2 Bolted Connections

UNIVERSITY OF SHEFFIELD ,
Department of Civil and Structural Engineering
COLUMN DESIGN
FOR
AXIAL COMPRESSION
AND
END ROTATION
by Charles MacIan King
A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy
June 2010
SUMMARY
A column design method has been developed for use in braced frames with discontinuous
columns using flexible cap and base plates and floor beams that are either simply supported
or continuous. The method is intended to be used with shallow floor construction with
concrete or steel/concrete composite slabs in which the floor slab occupies the depth of the
floor beams and is fully grouted to the beams so that the slab restrains the full depth of the
beams and was developed to simplify the design of discontinuous columns in frames using
Corus ASB type floor construction without resorting to methods using nominal moments.
Floor beams are therefore designed to carry the floor loads without interaction with the
columns. Columns are designed to resist the floor beam reactions with column end-rotations
equal to the slope of the floor beams at the top or bottom of the column, whichever is the
greater.
The method incorporates the elasto-plastic behaviour of columns subject to axial

compression and large end-rotations. The design procedure uses rectangular stress blocks of
classic plastic cross-sectional resistance and includes the end-rotations of the columns and
the effects of imperfections. Therefore the approach is similar to classic second-order rigid-
plastic analysis but with the important additions of actual end-rotations and an initial
imperfection. The method has been verified by physical tests on full-scale columns and finite
element analysis using non-linear geometry and material properties.
The design method

1. is derived from simple geometry and simple statics
2. accounts for initial imperfections by deriving a design value from the strut resistance
of whatever design code is specified
3. is limited to sections that cannot be affected by torsional, lateral torsional or
torsional flexural buckling
4. is not applicable to sections with thin walls
5. has been verified for use with hot-finished square hollow sections by testing full-
scale columns
The design method is defined in detail and illustrated by a worked example so that it can be
used in normal design practice. Connection design and bracing requirements are discussed to
enable the column design to be used to design complete frames.
ii
ACKNOWLEDGEMENTS
I would like to thank Dr J B Davison for his energy and time given in supervision and
encouragement.
The laboratory tests were funded by both the Construction & Industrial and the Tubes
divisions of Corns and I would like to thank Corns for this funding without which the tests
would not have been possible.
The PhD was partially funded by the SCI and I would like to thank the SCI for this funding.
These tests were conducted in the laboratories of the Department of Civil Engineering of the
University of Sheffield. These were organised by Dr J B Davison and the rig was designed,
constructed and operated by Mr S Waters. It is to the great credit of Mr Waters that very
smooth curves were recorded on the falling branch of the test, including an unloading and re-
loading cycle. I would like to thank both Dr Davison and Mr Waters for doing these tests
with only limited input from me.
I especially thank my wife Stephanie for her generous support and encouragement
throughout all the years of the project.
I would also like to thank following for their contributions:
Dr A R Gent for his original papers on elastic-plastic column stability

Dr Arun Kamtekar for pointing me to the Gent papers
Tom Cosgrove, for suggesting that I should conduct the research as a PhD
David Brown, for his support of the project and help in securing funding and travel time to
Sheffield
Charles King
iii
DECLARATION
Except where specific reference has been made to the work of others, this thesis is the result
of my own work. No part of it has been submitted to any University for a degree, diploma or
other qualification.
Charles MacIan King
iv
CONTENTS
Page No.
SUMMARY ii
ACKNOWLEDGEMENTS iii
DECLARATION iv
FIGURES ix
LIST OF SYMBOLS xv
1 BACKGROUND 1
1.1 Introduction 1
1.2 Braced frames 1
1.3 Beam and column arrangements in traditional braced frames 2
1.4 Braced frames with discontinuous columns 3
2 REVIEW OF DESIGN METHODS 6

2.1 Introduction 6
2.2 Design of traditional braced frames 7
2.3 Orthodox design of braced frames with discontinuous columns 10
2.4 Unorthodox design methods for braced frames 24
2.5 Summary of design methods, orthodox and unorthodox 32
3 THE NEW DESIGN APPROACH 33

3.1 Introduction 33
3.2 Behaviour of columns in reality 33
3.3 New column design approach 36
3.4 Development of a method to apply the new design approach 37
3.5 Effects of breadth to thickness ratios of the walls of the members 54
3.6 Range of column end-rotations 54
3.7 Summary 57
4 ANALYSIS TOOLS 59
4.1 Introduction 59
4.2 Finite slice model 63
4.3 ABAQUS finite element model 65
4.4 Summary 69
v
5 FULL SCALE LABORATORY TESTS 70
5.1 Introduction 70
5.2 The test programme 71
5.3 Test rig 72
5.4 Instrumentation 74
5.5 Test duration 76
5.6 Material properties 76
5.7 Test results 77
5.8 Effect of wall thickness 97
5.9 Summary 98
6 COMPARISON OF TESTS WITH ANALYSIS 100

6.1 Introduction 100
6.2 Correlation factor for test v Abaqus 106
6.3 Straight members with asymmetrical eccentricities 110
6.4 Straight members with symmetrical eccentricities 114
6.5 Effects of breadth to thickness ratios of wall 132
6.6 Correlation factor, C, , used for the design model 136
6.7 Summary 137
7 BREADTH TO THICKNESS LIMITS 139

7.2 Parametric study 143
7.3 Calibration of Abaqus with laboratory test results 146
7.4 Design limits 146
8 VALIDATION OF THE DESIGN MODEL BY A PARAMETRIC STUDY AND

CALIBRATION 150
8.2 Load-rotation path sensitivity 151
8.3 Planes of rotations 155
8.4 Similarity with different slendernesses 155
8.5 Confirmation of the proposed design model 157
8.6 Conclusions 163
9 DESIGN METHOD 165

9.1 Scope of application 165
9.2 Column design model 166
9.3 Beam design 170
vi
9.4 Frame effects 175
9.5 Connection design fundamentals 181
9.6 Avoiding disproportionate collapse 187
9.7 Connection design to avoid disproportionate collapse 190
9.8 Alternative methods to satisfy the regulation 195
10 WORKED EXAMPLE 196

10.2 Structural arrangement 196
10.3 Loading 198
10.4 Slab 198
10.5 Main beams 200
10.6 Column 205
11 CONCLUSIONS 212
11.1 Summary of what has been achieved 212
11.2 Suggestions for further developments 213
APPENDIX A SHS SECTION PROPERTIES 223

A.2 Axial not exceeding 50% squash load 224
A.3 Axial compression exceeding 0.5 Npl • 227
APPENDIX B BS 5950 CALCULATIONS 231

B.1 Buckling resistance of Square Hollow Sections 231
B.2 Single curvature 232
APPENDIX C FINITE ELEMENT MODEL DETAILS 234

C.1 Introduction 234
C.2 Full model 234
C.3 Half model 244
APPENDIX 0 STRUT RESISTANCE FROM INITIAL IMPERFECTIONS 247

0.2 Equilibrium 247
0.3 Derivation of strut formula 248
APPENDIX E TRANSVERSE BENDING OF WALLS 251

E.1 Introduction 251
E.2 Second order elastic behaviour 253
E.3 Elastic behaviour bending about X or Y axes 254
vii
E.4 Effects of plastic flow 256
E.5 Elastic-plastic behaviour bending about X or Y axes 264
E.6 Summary 269
APPENDIX F FINITE SLICE SPREADSHEET 271

F.1 Outline of spreadsheet calculations 271
F.2 Calculation of the 'line of thrust' 272
F.3 Generation of the curvature and the deflected shape 273
FA Calculation of the moment induced by curvature 281
APPENDIX G TABLES OF RESULTS OF PARAMETRIC STUDY 294

G.3 1.5m columns 295
GA 3.0m columns 299
G.5 6.0m columns 305
APPENDIX H IMPROVED ECONOMY WITH VARIABLE IMPERFECTION 308
APPENDIX I DEFORMED SHAPE OF TEST SPECIMEN 311
APPENDIX J CMK 2007 SSRC MODEL 312
viii
FIGURES
Figure 1.1 Typical braced frame 2

Figure 1.2 Columns and beams in a traditional frame 3
Figure 1.3 Columns and beams in a typical frame with discontinuous columns 4
Figure 1.4 Typical column-beam connection 5
Figure 2.1 Beam-column connection with thin end-plates and low axial compression 7
Figure 2.2 Typical column with a beam connected on one side 8
Figure 2.3 Typical column with beams connected on both sides 8
Figure 2.4 Beam on a thin cap plate on a 140 x 140 x 10 SHS 11
Figure 2.5 Column loading and bending moment diagram for double curvature 14
Figure 2.6 Column loading and bending moment diagram for single curvature 15
Figure 2.7 Typical end plate deformations 16
Figure 2.8 Force distribution at a flexible column end plate 18
Figure 2.9 Plastic cross-section showing strips resisting bending moment 20
Figure 2.10 Moment shedding from increasing axial load 25
Figure 3.1 Single curvature with small end-rotation 35
Figure 3.2 Single curvature with large end-rotation 35
Figure 3.3 Single curvature bending in the column 37
Figure 3.4 Double curvature bending in the column 37
Figure 3.5 Triangular approximation to the bending in the column 39
Figure 3.6 Deformation with increasing end rotation 40
Figure 3.7 Comparison of typical actual v plastic stress distributions 41
Figure 3.8 Design bending moment in the column including imperfection 44
Figure 3.9 Comparison of actual v plastic stress distributions at low end rotations 45
Figure 3.10 Plastic cross-section to calculate Nb from Mx only 47
Figure 3.11 Plastic cross-section to calculate Nb from Mx and My 48
Figure 3.12 Mid-height displacements - rotations applied about rectangular axes 49
Figure 3.13 Mid-height displacements - rotation applied at 45 to rectangular axis 50
Figure 3.14 Spread of plasticity - column displacements in X=Yplane 51
Figure 3.15 Mid-height displacements - rotation at 22.50 intervals 52
Figure 4.1 Imperfection and deflections in one plane only 60
Figure 4.2 Deflections in two planes 61
Figure 4.3 Imperfection and deflections in the X = Y plane 62
ix
Figure 4.4 Parabolic residual stress pattern (on all sides) 64
Figure 4.5 Finite element model mesh of full model 66
Figure 4.6 Finite element model mesh of half model 67
Figure 4.7 Bi-triangular residual stress pattern (on all sides) 68
Figure 5.1 Test rig with column in position 72
Figure 5.2 Shoe at bottom of test column 73
Figure 5.3 Spring grip frame, the "thing", for LVDT connections 74
Figure 5.4 Unloading cycle with no friction in the bearings 77
Figure 5.5 Unloading/re-Ioading cycle with no friction and no Bauschinger effect 78
Figure 5.6 Unloading/re-Ioading cycle with no friction in the end fittings 78
Figure 5.7 Unloading/re-Ioading cycle test kc3 79
Figure 5.8 Rising and falling branches 79
Figure 5.9 Bending moment for ram extension 80
Figure 5.10 Rising and falling branches with unloading/re-Ioading cycle 81
Figure 5.11 Bending moment for ram retraction 81
Figure 5.12 Load v mid-height displacement (in-plane) 82
Figure 5.13 Load v mid-height displacement (in-plane), zero to 8mm 82
Figure 5.14 lOmm wall SHS, mid-height displacements v 3mm half-sine 84
Figure 5.15 6.3mm wall SHS, mid-height displacements v 3mm half-sine 84
Figure 5.16 5mm wall SHS, mid-height displacements v 3mm half-sine 85
Figure 5.17 Load v mid-height displacement (out-of-plane) 86
Figure 5.18 Load v mean end-rotation 87
Figure 5.19 Normalised load v mean end-rotation 88
Figure 5.20 kc3 end rotations 89
Figure 5.21 kc3difference from mean end rotation 89
Figure 5.31 kc8difference from mean kc8 end rotation 94
x
Figure 5.34 ketO end rotations 96
Figure 5.35 ketO difference from mean end rotation 96
Figure 5.36 kc6 and ketO end rotations 97
Figure 6.1 End rotations from eccentric load at top of column 103
Figure 6.2 End rotations from eccentric load at bottom of column 104
Figure 6.3 Typical end-rotation v load from tests 106
Figure 6.4 Mean end-rotation v load: test v Abaqus 107
Figure 6.5 Top of the unloading/reloading cycle - kc3, test & Abaqus 108
Figure 6.6 End-rotations at top of unloading path - kc8, test & Abaqus 109
Figure 6.7 End-rotations - kc7, test & Abaqus 110
Figure 6.8 End-rotations - kc8, test & Abaqus 111
Figure 6.9 Asymmetric eccentricity of loading for test kc7 113
Figure 6.10 End-rotations - kc3, test mean & Abaqus 115
Figure 6.17 End-rotations - kc 10, test mean & Abaqus 119
Figure 6.18 Initial end-rotations - kc3, test mean & Abaqus 119
Figure 6.25 Initial end-rotations - kc 10, test mean & Abaqus 123
Figure 6.26 Correlation factors for rotation 125
Figure 6.27 Effect of out-of-plane displacements - kc9, test & Abaqus 127
Figure 6.29 possible initial imperfections - kc7 129
Figure 6.30 Unloading/reloading cycle - kc7, test 130
Figure 6.33 Initial imperfections - effect of residual stresses kc7 132
Figure 6.34 Effect of local buckling - rotation capacity from tests 133
xi
Figure 6.35 Effect of local buckling - kc5, test & Abaqus 134
Figure 6.36 Effect oflocal buckling - kc9, test & Abaqus 134
Figure 6.37 Effect oflocal buckling - kc6, test mean & Abaqus 135
Figure 6.38 Effect oflocal buckling - kc1 0, test mean & Abaqus 135
Figure 7.1 Out-of-plane loads on curved walls 139
Figure 7.2 Transverse bending of walls, both sides in compression 140
Figure 7.3 Transverse bending of walls, one side compression, one in tension 140
Figure 7.4 Increased curvature of wall towards the middle of the wall 141
Figure 7.5 Column in double curvature 143
Figure 7.6 140x 140 SHS in single curvature in rectangular plane 144
Figure 7.7 140x 140 SHS in single curvature at 45° to rectangular plane 145
Figure 7.8 Limiting wall breadth to thickness ratio from Abaqus 145
Figure 7.9 Wall slenderness limitations, test v Abaqus 146
Figure 7.10 Wall slenderness limitations, overall breadth/thickness, Bit 148
Figure 8.1 Load paths for a column resisting axial and bending 152
Figure 8.2 Load paths for a column subject to axial and end-rotation 152
Figure 8.3 Load paths for imposed rotation in the plane of the initial imperfection 153
Figure 8.4 All column lengths, end-rotations about a rectangular axis 156
Figure 8.5 Comparison of imperfections "in plane" with "out of plane" 157
Figure 8.6 Design bending moment in the column including imperfection 158
Figure 8.7 1.5 m columns, eO required 160
Figure 8.8 3.0 m columns, eO required 160
Figure 8.9 6.0m columns, eO required 161
Figure 8.10 Abaqus v CMK. 2009 design model, 3.0m height 162
Figure 8.11 Abaqus v CMK 2009 design model, 1.5m height 162
Figure 8.12 Abaqus v CMK. 2009 design model, 6.0m height 163
Figure 9.1 Instabilities to be avoided in beams at top and bottom of columns 165
Figure 9.2 Wall slenderness limitations, overall breadth/thickness, Bit 167
Figure 9.3 Design eccentricity including imperfection 168
Figure 9.4 Plastic cross-section to calculate No from My only 169
Figure 9.5 Simply supported beams (connections at column face) 172
Figure 9.6 Cantilever and suspended span 173
Figure 9.7 Continuous beams 174
Figure 9.8 Bending moment diagrams used to calculate slopes at columns 175
Figure 9.9 End-rotations assumed where there is no sway 176
Figure 9.10 End-rotations modified by sway 177
Figure 9.11 Columns not deflected to slope of beams 179
xii
Figure 9.12: Column end-rotation equal to beam slope, simply supported beams 180
Figure 9.13 Column end-rotation equal to beam slope, continuous beams 180
Figure 9.14 Typical thin end-plate connection 183
Figure 9.15 Bending moments in end-plate 184
Figure 9.16 Typical beam-column connection for plastic rotations in the beam 186
Figure 9.17 Model assumed in BS 5950-1 tying 191
Figure 9.18 Example beam-column connection 193
Figure 9.19 End-plate yield line patterns 194
Figure 10.1 Structural arrangement 197
Figure 10.2 Slab variable load and bending moment diagram 199
Figure 10.3 Slab variable load for max reaction on main beams 200
Figure 10.4 Two-span slab 200
Figure 10.5 Main beam load and bending moment 201
Figure 10.6 Plastic stress blocks in column 208
FigureA.1 Plastic stress distribution, low axial load 224

FigureA.2 Plastic stress distribution, high axial load 228
Figure C.1 Node numbering of comer nodes of SHS 234
Figure C.2 Imperfections and nodes to apply eccentric load 235
Figure C.3 Node numbering of rigid bodies at ends 236
Figure C.4 Wall 1, node and element numbering 237
Figure C.8 Node numbering at mid-height and inclinometers 241
Figure C.9 Node numbering of SHS - half model 244
Figure C.1 0 Imperfections and eccentric load application - half model 245
Figure C.II Node numbering at mid-height and inclinometers - half model 246
Figure 0.1 Imperfections; eo at zero load and CN at load N 248
Figure E.l BMD with equal stresses in the compression and tension flanges 255
Figure E.2 Schematic comparison of (elastic strain) v (elastic strain + plastic flow) 260
Figure F.l Bending moment in column 271
Figure F.2 Displacements of nodes from curvature at node "i" 276
Figure F.3 Correction of displacements 277
Figure F.4 Member shape at ends 278
Figure F.5 End-shape from curvature in end sections 280
Figure F.6 Residual stress distribution 281
xiii
Figure F.7 Average compression stress 282
Figure F.8 "Elastic" bending stresses 283
Figure F.9 Stress diagram for one wall of width 2X 286
Figure F.IO Possible stress distributions 288
Figure F.ll Correction of axial force resisted 289
Figure F.12 Calculation of bending moment resisted 292
Figure Gl Mid-height displacements - rotations at 22.5° intervals 294
Figure H.l eO for failure load for rotations in a rectangular plane. 309
Figure H.2 eO for failure load for rotations not in a rectangular plane 310
Figure 1.1 Deflected shape of test columns 311
xiv
LIST OF SYMBOLS
The symbols listed below exclude the symbols from Appendices F, H and J.
A gross cross-sectional area
amplification factor
A 100 amplification factor at a distance of 100 mm from the point of
maximum curvature
Amax amplification factor at the point of maximum curvature
Ab area required, if stressed to yield stress, to resist the buckling
resistance of a pin-ended strut
As tensile stress area of bolt (EN 1993-1-8)
B external breadth of SHS
B breadth of a solid section (eg slice of wall)
b distance mid-plane to mid-plane of walls of SHS
breadth of the external flat face of SHS, between the rounded
comers (BS 5950)
breadth of section (AISC)
breadth of cap-plate or base-plate
C curvature of member
maximum member curvature at the cross-section under
CErn
consideration
C IOO maximum curvature of member at a distance of 100 mm from the
point of maximum curvature
Cmax maximum curvature of the curvature
curvature when longitudinal strain is at 1.0 yield strain
c flat portion of compression element for classification of
compression elements (EN 1993-1-1)
half-depth of plastic stress block in "webs" resisting axial force
half-breadth of plastic stress block in "flanges" resisting bending
moment
correlation factor
D Depth of section (BS 5950)
depth of a solid section (eg a slice of wall)
factor relating transverse stress in wall, 02, to the curvature C in the
xv
equation 02 = DC
d 4
factor in formula for deflection 0 = d pb
E1wall
Depth of wall available to resist bending in plastic cross sectional
analysis
E Young's modulus
design eccentricity of load application on column from neutral axis
e actual eccentricity of load application on column from neutral axis
Eccentricity of column at mid-height from end-to-end line
edge distance of bolts in cap-plate or base-plate
initial imperfection along one rectangular axis

initial imperfection in the plane of buckling,
geometrical imperfection increased to give strut resistance reduced
by both the geometrical imperfection and the residual stresses
eccentricity of load at bottom of column
Additional eccentricity of column at mid-height from end-to-end
line to represent the effect of imperfections
amplified geometrical imperfection in the plane of buckling
design deflection along one rectangular axis
imperfection at mid-height of a column that makes the column
resistance at zero end-rotation equal to the buckling resistance of a
pin-ended strut
eccentricity of load at top of column
Cr x end-rotation calculated by Abaqus
e theta
F factor relating longitudinal strain in wall, E), to the curvature C in
the equation E\ = FC
Fe Applied axial compression
FtRd bolt tensile resistance (EN 1993-1-8)
Fy yield stress (AISC)
fmax maximum compressive residual stress
fr maximum compressive residual stress
fu minimum specified ultimate tensile strength (EN 1993-1-1)
fub minimum specified ultimate tensile strength of bolt (EN 1993-1-8)
fwvd design strength of fillet weld (EN 1993-1-8)
fy minimum specified yield stress (EN 1993-1-1)
xvi
factored yield stress which is the minimum specified yield stress
reduced by the appropriate material factor for cross-sectional
resistance in the specified design code.
For example, using
EN 1993-1-1, fyd = f/YMo,
BS 5950-1, fyd = Py,
AISC, fyd = Cl>F y,
h Height of column
I Second moment of area of member
second moment of area/unit width of wall of SHS
radius of gyration (EN 1993-1-1)
k axial stress ratio
k2 bolt resistance factor (EN 1993-1-8)
L Member length
La distance from face of column to bolts in cap-plate or base-plate
Lcr buckling length (EN 1993-1-1)
LE Nominal effective length of member (BS 5950)
[elf Half-wave effective length
M Applied moment
Mb moment at bottom of column
Mbs Moment of resistance, major axis, for simple construction
(BS 5950)
M Applied moment
Mcx Cross-sectional moment resistance about the major axis (BS 5950)
MEd Moment in member (EN 1993-1-1)
Mext Externally applied moment on column, for example from wind load
MLT Moment about major axis causing lateral torsional buckling
(BS 5950)
Mpl,Rd plastic moment of resistance of cross-section subject to bending
alone (EN 1993-1-1)
Mpr plastic moment of resistance of section reduced by axial force
Ms Reduced plastic moment of resistance of cross-section when
resisting axial compression equal to the flexural buckling resistance
of a pin-ended strut
Mt moment at top of column
Mx Applied moment, major axis (BS 5950)
xvii
My Applied moment, minor axis (BS 5950)
m Moment gradient factor (BS 5950)
mLT Moment gradient factor applied to major axis moments for lateral
torsional buckling (BS 5950)
mx Moment gradient factor applied to major axis moments for flexural
buckling (BS 5950)
N Axial force
Nb buckling resistance of a pin-ended strut using the specified design
code
Nb,Rd buckling resistance of a pin-ended strut (EN 1993-1-1)
Ncr Elastic critical buckling load, Euler load
NEd Axial force in member (EN 1993-1-1)
Npl squash load of section
Nu Load at point of unloading measured in the laboratory tests
No resistance from new design model with imposed end-rotation 9
Ne,Rd resistance from new design model with imposed end-rotation

9 (EN 1993-1-1)
n number of yield strains
p axial compression resisted
tension in bolts in cap-plate or base-plate
Axial compression resistance (BS 5950)
Axial compression resistance, buckling about major axis (BS 5950)
radial pressure on wall of SHS due to curvature and longitudinal
compressive stress
inward radial pressure on wall of SHS due to curvature and
longitudinal compressive stress
outward radial pressure on wall of SHS due to curvature and
longitudinal compressive stress
radial pressure on wall of SHS due to curvature and longitudinal
tensile stress
p radial pressure on wall of SHS due to curvature
PRC radial pressure on wall of SHS due to curvature
PRC, I 00 radial pressure on wall of SHS due to curvature at distance of
100 mm from the point of maximum curvature
PRC,max radial pressure on wall of SHS due to curvature at distance the

point of maximum curvature
xviii
py yield stress (BS 5950)
Q prying force on the column base-plate or cap-plate from tension in
the bolts connecting the column to the beams
R radius of curvature of loaded SHS
r radius of gyration (BS 5400-3)
plastic modulus, major axis (BS 5950)
plastic modulus, minor axis (BS 5950)
wall thickness
thickness of cap-plate or base-plate
V Vertical shear in beam (BS 5950)
V Vertical shear on the column base-plate or cap-plate from tension in
the bolts connecting the column to the beams
VH Horizontal shear (sway shear) across column
W total uniformly distributed load on member
We elastic modulus in the plane of buckling
WpI plastic modulus of section (EN 1993-1-1)
w uniformly distributed load (force/unit length)
y distance from elastic neutral axis to the extreme fibre of the cross-
section in the plane of buckling
Ym displacement of member from end-to-end straight line
Z elastic modulus (BS 5950)
Zy Elastic modulus, minor axis (BS 5950)
a non-dimensional imperfection factor (EN 1993-1-1)

~ Ratio of end moments on member (BS 5950)
~w strength factor for fillet welds (EN 1993-1-8)
'YMO Partial safety factor for cross-sectional resistance (EN 1993-1-1)
'YMI Partial safety factor for buckling resistance (EN 1993-1-1)
'YM2 Partial safety factor for net area tension resistance (EN 1993-1-1)
Partial safety factor for bolt and weld resistance (EN 1993-1-8)
deflection,
displacement of wall at mid-width of wall
alOo displacement at mid-width of wall at 100 mm along the axis of the
column from point of maximum curvature
amax displacement at mid-width of wall at point along the axis of the
column where the curvature of the column is a maximum
xix
total displacement at mid-width of wall, including effects of
plasticity
Ot,IOO total displacement at mid-width of wall, including effects of
plasticity, at 100 mm along the axis of the column from point of
maximum curvature
Ot,max total displacement at mid-width of wall, including effects of
plasticity, at point along the axis of the column where the curvature
of the column is a maximum
&:t. &:2, 3£3 plastic flow vector (Calladine)
&:2b plastic flow vector for plastic strain increasing the curvature of the
wall
£ yield stress factor
~ ~275 in BS 5950
Py
~ ~235 in EN 1993-1-1
fy
£2b plastic strain that increases the curvature of the walls
£2bc elastic bending strain that increases the curvature of the walls
Ey yield strain
11 non-dimensional imperfection factor (BS 5400)
S End-rotation of column
and/or slope of beam
~ End-rotation at bottom of column
SeA Mean value of end-rotation calculated by Abaqus at the load Nu
equal to the load in the laboratory tests at the point of unloading
Mean value of end-rotation measured in the laboratory tests at the
point of unloading
End-rotation at top of column
Lcr .
slenderness of member = - In EN 1993-1-1
i
and = L in BS 5400
r
normalizing factor for slenderness = 1t fK (EN 1993-1-1)

Vfy
xx
nonnalized slenderness = - A~
Al
= -y
Ocr
cro yield stress (Canadine)

stresses along orthogonal axes 1, 2 and 3 (Canadine)
cr' I, cr' 2, cr' 3 deviatoric stress vector (Canadine)
crc Longitudinal compressive stress in wan
crmax Maximum elastic stress in column

Mean axial stress
Elastic bending stress at extreme fibre
elastic critical buckling (Euler) stress
yield stress
shear stress
sway angle of frame (EN 1993-1-1)
factor in buckling resistance calculation
=0.5l1 + a~ - 0.2)+ ~2 J(EN 1993-1-1)

Joint rotation (Gent & Milner)
reduction factor for buckling resistance
xxi
xxii
1 BACKGROUND
1.1 Introduction
A new form of braced frame has appeared in Britain for residential construction and has
already been used for frames up to 14 storeys high. In these frames, the columns are
discontinuous. Each column segment is only one storey high and is fitted with end plates to
bolt to the beams below and above. Columns are square hollow sections with the smallest
possible size so they can be hidden in the thickness of the walls. The beams are continuous,
passing uninterrupted through the column lines, giving the efficiency of continuous beams
with minimum fabrication cost. Due to continuity of the beams across the line of the
columns, some rotation at the top and bottom of the column is likely to be induced under
certain loading arrangements resulting in curvature of the column. This may reduce the
resistance of the column below that of an equivalent pin-ended strut. Traditional steel frame
construction does not use discontinuous columns.
This thesis describes the development of a method for calculating the strength of columns in
"braced" frames with discontinuous columns. This new method can give more economical
column sizes than other methods and is easy to apply by designers by use of a simple column
resistance model.
1.2 Braced frames

