the
6082, because of its lower proof stress. When, however, the comparison is
made non-dimensionally (in terms of C,) the difference between the two,
although significant, is much reduced. In the low to medium range of f the
tests have C, for the 5083 lying some 10-15 % below that for 6082, although
the theoretical results show a difference of only a few percent.
It should here be noted that, for a given n value, the supposedly non-
dimensional plot of C,, against f is not quite independent of a,,,tending to
shift up and down slightly with og,,.All the theoretical curves presented in
Fig. 9 are based on o, ,= 200 N/mm?. Had the correct value been used
(69.. ~ 290 N/mm? for the 6082, and a, ,~ 180 for much of the 5083) the
effect of C, would have been small, pushing it up very slightly for the 6082
and down for the 5083.

7.4. Effect of Transverse Welds

Included in Table | are results obtained for specimens containing a central
transverse weld. It is seen that such a weld causes no loss of strength for
 a ——-— aLINId dIuLS
TIGHIS AYOFHL Cp© “corre
u gine
g Q/y= ‘IOO°O
ON OTaM
oy ayy= "SCO"O
ON GTaM Isa TWLNAW
a) S
= 5S So S
x
3
S 53

WSSSSEEOS es
Sa cre-s3c---— S
3S
& 1S)i)
nH
=2 s

0°0 S°o o°t Sia 0°? S°%

“By “[] UOSLIedWOD

UDsEM}0q [eoNoJODY]
pue [eJUOUTIIodxe
sJewIN[N ‘syWueNs
(B) ‘W-¢Egos ‘popjem-uou
(q) “41-7809
S
a

(2)
(P)
“POP[oM
:PePiem
“W-€80¢
“41-7809
:Ppoplom-uoU
 ‘pluoo—TT “By
sve oz set
D. S. Mofflin and J. B. Dwight

O = 3/5
100°O = a/¥
aT3M ON ‘SCO'O = a/y ~~}
"INLNAW TYR IXS A
aQTaM ON “100°0O = Q@/9 °

2LINIA
XYOMILL STINIS

dINLS

-——-—
2 ar ee

Oo
420
 ——-—~ aLINIG d1luLs
FIdWIS ANOGHL ie) z°o= 0oz
‘ =u S

g q/y= ‘SOO°O TTWWS Glam

adXa
ITY TWLNAW
Buckling of Aluminium Plates in Compression

Ss‘
“By
| ‘pluod—]
421
 422

-—- 2LINIG dIQLs =o He

@aIGWIS AUORHL yon Sue

»@ WwW» ‘S00°O TIVWS U14dM

~® ay ‘sos DUVI OTAM
| TWINdIuauxg
. ‘

e
“Se. °

cae °
Se
= ie e q/¥= $00°O
ee ° 3/2= €
Sse HLIA ONINSLIOS
ASN e

Sa
D. S. Mofflin and J. B. Dwight

s°0 ov! pel 0°2 Se

“Bi Piuoo—T]
 vt
"
=)

Vv
°
"
Buckling of Aluminium Plates in Compression

0°O s*o0 o°t s*t 0°?

“BY “ZI yun suostiedwios

JOJ speusjew
JO JUIIOYIP
U “SONIPA
423
424 D. S. Mofflin and J. B. Dwight

5083 plates, but does have an appreciable effect for 6082—as one would
expect, due to the HAZ softening which is produced right across the plate.
The reduction on a,, was of the order of 30 % for 6082 plates containing no
other weld, and about 20 °% for plates which already had longitudinal edge
welds. These decreases seem remarkably small, when one considers that
0... for the HAZ material is about half that for the parent metal.

7.5. Comparison with Steel

Figure 12 shows non-dimensional plate strength curves (C, vs. /), based on
the finite strip theory, plotted for different values of the Ramberg—Osgood
index n. They cover unwelded plates having small and large imperfection
(wo/b = 0.001 and 0.005 respectively). The curves are applicable as follows
(approximately):

n= hot finished structural steel (elastic-perfectly plastic);

n=25 fully-heat-treated aluminium such as 6082-TF;
n=5 annealed non-heat-treatable aluminium (5083-0)—
pessimistically.