Braced frames are the most common type of frames used in UK beam and column steel
frames. A typical braced frame is shown in Figure 1.1. The distinguishing feature of a
braced frame is that the structure has a discrete system of parts that stabilise the structure
against failure in a sway mode. The bracing in a "braced frame" need not be diagonal
bracing as shown in Figure 1.1, but can be provided by any other stiff structure. Common
alternatives are concrete shear walls, concrete cores and stiff portal frames.
Figure 1.1 Typical braced frame
1.3 Beam and column arrangements in traditional

braced frames
In traditional braced frames, the columns are continuous and the beams are connected
between the columns. This is shown in Figure 1.2. In the UK, the connections between the
beams and the columns are almost always intended to transfer vertical shear only from the
beam to the column. This is because these connections are cheaper to fabricate than
connections that also transmit bending moment and the cost of the connections is a
significant proportion of the cost of a steel frame. In 1993 the cost of forming the
connections was quoted as 30% of the cost of a typical steel frame even when the cost of
handling is excluded [Fewster et al 1993], and although the cost of steel might have
increased even more than labour costs for skilled labour, it is clear that the connections
contribute a major part of the cost of steel frames.
2
Figure 1.2 Columns and beams in a traditional frame
1.4 Braced frames with discontinuous columns

A typical frame using discontinuous columns is shown in Figure 1.3. Much larger frames
have been constructed, up to 14 storeys high. Each column piece is only one storey high. To
provide a shallow construction depth, the floor is supported on the bottom flange of
asymmetric beams with a wider bottom flange. The floor may be composite construction
using deep profile decking or may be pre-cast concrete. . Being built-in between the beams,
the floor stabilises the beams, avoiding the potential local instability of the beams caused by
the discontinuity of the columns, as shown in Figure 9.1.
Potential advantages of this type of construction are:
1. Shallow floor construction.
2. Beam continuity achieved with inexpensive connections.
3. Slender columns that can either be hidden in walls or are oflow visual impact if not
hidden.
4. Safe and easy crane hook access when lifting in pre-cast concrete floor units or
metal decking because the columns do not extend above the floor beams until the
next storey is erected.
3
Potential disadvantages are:
1. Storey height column lengths give more individual column pieces to lift, so more
crane time for column erection.
2. Continuous beams give greater pIece weights for the beams, so the crane
requirements might be increased.
3. Column piece labelling must be carefully controlled if sections with different wall
thicknesses but of the same column size are used.
Figure 1.3 Columns and beams in a typical frame with discontinuous columns
Typical column to beam connections are shown in Figure 1.4. The column end plates are
welded to the columns and bolted to the beams. Figure 1.4 does not show the web stiffeners
that are used to prevent web buckling and web sway buckling. Where there are no secondary
beams orthogonal to the main floor beams at the columns, steel strut/ties are provided for
lateral restraint to the columns. These strut/ties are placed within the depth of the floor beam
and connect the columns on one line of beams to the corresponding column on the adjacent
line of beams as shown in AD 281 [SCI 2005a] and AD 283 [SCI2005b].
4
Figure 1.4 Typical column-beam connection
(floor and stiffeners to the beam web omitted for clarity)
5
2 REVIEW OF DESIGN METHODS
2.1 Introduction
The design methods available in the literature are reviewed in this Chapter. These methods
include the current UK. method for designing traditional braced frames and methods
available for the design of braced frames with discontinuous columns.
Design methods comprise the complete process of selecting a fonn of construction,

calculating the relevant forces and moments, choosing appropriate members and performing
the necessary calculations to verify the design. A design method is not simply the
verification of the resistance of a member. Therefore, the design methods reviewed below
are from BS 5950-1 because it and its predecessor BS 449 provide advice on how to apply
resistance checks to different fonns of construction such as cap-plates and columns in
"simple construction". The Eurocodes do not provide this type of advice and the advice
currently available for use of the Eurocodes in these fonns of construction is based on the
experience gained from earlier codes such as BS 5950-1 and BS 449.
There are two issues of special importance in frames with discontinuous columns.
The first issue is the stiffness of the column-beam joint. At the top of the building, the axial
compression in the columns is small. If relatively thin column end-plates are used, the
connections are flexible. In these cases, illustrated in Figure 2.1, the beam can rotate relative
to the columns. This results in higher sagging moments in the beams than calculated in a
rigid frame analysis. Either the analysis needs to include the joint flexibility or thicker end-
plates must be used to reduce the flexibility of the connections. At the bottom of the
building, the axial compression is high and this compression clamps the columns and beams
so that very little rotation of the beam relative to the column is possible. Therefore, to use
elastic analysis of the continuous frame, the designer must either determine the stiffness of
the joints (which means including the effect of axial compression) or specify end-plates so
thick that the joint is sensibly rigid even for low axial compression.
6
Figure 2.1 Beam-column connection with thin end-plates and low axial compression
The second issue is the effect of bending moments in the columns on the compression
resistance. In a rigidly jointed frame, the bending moments in the columns calculated by
elastic analysis can be so high that they cause a significant reduction in the resistance to axial
compression. To compensate for this, larger column areas are required. Where the larger
area is achieved by larger cross-sectional dimensions, the bending stiffness of the column is
greater, attracting even more bending moment from the frame. This may lead to heavy
columns, in contrast with the common preference of occupiers to have small column cross-
sections, either to allow the columns to be hidden in the walls or to limit the visual impact of
the columns when exposed.
2.2 Design of traditional braced frames

The design process for traditional braced frames, commonly known as "simple
construction", is given in BS 449 [BSI 1969] and BS 5950-1 [BSI2000a]. The process is
described for application to the Eurocodes in the Access-steel document SN005a-EN-EU
[Access-steel SNOO5a 2005]. The process is as follows:
1. The beams are designed as simply supported beams spanning between the columns.
2. The effects of pattern loading are ignored and the columns are designed assuming that
all the beams are fully loaded.
3. The columns are designed for the axial compression, from the sum of the shears from
the beams plus the column weight, and a bending moment. In UK practice, this bending
moment is calculated on the assumption that the beam shear, V, is applied to the column
at an eccentricity of 100mm from the face of the column to which the beam is
connected. A typical case for a column with beams connected to one side is shown in
7
Figure 2.2. A typical case for a column with beams connected to both sides is shown in
Figure 2.3.
4. Any relationship between the beam end rotations and the column rotations is ignored
because the beams are assumed to be pin ended.
D/2
D~~
0
0
0
0
0
i~
:
0
0
0
0
0
0 :
v
- ,...0
... - t- o
i~
- t- 0 :
0
0
0
:
0
0
0 i
::
0
0
0
0
•
Figure 2.2 Typical column with a beam connected on one side
0/2
D~~
:T
o
o
o
o
o
o
o o
o o
1====_=rI~ ~1'Fr-=====I
- t- -+r-
- 1- -tr--
I====:::::q: o
:1J=====l
o
o o
o o
~o
Figure 2.3 Typical column with beams connected on both sides
It can be seen from Figure 2.2 that a column with a beam connected on one side is designed
to resist, in addition to the axial compression, a bending moment applied to the column of
V(100 + 0/2) caused by the beam shear V. It can be seen in Figure 2.3 that a column with
beams connected on both sides is designed to resist, in addition to the axial compression, the
nett moment (VI - V2)(100 + 0/2) caused by the beam shears VI and V2 • Where the beam
shears V I and V2 are equal, the design moment is zero. The design method in BS 5950-1
[BSI2000a] is made as simple as possible by designing only for the case of maximum load
on all of the beams, so avoiding extra load cases from pattern loading.
8
To investigate the efficiency or inefficiency of this method of design and to allow a
comparison with the design of discontinuous columns, a calculation for a typical external
column is made below.
Assuming typical residential construction with columns on a 6.0 m x 7.5 m grid and a total
factored floor load of 10 kN/m2, the beam reaction is
v = 0.5(10 x 6.0 x 7.5) = 225 kN Eq2.1
Assuming a 140 x 140 Square Hollow Section (SHS), the total nominal applied moment on
an external column is given by:
Mx = 225(0.l00+0.l40/2) = 225 x 0.170 = 38.25 kN-m Eq2.2
This total moment is resisted in part by the column segment above the beam and the column
segment below the beam. Assuming equal storey heights above and below, the moment on
each segment of column is 38.25/2 = 19.1 kN-m.
The interaction equation in Clause 4.7.7 of BS 5950-1 [BSI 2000a] for use with traditional
simple construction is:
Eq2.3
Mbs is the buckling resistance moment for simple columns which in the case of a square
hollow section is equal to the plastic moment of resistance because square hollow sections
are not prone to lateral torsional buckling. Therefore
Mbs = Py X Sx = 355 x 246 x 10-3 = 87.3 kN-m Eq2.4
In the absence of minor axis bending, My, the equation can be re-written as follows:
Eq2.5
and re-written again as
Eq2.6
Assuming the column height is 3.0 metres, the column resistance Pc according to the SCI
Blue Book [SCI 2001] for a 140 x 140 x 10 SHS in S355 steel is 1550 kN:
9
Fe ~Pe(l- Mx J=1550(1-19.1)
Mbs 87.3
= 1550(1-0.219) = 1550xO.781 = 1211kN Eq2.7
This is a reduction to 78% of the strut buckling capacity. It should be noted that this is the
value for an external column. An internal column with equal span beams on opposite sides
of the column and equal loads on the beams may be designed on the assumption that there is
no moment applied to the column according to BS 5950-1.
The above summarizes column design for buildings using continuous columns and
traditional UK "simple construction".
2.3 Orthodox design of braced frames with

discontinuous columns
2.3.1 Currently available design methods
The "simple construction" design method described in Section 2.2 is not appropriate for
frames with discontinuous columns because of the structural form. In these frames, the
beams are not even approximately pin-jointed at the face of each column. Therefore the
method described in Section 2.2 is not applicable.
The design methods that are currently in use in design offices for multi-storey braced frames
and that could be used for braced frames with discontinuous columns are:
1. Simple buildings using the method for a beam supported on a cap plate
2. Rigid jointed frame using elastic analysis
3. Flexibly jointed frame using elastic analysis
4. Rigid jointed frame with plastic analysis using member checks from current design
codes
These methods are discussed below. The discussion is illustrated by calculations using
140 x 140 x 10 SHS sections in all cases. One section size is used throughout to simplify
comparison of results for different methods of calculation. The method of calculation is not
affected by the size of section used.
10
2.3.2 Beam on a cap-plate - simple approach
The simplest design approach is to design the structure using the traditional methods for a
beam on a cap plate, as shown in Figure 2.4. This design method is given in both BS 449
[BSI 1969] and BS 5950-1 [BSI2000a]. BS 5950-1 Clause 4.7.7 says:
"The nominal moments applied to the columns by simple beams or other simply supported
members should be calculated from the eccentricity of their reactions, taken as follows:
For a beam supported on the cap plate, the reaction should be taken as acting at the face of
the column, or edge of packing if used, towards the span of the beam."
This is shown in Figure 2.4 in which the column cap plate projecting beyond the face of the
140 x 140 x 10 SHS column is assumed to be too thin to support the reaction of the beam.
Figure 2.4 Beam on a thin cap plate on a 140 x 140 x 10 SHS
The effect of the cap plate design method on the calculated column resistance is evaluated
below for the typical case ofa 140 x 140 x 10 SHS with a storey height of3.0 metres.
The column is checked by using the equation in BS 5950-1 :2000 Clause 4.7.7 because that is
the clause where the cap plate design guidance is found. This equation is:
Eq2.8
From Figure 2.4, Mx is equal to the axial load in the column multiplied by the eccentricity
which is half the column width.
Mbs is the buckling resistance moment for a simple column and is equal to the plastic
moment of resistance for a square hollow section.
In the absence of minor axis bending, My, the equation can be re-written as follows:
Eq2.9
II
Eq 2.10
Substituting the values for this example:
F < 1 = 3
1
c - (_1_ + 0.070) 0.645 X 10- + 0.802 X 10-3
1550 87.3
1 = 691kN Eq 2.11
1.447 x 10-3
This is a reduction to 69111550 = 45% of the strut buckling capacity.
The above applied the cap-plate design method to an internal column without unbalanced
loading on the beams. If this method were applied, all columns would be very expensive.
2.3.3 Beam on a cap-plate - more exact approach

To try to find a more economical value of calculated resistance, a designer could use the
more detailed checks in BS 5950-1 :2000 Clause 4.8.3, as shown below. This requires checks
on two different types of resistance:
1. Cross-sectional resistance.
2. Buckling resistance.
The buckling resistance checks are made below for cases of both single curvature and double
curvature to show the effect of moment gradient.
The calculations are made using the equations for a square cornered section equivalent to a
140 x 140 x 10 SHS developed in Appendix A, which gives the reduced section properties
and resistances for square hollow sections with classic rectangular plastic stress blocks. The
equivalent section has a perfectly square centreline, avoiding the complexities of rounded
comers with varying thickness, with the dimensions and thickness calculated to give the
same area and second moment of area as the real section. For this reason, the overall
dimension taken in this worked example is 129.4 mm, which is the distance between the
centre-lines of the walls of the equivalent section instead of the 140 - 10 = 130 mm that
might be expected for a 140 x 140 x 10 SHS.
12
Firstly, find the compression resistance from the cross-sectional resistance checks:
Squash load for a 140 x 140 x 10 SHS = 355 x 5090 x 10-3 = 1807 kN Eq 2.12
Assuming that the beam bears in line with the external face of the column, the eccentricity of
10ade=70mm
Try case for axial compression < 0.5 x squash load and using the formula in Appendix A.2.2,
Eq 2.13
(The calculated half-depth depth Ca of the compression stress-block is less than 129.412 =
64.7 mm, confirming that the axial compression is less than half the squash load.)
The axial compression resisted = P = 4tcpy = 4 x 9.83 x 62.1 x 355 = 867 kN. Taking the
strut buckling capacity = 1550 kN from the "Blue Book" [SCI 2001], this is a reduction to
867/1550 = 56% of the strut buckling capacity.
The moment resisted with the assumptions made above is:
P e = 867 x 0.070 = 60.7 kN-m Eq2.14
To investigate if this end moment might occur, the end-rotation necessary to induce this
moment is calculated by traditional elastic analysis:
Assuming the pattern of loading shown in Figure 2.5, the end rotation of the column from
the end moments of 60.7 kN-m would be:
6
e = ML = 60.7x10 x3000 = 0.0104 Eq 2.15
6EI 6x205000x14.2x10 6
This end slope is smaller than the end rotation of many beams, so it is probable that the
moment and axial load combination calculated above would occur in external columns in
many frames. The rotation at the internal support of continuous beams might be less than
this, in which case the moment would not be so large and the coincident axial resistance
would be higher than the 867 kN calculated above. However, sway deformations of the
structure from notional horizontal forces or wind loads will induce some end rotation.
13
Secondly, find the compression resistance from the buckling resistance checks assuming
double curvature as shown in Figure 2.5.
Figure 2.5 Column loading and bending moment diagram for double curvature
The check is made to BS 5950-1:2000 [BSI 2000a] Clause 4.8.3.3.3(a). Only the first of the
three checks in 4.8.3.3.3(a) is needed, the check for major axis in-plane buckling, because
the section is an SHS. The equation in 4.8.3.3.3(a) is:
Eq 2.16
Pex = 1550 kN from the Blue Book [SCI 2001]

Mx = 60.7 kN-m calculated above
m = 0.4 for double curvature from BS 5950-1 :2000 Table 26
Eq 2.17
. F < (1-
•• e - (
mr.'t~' JJ
m M
p =
ex
1-0.278
1+0.5xO.278
1550kN
Eq 2.18
1+ 0.5 x x
Mex
:. Fe :::;; 0.722 1550 =0.634x1550 =983kN Eq 2.19

1.139
Therefore, in cases ofdouble curvature, the column resistance is reduced to 63% of the strut
buckling capacity.
14
Thirdly, find the compression resistance from the buckling resistance checks assuming
single curvature as shown in Figure 2.6
P ex = 1550 kN from the Blue Book [SCI 2001]
m = 1.0 for single curvature
Eq2.20
rnxMx
:. F S;
( l-
Mex
J p = 1- 0.695 1550
e (
1+ 0.5
rn M
x x
J ex 1+0.5xO.695
Mex
= 0.305 1550 = 0.226x1550 = 350kN Eq 2.21

1.348
Therefore, in cases of single curvature, the column resistance is reduced to 23% of the strut
buckling capacity.
Figure 2.6 Column loading and bending moment diagram for single curvature
The above calculations show that the cap plate method will give low values of calculated
column capacities, with only 56% of strut buckling capacity from the cross-sectional
resistance check and a possible reduction to 23% for single curvature in the column from the
buckling resistance check.
15
The above calculations have ignored the practical issue of cap-plate stiffness. The England
and Wales Building Regulations [ODPM 2004] require vertical tying in all structures of
Class 2B and Class 3, as explained in Section 9.6, so the column end plates and bolts must be
designed to resist significant tension forces. Being designed to resist these forces, the
column end plates will be sufficiently thick to have significant stiffness, so the bending
moments in the columns will be increased. The increase could arise from the stiff end plate
moving the point of application of the load out beyond the face of the column. It could also
arise from the bolts on the "tension" face of the column-beam connection carrying a tensile
force, so that the connection is a rigid connection even where the axial compression is low.
The deflected form of a typical plate is shown in Figure 2.7. The effect of end plate
thickness would have to be included in the calculations .
.."1.
I: !':
I' I '
Ii i i
I' , :
I
Ii ! :
"" I
<i
I
I 1
I
I
~
Figure 2.7 Typical end plate deformations
2.3.4 Rigid jointed frame using elastic analysis

It is possible to design the frame as a rigid jointed frame. Elastic analysis predicts large
bending moments in the columns. These are caused by the end rotation of the columns
enforced by the rotation of the beams at the beam-column connections. Using current steel
design codes, large bending moments reduce the axial compression resistance of the
columns, making it undesirable to have rigid joints. In fact, the bending moments will be
less than those calculated by the commercial software commonly used in design offices.
This is because commercial software assumes elastic behaviour of the members but, in
reality, the members will yield in some zones. This yielding reduces the bending moments in
the columns in braced frames because it reduces the stiffness of the members. (This is not
the same for the sway moments in un-braced frames because the sway moments are resisting
applied horizontal loads.)
16
It is seen in the calculations for the cap plate method in Section 2.3.3 above that for double
curvature bending in the columns, which means the beams above and below the column
deflect in the same way, the column capacity is reduced to 56% of the strut buckling capacity
for beam rotations as little at 0.0104 radians (0.6 degrees). This shows that rigid frame
design with elastic analysis will often give big reductions in calculated column capacity.
Designs with large bending moments in the columns are also undesirable because
transmitting large bending moments between the columns and the beams is expensive. This
is because they are transmitted by local bending of the beam flanges. These may need
extensive stiffening, depending on their thickness and the geometry of the end-plates on the
columns.
2.3.5 Flexibly jointed frame using elastic analysis

To reduce the bending moments in the columns calculated by elastic analysis methods, the
end-rotations of the column must be reduced. This is possible if the analysis can account for
the flexibility of the beam-column connections explicitly.
The moment transmitted through a simple end plate of a column is not simply a function of
the stiffness of the column and the stiffness of the end plate. The purpose of flexible end
plates is to avoid transmitting the full rotation of the beam to the column. However, the
difference between the rotation of the beam and the rotation of the column opens a gap
between the members as shown in Figure 2.8. This makes the connection similar to the cap
plate connection shown in Figure 2.4, but with some additional moment caused by the
bending of the end plate from the bolts through the plate.
Therefore, the bending moment transmitted is a function of both the axial load and the width
of the column, and the stiffness of the end plate, making analysis with flexible end-plates
complicated. The moment from the load at the face of the column is largely unaffected by
the magnitude of any difference of rotation of the beam and the column, but it is directly
affected by the direction of the rotation. The moment from the end plate stiffness is directly
affected by the difference of rotation of the beam and the column. At small beam rotations
or large column axial forces, the axial force in the column will be sufficient to press the
column and the beam together, so the joint will act as a rigid joint. It requires relatively
sophisticated software to model this because it needs capabilities like "gap" elements.
Without "gap" elements, the designer would need to calculate the stiffness after making a
preliminary calculation of the axial forces and the beam rotations and then re-analyse the
17
structure. Even then, it is possible that the stiffnesses and bending moments would be found
to be only approximately consistent. Such approximations are not rare in elastic global
analysis in which section sizes are modified as a result of the analysis, but they sometimes
add to the burden of the designer to demonstrate to the checking engineer that the analysis
and design are adequate.
Large arrows indicate compression
Small arrows indicate tension
Figure 2.8 Force distribution at a flexible column end plate
2.3.6 Rigid jointed frame using plastic analysis

Braced frames may be designed as rigidly jointed frames using plastic analysis together with
member checks from design codes. BS 5950-1:2000 [BSI2000a] Clause 5.7.2
Independently braced frames is applicable to the design of such frames because they fulfil
the bracing requirements. The ASCE manual Plastic Design in Steel [ASCE 1971] includes
the plastic design of rigidly jointed braced frames and EN 1993-1-1 includes provisions for
plastic design.
Where either the slope of the beams is large or the columns are stocky, the moment in the
columns will reach the plastic moment of resistance. Where the axial compression is low,
the load path on the tension side of the beam-column connections would need to be strong to
transmit this moment to the column. However, where the axial compression is large, the
columns will yield extensively so that the ends will rotate until they bear against the beam
across the whole of the column cross-section. This will occur because the column load will
act at the face of the column if there is a difference in the end rotation of the column and the
rotation of the beam, as shown in Figure 2.4. This induces a bending moment that will cause
the end of the column to rotate.
Plastic design is a very attractive design philosophy for this type of frame because it does not
regard the formation of a plastic hinge as "failure" of the frame. The potential advantage of
18
this in the design of a braced frame is that as the axial compression increases, the plastic
moment of resistance is reduced. Potentially, this allows the column to be efficient at
resisting axial compression by not being required to resist high coincident moments.
The potential improvement in calculated column capacity using plastic analysis is shown
below for the case of a 140 x 140 x 10 SHS. In this example the plastic section properties
are calculated not using the centre-line assumption of Annex A. This alternative method is
used so that this method is demonstrated.
Firstly, the reduced plastic moment is calculated.
The axial compression is 1095 kN, which is less than the squash load of 1807 kN. The area
not used to resist axial compression has a squash load of 1807 - 1095 = 712 kN. In classic
theory of plastic cross-sections, this is represented by a strip, with depth ds, either side of the
section, as shown in Figure 2.9 for bending about the yy axis. Therefore the depth of each
strip either side of the area resisting compression is:
3
d = !(712X10 = 7.16mm
s 2 355x140
J Eq2.22
and the plastic moment of resistance reduced by coexistent axial compression is:
712
-(0.140-0.00716) = 356xO.13284 = 47.29 kN-m Eq 2.23
2
(Altenatively, the stress blocks in Figure A.2 might be used.)
19
b
cis
·I~· y
I
I
i
------j------
.:1. d,
+'-'-1-'-'- b
I
I I
I I
I I
I I I
I I
Y I I
I
m=-------,~-J-----r-,;-;:
plastic stress diagram
= area resisting bending moment
Figure 2.9 Plastic cross-section showing strips resisting bending moment
Secondly the buckling resistance is checked.
The check is made to BS 5950-1:2000 [BSI 2000a] Clause 4.8.3.3.3(a). Only the first of the
three checks in 4.8.3.3.3(a) is needed, the check for major axis in-plane buckling, because
the section is an SHS. The equation in 4.8.3.3.3(a) is
Eq2.24
Pcx = 1550 kN from the "Blue Book" [SCI 2001]

For double curvature with equal and opposite end moments, m = 0.4
1095 + OAX47.29(1 +0.5 1095 )

1550 87.3 1550
=0.7065 + 0.2167(1 + 0.3532) =0.7065 + 0.2932 =1.00 Eq2.25
In this case, the column resistance is reduced to 1095/1550 = 71% of the strut buckling
capacity. This is a higher calculated resistance than found by the other methods. It is higher
20
than the cap plate method firstly because that method assumes that the load is applied at the
face of the column, so the cap plate method gives a higher calculated applied moment for a
given axial load, and secondly because the interaction formula accounts for combined axial
compression and bending more accurately than the simpler formula.
The above calculations conform to BS 5950-1, but calculations to Eurocode 3 follow the
same principles.
2.3.7 Sensitivity of plastic design to moment gradient

Plastic design checks are sensitive to the bending moment gradient. Because the section is
allowed to have extensive plasticity, even small additional bending moments from strut
action may be sufficient to cause the member to buckle. This is shown below for the same
140 x 140 x 10 SHS. In this example, the applied axial load is assumed to be 910 kN. This
is just above 50% of the squash load of 1807 kN. Therefore the area resisting the bending
moment remains in the flanges of the section.
Firstly, the reduced plastic moment is calculated.
The axial compression is 910 kN, which is less than the squash load of 1807 kN. The area
not used to resist axial compression has a squash load of:
1807 - 910 = 897 kN
Therefore the depth of each strip either side of the area resisting compression as shown in
Figure 2.9 is:
3
d =.!.(897 10 J= 9.02 mm
X
s 2 355 x 140
Eq 2.26
Therefore the bending resistance is:
897 (0.140 - 0.00902) = 448.5 x 0.13089 = 58.70kN-m Eq2.27

2
Secondly the maximum moment gradient factor, m, is found.
Checking the buckling resistance to BS 5950-1 :2000 Clause 4.8.3.3.3(a):

Pex = 1550 kN from the "Blue Book" [SCI 2001]
Fe = 910 kN defined above
21
Eq2.28
1
910 )
(
= 1550 87.3 = 0.4131.487 = 0.475 Eq2.29
1+0.5 910) 58.7 1.294
( 1550
This demonstrates how the calculated column capacity is reduced from 1095 kN = 71 % of
strut capacity to 910 kN = 59% of strut capacity by a change of moment gradient factor from
0.4 to 0.475. This is the moment gradient factor for the end moment ratio, p, of -0.625
according to BS 5950-1:2000 Table 26. This means that the end moments are opposite (so
the column is in double curvature) and the lesser end moment has a magnitude of 62.5% of
the greater end moment. The sensitivity to bending moment gradient means that the method
gives inefficient designs for columns in single curvature.
A problem arises using this design approach in the case of uniform single curvature. Uniform
single curvature is not a common design case in traditional UK. simple construction, but only
because BS 5950-1 Clause 4.7.7 says that it is not necessary to consider pattern loading.
This is because traditional simple construction uses beams that are effectively simply
supported, spanning between the column faces. However, in rigidly jointed frames, pattern
loading is required by both BS 5950-1:2000 (see Clause 5.7.2) and by the ASCE manual
[ASCE 1971]. A member that has equal and opposite moments at the ends is in uniform
single curvature. If the end moments were the plastic moment of resistance, even if that is
the plastic moment reduced by coexistent axial load, it is not possible to resist axial
compression. This is because the member is fully plastified over its entire length, so that
there is no resistance remaining to resist the strut action bending moments. The inability of
the method to accept single curvature bending moments of nearly constant magnitude means
that the method cannot be applied successfully to all frames.
22
2.3.8 Summary of orthodox methods
The advantages and disadvantages of the orthodox methods are:
Simple buildings using the method for a beam supported on a cap plate.
The method is simple to understand. The application is complicated by the practical effects
of end plates of sufficient thickness to provide tying capacity to satisfy building regulations.
This necessitates the addition of a moment from the end plate bending arising from the
difference between the beam slope and the column end-slope. The calculated column
capacities are relatively low.
Rigid jointed frame using elastic analysis.
The method is simple to understand and simple to apply through normal commercial frame
analysis software. Where column end moments are large, the calculated column capacities
are relatively low and there may need to be relatively expensive connection details to be
consistent with the analysis model.
Flexibly jointed frame using elastic analysis.
To analyse such frames rigorously requires considerable effort, making them economically
unattractive to design offices. To analyse them approximately may lead either to
underestimate of the column moments, giving unconservative design, or to overestimating
the column moments, giving design like a rigidly jointed frame. Because the method lies
somewhere between the cap plate method and the rigid jointed method, it is expected that the
calculated column capacities will be relatively low.
Rigid jointed frame with plastic analysis.
BS 5950-1:2000 [BS 2000a] Clause 5.7.2 Independently bracedframes allows plastic design
of such frames because they fulfil the bracing requirements. This clause allows the plastic
hinges to develop in the columns because the frames are braced against sway. Where the
columns are in double curvature with almost total reversal of moment, the method is more
efficient than other methods. However, where the columns are in single curvature, or nearly
single curvature, the method predicts low resistance to axial load.
23
2.4 Unorthodox design methods for braced frames
2.4.1 Unorthodox methods that might be used
There are other design methods that might be considered for braced frames with
discontinuous columns. These are:
1. Gent & Milner's 1966 & 1968 column research
2. Wood's 1973 plastic design method
3. Alpha-pin from the University of Sheffield
4. AD 283 from the SCI
These methods are discussed below.
2.4.2 Gent & Milner's 1966 & 1968 column research
Gent published a paper [Gent 1966] in 1966, followed by a second with Milner [Gent &
Milner1968] in 1968, describing tests on small scale steel I-section columns subject to an
initial end-rotation and then to increasing axial compression while the end-rotation remained
applied.
In the tests, end rotations were imposed at the two ends of the columns by moments applied
to a spring system which simulated the action of beams in a frame and applied end moments.
This system allowed the moment to reduce as the column ends rotated, just as the end
moments of a fixed-ended beam reduce if the end restraints are allowed to rotate. Initially the
column had no axial load applied. The axial load was then increased and the end moment
resisted was measured. The experiments showed that as the axial load was increased, the
yielding of the column allowed such large end rotations that that the columns "shed" the
moments, as shown in Figure 2.10.
24
Axial compression
Figure 2.10 Moment shedding from increasing axial load
As the axial compression was increased the end moments reduced to zero and then changed
to acting in the opposite direction to some small amount before the member failed by
flexural buckling in the plane orthogonal to the plane of the web of the column, even when
the end rotations were applied in the plane of the web.
It is important to understand that the end moment was applied by the spring system
simulating the behaviour of a beam in a frame. It was not applied as an eccentric load on the
column. Application as an eccentric load does not allow the moment to reduce as the column
ends rotate and this is the design case assumed in codified checks of resistance to combined
axial compression and bending.
Gent and Milner's tests showed that, at Ultimate Limit State in braced frames with rigid
beam-column connections, the column behaviour is significantly different from elastic
behaviour. The reason is simply that, by definition, the effects of plasticity are not included
in elastic analysis. In reality, the stiffness of the members is modified by the extent of the
plasticity. The reduced column stiffness limits the bending moment in the columns and also
allows more severe curvature in the columns. When moments are applied to the columns by
the beams, the reduced column stiffness allows increased rotation of the column ends,
tending to shed the applied moments provided that the beams can resist the moments shed by
the columns. Gent and Milner observed "that even under biaxial bending restrained columns
have a remarkable capacity to sustain high axial loads by shedding end moments".
In Gent 1966 Part II: An approach towards a rational and economic design method for
office type structures, Gent wrote that "By considering limiting cases in this way, the design
25
of the beams and the columns could largely be divorced". The papers propose a possible
approach to design, but it is not developed into a complete method. The design approach
uses the elastic critical buckling load of the part of the column that remains elastic, but does
not consider residual stresses. The papers refer to the issue of geometrical imperfections
arising from the curvature induced by the imposed end-rotations and the effect on the
compression resistance, but a method to consider different magnitudes of imperfections is
not given. Therefore, this research gives very important insights into the behaviour of
columns in braced frames, but it does not give the designer a complete procedure.
2.4.3 Wood's 1973 plastiC design method

Wood published a method [Wood 1973] devised for frames in which the columns are
assumed to be connected at each floor level by rigid connections to both major and minor
axis beams. The beams on the major axis of the columns are designed plastically and the
beams on the minor axis of the columns are designed to remain elastic. In columns resisting
a large bending moment, the column has an extensive plastic zone which has zero stiffness
because of the plasticity. The compression resistance of the column is calculated by finding
the portion of the cross-section that remains elastic and calculating the Euler buckling load
of this elastic section. This is the same philosophy as used by Gent to estimate the resistance
of columns.
There are problems in using this method.