It is seen that for the large imperfection the three curves are virtually
indistinguishable. At the smaller imperfection there is only a slight
difference between the n = «© and n =25 curves, but an appreciable drop
tonn = 3:
The important question is: ‘can a unified C,—f design curve be used for
all materials? The answer appears to be a limited yes, with some exception
made for materials having a very rounded stress-strain curve. It is probable
that cold-reduced light-gauge steel has a g—-e curve comparable to that of
6082-TF in terms of n value, so that a unified plate strength curve could
cover light-gauge as well as hot finished heavy gauge steel. But this
conclusion is tentative at this stage bearing in mind the considerably greater
difference between 6082 and 5083 which has been observed experimentally.

8. CONCLUSIONS

(1) The finite strip theoretical model gives reasonable agreement with
tests.
(2) The simple theory gives remarkably good results, bearing in mind
its relative crudity.
 Buckling of Aluminium Plates in Compression 425

(3) The tests showed an appreciably lower strength for 5083 plates in
the low to medium range of b/t, as compared with 6082—when
plotted non-dimensionally (C, vs. B). The theories predicted a
much smaller difference.
(4) Plates with longitudinal edge welds have an appreciably reduced
strength due to residual stress and (for the 6082) HAZ effects.
(5) Transverse welds had negligible effect on the strength of the 5083
plates, but reduced the strength of the 6082 plates by some 20-30 %
due to HAZ softening.
(6) Theoretical comparisons suggest that a common non-dimensional
design curve (C,, vs. £3) might possibly be acceptable for unwelded
heavy gauge steel, light gauge steel and heat treated aluminium—
provided the levels of out-of-flatness are comparable.

Acknowledgements
The work formed part of a Ministry of Defence programme of research on
structural aluminium at Cambridge University Engineering Department,
funded jointly by the Admiralty Marine Technology Establishment and by
the Military Vehicles and Engineering Establishment. A major role was
played by Mr M. J. Free in the Structures Laboratory at Cambridge.

APPENDIX

The displacement functions used in the finite strip method for describing
the deflected shape within each strip were (see Fig. 6):

u = f(y) sin (2n¢) + ae(z — ¢)

v =f"(n)cos(2n€) + bve(n — 3) + e(n — 2)
w = f"(n) sin (16)
where
S'(n)=
(CU — nyu, + nu,
F°(y) =U —n)v, +702
fn) = (1 — 3n? + 2n?)w, + (4 — 2? +0°)b9,
+ (3n? — 2n3)w, + (n° — n7)d,0,
and
n=y/b,
€ =x/a
v = Poisson’s ratio
é =applied end shortening/a
426 D. S. Mofflin and J. B. Dwight

REFERENCES

. Moxuam, K. E. and BRADFIELD, C. D., ‘The strength of welded steel plates

under in-plane compression’, University of Cambridge, Department of
Engineering, Report CUED/C-Struct/TR.65, 1977.
. CRISFIELD, M. A., ‘Large deflection elasto-plastic buckling analysis of plates
using finite elements’, Department of the Environment, Transport and Road
Research Laboratory, Report LR593, 1973.
. FRIEZE, P. A., DOWLING, P. J. and Hosss, R. E., ‘Ultimate load behaviour of
plates in compression’, Steel Plated Structures, Proceedings of Conference,
London, 1976, published by Crosby Lockwood Staples.
. HARDING, J. E., Hopps, R. E. and NEAL, B. G., ‘Ultimate load behaviour of
plates under combined direct and shear in-plane loading’, Steel Plated
Structures, Proceedings of Conference, London, 1976, published by Crosby
Lockwood Staples.
. Littie, G. H., Rapid analysis of plate collapse by live-energy minimisation,
International Journal of Mechanical Science, 19 (1977) 725-44.
. Rocers, N. A. and DwiGut, J. B.,“Outstand strength’, Steel Plated Structures,
Proceedings of Conference, London, 1976, published by Crosby Lockwood
Staples.
. MorFLin, D.S.and DwiGut, J. B.,‘Tests on individual aluminium plates under
in-plane compression’, University of Cambridge, Department of Engineering,
report CUED/D-Struct/TR.100, 1983.
. Morr iin, D.S., A Finite Strip method for the collapse analysis of compressed
plates and plate assemblages’, University of Cambridge, Department of
Engineering, report CUED/D-Struct/TR.101, 1983.
. Morr in, D. S., ‘Results obtained from a Finite Strip method for the ultimate
compressed strength of plates and plate assemblages’, University of
Cambridge, Department of Engineering, report CUED/D-Struct/TR.102,
1983.
. CALLADINE, C. and MorFFLin, D. S., ‘A simple method for predicting the
collapse strength of individual plates in compression’, University of
Cambridge, Department of Engineering, report CUED/D-Struct/TR.103,
1983.
. Dier, A. F. and Dow ina, P.S., ‘Aluminium plated structures—plates under
various load systems’, Imperial College, Civil Engineering Department,
CESLIC Report APS2, 1981.
. Littte, G. H., Collapse behaviour of aluminium plates, /nternational Journal
of Mechanical Science, 24, No. 1 (1982) 37-45.
. RAMBERG, W. and OsGoop, R., ‘Description of stress-strain curves by three
parameters’, NACA, Tech. Note 902, 1943.
. BRADFIELD, C. D., Tests of plates loaded in in-plane compression’, Journal of
Constructional Steel Research, 1, No. 1 (1980) 27-37.
. DwiGut, J. B., ‘Collapse of steel compressed panels’, Developments in Bridge
Design and Construction, Proceedings of Conference, 1971, published by
Crosby Lockwood Staples.
 Buckling of Aluminium Plates in Compression 427