1. The method uses a calculation of column resistance that is not connected to any
other codes that designers or checkers might recognise. This is a great disadvantage
for a design method because those asked to authorise the design do not have the
comfort of extensive use of the method. Instead, it conflicts with the normal code
design methods.
2. The method has not been verified against many types of columns, so there is a
question of confidence in the results. The columns tested by Gent were I-sections
but were small laboratory test specimen with flange and web breadth to thickness
ratios far less than in common structural sections. This gave them unusually high
torsional stiffness and unusually high resistance to local buckling.
3. Imperfections are recognised in normal modem codes by using a buckling curve

appropriate to the type of member. Wood's method does not explicitly recognise the
effects of imperfections. This might be because the method is suitable for even the
worst level of imperfections, but it is not clear.
26
2.4.4 Alpha-pin from the University of Sheffield
The Alpha-pin Method [Gibbons 1990, Kirby et al 1992, Gibbons et al. 1993, 1993a,] was
developed at the University of Sheffield with extensive analytical studies and full scale tests.
The method was developed for braced frames in which
1. The columns are continuous and the beams are connected between the columns as
shown in Figure 1.2 and
2. There is some stiffness in the beam-column connections
The method allows the columns to be designed for the axial compression alone, taking the
effective length of columns as 1.0 times the height between floors. The design moments on
the column are taken as zero even where the beam column connections are stiff. The
explanation for the neglect of the applied moments is that the detrimental effects of the
transmitted moments resulting from connection stiffness are outweighed by the beneficial
effect of the restraint afforded by the beam. The method is very attractive because designers
often need to have beam column connections with significant stiffness. The connection
stiffness may be explicitly desired either to reduce the risk of instability during erection or to
benefit from some continuity of the beams for greater strength and stiffness. In other cases,
the connection stiffness may result from the geometry of the connections, either where the
connections must be designed to resist large tying forces to avoid disproportionate collapse,
or where diagonal bracing is connected into the beam-column connection, resulting in a large
beam-column connection.
For internal columns of buildings with a regular column grid, the Alpha-pin method is
identical in its effect to the UK "simple construction" method given in BS 5950-1 :2000
[BSI 2000a] because in these cases the nominal moments are equal and opposite, giving the
same results.
The Alpha-pin method was developed for frames with continuous columns as used in frames
of traditional simple construction and the analysis and physical testing were conducted on
this type of frame. Therefore this method is not immediately applicable to frames with
discontinuous columns.
2.4.5 AD 283 from the SCI
The construction of large frames with discontinuous columns of up to 14 storeys in height

attracted interest and concern. The Steel Construction Institute published a series of
27
Advisory Desk articles in the magazine New Steel Construction, AD 281, AD 283, AD 285,
AD 288 and AD 292 [SCI 2005a, 2005b, 2005c, 2005d and 2005e], to give guidance on the
use of discontinuous columns in residential construction. The following issues needed to be
addressed:
1. Application of the method needs to be as simple to apply as possible, because
complex methods either dissuade designers from using the method or encourages
designers to do something simpler that might not be safe.
2. The method needs to lead to member sizes that can be seen to be reasonably
economic. If they do not, either the structures will never be built or designers will
devise their own methods that give more economical members but might not be safe.
3. The method ought to agree with other design methods where possible, or ought to be
similar in philosophy.
In preparing guidance, the following information was reviewed:

1. Some discontinuous columns were reported to be designed assuming that beam-
column connections acted like "nominally pinned" bases. This means that they were
designed with a low connection stiffness and without the eccentricity of forces
shown in Figure 2.4. This would lead to underestimating any bending moments on
the columns.
2. The Alpha-pin method suggests that the columns might be designed for the axial
load alone, neglecting any bending moments.
3. Gent and Milner, [Gent and Milner 1968], contain some comments on rotation
values that suggested caution in adopting Alpha-pin completely, especially as the
experience of application of the method is limited. For example, Paragraph 21
includes the comment, "The primary factors emerging as controlling performance
under these conditions are the minor axis elastic critical load of one flange and the
joint rotation R to relax the end moments. Provided the <p..& values are not extreme
the effect of bending actions is secondaty, rather analogous to imperfections in an
axially loaded column,_and the type of performance and ultimate axial load capacity
is primarily controlled by the minor axis stability of one flange."
4. BS 5950-1 [BSI 2000a] gives guidance for designing beams on cap-plates and it is
desirable to produce guidance that is in harmony with the code.
5. Designers expect to have some reduction in column capacity due to bending

moment, especially in a frame that is almost a rigid jointed frame.
28
6. Designers are accustomed to having to apply a nominal moment which is recognised
to be approximate, as in the "simple construction" method shown in Figure 2.2 and
Figure 2.3.
The resulting method of column design, given in AD 283 [SCI 2005b] uses a nominal
moment, but derived differently from the traditional methods used in the design of columns
in traditional "simple construction". In the AD, the moment is calculated from the total
reaction from the beam (ie the sum of the shears from the beam extending both sides of the
column) acting at the face of the column. This differs from traditional "simple construction"
in which equal spans, equally loaded, give zero net moment on the column. It also differs
because the moment calculated is applied to the column without sharing between the column
above and the column below. However, as in the traditional method, the moment is not
expected to be an accurate description of the structural behaviour. Rather, it may be viewed
as a reduction in column resistance to allow for applied moments and/or imposed curvatures
that might occur and that might reduce the capacity of the column.
The AD was written for use in buildings in which the columns are partially restrained by the
floor slab against rotation about the axis along the floor beams. This restraint arises from the
stiffness of the floor slab which is constructed so that it occupies the full depth of the beams.
This is often the case in modem residential construction and this is the type of construction
which had provoked the interest in discontinuous columns. The design rules in AD 283
require that the designer uses an effective length of the columns that is the centre to centre
height from floor to floor. This means that the effective length factor for the columns is
taken as 1.0. This contrasts with conventional design in which the effective length factor in
this case would normally taken as 0.85 or even 0.7 following either the guidance in
BS5950-1 [BSI 2000a] Table 22 "Nominal effective length LE for a compression member" or
similar guidance inBS 449 [BSI 1969]. The AD uses 1.0 because the extensive plasticity
required to allow moment shedding reduces the stability of the column. Therefore, AD 283
method allows design using the nominal moment, which is less than the moment calculated
in conventional rigid frame analysis, but this can only be done if the buckling resistance of
the column, Pc, is calculated using the effective length factor of 1.0 to allow for the extensive
plasticity. This needs to be remembered when comparing the results of design to the AD
with the results of other methods.
The AD uses the column check in BS 5950-1:2000 [BSI 2000a] Clause 4.7.7, which is the
clause for the design of columns in traditional simple construction. This check was chosen
29
both because of its simplicity and because it is this check that designers expect to use when
checking columns in braced frames. The method is demonstrated in the following
calculations.
Firstly, for a typical exterior column in residential construction

Exterior column, beam span = 6.0 m, column spacing = 5.0 m
Assuming 4.0 kN/m2 for self-weight x 1.4 = 5.6 kN/m2
1.5 kN/m2 for partitions x 1.4 = 2.1 kN/m 2
2.0 kN/m2 for live load x 1.6 = 3.2kN/m2
Total = 1O.9kN/m2
Reaction on column from beam = 10.9 x 5.0 x 6.012 = 327/2 164kN
Load from external wall = 2.5 x 5.0 x 1.5 = 19kN
Total 183 kN
Using the assumption ofa nominal design moment,
Design moment = reaction at floor x half-width of column
= 183 kN x 0.070 = 12.8 kN-m Eq2,30
F M M
Using the equation from 4.7.7, _c+ __x +--Y-:51 Eq 2.31
Pc Mbs PyZy
My = 0, so Fc ~ 1- M x = 1- 12.8 = 1- 0.147 = 0.853 Eq2.32

Pc M bs 87.3
Therefore the maximum value of Fc = 0.853 x 1550 = 1322 kN Eq2,33
Secondly, for a typical interior column in residential construction

Assuming the same floor loading as above and that the beam is continuous across the
column,
The reaction on the column from the beam is:
(5/8) xl0.9x5.0xI2.0 = (5/8) x654 = 409 kN Eq 2.34
Using the assumption of a nominal design moment,

Design moment = reaction at floor x half-width of column
= 409 kN x 0.070 = 28.6 kN-m Eq 2.35
Eq 2.36
30
My = 0, so Fe ~ 1- M x = 1- 26.8 = 1- 0.307 = 0.693 Eq2.37
Pc M bs 87.3
Therefore the maximum value of Fe = 0.693 X 1550 = 1074 kN Eq2.38
To find the reduction in capacity below the capacity of a pin-ended strut as above
Pc = 1550 kN
Therefore 1322 kN above is a reduction to 1322/1550 = 85% strut capacity

and 1074 kN above is a reduction to 1074/1550 = 69% strut capacity
The method is known by the authors to be conservative in many cases (such as for internal
columns with symmetrical beam arrangement and good fit-up between the columns and the
beams). However, given the simplicity of the method, this is necessary so that it is safe
when used for highly asymmetric cases (the worst of which are exterior columns) or in cases
where the lack-of-fit between columns and beams may cause unexpected moments.
2.4.6 Summary of unorthodox methods

The advantages and disadvantages of the unorthodox methods are:
1. Gent & Milner's 1966 & 1968 column research is not a complete design method, even
though it provides a very good explanation of the behaviour of columns in braced
frames. The approach to checking column resistance used is that of the critical buckling
of the member with the stiffness reduced by plasticity. This is a good method for
understanding trends and was the best approach available at that time. However,
calculation of the critical buckling load of a member with its stiffness reduced by the
plasticity developed in a column with end rotations is a complex task. In addition, the
method is very different from current codes, neither considering initial imperfections
explicitly nor benefiting from extensive calibration. Therefore it is not sufficient as a
design method.
2. Wood's 1973 plastic design method is interesting because it is a method that broke away
from elastic analysis of columns in braced frames. It suffers the same major
disadvantage as Gent and Milner's work in that the approach to checking column
resistance used is that of the critical buckling of the member with the stiffness reduced
by plasticity. With the limited test justification of the factors used, the absence of
explicit treatment of imperfection and a method that is not clearly related to modern
codes, few designers and fewer checkers could agree to use of the method.
3. Alpha-pin from the University of Sheffield offers outstanding simplicity and economy,
but it was developed on frames in which the columns are continuous and has not been
31
demonstrated on frames with discontinuous columns. There is some concern that where
discontinuous columns are used with long spans of shallow beams, the method might be
optimistic because of the end slope induced by the beams.
4. AD 283 from the SCI gives a simple method, but it is clearly conservative in many
cases. In the case of internal columns at the internal support of continuous beams, it is
very conservative. The method is also imperfect because the nominal moment used to
reduce the capacity of the column is only loosely related to the end-rotation induced by
the slope of the beams.
2.5 Summary of design methods, orthodox and

unorthodox
None of the methods currently available are ideal for the design of discontinuous columns in
braced frames. A new design method is required because, firstly, frame analysis with
varying joint stifthess is too complex and, secondly, methods based on nominal moments are
uncertain and often give very conservative approximations. The Alpha-pin method and the
research by Gent and Milner make it clear that much greater economy of material is possible.
The Alpha-pin method suggests that this should be possible with a simple method that would
make the design process more economical also. The studies presented in this thesis develop
such a new design method for discontinuous columns in braced frames.
32
3 THE NEW DESIGN APPROACH
3.1 Introduction
Braced frames with discontinuous columns are attractive in residential construction because
the columns are small in section and can be concealed within the thickness of the walls
without losing considerable floor area by using unreasonably thick walls. Because the
intention is to minimise the column section, the design process will not try to use moments in
the columns to relieve the bending moments in the beams, so the beams will be designed to
carry all the bending moments from the floor loading. Therefore, an acceptable process of
design would be to:
1. design the beams independently from the columns
2. design the columns to suit the beams, without iterations of analysis and without relying
on the beams for restraint
Such a design process would be very simple for designers. The beams would be designed
either as simply supported single spans or as continuous beams on knife edge supports.
Then the columns would be designed for the axial compression and end slopes equal to the
slopes of the beams.
In braced frames, sway effects (storey shears) are resisted by a "bracing system" which may
be one of many systems such as cross-bracings, stiff portals or a concrete core. The columns
that are braced are not required to resist any bending moments other than those arising from
gravity loads on the beams.
The new design method is developed using the symbols from Eurocode 3 because the
Eurocodes are to become the main design code in the UK and throughout the European
Union during the next decade.
3.2 Behaviour of columns in reality

The starting point for the new design approach is that behaviour of columns in reality is
different from the behaviour commonly assumed by designers using elastic analysis
techniques. This is reported in Gent [Gent 1966] and Gent and Milner [Gent & Milner 1968]
and is presented below.
33
The behaviour was investigated by non-linear finite element analysis as described in
Chapter 4. Examples of the behaviour are shown in Figure 3.1 and Figure 3.2. These show
the axial force and coincident end moments resisted by hot-finished square hollow sections
(SHS) subject to an increasing load combination of axial compression and end rotations.
The graphs are for 140x140xlO sections 3.0 metres long with equal and opposite end
rotation (single curvature) and are for load cases in which the axial compression and the end-
rotations are increased linearly until the maximum axial compression is reached. The
horizontal axis is the end-rotation in radians. The vertical axis is the proportion of the strut
buckling resistance (for axial load) and the proportion of the plastic moment (for bending
moment).
At lower levels of load, the bending moments increase almost linearly. The increase is not
entirely linear because second-order effects reduce the effective stiffness of the member. At
higher levels of load, the end moments reduce because of increasing plasticity. These graphs
show:
1. The columns need not be designed to resist the bending moments calculated by
elastic analysis, provided that the beam design is not relying on the moments in the
columns to resist the loads on the beam (for example at stiff external columns
resisting hogging moments in the beam).
2. The end moment might not reduce to zero before overall buckling occurs, as in
Figure 3.1. This shows the load case in which the maximum load of the column
occurs at 1404 kN with end rotations of 0.013 radians.
3. The end moment might change to the opposite sign as in Figure 3.2. This shows the
load case in which the maximum load of the column occurs at 1109 kN with end
rotations of 0.041 radians. In this case the end moments are increasing the stability
of the column. Given that the design approach proposed in Section 3.1 is to separate
the column design from the beam design, the increase in axial resistance beyond the
point of zero end moment will not be employed. This might be seen as uneconomic,
but reliance on reversed end-moments requires that the beam-column connections
have sufficient strength and stiffness to mobilise the end moment. Such strong and
stiff connections are relatively expensive, so not using end-moments is economic.
34
3m equal & opposite end rotns
At 0.013 radians, max axial = 1404 kN
Q)
()
c
~
'iii 0.8 I
!!!'O
-0:::
c ()
V
~ ~I 0.6 /
0-
E~
oc ~ I 0.4 /
.2 0
"''0
mo:::
0...0 1 0.2 I
~
Ez
8:0
'Owl
cZ
0 V 1\
0 0.01 0.02 0.03 0.04 0.05
8.
e
Cl..
End rotation (radians)
-+-Axial __ End moment
Figure 3.1 Single curvature with small end-rotation
Plot of proportion of compression resistance and moment resistance with

increasing end-rotations
3m equal & opposite end rotns

At 0.041 radians, max axial = 1109 kN
0.8
0.6
0.4
0.2
o
O. 5
-0.2
End rotation (radians)
--Axial --... End moment
Figure 3.2 Single curvature with large end-rotation
Plot of proportion of compression resistance and moment resistance with

increasing end-rotations
35
3.3 New column design approach
After studying the work by Gent and Milner [Gent 1966], [Gent & Milner 1968], and the
development of Alpha-pin [Carr 1993] it was decided that a new design philosophy could be
adopted for braced frames with discontinuous columns. This new design philosophy exploits
this phenomenon of "moment shedding" demonstrated by Gent.
This new design approach is radically different from current approaches to column design. It
uses column behaviour that is very different from that at "working loads" because at working
loads there is little or no plasticity in the columns.
Step 1
The beams are designed independently of the columns. Single span beams are designed as
simply-supported beams. Multiple-span beams are designed as beams on knife-edge
supports. The beam slopes at the supports must be calculated for use in the column design.
If the design is to BS 5950-1 [BSI 2000a] or to the ASCE manual [ASCE 1971], the designer
should use "pattern loading" to find the worst loading condition for column buckling, as
specified in Clause 5.6.2. However, if the design is to the Eurocodes and the structure is a
building, the designer would be free to ignore "pattern loading" from the provisions of
EN 1991-1-1 [BSI 2002b, CEN2002b] Clause 6.2.2( 1).
Step 2
The columns are designed to suit the beams and are assumed to have no effect on the beam
design. Therefore the columns are designed to resist the design axial compression while
satisfying the following criteria:
1. The rotation at each end of the column is equal to or greater than the rotation of the
beam to which it is connected.
2. The column does not require bending moments in the beams to maintain the equilibrium
of the column. For example, however slender the column, it is designed assuming
single curvature bending, as Figure 3.3, and not double curvature bending, as Figure
3.4, so that the design process remains as simple as possible. The assumption of single
curvature will not be consistent with the actual column deflections, but the assumption
is convenient and is conservative for calculating the overall buckling resistance.
36
Figure 3.3 Single curvature bending in the column
Figure 3.4 Double curvature bending in the column
3.4 Development of a method to apply the new design

approach
3.4.1 Introduction
The new design philosophy needs a new design method that is simple to apply either by
pencil and paper or by coding into software. It is always desirable to have a method that can
37
be applied by pencil and paper because it enables designers to make checks independently
from software to confirm the output from software. It also means that coding into software
should be simple and therefore reliable. The method must be able to calculate the axial
compression that can be resisted by the column under enforced rotation at its ends but with
no restraint provided by the beams, ie zero end moments.
3.4.2 The new column design model
To allow fabrication with thin end-plate connections, the design model assumes that the
beams do not provide rotational restraint to the columns. For simplicity, the model assumes
equal but opposite end rotations (single curvature) because this is both the weakest case and
the simplest case to consider.
Numerous cases of end rotation were studied using non-linear elastic-plastic analysis as
reported by King [King 2006]. In the course of the study, three points became clear.
Firstly, the bending moment diagram in the column is defined by the deflected shape of the
column. Therefore, where there is no reversal of moment at the ends and the column is
initially straight, the deflected shape of the column always lies between the original unloaded
shape and the tangents to the ends of the deflected shape as shown in Figure 3.5 (a). The
actual bending moment diagram is shown in Figure 3.5 (b) if the end moments are zero.
There are cases where there are end moments that are destabilising (ie of the same sign as the
maximum end moment) but where such moments occur, the curvature at the ends of the
columns result in less curvature at mid-height of the column for a given end rotation, so
reducing the maximum moment in the column. If the maximum deflection of the column is
e, as shown in Figure 3.5 (a), then the actual maximum bending moment in the column is Ne
as shown in Figure 3.5 (b). Since the actual maximum bending moment diagram is always
less than the maximum from the simple triangular diagram shown in Figure 3.5 (c), the
maximum applied moment is not greater than M = N {9(h/2)}.
38
N
6(hl2 1 M =,N{6(h/2)}
M,=Ne
I I I
I I I
I I I I
I I
I
hl2 I
I
I
I
I I
I
I
I I
I
I I I
I
I : \:,' I
I
- - - J - - - -'f. -----
I ,'1\
I I'
end-rotation 6
I " \', ,,
I
,, hl2
, ,,
N
(b) (c)
(a) Actual bending Triangular bending
Load and shape diagram
moment diallfam moment diagram
Figure 3.5 Triangular approximation to the bending in the column
Secondly, the shape of the column changes as the magnitude of the imposed end rotation
increases. The shape becomes less like a sine curve because much sharper curvatures
develop at mid-height as a result of plasticity. Above and below mid-height, the deformed
shape of the column approaches the tangents to the end-slopes. Figure 3.6 shows the
deformed shape and tangents to the ends for both small end rotations and large end rotations.
The small end rotation case in Figure 3.6 (a) corresponds to the column behaviour near the
maximum strut compression resistance. The large end rotation case in Figure 3.6 (b)
corresponds to the column behaviour in the post buckling range, at deformations greater than
at the maximum strut compression resistance. Figure 3.6 (b) uses a solid line for the actual
shape and uses a dotted line to show the shape in (a) but amplified to give the same end
rotations to show the difference in shape arising from the tighter curvature at mid-height in
the case of large end-rotations.
39
N N
, ,,
,, ,
, ,
\
\,
,,
,,
~
\
, \
,,
\
\
\
\
\
\
\
\
(a) (b)
At maximum strut At larger end rotations than at
resistance maximum strut resistance
Figure 3.6 Deformation with increasing end rotation
Thirdly, at higher end rotations, there is extensive plasticity across the section. In these
cases, the stress distribution mid-height of the column approaches a classic rectangular
plastic stress distribution, but the stress-blocks are not entirely rectangular due to the finite
rotation, as shown in Figure 3.7 (a). For a given axial compression, N, the classic plastic
stress distribution shown in Figure 3.7 (b) gives a larger moment of resistance than the actual
stress distribution in (a). This is shown in Figure 3.7 (c) as the difference between the
typical actual and plastic stress diagrams.
40
I I I
I I I
Cross section Cross section Cross section

•I
I
I I
I I -~
I
I
I
I
I _f---
fy
I
(a) (b) (c)
Actual stress Clas ic plastic Difference between
diagram tress diagram typical actual and

plastic stress diagrams
Figure 3.7 Comparison of typical actual v plastic stress distributions
The simplified triangular bending moment diagram in Figure 3.5 (c) over-estimates the
bending moment applied to the column. The classic pIa tic stress diagram in Figure 3.7 (b)
over-estimates the bending moment resisted by the column. The two over-estimate tend to
cancel each other, providing a possible a design model which is investigated by a parametric
tudy in Chapter 8.
UNIVERSITY
OF SHEFFIELD The combination of the triangular bending moment shown in Figure 3.5 (c) and the classic
, IRRARY plastic stress distribution shown in Figure 3.7 (b) provide the basis for a simple design model
which predicts the column resistance by equating the plastic moment of resistance at mid-
height to the applied bending moment at mid-height. In the new design model, only single
curvature is used, so the shape of the column is always within the tangents to the end of the
column. Therefore the upper bound of the deformed shape is limited to this, as shown in
Figure 3.6. Therefore the approach is similar to classic second-order rigid-plastic
[Home 1963] analysis .
If the effect of imperfections is ignored, this gives a design equation derived from moment
equilibrium at mid-height:
41
Eq3.1
where MN,Rd is the design plastic moment resistance reduced due to axial force NEd.
However, the above equation ignores imperfections and results in the axial compression
resistance being the squash load of the strut, NeRd, for zero end-rotation instead of the strut
buckling value, NbRd.
3.4.3 How the new design model allows for imperfections
The compression resistance of real columns is affected by geometric imperfections and

residual stresses. The amplification of the geometric imperfection depends on the stiffness,
EI, of the column as seen in the classic elastic equation
amplification factor = 1N Eq3.2

I--
Ncr
which, for a pin-ended strut, is
amplification factor = 1N Eq 3.3
1- (1f:;/)
The stiffness is reduced by the extent of the zones of yielding, which is affected by the
residual stresses. Calculation of the resistance of columns is complicated because it must
allow for these effects. Design codes include the effects of imperfections in the calculated
design resistances. Most codes rely on formulae that give a "best fit" to test results.
Both geometric and residual stresses can be treated by a design value of geometrical
imperfection greater than the geometrical tolerance, as Perry-Robertson in Annex G of
BS 5400-3 [BSI2000b]. This approach calculates the bending moment in the strut arising
from the deflection which is calculated from the amplification of the design value of the
initial imperfection. The BS 5400-3 formula is derived from first principles in Appendix D.
The effect of imperfections is most pronounced in columns with high axial compression and
small bending moments. This is because cross-sectional bending resistance is not affected
by residual stresses provided the elements of the cross-section are not affected by local
buckling. In cases where the bending moments induced in a column are higher than the
42
strut-action bending moment that occurs at maximum strut resistance, the bending moment
caused by the initial geometrical imperfection and amplification is only a proportion of the
total bending moment. Therefore, the effects of initial geometrical imperfections and
residual stresses are of lesser importance in cases of higher bending moments than the strut-
action bending moment that occurs at maximum strut resistance. This means that at higher
column end-rotations, the column resistance becomes less sensitive to initial imperfections.
The design model can be adjusted by use of an appropriate additional moment to represent
the effect of initial imperfections. Figure 3.8 (a) shows the same column as in Figure 3.5 (a)
and Figure 3.8 (b) shows the design bending moment diagram taken as uniform over the
entire height of the column with the maximum value from the triangular construction in
Figure 3.5 (c). Figure 3.8 (c) shows a second curved dotted line, to represent initial
imperfection of the column, and straight dotted lines describing an additional bending
moment of magnitude Nej resulting from the initial imperfection. The design bending
moment is drawn as if it is uniform over the entire height of the column. This will not occur,
but it is shown like this to discourage designers from using non-uniform sections because
only uniform members will be considered in this report. The design simplification of single
curvature is safe for uniform members, but might not be safe for varying cross-sections,
because any cases of double curvature might result in high moments applied to sections that
are weaker than the section at mid-height.
The compression resistance, Nb , may be found from the classic rectangular plastic stress
distribution, so the approach is similar to classic second-order rigid-plastic [Horne 1963]
analysis but with the important additions of actual end-rotations and an initial imperfection
to give the correct failure load rather than the over-estimate calculated by classic second-
order rigid-plastic analysis. The initial imperfection of a column may be either in the same
plane as the column end-rotations or in another plane, so both possibilities have been
investigated in the parametric studies in Chapter 8.
43
N
M = N..8(h12) + Nej
~.,.oJ {9(h/2)}
I I
N9(h/2) "I ~e'1
I I
I I
I I I I
I I I I
I I I
I I I
I
I " h/2 I
I I I
I
I I"
I
I I
I
\ \
\ \
\ \
end-rotation 9 \
\
\
\
\
\
h/2 \
\
\
\
\
\
\ \
\ \ \
\ \ \
(c)
(b)
N Design bending moment
Design bending moment
(a) diagram
Load and shape diagram diagram from triangular
with additional moment
construction
Nej to allow for
imperfections
Figure 3.8 Design bending moment in the column including imperfection
The new column design model derives the initial imperfection from the code strut resistance,
so the design model will agree with whatever code is specified for the design of the structure.
Therefore, when the effect of imperfections is included, the design equation (derived from
moment equilibrium at mid-height) becomes:
Eq 3.4
where MN,Rd is the design plastic moment resistance reduced due to axial force NEd
and ej is the design value of imperfection.
44
3.4.4 Defining the design value of the initial imperfection
The model is demonstrably most in need of correction by additional imperfections in the

case of very low curvature, as shown in Figure 3.9. In this case, the difference between the
actual stress diagram and the classic plastic stress diagram is large.
r-
fy
fy
Actual stress Classic plastic Difference between
diagram stress diagram actual and classic

plastic stress diagrams
Figure 3.9 Comparison of actual v plastic stress distributions at low end rotations
Where a column is between beams that have no rotation, the resistance of the column can be
expected to be not less than the buckling resistance of a pin-ended strut. A pin ended sout
has some end-rotation at maximum resistance, but this end-rotation is small and mayor may
not be in the plane of the rotation of the beam. It may be taken as zero for simplicity. It is
necessary that the new design model can be applied to struts with any magnitude of end
rotation . Therefore the additional imperfection must be such that the calculated resistance at
zero end-rotation is equal to the strut resistance. TIlls allows the new design method to be
calibrated to whatever design code is specified by the calculation of an imperfection, es, that
makes the column resistance at zero end-rotation equal to the pin-ended strut resistance of
the specified code.
For a strut with a gross cross-sectional area, A, the imperfection, es, may be calculated as
fo llows:
1. Calculate the buckling load of the column, Nb , as if it were a pin-ended strut using
the specified design code.
2. Calculate the area, Ab , stressed to the factored yield stress, fyd , required to resist the
pin-ended strut load, Nb , so
Eq 3.5
45
where the factored yield stress, fyd, is the minimum specified yield stress reduced by
the appropriate material factor for cross-sectional resistance in the specified design
code, for example using
EN 1993-1-1, fyd = f/'fMo, Eq3.6
BS 5950-1, Eq3.7
AISC, Eq3.8
3. Calculate the reduced plastic moment of resistance, Mpr, of the section assuming the
remainder of the area of the section, (A - Ab) is available to resist bending at yield
stress in tension and compression, and
4. Calculate the imperfection, es, at axial load Nb from the assumption that
Eq3.9
Eq 3.10
It is clear that this imperfection es is the appropriate value for ej of Figure 3.8 (c) at zero end-
rotation. As explained above, for larger end rotations, the actual stress diagram becomes
more like the classic plastic stress diagram, so it is expected that the imperfection will be
reduced at higher end rotations. This is investigated in the parametric studies reported in
Chapter 8 and confirmed by the data plotted in Section 8.5.
Therefore, when the effect of imperfections is included so as to give the pin-ended strut
buckling resistance at zero applied end-rotation, the design equation (derived from moment
equilibrium at mid-height) becomes:
Eq 3.11
where MN,Rd is the design plastic moment resistance reduced due to axial force NEd.
and es is the design value of imperfection derived from the buckling of a pin-ended strut
using the design code specified for the particular structure.
46
3.4.5 Different planes of end rotation
The design model has been proposed in Sections 3.4.2 and 3.4.3 without defining the plane
of the column end-rotations.
Examples are shown in Figure 3.10 and Figure 3.11 in which the areas resisting axial
compression are shown by single diagonal shading and the two areas resisting bending
moment (one in compression and the other in tension) are shown, one with double shading
and the other without shading. Figure 3.10 shows the case of a section with bending about
the X axis alone resisting an axial load slightly greater than half of the squash load. Figure
3.11 shows a section with bending about both the X axis and the Y axis and resisting an axial
load of about half the squash load.
Without detailed examination of the deformations of the column, it appeared logical that the
bending resistance should be calculated from the properties of the section bending in the
plane of the end-rotations. For example, for column end-rotations in the V-plane, a
rectangular plane, the stress blocks would be as in Figure 3.10 but for end-rotations in the
X = Y plane, the stress blocks would be close to those shown in Figure 3.11.
•
I
I
I
I
-------+-------- ---.x
I
i
I
Figure 3.10 Plastic cross-section to calculate Nb from Mx only
47
y
•I
I
i
I
._._._ . .j....._._._. ;i, .---. x
Figure 3.11 Plastic cross-section to calculate Nb from Mx and My
A study was made of the deformations of the column using the Abaqus model described in
Section 4.3 and this showed that application of the moment in the plane of the end-rotations
is not necessarily appropriate. This is because the behaviour is governed by the loss of
stiffness of the section due to plasticity and the direction of the initial imperfection so that
the direction of the initial imperfection dominates the displacements at mid-height of the
column when extensive plasticity has developed.
In an SHS column, where the initial imperfection is in one of the rectangular planes of the
column and the applied rotations in the same plane, the mid-height displacement of the
column is also in that rectangular plane. This is shown by the plots of "0 away" and
"0 towards" in Figure 3.12 in which the initial imperfection is in the X=O plane (ie in
direction +Y). When the rotation is in the rectangular plane perpendicular to the initial
imperfection, the mid-height displacement is between the rectangular planes, but close to the
rectangular plane. This is shown by the plots "90 left" and "90 right" in Figure 3.12.
48
Mid-height displacements
;---1 - +~-- ,
1-1-- --T- - ! ---- l4&-- 1 -- r----· - ---1
~
,
--I
;- -j--~- ~-- ~
I
H.f L . . ~
I
-
~Oaway
, I ~~
- 0 towards
E
.§.
J.
-
l-I
I ' ~90left
:s! '¥l-'~'
i I I I I
> f---j - ---,- t- - -1
I ~90right
1
I
- f -- r ,I __ 1 ~-.
I
-
I - - r
I
- I
I --
--+ i- --
,
i
I
i_ _1 ___ 1 L ___ _J
!
X direction (mm)
Figure 3.12 Mid-height displacements - rotations applied about rectangular axes
from Abaqus analysis
Where the initial imperfection is in a plane at 45° to the rectangular planes of the column,
referred to later as the X=Y plane, and the applied rotations are in the same plane (ie in the
X=Y plane), the displacements are in the X=Y plane. However, where the initial
imperfection is in one of the rectangular planes of the column and the applied rotations are in
the X=Y plane, the initial displacements are in the X=Y plane, but then the displacements
develop out of the X=Y plane, approaching a locus perpendicular to the X=Y plane. This
can be seen in Figure 3.13 which shows the locus of the mid-height when the applied
rotation is towards the initial imperfection (in the X=O plane) and also when it is away from
the initial imperfection.
49
Midheight displacements, 1.5m column
~ ----r1£-- ~- --- r----I
-~~~-~
I
·-------W-~~
.1 i
~ Rotn away from impn
- S - Rotn towards impn
Xdirection (mml
Figure 3.13 Mid-height displacements - rotation applied at 45 to rectangular axis
from Abaqus analysis
The reason for this behaviour was seen in the stress contour plots of Abaqus which showed
that the displacements remain in the X=Y plane while the plastic zone remained less than
30% to 40% of the area of the section, but became out of plane when the plastic zones
became more extensive. This is shown diagrammatically in Figure 3.14 which shows the
cross-section of a column at mid-height. The cross-section is is shown with an initial
imperfection in the Y=O plane,
firstly with a small plastic zone and
secondly with the plastic zone exceeding 50% of the area including the corners at A and B.
While the plastic zone is small, there is little reduction in stiffness to resist displacements in
the X= -Yplane, so the deflections remain in the X=Yplane. As the plastic zone increases, it
spreads across the corners at A and B so the stiffness to resist displacements in the X= -Y
plane is greatly reduced and the displacements increase in the direction of the initial
imperfection.
50
Y
~~
A
I
/
/
/
/
//
/
/
/
/
/ 1-
// ~ B
1- /
/
/
.
/
/
/
/ II""" x
/ ~
~Plastic zones shown shaded

/
/
Figure 3.14 Spread of plasticity - column displacements in X=Y plane
Where the initial imperfection is in a rectangular plane and the rotations are applied in other
planes, similar behaviour is observed, as shown in Figure 3.15. This observation is
important because it shows that the behaviour of the column at mid-height is not a simple
function of the planes of the applied end rotations.
51
~45towards
___ 45 away
-a-22away
.... i
-e-22 towards
~67away
~67towards
! J -+-0 away
1- L +
- 0 towards
~- -~-
l
I
·1 ...... 90 left
~90right
X direction (mm)
Figure 3.15 Mid-height displacements - rotation at 22.5° intervals
applied at 0°,22.5°,45°,67.5° and 90° to Y=O plane with imperfection in Y=O

plane.
Having seen that the column displacements might not remain in the plane of the applied end-
rotations, it was observed that the design model might be kept simple by using bending about
one rectangular axis only. This is verified by the parametric study in Chapter 8.
For simplicity the properties may be calculated assuming that they are concentrated along the
mid-plane of the walls. An example is shown in Figure 3.10 in which the areas resisting
axial compression are shown by single diagonal shading and the two areas resisting bending
moment (one in compression and the other in tension) are shown, one with double shading
and the other without shading.
3.4.6 Different end rotations at top and bottom
The design model proposed above assumes equal but opposite end rotations at the top and
bottom of the column. In most structures and under most load cases, this will not occur. End
rotations are generally not equal and opposite, but a design model allowing for this requires
52
uneconomic beam-column connection details and undesirable complexity of calculation to
justify it, as explained below.
Connection details
One of the goals of this design approach is to allow simple beam-column connections made
with thin column end-plates. Where columns are in double curvature, the end moments are
usually too big to be resisted by thin end-plates, so the exploitation of the double curvature is
in conflict with this goal.
Complexity of calculation
If the reversal of end moment is exploited to reduce the column deflections, the effect of the
column end moment must be calculated both for the beams and for the columns. Another
goal of this design approach is to allow the simplest possible analysis of the beams by
making the analysis separate from the column behaviour, so incorporating the lesser end-
rotation is undesirable.
Defining a simple column design model of the column in the elasto-plastic domain becomes
extremely difficult for deformed shapes other than the symmetrical shape arising from equal
and opposite end-rotations. This is especially difficult because the effects of plasticity so
alter the stiffuess of the member along its length that the behaviour is far from simple to
predict.
Using the assumption of single curvature with end-slopes equal to the greatest slope at either
end is always conservative. The conservatism might be questioned as uneconomic, but, the
design method has the potential to allow much more economical columns than traditional
methods available in columns with significant end-slopes.
The issue that does need to be considered for cases of double curvature is the issue of wall
stability. Double curvature gives greater curvature for the same end rotation, so the wall
stability is of greatest concern under double curvature.
3.4.7 Effect of lateral loads

Lateral load on the column increases the bending moment applied to the column. Because
the design model assumes that the beams do not provide resistance to end-moments in the
columns, the axial compression resistance cannot exceed the resistance of a pin-ended strut
with lateral loading. This must always be checked.
53
Therefore the design equation derived from moment equilibrium at mid-height becomes:
Eq 3.12
but NEd ~ resistance of a pin-ended strut with the the same lateral loading
where MEd is the applied moment due to lateral loads on the column,
MN,Rd is the design plastic moment resistance reduced due to axial force NEd.
and es is the design value of imperfection derived from the buckling of a pin-ended strut
using the design code specified for the particular structure.
3.5 Effects of breadth to thickness ratios of the walls of

the members
Local deformations may reduce the buckling resistance of columns of either "open sections",
such as I-sections or H-sections, or hollow sections. The deformations might be caused
1. by local buckling of the parts in compression,
2. by radial loads caused by curvature of the member, or
3. by a combination of both 1 and 2 above.
This is recognised in most steel design codes by limiting the breadth to thickness ratios of
the elements through a procedure commonly called "classification" of the cross-section.
The breadth-to-thickness ratios required for the new design method were investigated
both in the full-scale laboratory tests (see Section 5.8) and in a parametric study (see Section
7.2). The laboratory test results were compared with the Abaqus analysis (see Section 7.3)
and this comparison was used to define a calibration function (see Figure 7.9) which was
used in Section 7 to define design limits for the breadth to thickness ratio shown in Figure
7.10.
3.6 Range of column end-rotations

It is important that the range of column end-rotations considered covers the range that might
occur in practice. The maximum column end-rotations are caused by the maximum slopes of
54
the beams and these occur at the ends of simply supported beams rather than continuous
beams. It must be remembered that the design method checks the columns at Ultimate Limit
State, so the column end-rotations must be calculated at Ultimate Limit State also, not at
Serviceability Limit State.
The calculations below are to the Eurocodes, following the guidance of SCI P 365
[SCI 2009]. Therefore the governing loading equation is 6.10(b) of EN 1990 [BSI2002a,
CEN 2002a], which gives the reduced partial safety factor on permanent loads of 1.25
together with the partial safety factor on variable loads of 1.50.
For a typical simply supported beam of longer span in residential construction

Beam span = 7.5 m, column spacing = 5.0 m
Assume
4.0 kN/m2 for dead x 1.25 = 5.00 kN/m2
1.5 kN/m2 for partition walls x 1.25 1.88 kN/m2
2.0 kN/m2 for live x 1.5
Total 9.88 kN/m2
Totalloadlm on beam = 5.0 x 9.88 = 49.4 kN
Firstly, find the beam size:

The bending moment in the beam is
Eq 3.13
The resistance of a 280 ASB 74 is
Eq 3.14
Therefore use a 280 ASB 74
Secondly calculate the end slope at Ultimate Limit State

The end slope assuming elastic behaviour is
_1 we
3 3
=_1 49.4 xl 0 X 7500 =0.0339 Eq 3.15
24 EI 24210000x121.9x10 6
Therefore slopes of around 0.034 radians can be expected.
55
The above is only one example, but the order of magnitude remains valid for most practical
cases because as the span increases, the beam size must be increased to provide sufficient
bending resistance and the increased beam size provides greater stiffness. There also two
other limitations that arise from code limits on deflections and from floor stiffiless
requirements to avoid unsatisfactory vibrations of the floor.
Deflection related limitations arise from the relationship between end slope and deflection
under elastic conditions. Taking the case of a simply supported beam, the deflection is given
by
3
5 WL
a=--
384 EI
Eq 3.16
The end-slope is given by
_ 1 We _ 1 384 5 WL3 1 _ 1 384 s: 1 _ 16 0

e- - - - - - - - - - - - - - - - - u - - - - Eq3.17
24 EI 24 5 384 EI L 24 5 L 5 L
For brittle finishes, a common case for shallow floors in domestic construction, the
deflection limit in BS 5950-1 :2000 is Ll360 under un-factored live load alone. If the live
load is approximately 1I3rd of the total load, which is possible for domestic loading with
concrete floors, and the factored load is 1.5 times the working load, then the end-slope at
ULS is:
e = 16 L/360 x-x1.5
3 16 1
=---x3xl.5= .
0.040 radians Eq 3.18
5 L 1 5 360
This suggests that the range of end-rotations needs to extend to about 0.040 radians.
Vibration related limitations arise from the need to avoid very low natural frequencies in
floor so that the dynamic response is limited to acceptable levels. Although it is now
recognised that the dynamic response of floors depends on more than the natural frequency
alone, it is advisable to design a floor to have a natural frequency above 3 Hz so that the
natural frequency is separated from the frequency of walking activity. The natural frequency
depends on the second moment of inertia of the beams, I, the span and spacing of the beams
and the self-weight of the floor. A numerical investigation of this effect was conducted for
both domestic and commercial floor loadings with floors designed with ASB beams and
concrete slabs. The beam design was checked both for cross-sectional resistance and for
natural frequency and showed that the beams have end slopes of about 0.034 radians when
the natural frequency is limited to 3 Hz, whatever span of beam is used. This appears to give
56
a limit of beam slope even lower than the limit from deflections of brittle finishes described
above and applies to beams without brittle finishes.
In addition to the slopes of the beams, the column end-slope is affected by sway of the
building, so some allowance for a relatively small inclination from sway must also be made,
probably of the order of 0.005 to 0.010 radians. This can be deduced from the SLS limit of
h/300 limit on sway. Taking ULS wind = 1.5xSLS wind, then ULS sway = 1.5 x(h/300) =
h/200 = 0.005h, so the inclination of the columns = 0.005 radians. However, this has ignored
second-order effects that must be considered at ULS, so the inclination might be greater than
0.005 radians, but not greater than 0.010 radians.
It should be noted that where continuous beams are used, the beam slopes can be expected to
be less than calculated above for simply supported beams. Therefore it was concluded that
the design method must allow for column end-rotations up to 0.040 to 0.050 radians and so
the laboratory testing and parametric studies should cover this range.
3.7 Summary
A design model has been proposed for square hollow sections (SHS) that uses the moment
shedding behaviour recorded by Gent [Gent 1966]. The model uses classic plastic cross-
sectional resistance calculations with rectangular stress-blocks together with rigid-plastic
behaviour modified by application of an imperfection. The imperfection is derived so that at
zero applied end-rotation, the resistance to axial compression is equal to the resistance of a
pin-ended strut.
The model assumes that the beams behave as beams on knife-edge supports without any
affect from the columns. The columns are designed with equal and opposite end-rotations
top and bottom (ie single curvature), taking the end-rotation equal to the greater of the beam
rotations at the top or bottom of the column. The model assumes no stabilizing action is
provided by floor construction (beams and/or slabs).
The breadth to thickness ratio of the wall of the SHS has been identified as an issue that
should be studied to avoid local instability. Initial calculations show that the limits in current
codes are not sufficient for columns with higher end-rotations designed with the new
method.
57
Parametric studies are required to confinn two aspects of the new design model:
1. To confinn that the design value of the initial imperfection can be reliably taken as
the strut imperfection, es, proposed in Section 3.4.4.
2. To establish what breadth-to-thickness ratio is required for different column end-
rotations.
These parametric studies were conducted using Abaqus finite element software calibrated
with the results of laboratory tests conducted with full-size columns, as described in the
following chapters.
The bending moment calculated in the design model comprises two components. The first is
the moment due to end rotation caused by the beam rotations. The second is the component
to allow both for imperfections (geometrical and residual stresses) and the difference
between the actual stress distribution and the classic rectangular plastic stress blocks used in
the model. The first is always conservative because it assumes that the column is rigid
between the beams and the hinge at mid-height. The second is always conservative because
the actual stress distribution becomes closer to the rectangular plastic stress distribution at
end-rotations higher than occur at the pin-ended strut buckling condition at which the
imperfection is calibrated. Therefore the design model is expected to be conservative.
58
4 ANALYSIS TOOLS
4.1 Introduction
4.1.1 General
The analytical studies reported in this thesis were performed in two phases. The first phase
used a finite slice model (developed by the author and described in Appendix F) to confirm
the general principles of the design approach which led to the initial version of the design
model [King 2006]. The second phase included both a programme of testing of full-scale
columns, reported in Chapter 5, and a parametric study using ABAQUS finite element
software that was used to confirm and refme the design model. The results of the full scale
tests were compared with ABAQUS analysis of the test specimen as reported in Chapter 6.
The results of these comparisons were used to refine the design model by applying an
appropriate calibration factor to the results of the parametric study.
The analyses included the effects of non-linearity of both geometry and material behaviour
and also the effects of residual stresses and of initial geometrical imperfection. Throughout
the analyses, it has been assumed that the values of resistance calculated from the
appropriate structural code are correct when the straightness tolerance is as defined in
Section 4.1.2. This assumption was used to derive the appropriate magnitude of residual
stress. The effect of initial imperfection in a plane that is not a rectangular plane was
investigated for a few cases.
4.1.2 Initial geometrical imperfections
The initial geometrical imperfection is taken as a half sine curve. Imperfections are taken as
acting in only one plane at a time for frame analysis in both BS 5950-1 [BSI 2000a] and
ENI993-1-1 [BSI2005a, CEN 2005a]. Therefore, in these analyses, the initial imperfection
was generally applied in one plane at a time. Figure 4.1 shows the case when the initial
imperfection is in the same rectangular plane as the deflection due to the end-rotations of the
column imposed by the rotations of the beams. This results in deflections in only one plane.
59
y
Initial
i
Imperfection I , Deflection
!4 .,4 .:
!: :
! -~-- ffi-.'
-'-'-!'-~-"~'-'-'- -:. X
~ I
I
~:--
I I I
'-1!t ~I
Sectional Plan AA
A A A
Elevation BB Elevation CC
Figure 4.1 Imperfection and deflections in one plane only
Figure 4.2 shows the case when the initial imperfection is in the rectangular plane orthogonal
to the deflection due to the end-rotations of the column imposed by the rotations of the
beams. This results in deflections in the two rectangular planes.
60
y
tj4 Deflection
I ... :.
.,. ,
,,
.
Initial
Imperfection
J- LJ:l ----- ____ ~ ~G
I
, I '
L. - . - -
_~~_~ ~ ~ ~tDeflection
'
,
I
'-'"""j'-'-'-'-'-'-'-'-+X $.I
k.- A
Sectional Plan AA
I
I
I
,,,
I
I"
,,
,:
Elevation BB Elevation CC
Figure 4.2 Deflections in two planes
The magnitude of initial imperfection used is length/lOOO. This magnitude corresponds to

the principal manufacturing tolerance in BS 5950-2 [BS! 2001] and in the National
Structural Steelwork Specification [BCSA 2002]. The manufacturing tolerance allows a
6mm deviation from straight in all members. In a 3 metre long member, this is length/5oo
which is much greater than length/lOoo. However, this 6mm deviation would result in
column resistances significantly below the buckling curve for hot-finished SHS. Therefore,
a choice has to be made between:
1. analysis with the worst tolerance (in addition to residual stresses), or
61
2. analysis with a tolerance (in addition to residual stresses) that agrees with the strut
buckling curves in the design code.
It was decided to conduct the analysis to agree with the strut buckling curves in the design
code BS 5950-1 [BSI 2000a] so that the design method converges with the code method for
columns which have no imposed end rotation.
The effect of an initial imperfection not in a rectangular plane was investigated for a few
cases in the Abaqus parametric study reported in Chapter 8 by using an imperfection in the
X = Y plane and end-rotations in the X = Y plane resulting in deflections in X = Y plane as
shown in Figure 4.3.
Y
/ X=Yplane
t /
/
/
/
l21/
/" "
,,/. "
~,--/-;
, . :.J<:
~----.,
',1/' "
-"1'~-"-"-""",
/ ,,
,,
,
\' I
, I
~
!\ ,,
\
Sectional Plan AA \
\
\
Deflected column \
Column with initial imperfection
Elevation BB
Figure 4.3 Imperfection and deflections in the X = Yplane
62
The finite element model used the same residual stress as previously derived and the
magnitude of the initial imperfection was chosen so that the resistance of a pin-ended
column with the imperfection in the X = Y plane was equal to the resistance of the same
column with the imperfection of lengthl1000 in one rectangular plane. This magnitude was
found to be 0.75 xlengthllOOO in both rectangular planes, which in the X = Y plane is
.J2 xO.75 X length /1000, which is 1.06 xlength /1000 in the X = Y plane.
4.2 Finite slice model

4.2.1 Description of analysis tool
The first step in developing the new design model was to find an analysis tool that can
analyse columns at different end rotations very accurately over a wide range of members and
end rotations. Study of the results from these analyses stimulates ideas for possible simple
design models and shows potential complications. An analysis tool was developed for
Square Hollow Sections which are the column sections most commonly used in frames with
discontinuous columns. The tool is a multiple-page Excel spreadsheet which finds the
compression resistance by a series of iterations. This considers flexural buckling about both
principal planes. It does not consider any torsional deformations, but these do not affect
normal hollow sections. The analysis divides the length of the member into 30 sections, so
that there are 31 cross-sections at which analysis is made. The input data includes the length,
the yield stress, the section properties, residual stresses and the geometrical imperfection
taken as a half-sine curve along the length of the member in either of the principal planes, or
in both principal planes if required. The section properties are applied in the form of wall
width and thickness that give the area and inertia of the specified SHS but allows the cross-
section to be simplified to four walls of constant thickness with perfectly square corners.
The model and the analysis spreadsheets are described in Appendix F.
At each analysis cross-section, the stress distribution is found for equilibrium with the
applied axial compression, N, at the deformation arising from the current curvature. To date,
the stress distribution includes the effects of plasticity in compression only. Therefore, the
equilibrium stress distribution currently must be checked to ensure it does not exceed yield
in tension.
63
4.2.2 Residual stresses
The residual stress distribution used was parabolic as shown in Figure 4.4. The parabolic
distribution was used because the finite slice had been developed originally for I-sections
and parabolic distributions have been used for these sections by various researchers
[ECCS 1976, Galambos 1998]. For the preliminary investigation of this design method
using structural hollow sections, it was decided to use the finite slice analysis software using
the same form of distribution as I-sections to avoid delay. This distribution was assumed to
exist in all four sides of the SHS. The magnitude of the maximum compressive residual
stress, /rmax, is 10% yield stress. This magnitude was chosen because it gave the best
agreement between calculated column resistance and the column resistance calculated using
BS 5950-1 :2000 for hot-finished SHS when using the initial geometrical imperfection of
lengthll000.
~
'CQ[Q]] IIIIIIIJI[DJY
.A Jr
Figure 4.4 Parabolic residual stress pattern (on all sides)
4.2.3 Validation of the column analyses
Validation of the analyses was done by the following steps:
1. Validation of elastic range by checking the elastic critical buckling load for known
modes with Sine curve displacements.
2. Validation of elastic behaviour by comparison with the BS 5950-1 strut buckling curve
using the code value of the Perry-Robertson imperfection and with the residual stress set
to zero.
3. Validation of the elastic-plastic range for a column in single curvature with severe end-
slopes by pencil and paper checks. Firstly, the input curvature was taken from the
second-order elasto-plastic analysis spreadsheet and used to calculate the deflected
form. Secondly, the equilibrium of the sections was checked by drawing the strain and
stress diagrams on graph paper and calculating the bending moment and axial force
resisted. Finally the equilibrium of the axial force at the eccentricity of the deformed
shape was compared with the bending moment resisted.
64
4.3 ABAQUS finite element model
4.3.1 Change from finite slice to Abaqus finite element analysis
The finite slice model was useful as an initial analysis tool. Firstly, creating the finite slice
analysis tool required detailed consideration of all aspects of the structural behaviour.
Secondly, the simplicity arising from using entire slices helped to identify the design model.
However, the finite slice model has disadvantages:
I. It cannot consider the effects of wall thickness and local out-of-plane deformations.
2. It requires a lot of time and attention to converge even in the simplest cases.
3. It is a "custom-made" analysis tool, so it does not have the credibility of an
established finite element analysis software package.
Therefore it was decided to change to a well-known finite element analysis software.
Abaqus was selected because it is used by the University of Sheffield and was used by
Professor Greiner and his co-workers in the derivation of the rules for calculating resistance
of columns to combined axial compression and bending for Eurocode 3 Part 1.1 [BSI 2005a,
CEN 2005a], described in detail in ECCS No 119 [ECCS 2006].
4.3.2 Analysis method
The analysis method is the finite element method. The analyses used geometric non-linear
and material non-linear solutions. The model used shell elements with the nodes and the
mid-thickness of the elements in the plane of the centre-line of the walls of the column. This
results in a model with square corners, which is slightly different from the tight radius
corners of hot-formed SHS sections. The ends of the model are connected to a "spider" of
rigid-body elements, whose legs radiate the point of intersection of the centre-line of the
column and the plane of the end of the column. Most analysis was conducted using a model
of the entire column as described in Section 4.3.3. The effect of mesh refinement was
checked by using a half-model that comprised a column cut longitudinally along the centre-
line of two opposite sides as described in Section 4.3.4.
4.3.3 Full model
The mesh of the finite element model is shown in Figure 4.5. The same proportions of
elements, along the length of the member, and the same number of elements were used for
all full models, whatever the member length. The element mesh of the walls divides each
wall into 6 elements across the width. Details of the model are given in Appendix C.
65
Figure 4.5 Finite element model mesh of full model
4.3.4 Half model
To check the sensitivity of the model to the element dimensions, the length of all the
elements was halved to produce a comparison of results. The number of nodes available is
limited to 1000 in the student edition of Abaqus being used. Therefore the model had to be
reduced to one half, as shown in Figure 4.6.
Because of the 1000 node limitation, the mesh lengths are halved except at the bottom slice
which remains at the original length because there are not enough nodes available. The
model was made full length but half width to allow the tests to be modelled with the different
eccentricity at each end. Details of the model are shown in Appendix C.
The output from the half-model showed no significant difference from the output from the
full model.
66
Figure 4.6 Finite element model mesh of half model
4.3.5 Residual stresses
For the finite element analysis, the pattern of residual stresses applied was slightly different
from that used in the finite slice model. The stress pattern was taken as straight lines as
shown in Figure 4.7. Straight line residual stress patterns were used by the authors of the
rules for axial and bending in EN 1993-1-1 [BSI 2005a, CEN 2005a] as may be seen in
Greiner et al [Greiner et al 1998]. The residual stress pattern used for structural hollow
sections in the development of the rules for Eurocode 3 was chosen to reproduce the
behaviour of cold formed sections, which is the method of manufacture used for the majority
of hollow sections in Europe. The cold forming of the comers necessitates high residual
stresses in the model. In contrast, this report considers hot-finished hollow sections, which
are designed with compressive strength from higher buckling curves because the hot-
finishing reduces the residual stresses. Having established that a residual stress of 10% yield
stress gave strut buckling results very close to BS 5950-1 in the first phase of the analysis, it
was decided to try this magnitude with a simple straight line stress distribution as used by the
by the authors of the rules for axial and bending in EN 1993-1-1 [BSI2005a, CEN 2005a]
for the webs of hot-rolled I and H sections. This pattern was assumed to exist in all four
sides of the SHS. The value of 10% yield stress was adopted because it gave the best
agreement between calculated column resistance and the column resistance calculated using
BS 5950-1 :2000 for hot-finished SHS when using the initial geometrical imperfection of
lengthllOOO. The residual stresses were introduced by the technique of applying a
temperature to the walls such that the intended residual stresses result.
67
Figure 4.7 Bi-triangular residual stress pattern (on all sides)
4.3.6 Comparison with design codes