16. DENSTON, R. J. and WuirteE, J. D., ‘An electrical demountable extensometer’,

University of Cambridge, Department of Engineering, report CUED/C-
Struct/TR.61, 1977.
17. WonG, M. P. and Dwiacut, J. B., ‘Longitudinal weld shrinkage in materials
having rounded stress-strain curve’, Joints in Structural Steelwork,
Proceedings of Conference, 1981, published by Pentech Press.
 Index

Acrylonitrile butadiene styrene, 359 Bending

Airbag test, 156 capacity of profiled sheeting, 157
Aircraft elements, 138-9
design, 3 in-plane. See In-plane bending
structures, 18 strengths of beam webs, 114
AISI Specification for the Design of unstiffened, 140
Cold Formed Steel Structural Bernoulli's hypothesis, 307
Members, 112-18 Biaxial compression of cylindrical
Allowable stresses design method, 19 shells, 192-5
Aluminium Biaxially stressed plates, 329-53
alloy, 14, 15, 17 analysis, method, 331-3
plates, buckling in compression, basic behaviour, 336-43
399-427 design considerations, 345-51
Amplitude modulation of short-wave elasto-plastic interaction curves, 337
buckling modes, 35-45 experiments, comparisons with, 343-5
ASME III Code, 199, 200, 205 generation of interaction curves,
Aspect ratio, 338, 341, 355-73 335-6
Automotive structural components, imperfections, effect of, 343
116-17 normalised interaction curves, 348-9
Axisymmetric elastic-plastic collapse of numerical results, 333-43
cylinders, 211-33 Bolted connections in cold formed
steel structures, 115
BOSOR 5 shell buckling computer
Beads, 1-20 program, 185
design codes, 18 Boundary conditions, 244
Beam-column analysis, 74-8, 84-8 Boundary constraint effect, 228-30
Beam-column model, 73-8 Box girder
Beam-column representation, 73-4 bridges, 331
Beams, lateral buckling of, 140-1 test arrangement, 394
429
430 Index

Britannia Bridge, 377 Buckling—contd.