The maximum load for a 140x140x1O SHS with yield stress of 355 N/mrn2, pin-ended and
3.0m long, calculated to BS 5950-1 is 1550 kN according to the SCI Blue Book (SCI 2001).
The section in Abaqus that is used to model a 140x 140x 10 SHS is a square section with the
dimensions from mid-plane to mid-plane of the walls of 130mrnx 130mrn. The walls
modelled with shell elements of thickness 10mrn and no radii at the corners. This section
has an area of 4x 130x 10 = 5200 mrn2 • The maximum load from the model when analysing
with the yield stress of 355 N/mrn2, initial geometrical imperfection of LlI000 and residual
stresses of 10% yield as Figure 4.7 is 1582 kN . When this load is multiplied by the ratio of
the model area to the nominal SHS area, it is reduced to (5090/5200)x1582 = 1549 kN. This
is almost exactly the resistance calculated from BS 5950-1 showing that the use of an initial
geometrical imperfection of Ll1000 and a residual stresses of 10% yield as Figure 4.7 is
appropriate.
Calculated to EN 1993-1-1, the pin-ended strut resistance is 1492 kN:
(
Lcr J (3000)
~=~= i = 52.7 = 56.9 = 0.745 Eq 4.1
~, (nJf,) (n 213~) 76.4
 =0.5~ + (l(~- 0.2)+ ~2 J= 0.5[1 + 0.21(0.745 -0.2)+ 0.745 2

] Eq4.2
= 0.5[1 + 0.114 + 0.555]= 0.835
68
1 1
X =[ + ~(' - X' l] ='0.835
~
+ ~(0.835' - 0.745' )
]= 0826
.
Eq 4.3
:. Nb Rd =AXfy =5090xO.826x355 =1492 kN Eq 4.4

• 1Ml 1.0
There are differences in calculated resistance between design codes, for example BS 5950
and EN 1993-1-1, but this is not important because the design method adjusts the resistance
according to the pin-ended strut resistance of the specified design code, as shown in Sections
9.2.3 and 9.2.4.
4.4 Summary
The analysis of columns was initially performed using a custom-made finite slice analysis.
After the initial analyses, Abaqus finite element software was used with a finite element
model of the entire column section. The sensitivity of the element shapes used was checked
with a model of half of a column but with double the number of elements along the length of
the model.
All the analyses incorporate initial geometrical imperfections and residual stresses which
gave buckling resistance equal to that from BS 5950-1. The design model in Section 9.2
adjusts the resistance to agree with whatever design code is specified.
69
5 FULL SCALE LABORATORY TESTS
5.1 Introduction
Tests were conducted on full scale columns
1. To validate the analysis used to develop the design model
2. To investigate the effects of wall slenderness
3. To give confidence in the design method
As reported in Chapter 4, the analysis was done both by an iterative finite slice model, and
also by Abaqus finite element analysis software using non-linear geometry with non-linear
material properties. While Abaqus is regarded as a generally reliable software package, there
is always scope for any analysis system to be applied in such a way that it does not model the
structural behaviour correctly. This is particularly true in models considering instability
where practical variations along a member of thickness, flatness, residual stress, yield stress
and stress-strain relationship might cause behaviour not foreseen in the model. While the
design model is broadly logical in its form, it required validation and calibration.
Early studies using elementary theory of plastic flow and simple slice models, as reported in
Appendix E, had identified that the curvatures in the columns would cause major out-of-
plane deformations in thinner walled columns. The proposed design model is based on the
gross section of the column, so these deformations might make the proposed design model
unsafe for columns with slender walls. These deformations are simulated by Abaqus, but
such effects might be affected by aspects of the member that are not included accurately in
the model, such as local geometric imperfections, variations in residual stresses or even
variation of the stress/strain characteristics of the material. Therefore it is wise to assess the
behaviour both by analysis and by full-scale testing.
The philosophy used in the proposed design method is entirely different to the philosophy to
which most structural engineers are accustomed. Historically, designers regarded the onset
of yield as defining the failure load. Since the development and application of plastic
methods of design in the 1950s, many designers are accustomed to designing with plastic
cross-sectional resistance, but many are not comfortable with plastic frame design. The
proposed design method is only an extension of the fundamental principle of plastic design,
that the designer is free to proportion a structure so that it develops the bending moment
70
diagram that is desired. However, the method is likely to be greeted with suspicion because
plastic design has not been applied in this way in design offices for members sensitive to
buckling. Therefore it was decided that laboratory testing of full scale columns was
desirable to give designers the confidence to use the method.
5.2 The test programme

The test programme is shown in Table 5.1. All specimen were Celsius 355 120x120 Square
Hollow Sections (SHS). Celsius is the brand name for structural grade hot-finished hollow
sections produced by Corus to EN 10210 in steel grade S355J2H.
As one of the reasons for the testing was to validate the modelling and analysis, it was
desirable to have more than one test so that repeatability could be investigated. The most
common application of this column design philosophy is expected to be with thick wall
columns to reduce the column section size to the minimum. Therefore there are more tests
on sections with lOmm wall thickness.
Another reason for full scale testing was to study the effect of slender walls. To allow the
wall slenderness effect to be compared with thick walled behaviour, the same size section
was used but with thinner walls. Two wall thicknesses were chosen. Two tests on each wall
thickness were conducted which was considered the minimum number of tests. More tests
are always desirable, but testing budgets are limited.
Table 5.1 Test programme
Test no. Specimen

kc1 120 SHS x 5mm pilot test
kc2 120 SHS x 5mm pilot test continuation
kc3 120 SHS x 10mm stable wall thickness
kc4 120 SHS x 10mm stable wall thickness
kc5 120 SHS x 6.3mm possible sensitivity to wall slenderness
kc6 120 SHS x 5mm wall expect sensitivity to wall slenderness
kc7 120 SHS x IOmm wall stable wall thickness
kc8 120 SHS x IOmm wall stable wall thickness
kc9 120 SHS x 6.3mm wall possible sensitivity to wall slenderness
kc10 120 SHS x 5mm wall expect sensitivity to wall slenderness
71
5.3 Test rig
The test rig was defined by the need to have full-scale tests and to be able to conduct
numerous tests for the budget available. The University of Sheffield laboratories have
numerous test rigs available. The tests required full-scale specimens, so the hydraulic two-
post rig, shown in Figure 5.1, was chosen because of the height of colunID that can be tested
and the load that can be applied. The laboratory staff have extensive experience in testing
with this rig. Although the rig is normally operated under load control rather than
displacement control, with skill the displacement can be controlled by limiting the flow of oil
from the pump to the jack. Using this control technique, very smooth curves were recorded
on the falling branch of the plot of load versus displacement.
Figure 5.1 Test rig with column in position
The section size of SHS 120 was chosen because the failure load was well within the
capacity of the jack in this rig. Therefore, even if the steel in the specimen was at the upper
end of the yield range of S355 steel, the tests would still be successful.
72
The laboratory staff manufactured end-fittings or "shoes" to fit to the ends of the specimen
without fastening. That avoided the expense of work on the specimen themselves, so the
plane SHS sections supplied could be used directly. The shoe at the bottom of the test
column is shown in Figure 5.2 and the shoe at the top of the column was imilar. They
articulated about a cylindrical pin. These shoes ensured that the deformation of the column
was in one predetermined plane. The use of these shoes introduced some friction which was
evaluated by the unload/re-Ioad cycles. The shoes added 75mm to each end of the column ie
the overall length of the sout is 2650mm from centre of roller to centre of roller.
Figure 5.2 Shoe at bottom of test column
73
5.4 Instrumentation
The instrumentation on each specimen was as listed in Table 5.2.
Table 5.2 Instrumentation on the specimen
Note: In-plane and out-of-plane are defined with reference to the intended plane of
displacement which is determined by the orientation of pins in the end shoes.
Top 1 in-plane inclinometer
I out-of-plane inclinometer
Middle 2 LVDTs in-plane, one on each side

1 LVDT out-plane
Bottom 1 in-plane inclinometer
1 out-of-plane inclinometer
The LVDTs were connected by wires to a pring-grip frame, known as the "thing", haped to
avoid displacement arising from local deformations of the walls. It is shown in Figure 5.3
at the end of a column to show how it fit the ection at a114 comer .
Figure 5.3 Spring grip frame, the 'thing", for LVDT connections
74
The inclinometers were attached directly to the specimen at 200mm from the centres of the
pins. Only in-plane rotations were"measured in the pilot tests, kcl and kc2, and tests kc8, 9
and 10.
The axial shortening was measured by the ram movement. Therefore these measurements
include the extension of the posts of the rig. The ram displacement within the jack was
measured by an externally mounted LVDT on the lower loading platten and the jack force
was determined by a pressure transducer fitted to the hydraulic circuit of the machine.
The test data was recorded to spreadsheets. The columns titled P were zeroed at the start.
The channel numbers in tests kc3 to kc I 0 were as follows, though P 8 and P 9 were not
recorded for tests kc 8, 9 and 10.
P I is the applied load in kN
P 2 is the in-plane rotation at the top (milliradians)
P 3 is the in-plane rotation at the bottom (milliradians)
P 4 mid height out-of-plane displacement (mm)
P 5 mid height in-plane displacement, rear (mm)
P 6 mid height in-plane displacement, front (mm)
P 7 ram movement (axial shortening), (mm)
P 8 is the out-of-plane rotation at the top (milliradians)
P 9 is the out-of-plane rotation at the bottom (milliradians)
75
5.5 Test duration
The duration of the tests is listed in Table 5.3. The strain rate was as slow as reasonably
practical with the test equipment used.
Table 5.3 Duration oftests
Time start to maximum load maximum load to end Total time

(mins) (mins) (mins)
Test
kc3 9.4 3.3 12.7
kc4 13.8 5.0 18.8
kc5 10.5 4.3 14.8
kc6 10.5 3.6 14.1
kc7 11.7 14.7 26.4
kc8 9.9 7.4 17.3
kc9 10.2 6.7 16.9
kclO 7.7 2.4 10.1
5.6 Material properties

The yield stress of the test columns were measured from coupons cut from most of the
specimen after the tests were completed. The results were as follows:
Table 5.4 Material properties
Specimen Thickness Yield UTS Elongation

(mm) (N/mm2) (N/mm2) %
kc1l2 5 396 548 27.5
kc3 10 375 530 35.0
kc4 10 390 530 35.0
kc5 6.3 428 561 25.0
kc6 5 389 540 24.0
kc7 10 not known not known not known
kc8 10 375 514 36.5
kc9 6.3 431 549 32.0
kcl0 5 402 533 32.0
76
5.7 Test results
5.7.1 General
The test results proved to be broadly as predicted with all the tests showing a long falling
branch after maximum resistance.
Different aspects of the results are discussed below. The general form of the results was very
similar for all the tests, as shown by Figure 5.12, Figure 5.18 and Figure 5.19, except for the
loss of strength due to local buckling in the thinnest wall sections at end rotations greater
than 40 milliradians.
The tests on the 10mm, the 6.3mm and one 5mm wall thickness specimen included an
unloading and re-Ioading cycle. This forms a hysteresis loop as described below.
5.7.2 The unloading/re-Ioading cycle
If the test were conducted at an infmitely slow strain rate and if there were no friction in the
bearings, the unloading plot would be as shown in Figure 5.4. Under the same conditions
plus the condition of no Bauschiner effect, the unloading/re-Ioading cycle would be as Figure
5.5. The Bauschinger effect causes a transition curve from the elastic reloading line to the
line of the falling branch as shown in Figure 5.6.
Figure 5.4 Unloading cycle with no friction in the bearings
77
B
A H
Figure 5.5 Unloading/re-Ioading cycle with no friction and no Bauschinger effect
axial load N
mid-height displacement
Figure 5.6 Unloading/re-Ioading cycle with no friction in the end fittings
The friction in the bearings at the two ends of the specimen causes moments that affect the
plots because the friction always opposes the rotation of the ends. This effect can be seen in
Figure 5.7 which shows the load v displacement plot of the unloading/re-Ioading cycle from
test kc7 which had the widest loop. The total width of the loop is 1.2 mm at its widest
between the unloading and loading paths, so the friction moment is equal to 0.6 mm times
the axial compression. At the maximum axial load of 1290 kN, this gives a moment of 0.77
kN-m which is very small compared with the plastic moment of resistance of 87.3 kN-m.
78
kc7i unloading cycle
950
900
850
!
J" 800
1
~
750
700
650
22 23 24 25 26 27
Mid-height dis placement (mm)
- k c 7 mld-helghtdlsp
Figure 5.7 Unloading/re-Ioading cycle test kc3
The effect of the friction moments on the bending moment diagram in the column is
determined by the movement of the ram. If there is no unloading cycle, in Figure 5.8, the
ram is extending throughout the test from zero load at A through the point of maximum load
at B and along the falling branch to Z.
Figure 5.8 Rising and falling branches
The friction in the end fittings resists the extension of the ram and the increasing deformation
of the column, so the bending moment in the column is of the form in Figure 5.9. This may
be understood as the sum of the bending moment from [the axial compression x the
displacement of the column] minus the bending moment from the friction moment = J..lNr as
79
~ M=J.lNr
@M=J.lNr
Figure 5.9 Bending moment for ram extension
The test curve including the unloading/re-loading cycle including the effect of friction is
shown in Figure 5.10. The un-loading path is shown vertical from C to D as the friction
moment reduces from opposing the buckling of the column to opposing the unloading. The
unloading path is then inclined to represent the elastic behaviour from D to E. The reloading
path starts vertically from E as the friction moment changes direction until at F the limiting
friction moment is reached and the reloading path commences linearly towards F, with some
small elasto-plastic effect expected due to the Bauschinger effect as the loading approaches
the path CZ.
On the unloading path from C to D, the friction moments at the end fittings prevent the ends
from rotating as the bending moment diagram changes from that in Figure 5.9 to that in
Figure 5.11. On the re-loading path from E to F, the friction moments again prevent the ends
from rotating as the bending moment diagram changes from that in Figure 5.11 to that in
Figure 5.9.
On the unloading path from D to E, the friction in the end fittings resists the decreasing
deformation of the column, so the bending moment in the column is of the form in Figure
5.11. This may be understood as the sum of the bending moment from [the axial
compression x the displacement of the column] plus the bending moment from the friction
moment = J.lNr as shown in Figure 5.11.
On the reloading path from F to C, the bending moment diagram is again as shown in Figure
5.9.
80
z
E
A
Figure 5.10 Rising and falling branches with unloading/re-Ioading cycle
.@ M=~Nr
~ M=~Nr
Figure 5.11 Bending moment for ram retraction
On the unload/re-Ioad cycle CDEFC, the difference in the displacement arises from the
difference in the bending moment diagrams in Figure 5.9 and Figure 5.11 which depends on
the friction moment acting in opposite directions.
5.7.3 In-plane displacements
Plots of load versus mid-height displacement in the plane of buckling are shown in Figure
5.12.
81
kc3 to kcl0j axial v mid-height displacement
1600
-a-kc3
1400
-6-kc4
1200
-e-kc5
Z 1000
:!.
~kc6
] 800
:!
~ 600 ~kc7
400 ~kc8
200 ~kc9
o
90 ~kclO
o 10 20 30 40 50 60 70 80
Mid-height displacement (mm)
Figure 5.12 Load v mid-height displacement (in-plane)
The mid-height displacements differ between tests during the initial loading up to the
maximum load. This is shown more clearly in Figure 5.13 which shows the displacements at
a larger scale. This shows that the tests cover a wide range of imperfections as might be
expected in practical construction.
kc3 to kclOj axial v mid-height displacement

1600
-a-kc3
1400
-6-kc4
1200
-e-kcS
- 1000
! ~kc6
1 800
-+-kc7
600 I j
400
! I i ~kc8
ILl
I '
200 I! j , I i I -e-kc9
! I,
_____L ___ L.
I
I I ___ LI
o ---- _~, ~ ~I ___ L ___ _
~kc10
-1 0 1 2 3 4 5 6 7 8
Mld-hel,ht displacement (mml
Figure 5.13 Load v mid-height displacement (in-plane), zero to 8mm
82
The target initial imperfection for the tests was length/750 which is the maximum geometric
imperfection expected to be permitted for hot-finished hollow sections in EN 1090-2
[BSI 2008, CEN 2008], the Standard giving straightness tolerances for the execution of
structural steelwork to Eurocode 3, in Tables 0.1.11, 0.1.12 and 0.1.15.
The straightness of each specimen was measured against a straight edge. The longest
available straight edge was 1.5m long, so only the central half-length could be checked in
detail. These checks showed that the straightness was better than the limit in EN 1090-2
[BSI2008, CEN 2008]which would be 2mm over a length of 1.5m using length/750 as the
limit or 1.5mm over 1.5m using length/l 000. Generally the deviation from straight was
found to be much smaller than 2mm, with values recorded from less than 0.05mm in 1.5m to
0.95mm in 1.5m on kc8.
The end-fittings used for the tests were designed to allow some adjustment of the position of
the column with respect to the pivots which are the points of application of load. This was
controlled by the horizontal screws that are visible in Figure 5.2 However, it is not easy to
control the position precisely.
It is possible to estimate the actual position of the columns in the rig from the plots of end
rotation versus load in the elastic range of loading. The estimates are approximate because
the end-rotation depends on the shape of the initial imperfection along the entire length of
the column. However, if the column were perfectly straight, the end positions are simply
calculated from the end rotations. It is also possible to estimate the initial bow of the column
from the change of end-rotations and the change of mid-height displacement with change of
axial load.
The mid-height displacements are plotted in Figure 5.14 for 10mm wall thickness, in Figure
5.15 for 6.3mm wall thickness and in Figure 5.16 for 5.0mm wall thickness. In these figures,
the test displacements are compared with the mid-height displacement expected from a
3.0mm half-sine curve initial imperfection assuming entirely elastic behaviour. The 3.0mm
value is used simply to give a comparison against an imperfection of the appropriate order of
magnitude. (A tolerance of length/750 on a length of 2650mm is 3.5mm, but friction would
reduce the expected deflections below those arising from the imperfection, making 3.0mm a
reasonable comparator.) From these figures, it can be seen that the test behaviour varies
significantly from that of a half-sine imperfection.
83
10mm wall SHS;mid-height displacement
comparison with 3mm half-sine impn
- k c 3 m~ht <isp
z
....
'"' - • - kc1 m~ht <isp
- -1<c7 mid....eight <lisp
----- kc8 m~ht <isp
--a-10 mm wal haIf-si~

impetfection (mm) = 3.0
·1 0 1 2 3 4 5 6 7 8 !) 10
MJd.hel&ht displacement (mm)
Figure 5.14 IOmm wall SHS, mid-height displacements v 3mm half-sine
6.3mm wall SHS;mid-height displacement

comparison with 3mm half-sine imperfection
1~90
I I I
I
I
I
I
-z
~
1
- • -k<~lIIid~lI'iKhl <Ihj.
'V
"
J! - kc9 mld-helght lisp
:iIe
C
-+-/).3mm wal halt silc
imperfection (mm) - ~.O
·1 0 1 2 J 4 5 6 7 8 !) 10
Figure 5.15 6.3mm wall SHS, mid-height displacements v 3mm half-sine
84
Smm wall SHS;mid-height displacement
comparison with 3mm half-sine imperfection
rl~T--~1 -I--
I I i r
! 1il T-r- - --l
~ • - keG mkUwlght dkp
I I i ..-7~-"t--t--T-+-_-. +-----~.~ kc10mid-Might lisp
foo~l --:
I . ~ 5 mm wal 1wIf-sine
I, I Impttftction (mm) - 3.0
-1 0 1 2 3 4 5 6 7 8 9 10
Figure 5.16 5mm wall SHS, mid-height displacements v 3mm half-sine
85
5.7.4 Out-of.plane displacements
It was intended that the out-of-plane displacements should be limited to small values relative
to the in-plane displacement by the design of the end fittings. The plots in Figure 5.17 show
that this was achieved for the large in-plane displacements along the falling branch. At the
point of maximum load, both the range of out-of-plane displacements and the range of ratios
of out-of-plane displacement to in-plane displacement was considerable, as shown in Figure
5.17 and Table 5.5 .
kc3 to kclO; out-ot-plane displacement
...... kc3
-6-ke4
..... keS
"'ke6
·3
~ke7
·5
"'ke8
I ;rkc9
-~-r----
·6
·7
I '
-.~--~-.~-- .___ ~I ___ ~ _ _ _ _ _ ' ~kel0
o 20 40 60 80 100
In-plane mid-hei,ht displacement (mm)
Figure 5.17 Load v mid-height displacement (out-of-plane)
Table 5.5 Ratio of out-of-plane to in-plane displacements at maximum load
Test kc3 kc4 kc5 kc6 kc7 kc8 kc9 kctO

Ratio 0.55 0.71 0.04 0.74 0.07 0.23 0.69 0.41
86
5.7.5 In-plane end rotations
Plots of load versus mean end-rotation are shown in Figure 5.18. This shows that the
load/end-rotation from all the tests have long falling branches as expected. The figure also
shows the design values of resistance of the sections when calculated to EN 1993-1-1 as pin-
ended struts.
The plots for the two tests on specimen with 5mm wall thickness show marked reductions in
resistance below the general trend due to local wall buckling. This was predicted, as
discussed in Chapter 7 and Appendix E. It is interesting to note that kc 10 has a higher
maximum resistance than kc6 but has a lower rotation capacity before the dramatic loss of
resistance at about 50 milliradians. This suggests that the lower yield of kc6 produced a
more uniform curvature which was lower than the curvature ofkcl0.
kc3 to kcl0; axial v mean end rotation

1600
I ..~ ~-=]~--l=
1500 -ke3
1400 ;
-ke4
1300 - -- ---1
1200 rtf1~~~~~~~~-~~-""'I~-~----'-~------i -ke5
. I
1100 -- ·------ --- ~
-ke6
Z- 1000 i
:!!. -ke7
"a 900
.2 800
700
-ke8
iii
600 -ke9
:i 500 -ke10
400
300 -EC3120·10
200 -EC3120·6.3
100
0 -EC3120·S
o 20 40 60 80 100
Mean end rotation (millirads)
Figure 5.18 Load v mean end-rotation
Plots with the maximum load normalised to 1000kN are shown in Figure 5.19. These plots
show how similar the behaviour is up to the rotation at which local instability of the wall
precipitates a drastic loss of resistance. It is difficult to identify the individual tests in Figure
5.19, but the important point is that all the tests are similar except the 5mm wall tests drop
significantly from about 40 milliradians.
87
kc3 to kclO;
normalised axial v mean end rotation
- ke3
1100
1000 - ke4
900
- ke5
800
2: 700
.x. - ke6
-c 600
'"
.S!
500 - ke7
"'iii
'x
« 400
ke8
300
200 ke9
100
0 kc lO
0 20 40 60 80 100
Mean end rotation (millirads)
Figure 5.19 Normalised load v mean end-rotation
5.7.6 Difference in rotations at each end
The least welcome observation was that the end rotations at top and bottom of the specimen
were different as shown below in Figure 5.20 to Figure 5.31. These figures show the end-
rotations at top and bottom of each test with a plot of the difference between the end-rotation
and the mean of the end rotations.
The companson of the tests with the analyses reported in Chapter 6 indicates that the
primary reason for the difference in the zone from zero to maximum load almost certainly is
sway of the rig, as discussed in that chapter. The stability of the column is governed by the
bending moment at mid-height, so the most reliable measure of end-rotation is the mean end-
rotation . In the zone beyond maximum load, the proportion of the member that is strained
beyond yield is very large, so there is little elastic stiffness remaining. Therefore, normal
expectations of structural behaviour are not fulfilled. With such extensive plasticity, the
probable reasons for the difference in end rotations are either the slight variation of stress-
strain curve or the slight variation of wall thickness along the member. Even slight
variations could cause asymmetry of the curvature about the mid-height of the member,
resulting in asymmetry of end-rotations. As a result of these observations, the analysis of the
results was made using the mean value of the end-rotations.
88
kc 3; load vend rotation
T- rT····j-··,-- T-l
I I I i I
T-t- r-l-
I
.). - j
I
I
II I II I
I
Z l~-j
i
--1 ----t- -t---1
~
s:
.2
.,.
::I
~,., ,L-~ : I _roln in-plane top P 2
Do
E
0
u
~. -t-~iI
.
:! I
I I
;
I
C
fa r
I
t +--1 -rotnin-plane bottom P 3
r
I II I
I
i
,~
r
I
. ~
l
_LJ
! ~. _L_ ., _~ _.1. _
-10 0 10 20 30 40 50 60 70 80 90 100 110 120

End rotation (mrads)
Figure 5.20 kc3 end rotations
kc 3; difference from mean end rotation
I E
- I
~ + -----; -
16-
o I
.. I
~ ~.~-+-+.-+-+-4-4~---~~~
GI
C
It -kc3 difference
GI
E from mean rain
E
...e
~ ~I~~-+~+--+4L~~--~---~~--r--+~
j t +~ ~
Mean end rotation (mrads)
10
Figure 5.21 kc3difference from mean end rotation
89
kc 4; load vend rotation
1600
1400
1200
z
"""c: 1000
o kc4 rotn il1-
'iii
III
QJ plane top
C. 800
E
o - kc4 rotn in-
~ 600 plane
~ bottom
400
200 -j-----;
o --- '-----
o 10 20 30 40 50 60 70 80 90 100 110 120

16
- kc4
.,. - - - t - -
o
o 10 20 30 40 50 60 70 80 90 100
Figure 5,23 kc4difference from mean end rotation
90
kc5; load vend rotation
1000
I I .. I
900 -----r---r- ----I
I I ,
800
Z 700
~
r::
0 600
1III
15. 500
E ~kcS test rotn in-plane top P 2
0
U
400 I
1 -kcS rotn in-plane bottom P 3

~ 300
200
100 t" .
0 - -I i
___1. _
0 20 40 60 80 100 120

16 r-
'iii'
I
.
1 14
!
I
- r--
12
j
.
~ 10
~ I
t- -
!
.
III
r:: 8 II
~ I
. I
- kcS difference from mean rotn
6
-.
E ----------t------j P3
~ I I
III
u
r::
III
III
4
2
--l-----
----1-----
- - - _ !-.---.~ --J
I
!I
I
!
I
! I
- ----r--------j
I i:
.
0
0 20 40 60 80 100
91
kc6: Smm wall thickness
load vend-rotation
800
700
GOO
z
:!.
c 500
0
·iii
11\
CII .....- keG rotn in-plane
Is. 400
E topP 2
0
--~~ -·-I-""~~
u 300
iii
~ 200 -+-_.
I -keG rotn in-plane

100 bottom P 3
0
0 10 20 30 40 50 GO 70 80 90 100 110 120
kc 6; Smm wall thickness,
-I II
"tI
I!
30
difference from mean end rotation
-S
E
25 ---- ~--
~... 20
-g
CII
; 15
CII
E
j 10
I
- keG difference from
mean rotn
CII
~ 5 +-----+---I----~~
!
:!Q +I
--~--
0
92
kc 7: load vend rotation
1400 ----,---,-
i I
I I
1200 --L-t-
I I
- 1000 --j--t-
-
Z
.¥
c
0
'Vi
800
_1 __...'
1/1 ~kc7 rotn in-plane top

Qj
"-
Q. 600 P2
E
0
u
400
ii
~ 200 -kc7 rotn in-plane
bottom P 3
0
o 10 20 30 40 50 60 70 80 90 100 110 120
- I II
'a
l!! 20
kc 7: difference from mean end rotation
---l-------.--
- E
18
16
I / I
II
14
12
I
10 - - - - - ----)---
6
If I -kc7 differer
mean rotn
L I
4
2 --- ~
o ~---- ------~--'----'-- _______J

I
o 10 20 30 40 50 60 70 80 90 100 110
93
1400
1200
- 1000
-
Z
.¥
c
0
'iii
800
~kc8 test rotn in-plane
'"GI
~
Q. 600 topP 2
E
0
u
400
iii
i 200 -kc8 rotn in-plane
bottom P 3
o 10 m m ~ ~ w ro ~ ~ 100110 1m
Figure 5.30 ke8 end rotations
-'"
~ 16
-E
c 14
~
o
r!I
12
I
~ ,t r
~
GI
C
8
"GI
E
E 6
o -kc8 difference from
....
~
4
mean rotn
GI
~
2! 2
~c o
o 10 m m ~ ~ w ro ~ ~ ~
Figure 5.31 ke8differenee from mean ke8 end rotation
94
1200
1000
-
-
~
Z
c
0
'iii
800
.
\/I
cu
Q.
600 -+-ke9 rotn in-plane top
P2
E
0 400
u
ii
~ 200 !
-ke9 rotn in-plane
-1..-. ___ L ___

I
~_
bottomP 3
0 ~
o 10 m ~ ~ ~ w m w ~ ~11O
-~
\ /I
14
-- r--
-..
E
oc 12
I
"'
~ 10
6
-kc9 difference from
4 m~nro~
o
o 10 20 30 40 50 60 70 80 90 100
95
1000
800
-
-Z
.¥.
c
0
'Vi
600
.a.
1/1
GI
400
~kc10 rotn
topP 2
in-plane
E
0
\,I
ii
~ 200
-kc10 rotn in-plane
bottom P 3
0
0 10 20 30 40 SO 60 70 80 90 100
Figure 5,34 kc 10 end rotations
-'a
1/1
I! 18 :~
---- '---~-T-~-~ ,------ - - -~-I
----
-'~
E
c
0
16
! -
!
I
I I
I I
I
...0.
1\1 14
12 I
~
'a
c I
GI 10 c~--- - ._. 1------ t--- -
c
1\1
QI
8 ----~
r----- I- J
E
E 6 ~ ~ -kc10difference from
0 iI"""" --
~
QI
\,I 4 I ~ ./ i
J mean rotn
.
C
GI
2 II \ )
~
Q 0
~ ____\~___ ~L ____ II J
....L~_
Figure 5.35 kclO difference from mean end rotation
96
5.8 Effect of wall thickness
The two columns with 5mm wall thickness, kc6 and kc 10, show a clear drop in axial load
arising from buckling of the wall. This is seen in terms of end-rotations in Figure 5.36. The
mid-height displacement is approximately 40 to 50 mm as seen in Figure 5.12.

kc6: load vend-rotation
800 r -- 800
700 700 f
600 600 I
I
I I~
! I i "
I~
500 500 I
I I I '\.
\
t1
:_z ~~-
400 400 -
\.
r ---r--
~...
:!
~
300
200
-
"
-~-- ,
f
'~t
- --+- ~
t
100
l' .---~~------
0 ! I
o _~L_
o 10 20 30 40 50 60 70 80 9010Cllm20
o 10 20 30 40 50 60 70 80 90 100
End rotation Imrllds)
- - rotn in-plane top

-rotn in-plane top
- - rotn in-plane bottom
- - rotn in-plane bottom
Figure 5.36 kc6 and kc10 end rotations
The two columns with 6.3mm wall thickness, kc5 and kc9, show only a very slight reduction
in axial resistance in the tests. There is a small increase in the downward slope of the axial
load versus mid-span displacement curve at a mid-height displacement from about 70 mm as
seen in Figure 5.12. This is seen in terms of end-rotations in Figure 5.24 and Figure 5.32.
These observations and analysis contribute to the formulation of a design limitation on wall
thickness later in this report.
97
5.9 Summary
Tests have been perfonned on full-scale columns with different wall thicknesses, using
120x 120 SHS sections. The column behaviour was generally as predicted both by analysis
including non-linear geometry and material properties and by the proposed design model.
Sections with 10mm wall thickness showed no wall instability over the full range of the
tests, whereas sections with thinner walls displayed wall instability.
98
99
6 COMPARISON OF TESTS WITH ANALYSIS
6.1 Introduction
6.1.1 General
The reason for the comparison of the tests with the Abaqus finite element model is to
confirm the structural behaviour assumed by the design method and to calibrate the results of
the parametric study made using Abaqus analysis so that the behaviour of the tests is
included in the design model.
The test specification required the columns to be installed into the rig with eccentricities
such that the mid-height bending moment would be similar to that occurring in columns with
the worst out-of-straightness allowed by current Standards. The test columns were found to
be very nearly perfectly straight, so the columns were installed in the rig with equal
eccentricity top and bottom of Ll750 which was the value of out-of-straightness being
discussed by European standardisation committee CENrrC250/SC3 at the time of the test
program.
6.1.2 Summary of method
The general form of the load-displacement curve measured in the tests is broadly similar to
the results generated by Abaqus if the initial imperfection is applied as either a straight
member at a uniform offset from the line of thrust between the rig bearings or a member
curved throughout as a half-sine curve. However, the test curves are slightly different,
especially in that many of them show distinct asymmetry of end-rotations even in the elastic
range. To achieve Abaqus analyses that are closer to the test curves, initially it was decided
to alter the initial imperfection to arrive at close agreement in the elastic range. The
imperfections required to make the Abaqus analysis agree with the test results can be
calculated with reasonable accuracy from simple structural mechanics for a first analysis.
The imperfections can then be refined from the results of this first analysis.
Because the measurements of the straightness of the specimen showed such small deviations
from straight, the Abaqus model was made straight. Some investigation of different initial
geometry was made, but the out-of-straightness observed before the tests was so small
compared with the eccentricity deliberately applied at the ends that it was decided to model
100
the members as straight. Compared with the effects of possible variations of friction in the
bearings, yield stress and section properties, the assumption of a perfectly straight member
has only a modest effect. The sensitivity to member curvature between the ends was
investigated, as reported in Section 6.4.4.
Abaqus was used to reproduce the maximum resistances found in the tests, assuming that
1. the stress-strain was the simple elastic/perfectly-plastic stress strain diagram as used
in the Abaqus parametric study,
2. the yield stress in the steel was equal to the test coupon 0.2% proof stress,
3. the residual stress is 10% of the nominal yield of 355 MPa with the same stress
distribution as used in the Abaqus parametric study,
4. the Young's modulus of the steel is 210000 MPa as in Eurocode 3 [BSI2005a,
CEN 2005a],
5. the behaviour of the test column was similar to a pin-ended member with the friction
effect equivalent to a reduction in the eccentricity of application of load.
The results of analyses using these asymmetric eccentricities are shown in Section 6.3 and
the key test and analysis data are shown in Table 6.2, The Abaqus analyses showed that the
eccentricities were far in excess of what is credible with such carefully made end-fittings
which allowed fine adjustment and in a laboratory with such long experience of testing.
On reflection, it was realised that the fmite stiffness of the relatively tall test rig must allow
some small amount of side-sway. This could be initiated by the axial load because it is
impossible to achieve perfect vertical alignment. The inclinometers use gravity as the
datum, so sway of the test rig will increase the inclination at one end of the column while it
gives an equal decrease in the inclination of the other end. Therefore, even if the bearings
were perfectly friction-free and the end-rotations of the column were equal relative to the
line of thrust between the centres of the bearings, the inclinometer output would indicate
unequal end-rotations. After this was noticed, the analyses of the tests were then conducted
with equal eccentricities top and bottom with the eccentricity chosen to give reasonable
agreement between the test and analysis for both the maximum load and the mean end
rotation. The results of analyses using these asymmetric eccentricities are shown in Section
6.4 and the key test and analysis data are shown in Table 6.3.
In an ideal world, differences between the Abaqus behaviour and the test results might be
resolved by adjusting the Abaqus model until the Abaqus results exactly mirror all the test
data including
101
1. maximum load
2. end-rotation at each load
3. in-plane deflection at mid-height at each load.
However, this requires a precise knowledge of the test specimen, including

1. stress/strain curve at all points including dynamic effects,
2. geometry including all irregularities,
3. distribution of residual stresses, and
4. a precise knowledge of the behaviour in the rig, including
• what friction is acting at what time,
• what is the position of the bearing components relative to the specimen,
• any inclination of the specimen or the line of thrust (from centre to centre of the
bearings)
• the sway behaviour of the rig
• the flexural behaviour of the columns of the rig when loaded because the
transducer recording the in-plane deflection at mid-height was attached to one of
these.
Very high precision is NOT required on these points for the falling branch of the tests (the
region of interest for these tests) because the deformations of the specimen are large in this
region.
The Abaqus output is from analyses in which the specimen is assumed to be perfectly
straight and offset by a constant eccentricity from a line between the centres of the bearings,
except in the analyses investigating the sensitivity of the analysis to curvature in the
specimen in Section 6.4.4.
The eccentricity is chosen so that the maximum resistance of the Abaqus analysis is close to
the maximum load recorded in the test, ideally within 1%. The yield stress was generally
taken as the yield stress from the coupon test.
6.1.3 Estimating the offset of the column ends from line of load
The effect of an offset from the line of load is to induce a moment in the member. This
causes curvature in the member which results in end rotations. The axial compression
102
increases the moments according to the deflected shape. The behaviour is non-linear, but the
non-lineatity is small at low levels of load.
When the non-linear effects are small, which is the case when the axial load is small at the
beginning of the test, the end-rotations approach the end-rotations of beams with end
moments calculated by elastic small-deflection theory. The bending moments and end-
rotations, assuming the effect of axial load is negligible, are shown in Figure 6.1 for a pin-
ended column with an eccentric load applied at the top. The bending moments and end-
rotations, assuming the effect of axial load is negligible, are shown in Figure 6.2 for a pin-
ended column with an eccentric load applied at the bottom.
In some tests, the curves of end rotation and mid-height displacement exhibited steps at low
levels of load, making it difficult to calculate the initial imperfection accurately.
I
I
I
I
I
End rot tions
I
I
I
I
I
I
I
N Bending
Loads Moment Deflections
Diagram
Figure 6.1 End rotations from eccentric load at top of column
(neglecting non-linear effects)
103
N
~Ne"u3EI
N
Bending
Loads Moment Deflections
Diagram
Figure 6.2 End rotations from eccentric load at bottom of column
(neglecting non-linear effects)
For a traight bar with negligible non-linear effects,

The end rotation at the top is
Eq 6.1
Eq 6.2
Eq 6.3
Eq 6.4
Similarly, the end rotation at the bottom is
Eq 6.5
Eq 6.6
The end rotations can be fOW1d in terms of d9/dN from
Eq 6.7
Eq 6.8
104
=> 2d9t!dN- d9ldN = 3ebL/6EI = ebL/2EI Eq6.9
=> Cb= 2x(2d9t!dN- d9ldN)x(EIIL) Eq 6.10
Similarly
2d9ldN = (4et + 2eb)L/6EI Eq 6.11
d9t!dN = (et + 2eb)L/6EI Eq 6.12
=> 2d9ldN- d9t!dN = 3etL/6EI = etL/2EI Eq 6.13
Eq 6.14
6.1.4 Cross-sectional area effects
The cross-sectional area of the model is slightly larger than the nominal area of the test
specimen to allow a simple model with the correct wall thickness, because the models use
the nominal wall thickness and assume square corners. The test columns, being hot-finished
sections, have curved corners with very tight external radii. The exact area of the specimen
was not measured, so the nominal area of the section is assumed. The analysis results were
multiplied by a reduction factor of 'nom/model', the ratio of the nominal area to model area,
from Table 6.1 before making comparisons with the test results either by plotting or by
calculations. Steel producers are very careful not to give away steel in their cross-sections
and there is an area tolerance on sectional area in the product specification. It is very
unlikely that the areas were greater than the nominal area.
Table 6.1 Cross-sectional areas
test wall model area nominal area nom/model

thickness (mm2) (mm2)
kc3,4,7,8 10.0mm 4xl0x(120-10) = 4400 4290 0.975
kc5,9 6.3mm 4 x6.3 x(120-6.3) = 2865 2820 0.984
kc6,10 5.0mm 4 x 5x(120-5) = 2300 2270 0.987
105
6.2 Correlation factor for test v Abaqus
6.2.1 Introduction
As shown in Figure 6.10 to Figure 6.17, the Abaqus analysis predicts behaviour close to that
measured in the tests, but not identical. The relationship between the Abaqus analysis and
the test results is expressed as a correlation factor on the end-rotations. This factor is
calculated from the top of the unloading/reloading cycle and is the ratio of the end-rotation
measured in the tests to the end-rotation predicted by Abaqus. The correlation factors are
used to adjust the results of the Abaqus parametric study when defining the design model.
The top of the unloading/reloading cycle can be found from the plot of the mean end-rotation
v load. A typical test plot of is shown in Figure 6.3. The maximum load is at B. The top of
the unloading/reloading cycle is at C. This point C is the point at which the test operator
began to unload the specimen, so the end-rotation ceases to increase while the load is
decreasing. At this point, referred to later as the "static" point, the strain rate is zero and the
friction moment is about to decrease. The load at this point is designated Nu, the load at the
point of unloading.
axial load N
B
A mean end-rotation
Figure 6.3 Typical end-rotation v load from tests
106
6.2.2 Calculating the correlation factor Ct
The unloading/reloading cycle, described in Chapter 5, provides a point at which the test
results can be compared with the Abaqus results. The method used to calculate the
correction factor is to compare the mean value of the end-rotations at the top of the
unloading/reloading cycle with the end-rotation predicted by Abaqus at the same load.
The mean value of end-rotation in the laboratory tests at the point of unloading, designated
Oct. is shown in Figure 6.4. The value of OCt is found from the test results at the load Nu•
The value of mean end-rotation predicted by Abaqus, designated OCA , is found by
interpolation to Nu from the adjacent points from the Abaqus output.
The correlation factor, Cf, is the ratio of the rotations,
cf= OCt / OCA Eq 6.15
axial load N
Abaqus curve
test curve
A Ct OCA mean end-rotation
Figure 6.4 Mean end-rotation v load: test v Abaqus
The key points of test kc3 are shown in Figure 6.5. The point of maximum end-rotation is
easily seen and can be verified by inspecting the tabulated test results. The difference
between the test rotation and Abaqus rotation at this load is also seen in Figure 6.5. It can be
107
seen that for kc3, the Abaqus prediction of rota6on is very slightly less than the test result at
that load. The opposite is seen in Figure 6.6 which shows that for kc8 the Abaqus prediction
of rotation is greater than found in the tests.
The values of Cf are given in Table 6.2 for asymmetric eccentricities and in Table 6.3 for
symmetric eccentricities. The effect from partly curved imperfections on Cf is investigated in
Section 6.4.4 and the effect of residual stresses on Cf is investigated in Section 6.4.5. The
choice of value of Cf used to modify the parametric study is given in Section 6.6.
load vend rotation; test kc3 v Abaqus
1500
1400
1300
1200
1100 f
~1000
c 900
.~
.. 800
is. 700
-+- kc3 test mean end rotn
..e
E
600
500
~ 400 j ___ enlarged area
300
200
100
0
0 10 20 30 40 50 60 70 80 90 100
load vend rotation; test kc3 v Abaqus

enlarged area
1050
z -+- kc3 test mean end rotn

~ooo
c
o
.~
<11
Q. -+- Nu from rotations
E
o
u
~ 950
x
<{
- .. - Abaqus rotn at Nu
_ enlarged area
900
33 34 35 36 37
End rotation (mradsl
Figure 6.5 Top of the ullioading/reloading cycle - kc3, test & Abaqus
108
load vend rotation
test ke8 v Abaqus
1,500 T
1,400 + --r- !---j---+---+--t------f

1,300
1,200
_ 1,100
h:.a1lL.+r----1~
:1 t-
~Ab a qu s
kC8_2009mOld23_v2
~1,000 +
g 900
'iii
~ 800
a 700
-+- kc8 mea n end rotn
E
8
iO
'x 500
<t 400
600
[
l
300 t---
200
100
o
o 10 20 30 40 50 60 70 80
rf 1 90 100
...... enlarged area
load vend rotation

test ke8 v Abaqus
960
"""*- Abaqus
940 kc8_2009mOl d23 _v
2
~ kc8 mea n end rotn
Z 920
:.
c
0
'iii 900 t
'"~ _ Nu from
Q.
E
displacements
0
u
iO ~ Abaqus rotn at Nu
'x
<t 860
840
~ _ enl arged area
820
34 36 38 40 42 44 46
Figure 6,6 End-rotations at top of unloading path - kc8 , test & Abaqus
109
6.3 Straight members with asymmetrical eccentricities
6.3.1 Introduction
Analyses were made with different end eccentricities at the top and the bottom of each
specimen with the eccentricities adjusted until the end-rotations at 80% maximum load was
very close to the value from the test results. Analyses of all the tests were conducted with
perfectly straight members. The results are summarised in Table 6.2 together with the
correlation factor which is defined in Section 6.2.
6.3.2 Comparison of Abaqus analysis with test results
The key test results and the Abaqus analyses are listed in Table 6.2. The tests with the lowest
value of correlation factor were kc7 and kc8. The plots of load vend-rotations for the full
range of these two tests, together with the plots of load vend-rotations for the loading path
shown to a larger scale, are shown in Figure 6.7 and Figure 6.8. The plots of load v end-
rotations for the loading path show that the imperfections in the analysis give end-rotations
close to the test results.
load vend rotation load vend rotation

test kc7 v Abaqus test kc7 v Abaqus
1600 1600
1400 ,-~ ~ -- -~- ----~--- --- - ~ --~ -~~-- 1400
Z 1200 Z 1200 -
:!. :!.
.~ 1000 : g 1000
1. 800 ,-----
600 ,
'=~
§ 600
800 .
.--.- ~--+---+
:! :! , .
.ll 400- .ll 400 ------+--+ -~--~~-.
200 . 200
o o
o 10 o 10 ID ~ ~ ~ w ro ~ ~ 100 110 UO
Top and bottom end rotations (mradsl Top and bottom end rotations (mradsl
-kc7 rotn in-plane top -kc7 rotn in-plane top

_kc7 rotn in-plane bottom -.&-kc7_2009m04d17 _v2 top
"'!!!-kc7_2009m04d17_v2 top _kc7 _2009m04d17 _v2 bottom
_kc7 _2009m04d17 _v2 bottom _kc7 rotn in-plane bottom
Figure 6.7 End-rotations - kc7, test & Abaqus
lIO
load vend rotation
load vend rotation
test kc8 v Abaqus 2650mm
test kc8 v Abaqus 2650mm
1600 ,----. --- ---------~----- -- - -
1600 -~--~---.------ r-----··-
1400 1400 II
1200 '
i
i 1000 - - - - - # - - - #
IJ :
400
200
200
10
10 20 30 40 50 60 70 80 90 100 110
Top and bottom end _'ons (mrads)
Top and bottom end rotations Cmradsl
-I«:S test rotn In-plane top
-keS test rotn In-plane top
-I«:S rotn In-plane bottom -keS rotn in-plane bottom
-e-I«:S_200SmlOd24_vl top -.!r-keS_200SmlOd24_vl top
-1«:8_200SmlOd24_vl bottom -keS_200SmlOd24_vl bottom
Figure 6.S End-rotations - keS, test & Abaqus
111
Table 6.2 Abaqus with asymmetric eccentricities
Comparison of tests with asymmetrical model

Abaqus analysis with the classic elastic/perfectly-plastic stress/strain curve.
Test kc3 kc4 kc7 kc8
Wall thickness (mm) 10 10 10 10
Max test load (kN) 1328 1458 1290 1298
0.2% proof stress (MPa) 375 390 not known 375
area factor 0.975 0.975 0.975 0.975
yield stress (MPa) used in Abaqus 375 375 413 375
analysis note 1
residual stress (MPa) 35.5 35.5 35.5 35.5
Abaqus max load 1361 1458 1327 1310

area reduced Abaqus max load (kN) 1327 1422 1293 1277
difference in max loads (kN) -1 -36 +3 -21
Abaqus reduced N - test N
Abaqus et (mm) 2.25 3.20 10.39 7.92
Abaqus q, (mm) 2.87 -1.34 -0.59 -1.99
Abaqus mean e (mm) 2.56 0.93 4.90 2.97

Abaqus deviation from mean e (mm) -/+ 0.31 +/-2.27 +/-5.49 +/- 4.96
Correlation factor for mean end- 0.991 1.018 0.749 0.829
rotation, cf= OCt/OCA
note 1 Using the coupon stress as the yield stress with the classic
elastic/perfectly-plastic stress/strain curve, the analysis model
could not continue beyond the maximum load
note 2 Abaqus input files:
kc3:kc3- 2008mlOd19- v2
kc4:kc4_2008mlOd22_v3
kc7:kc7_2009m04dI7_v2
kc8:kc8_2008mlOd24_vl
The results for tests for kc5, kc6, kc9 and kclO were not calculated because it was concluded
from the results ofkc3, kc4, kc7 and kc8 that the tests had been affected by sway of the test-
rig, see Section 6.3.3, before these tests had been analyzed. Therefore there was no benefit
112
in completing the analysis of the other four tests using the assumption of asymmetric
behaviour. The low values of correlation factor, Cf, found for tests kc7 and kc8 are similar to
those reported in Table 6.3 and discussed in Section 6.4.2.
6.3.3 Asymmetry of specimen or of rig
Adjusting the Abaqus model to reproduce the end-rotations recorded leads to the
eccentricities eb and et given in Table 6.2. In the tests with the greatest differences between
end-rotations at the top and the bottom, tests kc7 and kc8, the eccentricities are not
believable considering the care taken in the assembly of the tests. The eccentricities for kc7
are shown in Figure 6.9.
mean eccentricity = 4.90
-0.59
Figure 6.9 Asymmetric eccentricity ofloading for test kc7
It should be noted that a significant eccentricity is required to cause only a very small
difference of end rotations. This is demonstrated by comparing Table 6.3 with Table 6.2. At
80% load, the end-rotations differ from the mean value by only +/- 0.9 mrads, but the
difference in end eccentricity needed to generate this is +/- 5.49 mm. However, 0.9 mrads
rotation from vertical would be generated in a rig height of 3.0 metres by a sway of only
113
0.9xlO·3 x3000 = 2.7 mm. This supports the theory that the rig swayed slightly, so the
rotations recorded relative to the vertical were not the end-rotations relative to the line of
thrust though the bearings.
Because of the high probability that sway of the rig had made the end-rotation test output
differ from the end-rotations relative to the specimen, it was decided to re-analyse based on
the assumption of equal end-rotations. The probable sway of the rig limited the use of the
mid-height deflection values because the LVDT was fixed to a leg of the rig. If the sway
involved a different stiffness between the top and bottom of the rig columns, which is likely,
the reading of the LVDT would be affected by the rig sway.
6.4 Straight members with symmetrical eccentricities

6.4.1 Introduction
Analyses were made with equal end eccentricities with the eccentricity adjusted until the
mean end-rotation at 80% maximum load was very close to the value from the test results.
Initially, the analyses of all the tests were conducted with perfectly straight members so that
correlation factors between analysis and test could be calculated. The results are summarised
in Table 6.3 together with the correlation factor, Cf , which is defined in Section 6.2. Then
the sensitivity of the results to effects of
1. out-of-plane deflections,
2. partly-curved initial imperfections,
3. residual stresses
were investigated to find the effect on the correlation factor.
6.4.2 Comparison of Abaqus analysis with test results
The plots of the Abaqus analyses listed in Table 6.3 are shown in Figure 6.10 to Figure 6.25.
Although in general the correspondence between test and analysis was very good, there were
some difficulties encountered in the analysis. Test kc4, for example, proved very difficult to
analyse. The minimum imperfection possible that would allow the analysis to pass the peak
load was 0.73mm with 375 MPa yield despite changing solution time-step parameters from
10's to 10-9• It can be seen from Figure 6.19 that the Abaqus plot is very close to the test plot,
so the results are acceptable. However no improvement was found possible. The analysis
would not run with yield at 390 MPa, the coupon 0.2% proof stress. Analysis runs
114
kc4_2009m03d06_vl to _vIO were devoted to using a stress/strain curve that had an elastic
limit at or below 375 MPa and a reduced modulus beyond, with a stress/strain graph passing
through 390 MPa, but a higher load and an analysis continuing beyond the peak load was not
achieved.
Analysis of tests kc3, kc4, kc8 and kc I 0 show analysis that is very close to the tests plots,
showing that the eccentricity in the analysis must be very close to the eccentricities in the
tests.
Test kc5 shows good agreement at 50% maximum load. Test kc6 did not have an unloading
cycle to allow a correlation factor to be calculated. The analysis of kc7 is not so close.
Regrettably the coupon test result for kc7 was not available, so the comparison with analysis
is uncertain. Test kc9 analysis is offset from the test results, but the analysis plot is 'parallel'
with the test plot. This effect might arise from a friction effect in the test rig or from the
stress/strain curve of the material being different from elastic/perfectly-plastic.
test 3; load vend rotation

v Abaqus 2650mm
1600
-I· - I
I
-kc3 test mean t'nd

rotn
-++-kc3_2009mOld09_
v2
200
o 10 20 30 40 50 60 70 80 90 100 110
Figure 6.10 End-rotations - kc3, test mean & Abaqus
115
load vend rotation
test 4 & Abaqus
1600
1400
1200 - kc4 test mean end

Z
.II:
rotn
j1000
f 800
CI. _ _ k(4_200~mOld18_v
E
4
8 600
:!
~
400
~- ----- --~~- -~~--r--
o 10 20 30 40 SO 60 70 80 90 100
load vend rotation

test 5 & Abaqus
1000
800 -
! -kcSrotnin-plane mean
g 600
-=a
GI
II---~+-~-I--
~ 400
:!
~
200
I
o ~_~-"--~--,--_l~_
116
load vend rotation
test 6 v Abaqus
800
600
Z
:!.
c
·i
~ 400 ~.-~-+--
CI.
E
S
:! -kc6rotn in-plane mean
~
200
o w m ~ ~ ~ w ro w ~
load vend rotation

test 7 v Abaqus
1600
1400
-kc7mean end roln

z 1200
oM
} 1000
~ 800
CI. ~kc7_2008m12d12_
E
0
u v2
600
:i
~
400
200
0
0 10 20 30 40 SO 60 70 80 90 100 110
117
load vend rotation
test 8 v Abaqus
1600
1400
1200 -kc8mean end roln

Z
~
j1000
:l!
~ 800 f;f;-.-+--t--+-I'-t--"""'E:-+-~------:--+--+--<---1
a. ~ kcB_2009mOldB_
E
o v2
u 600 1-f----+--t--+--+---f--,-+-......-1----.;!c::I"I!IIlIIa!~____1
:!
~
400
200 t---+--+-~+----+--+--+-----;----i---_t____;
Figure 6.15 End-rotations - ke8, test mean & Abaqus
load vend rotation

test 9 v Abaqus
1000.0 ---r ---- --T--'---r---T--T--;
1 I . I I
I i II I I . I
800.0 1
.
;11~.- . ------1 I
1
__ ~_
,
I
+_ _ ~
I
I
I
1
'
I
Z
:!. I I iii
S600.0
-=~
a.
~400.0 ------+-~----+- --- - ke9 rotn in-plane mean
:li
~
I I
200.0 1---+-_t_-t------+---'--i--r-----i---t----1----'
0.0 (I}---l._~_-"----l._--'-~-'------L_--'-_ _'___'---'
0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 100.00
Figure 6.16 End-rotations - ke9, test mean & Abaqus
1I8
load vend rotation
test 10 v Abaqus
1000.0 ---~-- ~ . - r --r---r- -T--·-~ -~-
I i tI
800.0 r--:--t--+--+----+I--+--+---+--+---1
z I
~
6600.0
-m I
~
]
E
8400.0 ··l-·· ---+----r-.~-.
I I I
I
i
- ke10 mean end rotn
200.0
0.0
-~-.tj-tJ- L
o 10 20 30 40 50 60 70 80 90
Figure 6.17 End-rotations - kc 10, test mean & Abaqus
test 3; load vend rotation

v Abaqus 26S0mm
1600 I---~ . ----.. . ~~r-··---·--·----·-·-l
1400
1200 f--------=~-+_--------j -kc3 test mean end

z
~
rotn
~1000 .~~.- ... -- . . . --.-~-
.S!
=
~ 800
0-
--"--- -_... -..... ---.-~.-.--------~-
~kc3_2009mOld09_
S
... 600
I
v2
:!
~
400
200
0
0 5 10
Figure 6.18 Initial end-rotations - kc3, test mean & Abaqus
119
load vend rotation
test 4 & Abaqus
1600
1400
1200 -kc4 test mean end

Z
~
rotn
I
i'Bi lOOO --- ----t- --- ---- -- .-- -- --
I
fCI. 800
__ kc4_2009mOld18_v
~ 4
u 600
:!
~
400
200
o ___ 1____ - ___________ _

o 1 2 3
load vend rotation

test 5 & Abaqus
1000
800
Z
:!! I -kc5 rom in-plane mean
.~ 600 ---I
I
"& !
8 400
u
:!Ie
c(
200
0
0 5 10
120
load vend rotation
test 6 v Abaqus
800
600
Z
:!.
5
.~
~ 400
CI.
g
u
ii - ke6 rotn in-plane mean
~
200
o~---------------~----
o 5 10
load vend rotation

test 7 v Abaqus
1600 , - - - - - - - - - - - - , - - - - - - - - - - ,
1400 f-----------+-----------1
1200 f----------+--------r:;;orl"'''--------1
Z
~
-kc7 mean end rotn
1~ 800
CI.
1000
E
8 600 1---------:il~~--+---------1
:!
~ 400
200
o~-------~-------~
o 5 10
121
load vend rotation
test 8 v Abaqus
1600
1400 1_I - __ ____1~

!
_______________________ ~~I
1200 -kcBmt'an end rotn
Z
.lit.
jlOOO 1--
i
~ 800
CI. ,""*""kcB_2009m01dB_
E
v2
B 600
!
~ 400 --------~- - - - -----L-------------~-----------
200
I
o - ----- ------ - __1 _______________________ _
o 5 10
Figure 6.23 Initial end-rotations - ke8, test mean & Abaqus
load vend rotation

test 9 v Abaqus
1000_0 r------------=~;;;;:::::;;~;;;;,;;;;;;:;;;;;;;;;;;;~
I
800.0
I - ------ -----~---- ------------- ----II

~600.0
"15.
~400.0 --+---- - -~---~~----
~
- ke9 rotn in-plane mean
! I
200.0 ~-~-------- --~---+---~-~------------------l
0.0
0
_~__________________ J 5
._ 10
Figure 6.24 Initial end-rotations - ke9, test mean & Abaqus
122
load vend rotation
test 10 v Abaqus
800.0
_600.0 f------~~--+_-------__I
z
:!.
.~
GI
~400.0
&
... -kclOmeanend rotn
ii
'j(
0(200.0
0.0
o 5 10
123
Table 6.3 Abaqus with symmetric eccentricities - Comparison of tests and model.
Abaqus analysis using the standard elastic/perfectly-plastic stress/strain curve.