British Standard 5400, 349-51 stresses—contd.
British Standard 5500, 205, 206, 237 perfect cylinders, 193
British Standard 5950, 145 predicted minimum, 199-200
British Standards Institution, 19 tests on cylindrical shells, 196-9
Brockhouse ‘Clasp’ system, 110 thin, isotropic, homogeneous, flat
Buckle propagation, 315 plates, 166
Buckled specimens, contour maps of, Bulb(s), 1-20
364, 369, 370, 372 design codes, 18
Buckling function, 3
aluminium plates in compression, unsymmetric, 9, 12
399-427 Bulbed channel sections of uniform
computed results, 416 thickness, 11
finite strip analysis, 409-12 Bulbed flanges, 8-9
finite strip method, 425
imperfection effects, 411-12
load-shortening curves, 416 CIDECT (Centre International pour le
plate strength Developpement et Etude de
definition, 417 la Construction Tabulaire), 23
results, 417 Circumferential band of radial loading,
simple analysis, 412-16 252-4
steel, comparison with, 424 CEifactonel 27128
stress-strain curve, 403 CLM approach, 133
test programme, 407-9 Coated aluminium, 144
transverse welds, effect of, 418-24 Coated steel, 144
welding effects, 411-12 Coated steel strip, 143
anisotropic plates, 167 CODAP Pressure Vessel Code, 191
coefficient, 4-6, 10-13, 15, 19, 128 Codes of Practice, 19, 144-5, 200
cold formed steel structural sections, Cold bending, 71
125-42 Cold finished profiles, 28, 29, 31
cylindrical shells subjected to biaxial Cold finished tubes, 32
compression, 192-5 Cold formed beams, design analysis of,
internally pressurised torispherical 136
shells, 181-2, 185-91 Cold formed industry scene, 93-110
mode design strategy, 100-3
global, 357 developments since 1960, 105-7
local, 357 storage building at Burtonwood,
pipes under combined bending and 94-100
external pressure, 315 Cold formed section(s), 100
plot, 361, 372 compression elements of, 126-32
l-stiffener panel, 362 design rules, 125-42
2-stiffener panel, 362 load and resistance factor design of,
3-stiffener panel, 363 116
stresses, 128 perforated elements, 112
flange/bulb combination, 2 structural behaviour, 113
lipped flange, 2 truss
observed and predicted, garage, 104
comparison of, 368-71 school, 104
 Index 431

Cold formed steel structural sections, Critica] stress —contd.

buckling behaviour of, 125-42 uniform thickness thin-walled
Cold formed steel structures, 111-23 channel or I-section, 5
bolted connections in, 115 Critical wave numbers, 38
continuing education programmes, Cylinders, axisymmetric elastic-plastic
117-18 collapse of, 211-33
research projects, 112-17 Cylindrical shells
stainless steel design manual, 115-16 buckling
UMR research, 111 biaxial compression, 192-5
Cold roof construction, 145-7 tests on, 196-9
Cold-rolling of plates, 71 local loads on, 235-72
Collapse partially filled, 255
load of submarine pipelines, 287
modes, 379
axisymmetric and asymmetric Damping factors, 78
elastic-plastic, 213 Deep cassette profile, 148, 156
pressure of out-of-round tubes, 295 Deflection
tests on mild steel sections, 385 calculations, 155-6
Column buckling function, 171
equations, 23 Design code requirements, 126
stress, 17 Distortion velocities, 276
Columns, overall buckling of, 133-4 Dynamic relaxation, 75, 77
Compatibility equation, 173
Composite foam filled panel, 108-10
Composite plates, stability, 167-8 Eccentrically loaded columns, 58, 60
Composite structures ECSC (European Coal and Steel
imperfections in, 165-78 Community), 23
instability of, 165-78 Edge reinforcement, |, 2, 18
Compressed plates, 82 Edge stiffeners, 128-31
Compression Effective slenderness ratio, 12
buckling of aluminium plates in, Effective width
399-427 coefficient, 381, 389
elements, behaviour of, 126-32 formula, 384
loading, 84 formulation, 126-8
system at nodes of central buckle, Elastic buckling, 1-13
52 design practice, 9-13
Compressional stiffness, 48, 49, 54 stiffened plates, 355-73
Compressive strain ratio, 84 stress, 13
Contour maps of buckled specimens, tests, 201, 203-5
364, 369, 370, 372 Elastic critical stress, 19
Corten steel cold formed sections, 103 Elastic-plastic buckling tests, 202
Creep Elastic-plastic collapse model, 214-21
behaviour ofthin-walled smooth pipe Elastic-plastic large deflection analysis
bends, 273-85 of axially loaded stiffened
flexibility factors, 282 plates, 69
Critical load factor, 42 Elastic-plastic large deflection
Critical stress behaviour of plates, 329
end compression, 3 Elastic stability theory, 3, 37
432 Index