Test kc3 kc4 kc5 kc6 kc7 kc8 kc9 kcl0
Wall thickness (mm) 10 10 6.3 5 10 10 6.3 5
Max test load (kN) 1328 1458 915 695 1290 1298 971 777
0.2% proof stress 375 390 428 389 note5 375 431 402
(MPa)
area factor 0.975 0.975 0.984 0.987 0.975 0.975 0.984 0.987
yield stress (MPa) 375 375 437 389 400 375 431 402
used in Abaqus note 1 note2
analysis
residual stress (MPa) 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5
Abaqus max load 1356 1481 930 703 1320 1333 991 773
area reduced Abaqus 1322 1444 915 694 1287 1300 975 763
max load (kN)

Abaqus red' d N - test -6 -14 +0 -1 -3 +2 +4 -14
N 0.5% 1.0% 0.0% 0.1% 0.2% 0.2% 0.4% 1.8%
(kN) note4
Abaqus et (mm) 2.68 0.73 5.27 4.3 4.9 3.08 3.15 2.55
Abaqus eb (mm) 2.68 0.73 5.27 4.3 4.9 3.08 3.15 2.55
Correlation factor 1.003 1.021 0.831 see 0.804 0.841 0.941 0.918
Cf= 8et18cA note3
note 1 Using the coupon stress of 390 MPa as yield with the elastic/perfectly-plastic
stress/strain curve, Abaqus could not continue beyond the maximum load.
note2 A yield of 428 MPa, higher than the 0.2% proof stress, was used in the analysis
because the maximum load was not achieved with an initial imperfection sufficient
to make the analysis rotations agree with the test rotations.
note3 No unloading cycle in tests so no point at which to calculate correlation
note4 Using the coupon stress as the yield stress with the standard elastic/perfectly-plastic
stress/strain curve, the test maximum could not be achieved
note5 0.2% proof stress not known
Abaqus kc3 - 2009m01d09- v2 kc6- 2009m03d20- vI kc9_2009m03d19_v1
input kc4- 2009m01d18 v4 kc7- 2009m03d15- v5 kc10- 2009m03d07- vI
files kc5_2009mO 1d27_v 1 kc8- 2009mO 1d23 - v2
124
The correlation factors, which are correlation factors on end-rotation at a given axial load,
are shown in Figure 6.26. The mean value is 0.91, the maximum is l.042 and the minimum
is 0.804.
The reason for these differences between the test results and the Abaqus analysis is not
known. Possibilities are:
1. Cross-sectional area of some sections might be less than nominal. The tandard for
hot-finished hollow sections, BS EN 10210-2:2006 [BSI 2006b], allows the
weight/metre to vary by +/- 6% and individual wall thickness to vary by as much as -
10%. This could explain some deficiency in measured resistance versus predicted.
2. Stress/strain characteristics at strains greater than yield strain cannot be expected to
be as simple as the bi-linear stress- train characteristics used in the model. It is
probable that the actual stress at 1.0 yield train yield is below the 0.2% proof stress,
so the resistance i less than the Abaqus analysis predicts simply because it assumes
the yield stress is equal to the 0.2% proof stress when the strain is equal to 1.0 yield
strain .
3. The lack of yield stress coupon data for kc7 causes some uncertainty. The lowest
value of correlation factor was calculated for test kc7 for which even the yield stress
was not available, so the yield was deduced from the load and deflection output from
the loading range of the test and the maximum load.
......
.... Correlation factors from tests
Q)
Q)
.Q
--
C/)
~
00
.,f
1.2
...... .....-
~ r--
...... ...0 r--
S ...v O.S - I--

_ r-- r-- _
-
0\ of
0
c:
0
~,
0 O.S -
;::
- - - - -
10
~
~
t.:=:
t: 0.4 I--
0
- - - - -
u
..... ,
0 0.2 t-- - - - f---- -
.....
(.)
c.S 0
0'
.....
0 ke3 kc4 ke5 keS kc7 keS keg ke10
.....
~ Test
Q)
t::
0
(.)
Figure 6.26 Correlation factors for rotation
125
6.4.3 Effect on the correlation factor of out-of-plane deflections
The tests were designed to study the columns deflecting in only one plane. There were also
some modest out-of-plane deflections. These were not expected to make a significant
difference to the results because the simple plastic model gives no difference in in-plane
bending resistance for small out-of-plane bending moments. However, to quantify the effect
of these on the correlation factor, it was investigated by analysis of the test case with the
greatest out-of-plane deflection. This test was kc9 which had out of plane deflections of the
order of 4mm. The out-of-plane deflection was induced in the Abaqus analysis by imposing
equal and opposite end rotations. This might not induce the same deflected shape as in the
test, but the magnitude of deflection at mid-height was similar. The sensitivity to out-of-
plane deflection was then assessed by comparing the correlation factor calculated without
out-of-plane deflection with the correlation factor calculated with out-of-plane deflection.
The out-of-plane displacements from the test and from the analysis are plotted in Figure
6.27. The mean in-plane end-rotations are plotted in Figure 6.28, showing that the effect of
the out-of-plane displacement is very small.
The values are as follows:
Abaqus zero out-of-plane displacement gives Cf= 0.941
Abaqus with out of-plane-displacements gives Cf= 0.937
This shows that the change of correlation factor, Cf, from zero out-of-plane displacement to a
4mm out-of-plane displacement is (1- 0.937)x 100% = 0.4%. Therefore it is concluded

0.947
that the out of plane displacements have no significant effect.
126
out-ot-plane mid-height displacement
v mean end-rotation
test 9 v Abaqus
E 5 -r-r-
-C
E
! '
E
QI
QI
\,l
4 ~~N:b.U---l---: ~Abaqus
III
1i.
III
I
=a 3
~ I I I
III
1i.
'5
i
I 2 -+-r-- T- f -
1
I
I
I
I
I
I
-test
! I Iii
.~
~
1 -~ t-+---+--+----t-
I i I
-a I
I
I
I
1
~
I
Ii !
~ ~~
I
L ___ _ 1___ _
I
o ____ ---.l _____ __ ____
Mean end·rotation (mrads)
Figure 6.27 Effect of out-of-plane displacements - kc9, test & Abaqus
load vend rotation

test 9 v Abaqus
1000 I 1 I I I
~Abaqus
Ii! I I I I ! with out-
-z
~
800
j-- ·-t+I··rJ-t-
I I Iii
I I i I
of-plane
c
0
Vi
III
I!!
600 --r--j---f---,-- '_L_+_ i ~Abaqus
zero out-
Q. " I of-plane
E I I I
0
\,l
400
ii
I i I I I i I
I' I I I , •
'I
-test
~ I I I I Iii i
-t--
!
200
-t,----j+, --r-1-t--i--~1
'1 I
I
1__ ___L __L_...L.J_~_L~i_J
I I Ii I
i,' --
0
o 10 20 30 40 50 60 70 80 90 100
127
6.4.4 Effect on the correlation factor of a partly curved initial
imperfection
The effect on the correlation factor of an initial imperfection that is partly curved was studied
using test kc7. The test was to investigate if the value of the correlation factor is
significantly affected by the choice either of the imperfection being a uniform eccentricity or
a varying eccentricity as in a curved member. Test kc7 was selected because
1. it has the lowest correlation factor
2. the eccentricity is greater than the target eccentricity of 3.5mm
The end eccentricity was taken as equal top and bottom. The magnitude was taken as the
target eccentricity less the eccentricity equivalent to the friction in the test rig bearings. The
eccentricity equivalent to the friction was found from the plot of mid-height displacement v
load for unloading/reloading cycle shown in Figure 6.30. At 800 kN, the displacement is
24.8 mm on the unloading path and 23.6 mm on the reloading path giving a range of 1.2 mm.
The friction is equivalent to half the range so the magnitude is 0.6 mm. Therefore the end
eccentricity was taken as:
target eccentricity - friction eccentricity = 3.5 - 0.6 = 2.9 mm Eq 6.16
The remainder of the imperfection was modelled as a sine curve as shown in Figure 6.29.
This component of the imperfection needs to be 3.01 mm to give the same value of mean
end-rotation at 80% maximum load as measured in the the test kc7. This makes a maximum
initial imperfection of 2.9 + 3.01 = 5.91 mm compared with 4.90 mm for the case of a
uniform imperfection (equal to a uniform initial eccentricity of a perfectly straight column).
Given the partly sinusoidal imperfection of 5.91 mm, the yield stress needs to be 411 MPa to
achieve the maximum load measured in test kc7. The plots of test kc7 and the Abaqus
analyses are shown in Figure 6.31 and Figure 6.32.
The correlation factors calculated for the two cases are
Uniform imperfection = 4.90 mm, Cf= 0.804,
(from Sheet "cf(2)" of kc7_AbaqusJactor_201 Om03d05.xls)
Part curved imperfection = 5.91 mm, Cf= 0.781,
(from Sheet "cf(3)" of kc7_AbaqusJactor_20lOm03d05.xls)
This shows the sensitivity is 0.804 - 0.781 = 0.023 from 3.01 mm Sine curve
128
The measurement of the test specimen prior to testing found that the out-of-straight over the
central 1.5 metres (the length of the longest available straight-edge) was only 0.1 mm to
0.2 mm in both the XX and YY planes. If the sinusoidal component of imperfection had
been 3.01 mm as in the analysis, the out-of-straightness would have been
3.01(I-Sin(1t(0.575/2.65» = 1.11 mm instead of the 0.1 to 0.2 mm measured.
Therefore, the reduction in cffrom the out-of-straightness of, say 0.15 mm,
is only 0.023 (0.15/3.01) = 0.0011, giving Cf = 0.804 - 0.0011 = 0.803. This is an
insignificant difference from Cf = 0.804. This leads to the conclusion that the value of the
correlation factor is not significantly affected by the choice either of the imperfection being a
uniform eccentricity or a varying eccentricity as in a curved member.
4.90. 1
t 5.9_1: 1
I
I
I
I
I
I
I
I :
2.90~3.01
Uniform eccentricity Varying eccentricity
Figure 6.29 possible initial imperfections - kc7
129
test 7 v Abaqus 2650mm
load v mid-height diplacement
1000
900
Z
~
·K1 800 -kc7 disp in-plane

mean
e
:i
~
700
600
22 23 24 25 26 27
Mid·height displacement (mm)
Figure 6.30 Unloading/reloading cycle - kc7, test
load vend rotation

test 7 v Abaqus
1600
1400 -test
Z 1200
~
c 1000
0
'iii
11\
QI ~Abaqus straight
a
E
800
0
1.1 600 ___ + ______ ------..1--------- _ L ____ --- ____ _
ii ! 1
~ 400 --+- -- - - - - - - - - - -• .1------------- r----- ---- ----I

I . i ~Abaqus part curved
200 - ----- ---- ------------~----------------- --- -----1
. I I
________ ~ __l _______ '-________.-J
0
0 5 10
130
load vend rotation
test 7 v Abaqus
1600
1400
-test
Z 1200
;.
c 1000
.2
..
III
III
QI
800 I -&Abaqus straight
Co
E
0
IJ 600 r
iii I
.. ~ ___.+___.L_.
~ 400 i
I ~Abaqus part curved

-l--- __1
200 .. t-
I
0
o 10 20 30 40 50 60 70 80 90 100
6.4.5 Effect on the correlation factor of residual stresses
The effect on the correlation factor of residual stresses was studied using test kc7 using the
partly curved initial imperfection as shown in Figure 6.33. Test kc7 was selected because it
has the lowest correlation factor. The partially curved initial imperfection was chosen
because the effect of imperfections is least where the length of the yielded zone is least and
this occurs where the initial imperfection is curved. The initial imperfection was the same as
use in Section 6.4.4 so that the loading curve was as recorded in the test.
To achieve the same maximum load as in test kc7, the yield stress had to be reduced to
400MPa. The correlation factors for the two cases are:
With residual stress 10% nom yield, Cf= 0.781,

(from Sheet "cf(3)" ofkc7_AbaqusJactor_2010m03d05.xls)
Zero residual stresses, Cf= 0.837,

(from Sheet "cf(4)" ofkc7_AbaqusJactor_2009m03d05.xls)
This shows the sensitivity of the correlation factor is 0.837 - 0.781 = 0.056 from residual
stress of 10% nominal.
131
The actual residual stresses in the specimen are not known. However, assuming the actual
residual stress was between 50% and 100% of the 10% nominal yield assumed in the
parametric studies, the correlation factor is higher than shown in Table 6.3 by somewhere by
up to 50% of 0.056 = 0.023. This shows that the correlation factors deduced in the table are
slightly conservative, but it was decided to use the values in the table because conservatism
helps to give confidence in new structural models.
5.91
Figure 6.33 Initial imperfections - effect of residual stresses kc7
6.5 Effects of breadth to thickness ratios of wall

The effect of the breadth to thickness ratios of was originally investigated by the theory and
formlae in Appendix E. These predicted that a 140x140 SHS with a 6.8mm wall thickness
would be stable with equal and opposite end-rotations (single curvature) of 40 milliradians.
The range of wall thicknesses of the test columns was chosen to study the effect of the breath
to thickness ratio. Scaling from the predictions for 140x140 SHS to a 120x120 SHS, the
wall thickness for stability at 40 milliradians would be 6.8x120/140 = 5.8 mm. Therefore
the walls of the 120x 120x6.3 SHS in tests kc5 and kc9 were expected to be stable at end-
132
rotations greater than 40 milliradians but the walls of the 120x120x5.0 SHS in tests kc6 and
kc 10 were expected to be unstable at 40 milliradians.
The load vend-rotation curves together with the Abaqus analysis results for tests kc5 and
kc9 are shown in Figure 6.35 and Figure 6.36. The tests kc5 show a very slight and gentle
reduction of load from about 70 milliradians. Examination of the deformed plots of the
Abaqus models showed that Abaqus predicts that local buckling develops from
78 milliradians in test kc5 and from 75 milliradians in test kc9.
The load vend-rotation curves together with the Abaqus analysis results for tests kc6 and
ketO are shown in Figure 6.37 and Figure 6.38. The tests show a major reduction of load
from about 55 and 40 milliradians. The Abaqus plots predict that the reduction should not
appear until near 55 milliradians, showing that the actual stability might be less than
predicted by the Abaqus analysis but greater than that predicted by the finite slice model and
simplified application of plastic flow theory in Appendix E. The rotation capacity from the
tests is shown as a bar chart in Figure 6.34. The minimum rotation achieved without
reduction of resistance due to local buckling is shown for the three wall slendernesses tested.
The wall slenderness is calculated as the distance between mid-planes of the walls divided
by the wall thickness, so for a 120x120xl0 SHS, bit = (120-10)/10 = 11. It should be noted
that for the 10mm wall columns (wall slenderness bit = 11), there was no drop of resistance
due to local buckling even at the maximum test rotation.
25 .-
,
l··-~--T·--·-
i '
~5 mm wall test
T i
I
--
20 l:.-- ------ ---
i ~6.3 mm wall test
---r -- --- t--- - ---- ------

I
I I
_ 1 0 mm wall test
i I
~ 10 ~- ---- l--+---- i
, I I I ~AppendixE
!i I
j
i
! I
- 'ii
5 -t------ ----- -t---+-+--- - +---t-----j
o 10 20 30 40 50 60 70 80 90 100
Figure 6.34 Effect oflocal buckling - rotation capacity from tests
133
load v mid-height displacement
test 5 v Abaqus
1000
-z
~
800
c 600 -test
0
"iii
III
f
a.
E 400
0
u
iii ~Abaqus
~ 200
o 10 20 30 40 SO 60 70 80 90 100
Figure 6.35 Effect oflocal buckling - kc5, test & Abaqus
load v mid-height displacement

test 9 v Abaqus
1000
_ 800
z
-
~
c
"~ 600
-test
~
~
u
400
iii ~Abaqus
~ 200
a
a 10 20 30 40 SO 60 70 80 90 100
Figure 6.36 Effect of local buckling - kc9, test & Abaqus
134
load vend rotation
test 6 v Abaqus
800
I
i
.... 600 i
z
---t-- ~"""-d--
---T-I
-
.lI:
c
o
I
I I
I
I I ~Abaqus
'iii
..
::l 400
Q.
- r-----l- -----~-- -J
I
E
o
u I -i I
ii I -test
~ 200
o --
o 10 20 30 40 50 60 70 80 90
Figure 6,37 Effect of local buckling - kc6, test mean & Abaqus
load vend rotation

test 10 v Abaqus
---"--l----i ----I-------
t
800 --- ------r -- --1--- 1- 1
I I ! I I I
I Ii! , i
Z 600 - IJLf--L.;
-l.L+-i~l-l- -.
~
c
o
'iii -test
'"~ 400
E
o
u
ii
II . i I I I ~Abaqus
~ 200
--rt-l-ttT
II I
i
I I I I
I
I
I
o
Figure 6.38 Effect oflocal buckling - ketO, test mean & Abaqus
135
6.6 Correlation factor, Cf , used for the design model
The values of the correlation factor, Cr, from Table 6.2 and Table 6.3 are shown in Table 6.4.
Table 6.4 Abaqus with symmetric eccentricities
Comparison of tests and symmetrical model.

Abaqus analysis using the standard elastic/perfectly-plastic stress/strain curve.
Test kc3 kc4 kc5 kc6 kc7 kcS kc9 kc10
Wall thickness (mm) 10 10 6.3 5 10 10 6.3 5
Asymmetric eccentricities
Abaqus et (mm) 2.25 3.20 note 1 note 1 10.39 7.92 note 1 note 1
Abaqus e., (mm) 2.87 -1.34 -0.59 -1.99
Abaqus et - e., (mm) 0.62 4.54 10.98 9.91
Correlation factor 0.991 1.018 0.749 0.829
(rotns) cf= 9t/9A

Symmetric eccentricities
Abaqus et (mm) 2.68 0.73 5.27 4.3 4.9 3.08 3.l5 2.55
Abaqus e., (mm) 2.68 0.73 5.27 4.3 4.9 3.08 3.15 2.55
Correlation factor 1.003 1.021 0.831 note2 0.804 0.841 0.941 0.918
(rotns) cf= 9t /9A
note I Not analysed with asymmetric eccentricities

note2 No unloading cycle in tests so no defined point at which to
calculate correlation
The misalignment between the lines of application of load, (et - ~), in the asymmetric
analysis ofkc7 and kcS (in Table 6.2) are not credible given the quality of the test laboratory
and the care taken to align the column correctly. Therefore it must be concluded that the
majority of the difference of end rotation, at least in the loading part of the test, was caused
by sway of the test rig. In the critical case of test kc7, II mm eccentricity reduced the
correlation factor Cr from O.S04 to 0.741, a reduction of 0.063. Given the special adjustable
shoes used, the maximum credible misalignment is around I mm.
136
Therefore the worst credible correlation factor is 0.804 - 0.063/11 = 0.798. This gives the
range of credible values of correlation factor from 1.042 for test kc3 (symmetric) to 0.798 for
kc7 (worst credible asymmetric). These were all deduced with a residual stress of 10% of
nominal yield stress, which is a likely maximum value, so it is improbable that this was the
value in the test columns. If the test columns had lower residual stresses, the actual yield
stress would be lower than applied in Abaqus to achieve the predicted maximum load. From
Section 6.4.5 it can be seen that if the residual stress were 8% nominal yield instead of 10%,
the correlation factor would be raised by 0.2xO.056 = 0.011, giving a likely minimum value
of correlation factor of 0.798 + 0.011 = 0.809.
Therefore, the value of correlation factor selected for the calibration of the design model was
taken as 0.80, giving a small margin below the lowest value and a significant margin below
the median of the range from 1.042 and 0.809 = 0.926.
6.7 Summary
The Abaqus model described in Chapter 4 was used to simulate the full-scale laboratory
tests. A correlation factor, Cf, was obtained using an elastic/perfectly-plastic bi-linear model
for the stress-strain characteristics. For each test, the model was correlated to the test both
by using the "yield" test coupon 0.2% proof stress as the yield stress of the model (wherever
it was available and the Abaqus model could complete the analysis with it) and by adjusting
the initial imperfection in the model to reproduce the test behaviour of the test from zero to
near maximum load (the "elastic" range of the test).
The load vend-rotation curves from the models were close to the load v mean-end-rotation
curves of the tests. The difference between the end-rotations from the tests and from the
analyses was established from the "static" point on the unloading cycle and a correlation
factor, Cr, was calculated as rotation-from-testlrotation-from-model at the load of the "static"
point in each of the tests. The mean value of the correlation factors was 0.914 with the
minimum of 0.804.
While 0.8 might be considered disappointing as a correlation of end-rotations from a finite

element analysis of a structure in which considerable elastic zones remain, it is considered
acceptable for the falling branch of a compression member which has been almost
completely plastified. The correlation between test and analysis looks much better when
examined from the more common point of view of loads. The lowest correlation factor for
137
end rotation was found from test kc7 which gave an axial load of 927 kN at the "static" point
compared with 1011 kN predicted by Abaqus at the same rotation. This gives a ratio of 0.92,
which is good for a falling branch in a test which plastified almost the entire cross-section of
the column over a considerable length.
The lowest credible value of end-rotation correlation factor, Cf = 0.8, was selected for
calibration of the design model.
138
7 BREADTH TO THICKNESS LIMITS
7.1 Introduction
7.1.1 Overview
The structural mechanics of the effects of curvature are shown in Figure 7.1, Figure 7.2 and
Figure 7.3. Figure 7.1 shows a section of an SHS that is curved. On the compression face, a
longitudinal slice of the wall is marked. On the right of Figure 7.1, this longitudinal slice is
shown in elevation. The longitudinal force on this section causes a radial component as
indicated by the radial arrows. The radial component is most easily considered as a radial
pressure, PRC = to'clR, where O'c is the longitudinal compressive stress in the wall of the SHS
at that longitudinal slice, t is the wall thickness and R is the radius of curvature of the
deflected member at the section under consideration.
t x a c/unit width
t x aclunit width
Figure 7.1 Out-of-plane loads on curved walls
The radial pressure, PRC, can only be resisted by transverse bending of the wall of the SHS,
as shown in Figure 7.2 and Figure 7.3. Figure 7.2 shows the radial pressures and the
transverse bending moment diagram from longitudinal compression in both curved walls.
Figure 7.3 shows the radial pressures and the transverse bending moment diagram from
longitudinal compression in the wall on the inside of the curve and longitudinal tension in
the wall on the outside of the curve. A special limitation on the breadth to thickness ratios is
needed, similar to the need for a special limitation for seismic design in cases with high
ductility demand.
139
~Co
PRCi
Loading Bending moment Deflected form
Figure 7.2 Transverse bending of walls, both sides in compression
P rT
PRC
Loading Bending moment Deflected form
Figure 7.3 Transverse bending of walls, one side compression, one in tension
7.1.2 Non-linear geometrical effects (second-order effects)

The deflected forms shown in Figure 7.2 and Figure 7.3 cannot be calculated accurately by
first-order analysis of the transverse bending, because the deflections induced by the
transverse pressure cause an increase in local curvature along the centreline of the wall, thus
causing an increase in transverse pressure. This is shown in Figure 7.4. If the wall remains
in the elastic range, this magnification can be calculated approximately using classic
magnification approaches.
140
t x crclunit width
Increased
curvature
Increased curvature
t x crclunit width
Figure 7.4 Increased curvature of wall towards the middle of the wall
7.1.3 Yield and non-linear material effects

In all common structural steel columns at maximum load, the most compressed elements are
at yield stress. In columns of modest slenderness with significant end rotations, the strains
are in excess of yield strain. For example, for a 140 x 140 x 10 SHS 3.0 metres long in
single curvature with a 0.040 radian end rotation, the most severe curvature is 2 x 10-4 mm- I
and the maximum compressive strain is approximately (5000/355) x yield strain = 14 x yield
strain.
Where a material exceeds yield, there will be plastic flow. The extent of the plastic flow will
depend on the strain beyond the yield strain. The effect of plastic flow is to increase the
local deflections of the walls. beyond the elastic deflections. This will increase local
curvatures caused by transverse bending of the walls and might precipitate failure of the
column. Plastic flow is explained with elegant simplicity by Calladine [Calladine 1969].
Detailed calculations are given in Appendix E, showing that the breadth-to-thickness ratios
need to be limited.
The issue was studied in the laboratory test program reported in Chapters 5 and 6 by testing a
variety of wall thicknesses for the same nominal size of square hollow section and also by a
parametric study by Abaqus reported in Section 7.2.
141
7.1.4 Breadth to thickness limits in codes
BS 5950-1 :2000 and Eurocode 3 use similar classifications, from Class 1 to Class 4, where
Class 4 have the most slender component parts. Class 4 sections have parts that are so
slender that these sections would not be suitable for the structures considered in this project
firstly because these structures need the smallest possible sections (to allow the sections to
be hidden in the wall thickness) and secondly because the design method requires plasticity
to allow "moment shedding". Therefore, Class 4 sections will not be considered further.
The classification systems in BS 5950-1 :2000 and Eurocode 3 are as follows:
• a member resisting compression alone needs to be Class 3,
• a member that is designed to develop the plastic moment of resistance must be Class 2
• a member that is designed to develop a plastic hinge and rotate must be Class 1.
Because the classification rules are intended for application in relatively simple design
models, there is some ambiguity about the range of application of Class 2 and Class 1. This
arises because most global analysis routines in design offices do not consider local plastic
strains, so do not calculate rigorously the curvature at points at which the plastic moment of
resistance has been attained. Instead, the curvature for Class 2 is limited to use with elastic
global analysis with a redistribution allowance of 10% - 15% and the curvature of Class 1
sections (or the rotation of the plastic hinge) is generally not limited in codes.
Because codes do not give explicit curvature limits appropriate to each Class of element
slenderness, it is not possible to decide whether the columns may be Class 2 or need to be
Class 1 if designed with the methods developed in these studies. There is even the
possibility that the breadth to thickness ratios need to be more severe than Class 1. There is
concern that common Class 1 limitations are not adequate for earthquake design due to the
severe curvatures developed. The issue is recognised in design guidance for earth-quake
resistant design, such as Dowrick, [Dowrick 1977]. The AISC Seismic Provisions for
Structural Steel Buildings [AISC 2005a] contains Table 1-8-1 which gives more severe
breadth to thickness ratios for certain members than required for normal design.
A similar condition to the case of single curvature arises in earthquake resistant design in
certain types of structures. Three of these types of structure are included in the AISC
Seismic Provisions for Structural Steel Buildings [AISC 2005a]. These are Special Truss
Moment Frames, Special Concentrically Braced Frames and Ordinary Concentrically Braced
Frames. In these frames, compression members with high demand for ductility have to
satisfy the Limiting Width-Thickness Ratios for Compression Elements given in Table 1-8-1.
142
For rectangular hollow sections, this gives the bIt limit of 0.64"(ElFy). Structures in the
USA without seismic loading would be designed to the AISC Specification [AISC 2005]
which has a slenderness limit of 1.12"(ElFy). Therefore, for S355 steel, bIt ~
0.64"(205000/355) = 15.4 for the flat segment of the wall for these types of seismic resistant
structures compared with 1.12"(2050001355) = 29.9 in the AISC Specification [AISC 2005]
and 28£ = 28"(275/355) = 24.6 required for Class 1 in BS 5950-1 :2000.
Taking B = b + 3t, as given in BS 5950-1:2000 Table 12, and using bit = 15.4 gives the
overall breadth B = 15.4t +3t = 18.41. Therefore the minimum thickness for a 140x140 SHS
is 140118.4 = 7.6 mm to satisfy the AISC Seismic Provisions. By comparison, to satisfy
BS 5950-1:2000 Table 12, Class 1 allows B = b + 3t = (24.6 + 3)t = 27.61. Therefore the
minimum thickness for a 140x140 SHS is 140/27.6 = 5.1 mm, demonstrating that more
severe bit ratios may be required than found in BS 5950-1 and EN 1993-1-1. The initial
studies of this issue made as part of this project, reported in Appendix E, confirmed that
much more onerous limits are required than appear in the BS and the EN.
7.2 Parametric study