Energy functional, 37, 38, 40, 41 Hutchinson’s modes for spherical

Euler loads, 49, 54, 59 shells, 40-1
European Recommendations, 125, 131,
145, 153-5, 158, 161, 163
Exponential modulating factor, 35 IBBC-TNO, 289, 292, 294
Ilyushin yield criterion, 78
Imperfection(s)
Failure analysis, 65
loads for plain channel columns, 134 approximate fabrication, 223-8
predictions for beams with slender composite structures, 165-78
webs, 139 effects, 411-12
Finite difference discretisation, 77 influence of, 221-3
Flame-cutting of plates, 71 localised, 35, 41-2
Flange—bulb combination, 2 ultimate strength of plate, effect on,
Flange-lip combination, 13 343
Flanged pipe bends, in-plane bending Imperfection-sensitive interactions,
behaviour under steady state 42-3
creep, 273-85 Initially imperfect plates, 167
Flexibility factors, 279-84 In-plane bending behaviour of flanged
Flexural stiffness, 48, 49, 54, 55, 62, pipe bends under steady state
63, 64, 133 creep, 273-85
Fourier expansion solution, 241-4 Instability of composite structures,
comparison with other solutions, 165-78
252=5 Intermediate stiffeners, 131-2
Fourier series, 170
representation
applied loads, 245-7 Kirchhoff hypothesis, 168
local loading system, 247-9

Lateral buckling of beams, 140-1

Galerkin technique, 167, 169 Limit state design, 2, 19
Lipped channel
columns, 58
Hangar door, framing for, 105 sections, 16-18
Header building, 94, 96, 97, 99 uniform thickness, 6, 7
Heat affected zone, 72 Lipped cruciform sections, 13-14
High strength sheet steels, 116-17 Lipped flanges, 2-8
Hollow sections with square or buckling stress of, 2
rectangular cross-sections, 23, Lipped I-sections, 14-16
25-30 Lipped sections, 2, 19
Hooke’s Law, 75 Lips, 1-20
Hoop stress resultants, 183, 184 design codes, 18
Hot finished profiles, 26, 27, 30 reinforcing beads, 8
Hot finished tubes, 32 symmetric
Hot-rolling built-in flanges, 5
eflecta yl hinged flanges, 4
sections, 72 Load-deflection equilibrium, 59, 60
H-profiles, 32 Load-end displacement curves, 171
 Index 433

Load-end shortening Periodic post-buckling paths, 39

curves for constrained plates, 389 Perry—Robertson formulation, 141
relations, 73 Perry—Robertson interaction
Load-out-of-plane deflection curves, curve, 136
173, 174 formula, 133
Local buckling Perry—Robertson type strut curve, 18
deflections, 50 Pin-ended, centroidally loaded, singly
overall member behaviour, 49 symmetric sections, 58
Local loads on cylindrical vessels, Plastic capacity utilisation, 136—7
235-72 Plastic design of submarine pipelines,
Local panel buckling, 36 287-311
Localised imperfections, 35, 41-2 Plastic flow, 72
Locally imperfect plane channel Plastic hinges, 162, 290
columns, 47-67 Plasticity reduction factors, 19
buckled shape, change in, 56-9 Plate buckling
equilibrium behaviour, 64 coefficient, 19
local buckle magnitude, 57-9 papers published on, 376
local form change on equilibrium research review, 375-98
behaviour, effect of, 60, 61, 62 slenderness, 23
overall equilibrium, 55-6 test arrangement, 390
results and discussion, 59-66 Plate panel(s)
theoretical approach, 50-6 analysis, 78
Lock seamed spiral tube, 105-7 parametric study, 81—4
London Airport, 103 strength, design recommendations,
Longitudinal loading, 243, 246, 272 88
Long-wave overall buckling, 42-3 Potential energy
functional, 37
rate function, 278
Prandtl-Reuss flow rule, 78
Metal Roof Deck Association, 144 Prandtl-Reuss relationship, 319
Mild steel sections, collapse tests on, Pressurised vessel with constrained
385 ends, 254-5
Modulating factor, 41 Profiled steel sheeting and decking,
Moment capacity 143-64
hat section, 137 bending capacity at internal support,
variation with 4/t ratio, 137 157
MRDA code, 153 calculation procedures, 153-6
capacity
end support, 158
resisting concentrated loading, 158
Orthotropic plain channel columns, 49 combined moment and reaction, 161
Orthotropic plates correlation of theory with test
initial imperfections, 169-75 results, 159-62
in-plane stresses or lateral loading, future trends, 162
168-76 mid-span bending under distributed
Out-of-balance forces, 75 load, 156
Overall buckling of columns, 133-4 present position, 144-5
Overail instability, 36 testing of, 156-9
 434 Index

Radial loading, 243, 246, 252-5, 261, Stiffened plate(s)—contd.

264, 266, 268 numerically predicted strengths for,
Ramberg—Osgood stress-strain curves, 79, 80
403 shakeout in, 69-91
Rayleigh—Ritz method, 36, 41, 42,52, 57 strength, design recommendations,
Residual stresses, 32, 70-3, 78, 188, 89-90
190, 360 stress-strain curves for, 85-6
see also Shakeout Stiffened profiles, 154-5
‘Richardson process, 77 Stiffener
Ring stiffening deformations, 211-33 dimensions, 132
Rocker element, 290, 306 requirements, 131]
Roof Storage building at Burtonwood, 94-100
decking, 147-9 Strain-hardening, 186, 188, 189
sheeting, 145-7 Strain rate, 278
Stress function, 171
Stress-relief, 72
Safety factors, 205, 206 Stress—strain curves, 73
Shakeout Strut tests, 17
stiffened plates, 69-91 Submarine pipelines
use of term, 73 collapse
Shallow initial curvature, 174 external pressure and axial force,
Shell buckling problems, 179-210 296-7
Shells, torispherical. See Torispherical external pressure and bending,
shells 297-303
Short-wave buckling modes load of, 287
amplitude modulation of, 35-45 inelastic response under external
approximate description of, 36-40 pressure and bending, 313-27
Short-wave local buckling, 42-3 loading sequence, influence of, 303-5
Slenderness plastic design of, 287-311
effect, 340, 341, 343 test programme, 292-4
ratio, 19 theoretical model, 290-2
Snap-buckling
imperfect plates, 172
shallow shells, 174 Tangential loading, 243, 246, 270
Southwell Plot, 392, 393 Tensile loading, 73
Spatial interlacing meshes, 77 Tension pre-load, 83, 88, 90
Spherical shell(s) Thin-walled hollow rectangular
Hutchinson’s modes for, 40-1 sections, 21-33
problem, 35 Three-pinned arches, 100
Stainless steel cold formed structural Torispherical heads, 185
design manual, 115-16 Torispherical shells
Stiffened panels, parametric study, 80-8 internally pressurised
Stiffened plate(s) behaviour, 182-4, 190-1
beam-column analysis, 84-8 buckling, 181-2
elastic buckling of high aspect ratio, residual stresses and shape
355-73 imperfections in, 188
experimental results, 79-80 Torsional flexural buckling, 134-6
numerical studies on, 79 Torsional stiffness, 5, 8
 Index 435

Trapezoidal profiles, 153-4, 160 Wall cladding, 150

Two-pinned arch, 101, 102, 103 Web crippling and bending, 114, 154
Web/flange ratio, 134, 136
Web strength of flexural members,
Ultimate compressive strength, 13-18 113-14
Ultimate load sensitivity, 65 Welded connections, 136
University of Missouri—Rolla (UMR), Welding effects, 71, 411-12
111

Yield stress, 31-2, 72

von Karman compatibility equation,
33), GY
von Karman large deflection plate
equations, 78 Z purlin, 101, 105, 107
Von Mises yield condition, 309, 320 Z section, 101
2) .-ee4e7 GO 1-yeea!
> ===
oe ee ee ee
ee rn
=; -_ oy
6 (ea :
Se oe Ves
moce
be Te i
TITLES OF INTEREST

PLATED STRUCTURES: STABILITY AND

STRENGTH
edited by R. Narayanan
6 x 9”, x + 262 pp., 138 illus.

COMPOSITE STRUCTURES—2
edited by I. H. Marshall
6 x 9”, xvi + 576 pp., 309 illus.

BEAMS AND BEAM COLUMNS: STABILITY

AND STRENGTH
edited by R. Narayanan
6 x 9”, x + 242 pp., 99 illus.

GROUTING IN ENGINEERING PRACTICE

Second Edition
by Robert Bowen
6 x 9”,ix + 285 pp., 41 illus.

ANALYSIS, DESIGN AND CONSTRUCTION

OF DOUBLE-LAYER GRIDS
edited by Z. S. Makowski
8 x 103”, x +414 pp., 271 illus.
WIND EFFECTS ON BUILDINGS, VOLUME 1
Design Applications
by T. V. Lawson
6 x 9”, xii + 344 pp., 152 illus.

AXIALLY COMPRESSED STRUCTURES:

STABILITY AND STRENGTH
edited by R. Narayanan
6 x 9”, x +312 pp., 180 illus.