The reduction of resistance of high breadth to thickness ratios is most pronounced when the
curvature is high. The design model assumes that the column is in single curvature for
calculation of overall buckling resistance, but many columns will be in double curvature as
Column in double curvature Beams
Figure 7.5 Column in double curvature
143
This effect was investigated by a parametric study using 140x 140 SHS sections of different
thicknesses. The analyses were made using enforced shortening of pin-ended columns with
the initial imperfection in the plane of the end-rotations. The effects with end-rotations in a
rectangular plane were investigated for columns of in single curvature of lengths 750mm and
1500mm to reproduce the behaviour of columns in double curvature of 1500mm and
3000mm length. The results are shown in Figure 7.6. From these it can be seen that the
wall stability is more demanding for 750mm than for 1500mm, but not enormously so.
SHS 140 single curvature in rectangular plane

1.0
··~r~-F 1
-]I( - SHS140 x10 x1500long
0.9
'0 O.S - ~l------ -i - M- SHS140 x6.3 x15001ong
~ I
I
E 0.7 - + - SHS140 x5 x1500long
:s
E 0.6
••E"' 0.5 - 1
_SHS140 x12 x7501ong
....
"'c
.2
0.4 -+-SHS140 x10 x7501ong
1::
0.3
8. -'-SHS140 xSx7501ong
...0
a. 0.2 -i -
0.1 I I ~SHS140x6.3x750Iong
0.0 -~- ]~-] _SHS140 x5 x7501ong

0 20 40 60 so 100 120 140
Rotation (milliradians)
Figure 7.6 140x 140 SHS in single curvature in rectangular plane
Similar analyses were conducted for end-rotation in a plane at 45 0 to the rectangular planes
as shown in Figure 7.7. From these it can be seen that the wall stability requirements for
end-rotations at 45 0 are almost identical to those for end-rotations in a rectangular plane.
144
SHS140x750mm single curvature in 45 0 plane
1.0
0.9 __ SHS140x127501ong
'tl
III 0.8
.2
E 0.7 ...... SHS140xlO 750long
:I
E 0.6 -
'xIII
...
E
O.S
0.4 --
-i------i-
f
, -
....-SHS140x87501ong
I .
0 )--__L ___
C 0.3
i .~ ..... SHS140x6.3750Iong
0
:eo 0.2
Q.
...0
0.1 -
.t- . j- " , ~SHS140x5 750long
Q. 0.0
0 20 40 60 80 100 120
End rotation (milliradians)
Figure 7.7 140x 140 SHS in single curvature at 45 0 to rectangular plane
The breadth to thickness ratio at which the column resistance falls below the stable wall
behaviour is shown in Figure 7.8.
design limits 750mm single curvature

Qj
III
30 I - -r-:
'8 25
E
III 20 Ir" t
:I i
17
.aCI: 15 r
I
....
.5 10 I
-1 ~Abaqusb/t
0
' i
l! 5
:a- 0 - -_.-
0 10 20 30 40 50 60 70 80 90
Figure 7.8 Limiting wall breadth to thickness ratio from Abaqus
145
7.3 Calibration of Abaqus with laboratory test results
In Chapter 6, the comparison of the laboratory tests with Abaqus analysis shows that Abaqus
may overestimate the end-rotations at which the walls remain stable. The ratio of worst test
result to Abaqus analysis result is 40mrads/55mrads for 120x 120x5 SHS and
70mrads178mrads for 120x120x6.3 SHS. The points are plotted in Figure 7.9 and the
equation of the straight line through these points is used as a reduction factor on the results
of the Abaqus parametric study to allow for the differences between real SHS sections and
the Abaqus finite element models. The reduction factor is 1.5175 - 0.03436(b/t), but it
cannot be greater than 1.0.
reduction factor: testlAbaqus

1.2
,...
~
~ 0.8
J!
1.0
-"
Y= -O.0343E x + 1.517 ~,
~ Test/Abaqus
~ 0.6 " ...

~ '(D
~ 0.4
CII
~
- e- reduction factor
0.2
0.0
o 10 20 30
bIt ratio
Figure 7.9 Wall slenderness limitations, test v Abaqus
7.4 Design limits

The design limits are defined using the results of the Abaqus parametric study shown in
Figure 7.8 together with the reduction factor of testiAbaqus shown in Figure 7.9. Another
factor in defining design limits is the relationship between calculated values of column end-
rotations and the values that may occur in real structures. In design offices, the deflections
of beams are commonly calculated by elastic analysis. Where a plastic hinge occurs in a
beam, the hinge would be assumed to occur at a point and the deflections would be
calculated assuming elastic behaviour between the points of the hinges. The effects of
146
spread of plasticity along the beam would not normally be considered. This produces some
slight underestimate of deflections. Therefore it is helpful for designers to have design
values of wall slenderness versus end-rotation that can be used with the end rotations
calculated by the common design office techniques.
While bit is a useful ratio for theoretical studies of wall slenderness, it is not so helpful for
designers because section sizes are given in terms of the overall width, B, not the distance
between mid-planes of walls, b. Therefore, the design limits in this report are given in terms
of BIt.
To ensure that the columns can sustain the end-rotation, the slenderness should not exceed
the limit defined by the "B\t limit" line shown in Figure 7.10. This is derived from the
parametric study in Section 7.2 with the allowable rotations reduced
I. by the reduction factor derived from the ratio of limiting end-rotation in the tests to
the limiting end-rotation from the Abaqus analysis of tests from Chapter 6, shown in
Figure 7.9 and
2. by a factor of 0.8 to allow for both the uncertainties of manufacture (eg wall
thickness) and of analysis.
This curve has been derived for columns of lengthlbreadth = 10.7 in double curvature.
Columns that have smaller ratio of lengthlbreadth can be expected to need thicker walls, but
it is considered unlikely that such stocky columns would ever be desired in structures
designed using this design model. The curve is
Bit = -14Log lO(9/1000) +38.5 Eq 7.1
where 9 is in radians (not milliradians).
147
design limits: LIB ~ 10.7
30
25
I
I
20 -+I
.., - + - BIt limit
'ai' 15
10 ~B/t = -14*LoglO(rotn) + 3S.5

5
o
o 10 20 30 40 50 60 70 SO 90
End rotation (milliradians)
Figure 7.10 Wall slenderness limitations, overall breadth/thickness, Bit
These requirements are more demanding than common code limits as demonstrated for
BS 5950-1 and Eurocode 3 below.
Requirements of 8S5950-1
For a hot-finished square hollow section designed to BS 5950-1 [BSI 2000a], the slenderness
limits for sections is expressed as bit = 28£ for Class land bIt = 32£ for Class 2, where b is
the width of the external flat face and £ = (275/yield)0.5 = (275/355)°·5 = 0.880.
Therefore, for the flat face, bit = 28xO.880 = 24.6 for Class land bit = 32xO.880 = 28.2 for
Class 2.
According to Annex A3 of EN 10210-2 [BSI 2006b, CEN 2006b], the external corner radius
may be taken as 1.5 times the wall thickness. Writing the slenderness limits in terms of the
overall width, B, of the section, these limits are:
BIt = 1.5 + 24.6 +1.5 = 27.6 for Class 1 and BIt = 1.5 + 28.2 + 1.5 = 31.2 for Class 2.
Requirements of Eurocode 3
For a hot-finished square hollow section designed to Eurocode 3 [BSI 2005a, CEN 2005a],
the slenderness limits for sections is expressed as cIt = 33£ for Class 1 and cit = 38£ for
Class 2, where c is the width of the internal flat face and £ = (235/yield)0.5 = (235/355)°.5 =
0.814.
Therefore, for the flat face, cIt = 33xO.814 = 26.9 for Class 1 and cIt = 38xO.814 = 30.9 for
Class 2.
148
According to AnnexA3 of EN 10210-2 [CEN 2006b, BSI 2006b], the internal comer radius
may be taken as 1.0 times the wall thickness. Writing the slenderness limits in terms of the
overall width, B, of the section, these limits are:
Bit = 1 + 1 + 26.8 + 1 + 1 = 30.8 for Class 1 and Bit = 1 + 1 + 30.9 + 1 + 1 = 34.9 for
Class 2.
149
8 VALIDATION OF THE DESIGN MODEL BY A
PARAMETRIC STUDY AND CALIBRATION
8.1 Introduction
The new design model is described in Section 3.4. A parametric study was conducted to
show the reliability of the model as a design tool over a wide range of slenderness and end-
rotations.
The study needed to investigate:
1. end-rotations in different planes
2. a range of slendernesses representing practical construction
The study did not need to investigate:

1. different imperfections and residual stresses because the design model is calibrated
to the specified design code, whichever that might be,
2. difference in end-rotations at the two ends of the column because the design model
uses the worst rotation at either end and assumes single curvature, which is the worst
case,
3. different wall thicknesses because the test results in Chapter 5 from columns with
different wall thickness and the analysis of the results in Chapter 6 shows that the
column behaviour is essentially the same for all thicknesses up to the point of local
wall buckling. This is confmned by the parametric study of breadth to thickness
limits in Chapter 7,
4. the design limits the wall breadth to thickness ratio to avoid local buckling because
these were investigated by the study reported in Section 7.
The study was conducted using Abaqus finite element software as described in Section 4.
The section chosen was 140x140xlO SHS in S355 steel. This section was chosen for the
following reasons.
1. The size 140x 140 was reported to be one of the largest used in the multi-storey
buildings with discontinuous columns described in Section 1. The extent of
plasticity caused by end-rotations is more pronounced in larger sections because the
end-rotations in the elastic range are less.
2. 10mm wall thickness in a 140x 140 SHS is thick enough not to suffer significant
deformations even at high plastic rotations.
150
S355 steel was chosen because this is the commonest grade of steel used for structural
hollow sections in Europe.
The 140xl40 x lO SHS was analysed for lengths of 3.0m, a representative length for typical
domestic construction, and also for lengths of 1.5m and 6.0m to give a wide range of
slenderness.
End rotations were applied as equal and opposite because that is both the worst design case
and the design case assumed in the design model.
8.2 Load-rotation path sensitivity

Because the columns become partially plastic, the resistance is expected to depend on the
sequence of load application. The Eurocode proposes that incremental loading should be
applied in the proportions of the ULS loads. For example, if a column is subject to
coexistent loading at ULS of IOOOkN axial compression and 50kN-m bending moment, the
load might be applied in 10 increments of IOOkN combined with 5kN-m. In Figure 8.1, this
is shown by the straight line OA.
There is an infinite range of alternative load increments. One possible path would be to
apply increments of axial load alone up to the maximum applied axial compression, IOOOkN
in the example, and then apply increments of bending moment until the maximum applied
moment, 50kN-m in the example, has also been applied to the column. This is shown in
Figure 8.1 by the lines OBA. Another possible path would be to apply increments of
bending moment alone up to the maximum applied moment, 50kN-m in the example, and
then apply increments of axial load until the maximum applied axial compression, IOOOkN
in the example, has also been applied to the column. This is shown in Figure 8.1 by the line
OCA.
151
N(kN)
B A
1000
c
o ~--------......- - M (kN-m)
50
Figure 8.1 Load paths for a column resisting axial and bending
Loading a column subject to axial compression and end-rotation presents the same
possibilities. However, the imposition oflarge end-rotations causes more extensive plasticity
than traditional column design for axial and moments, so the effect of the load-path might be
more pronounced. This could result in a different resistance to axial load depending on the
load path. Possible load paths for a column subject to 1000kN axial compression and 50
milliradians end-rotation are shown in Figure 8.2.
N(kN)
B A
1000 ~--------------------.-~ " I
o ~ _________ ~L..-
c __ 9 (milliradians)
50
Figure 8.2 Load paths for a column subject to axial and end-rotation
152
For the initial imperfection in a rectangular plane of the section and imposed rotations in the
plane of the initial imperfection, the load-rotation path used in the analysis is the case of a
column unloading as the distance between the ends is reduced by an imposed shortening.
The path is shown in Figure 8.3 as ODA. This is a slightly conservative case because the
extent of plasticity is expected to be greater at the load at D than at B, which marks the
maximum axial load in Figure 8.2
N(kN)
D
1000 i--....,.-----.....,j...... 4
. . ._ _ _ _ _ _ _ _ _~L..::..-- C e (milliradians)
o 50
Figure 8.3 Load paths for imposed rotation in the plane of the initial imperfection
The effects of the load paths OBA and ODA shown in Figure 8.3 were investigated for of the
initial imperfection in a rectangular plane and end-rotations in the same plane. The case of
140x 140x 10 SHS columns of 3.0m length was calculated at one of the points generated in
the parametric study. The point chosen was at approximately 30 milliradians firstly because
this is the range where the critical case is end-rotation in the plane of the initial imperfection
and secondly because this is expected to be a relatively common magnitude of end-rotation
for columns at the ends of beams. Analysis following the path OBA (applying axial load
alone up to the maximum and then imposing rotation) gives the end-rotation as 31.60
milliradians at the point at which the end-moment changes from destabilising to stabilising.
Analysis following the path ODA (imposed end shortening) gives the end-rotation of 30.15
milliradians. These results are shown in Table 8.1.
153
Table 8.1 End-rotn Axial compn
(radians) (kN)
Load path ODA 0.03105 1158.0
SHS2006ml0dl0_130bylO_ v2
Load path OBA 0.03160 1158.0
SHS2008m09d05 _out 140by 10_Nthenrotn_ vi 0
Interpolation to zero moment
The effects of the load paths OA, OBA and OCA were investigated for the initial
imperfection in a rectangular plane of the section, but with end-rotations orthogonal to the
plane of the initial imperfection as follows. The case of 140x 140x 10 SHS columns of 3.0m
length was calculated at two of the points generated in the parametric study. The points
chosen were at 8.6 and 13.1 milliradians firstly because this is in the range where the critical
case is end rotation orthogonal to the initial imperfection and secondly because this is
expected to be a common magnitude of end-rotation at interior columns. The results are
shown in Table 8.2. It can be seen that there is almost no difference, showing that the
sensitivity to the load path is very small.
Table 8.2 end rot compn

(radians) (kN)
Point at 13.13 milliradians
Proportional loading, path OA 0.01313 1404
SHS2006ml0dI3_130byl0_v3
Axial alone then end-rotation alone, path OBA 0.013068 1404
SHS2008m09d05_out 140by 10_Nthenrotn_ oplane_ v3
Rotation alone then axial alone, path OCA 0.01313 1404.8
SHS2008m09d06_ out140by 10J0tnthenN_ oplane_v2
Point at 8.6 miUiradians
Proportional loading, path OA 0.0086058 1483.4
SHS2006m 1Od 13_130by 10_v5
Axial alone then end-rotation alone, path OBA 0.0085815 1483.4
SHS2008m09d08_out 140by 10_Nthenrotn_ oplane_vi
Rotation alone then axial alone, path OCA 0.0086058 1483.5
SHS2008m09d08_out 140by 10JotnthenN_ oplane_vi
154
For the initial imperfection not in the rectangular plane of the section, the initial imperfection
was introduced in the X=Y plane with a magnitude that gave the same maximum resistance
for a pin-ended strut as for the case with the initial imperfection in the rectangular plane. It
was found that the case of a pin-ended column loaded by imposed shortening, as load-path
ODA of Figure 8.3, gave significantly lower resistances (at end-rotations greater than at
maximum load) than by proportional loading. Therefore further analyses with the initial
imperfection in the X=Yplane were made by proportional loading.
8.3 Planes of rotations

The analyses have been conducted to explore the behaviour for the combinations of initial
imperfection and imposed end-rotation in about several different axes as shown in Table 8.3.
Table 8.3 Planes of initial imperfections and end-rotations
Plane of initial imperfection Plane of end rotation Case

x=o x=o I
Y=O 2
Y=X 3
Y=Xltan22.So 4
Y=Xltan67.So S
Y=X Y=X 6
Cases 1 and 2 have been analysed for 140x140xlO SHS columns of l.5m, 3.Om and 6.0m
lengths. From Figure 8.4 and Figure 8.S, it can be seen that the behaviour is very similar.
Case 3 has been analysed for 140x140xl0 SHS columns of 3.0m and l.Sm length only
because the similarity of behaviour in the Cases 1 and 2 show that the Case 3 will be similar
for all lengths of columns. Cases 4, S and 6 were performed for the 3.0m columns because
3.0m length of 140x140xl0 SHS columns is expected to be most representative of the
proportions of the majority of structures in which discontinuous columns are used.
8.4 Similarity with different slendernesses

The similarity of behaviour at different slendemesses is shown by the plots of all three
column lengths in Figure 8.4 and Figure 8.S for end-rotations in a rectangular plane.
Tabulated data giving the analysis file and describing the "failure mode" of each analysis are
given in Appendix a3 for the I.S m length, Appendix a4 for the 3.0 m length and
Appendix as for the 6.0 m length.
155
At higher end-rotations, the resistances for end-rotation in the plane of the imperfection are
lower than for end-rotation out of the plane of the imperfection. However, at lower end-
rotations, the opposite is found because the imperfection produces an end-rotation in its own
plane. This may be seen in Figure 8.S. Considering end-rotation in the plane of the
imperfection, the maximum strut load occurs with a slight end-rotation in this plane, which
reduces the effect of any imposed end-rotation. Considering end-rotation out of the plane of
the imperfection, the maximum strut load occurs without any end-rotation out of the plane of
the imperfection, so there is no reduction in the imposed end-rotation.
Results of
Abaqus finite element analysis
--z
~
2000
--1.5m in-plane
g 1500 --1.5m out of plane

"iii --3.0m in-plane
CIt
11000 --3.0m out of plane
E • 3.0m at 46 degrees
8 500 +-+-:."t--+-+--+--t-=-t----t
--6.0m In-plane
• 6.0m out of plane
o .-~-r~-+~~~~
0.00 0.02 0.04 0.06 0.08

End rotations (radians)
Figure 8.4 All column lengths, end-rotations about a rectangular axis
156
Results of
Abaqus finite element analysis
-z 2000
-o
~
c
-+-1.Sm in-plane
Ui -+-1.5m out of plane

en
~ 1500
E ........ 3.0m in-plane
B
-- 3.0m out of plane
1000
0.00 0.02 0.04
End rotations (radians)
Figure 8.5 Comparison of imperfections "in plane" with "out of plane"
8.5 Confirmation of the proposed design model

8.5.1 Basic design model
The column behaviour described in Section 3.4.5 shows that the behaviour of the column at
mid-height, the governing behaviour of the column, is not a simple function of the applied
end rotations. This suggests that the resistance model need not include a detailed calculation
of the cross-sectional resistance to co-existent moments about both rectangular axes.
Instead, it might combine the co-existent moments about both rectangular axes to form a
"design moment" about one rectangular axis and apply this to a simpler resistance model
which includes only bending about one rectangular axis plus axial compression. This greatly
simplifies the equations necessary in the resistance model which makes the method much
more attractive to designers.
157
Study of the results of the parametric study shows that the design model proposed m
Section 3 and shown in Figure 8.6 is valid using
1. the resistance properties about a rectangular axis,
2. the end-rotation applied (even when not in a rectangular axis), and
3. the initial imperfection ej = es, the imperfection about a rectangular axis that gives
the buckling resistance of a pin-ended strut.
M = N {9(h/2) + ed
I
I
I
I I
N9(h/2) I I Ne'
I 1
I I
I I I
I I I
I I I
I I I
I
I
h/2 I
I
I
I
I I I
I I
I
I
I
\
\
end-rotation 9 \
\
\
\
\
h/2 \
\
\
\
\
\
\ \
'. '.
(b)
N Design bending moment
(a) diagram
Load and shape diagram
including Nej
Figure 8.6 Design bending moment in the column including imperfection
The deflection at mid-height of the column caused by the end slopes is shown in Figure
8.6(a). The displacement is calculated from the root of the sum of the squares of the end-
rotations in the two rectangular planes and this displacement is applied in the calculation of
the resistance as if it is in one of the rectangular planes. The appropriate value of
158
imperfection, ei, is added in the same rectangular plane. Therefore the check is made with
design deflection:
Eq 8.1
giving the design bending moment:
M = Ned = N {(h/2)9 + ei} Eq8.2
where ed is the design deflection along one rectangular axis of the SHS
9 is the end-rotation applied to the column (the root of the sum of the squares of the
end-rotations in the two rectangular planes)
ej is the imperfection along a rectangular axis of the SHS. In the design model,
ej = es which is the imperfection along a rectangular axis of the SHS to make the
resistance to axial compression equal to the pin-ended strut resistance.
The reliability of this model is easily shown by plotting the difference between (i) the
eccentricity at mid-height, (h/2)9, imposed by the end-rotation 9 alone and (ii) the
eccentricity at mid-height needed for the design model to give the correct failure load for that
end-rotation. This difference is the additional imperfection, ej. These are shown for hot-
finished 140x140xlO SHS in S355 steel in Figure 8.7 for columns 1.5m long, in Figure 8.8
for columns 3.0m long and in Figure 8.9 for columns 6.0m long. These plots are normalised
byes, which is different for different lengths of column. The values of es are:
for l.5m long, es = 2.80mm, so 9=40mrads gives {(h/2)9}/es = 10.7

for 3.0m long, es = 10.81mm, so 9=40mrads gives {(h/2)9}/es = 5.6
for 6.0m long, es = 92.5mm, so 9=40mrads gives {(h/2)9}/es = 1.3
These plots use the correlation factor, Cf, of 0.8 described in Section 6, to reduce the end
rotation so that 9 = 0.8 times the rotation predicted by Abaqus. It can be seen that the
additional eccentricity required, ej, is less than es in all cases and that the imperfection
required in the design model reduces as the end rotation increases, as deduced in
Section 3.4.4.
159
ei/es,1.5m
0.5
OIl
~o.o
G/
o
-0.5
-1.0
(h/2,9/es
Figure 8.7 1.5 m columns, ej required
ei/es,3.0m
1.0
~Oei/es
0.5 - -+-22.5 ei /es

I i
I I ~45ei/es
~ I , I
- '--1-1
I
~ 0.0 ~---t--
• t
r-r-
-*"67.5 ei /es
, I I I
~ Ii;
-- -----.,- +--I~ ~90ei/es
-<>5
-1.0
Ij I
L____ _ _____ L __ _...l__
I I
l :
i
~I____ L _ __J
; I
I
-e-eOX=eOY 45 ei /es
(h/2)9/es
Figure 8.8 3.0 m columns, ej required
160
ei/es,6.0m
1.0 ~I - - - - - - - T-- ---------T---------l
0.5
I
I
~---
I
I l
1- ----- -.---.----
- I
-S-Oei/es
I I
I
---j
I I
;
::::- 0.0
cu
~____ __ 1
~ --~
~90ei/es
-0.5 --- -- -- - - - - ---
I I
-1.0
1___ ._ __ _ _ l_ .1
(h/2)9/es
Figure 8.9 6.0m columns, ej required
8.5.2 Comparison of 2009 basic design model with Abaqus
The resistances from the design model were calculated using a total mid-height eccentricity
= es + (h/2)9 where 9 = cfxAbaqus end-rotation (with Cf = 0.8 as Section 6) to give a design

model resistance corresponding to each Abaqus resistance but with the Abaqus rotation
reduced by the correlation factor to correct the overestimate of rotation given by Abaqus
when compared to the test results, as found in Chapter 6. It can be seen from these plots that
the design model always underestimates the resistance of the column.
Results for the 3.0m long columns are plotted with in Figure 8.10, for the l.5m long columns
are plotted in Figure 8.11 and for the 6.0 m long columns in Figure 8.12. The axial
resistance vend-rotation calculated using BS 5950 as applied in Appendix B is shown for
comparison in Figure 8.10. The label "N max model" is used for two sets of data, one
indicated by a triangle having the end rotation in a rectangular plane and the initial
imperfection orthogonal to that plane, the other indicated by a square having the end rotation
in a rectangular plane and the initial imperfection in that plane. Both are labelled "N max
model" because one gives the maximum in one range of end rotations and the other gives the
maximum elsewhere.
161
Resistance, model v Abaqus, 3.0m column
1600 ...... Nmaxmodel
1400 -+- N max model
1200 -+- model N 45
1000 -B-Abaqusmax N
Z
.Yo 800 --0-- Abaqus max N
---
m
--+- Abaqus N 22
~ 600
400 "'*" Abaqus N 45

200 -+Abaqus N 67
~BS 5950 single

0
curvature
0 20 40 60
End-rotation (mrads)
Figure 8.10 Abaqus v CMK 2009 design model, 3.0m height

1800
&b
1600 ..... ~
~~ ..... Nmax model

1400 --...:
1200 ~ ~... -+-N max model
-
z~
1000
-'x
~
800
-& Abaqus max N
~ 600 -trAbaqus max N
400
"'*" Abaqus N 45
200
o
o 10 20 30 40 50
162
800
700
~
~
~ ,
~
~
Z
600
500
400
" ~ ....
............
""""" ~
~~
~"l
~
--- N max model
--e- Abaqus es at 0
~ Abaqus
e ~
es at 90
c; ~~
'x
c(
300 ~
-...Abaqus es at 45
200
100
o
o 10 20 30 40 50 60 70
8.6 Conclusions
The accuracy of the CMK 2009 design model compared with finite element analysis is
shown
in Figure 8.10, and for 3.0m columns,
in Figure 8.11 for 1.5m columns and
in Figure 8.12 for 6.0m columns
These figures are graphs of the maximum axial compression vend-rotation when there is no
restraint moment provided at the end of the column to increase the buckling resistance.
The figures all show the resistance of a 140x 140x 10 column which is a typical column size
in residential construction. A column height of 3.0 metres is a typical column height in
residential construction. The behaviour of stockier and more slender column sections is
shown by heights of 1.5 metres and 6.0 metres. The same cross-section has been used in all
of the figures to make comparisons easier.
It can be seen that the model gives safe predictions of the column resistance when the
proposed value of the imperfection, es, is used in the basic design model. This confirms the
prediction that the design model is conservative made in Section 3.7
163
164
9 DESIGN METHOD
9.1 Scope of application

9.1.1 General
The design model is appropriate for the types of beam and column structures comprising
both:
1. Shallow beams (eg Corus ASS) with the floor filling the depth of the beam and fully
connected or fully grouted to the beam to provide stability as AD 281 and AD 285
[SCI 2005a, SCI 2005c] and
2. Columns of hot-finished square hollow sections (eg Corus Celsius). The distinction
between hot-finished and cold formed structural hollow sections is important. Cold-
formed hollow-sections have different residual stresses and have not been calibrated.
web sway beam rotation sway and rotation
Figure 9.1 Instabilities to be avoided in beams at top and bottom of columns
There is the possibility of using the method with other floor systems, but only if the
instabilities shown in Figure 9.1 are avoided. Sway stability of the beam webs can be
assured by web stiffeners and torsional stability of the beams can be assured by transverse
beams and connections of suitable stiffness spanning at right-angles to the main beams.
165
9.1.2 Factors affecting the reliability of the design method
There are a number of issues that affect the reliability of the design method.
1. The method assumes that the end-rotations of every column are equal and opposite
at the top and bottom. This is improbable. Therefore the column resistance will
frequently be higher than predicted by the model.
2. Plastic deformations of beams cause bigger beam rotations than predicted by the
normal engineering approach which assumes elastic deformations. Therefore plastic
deformations must either be avoided or they must be explicitly included in the
calculations.
3. The load factors specified in design codes are calibrated so that there is an
acceptable improbability of failure due to the probability distributions of both
resistance and load even for single elements (such as a simply supported beam).
Therefore the reliability is higher when there are two elements from different
statistical populations. It is improbable that the yield stress of both the beam and the
column are as low as the specified minimum yield and it is even more improbable
that this would happen in a structure that is subject to maximum design load. The
column might be stronger or the beam might be stronger, providing increased
column resistance because there would be some resistance to column end moments.
The design model assumes that the columns cannot derive any resistance from the
beams because the beams are designed to be just strong enough to support the
applied loads from the floor and do not have any spare resistance to resist column
end-moments. However, it is statistically improbable that beams have no spare
capacity because of the statistical distribution of both the beam strength and the
beam loading. Therefore the column resistance will frequently be higher than
predicted by the model because the beams will provide some resistance to column
end-moments. (Even though the columns may have flexible base plates and cap-
plates, some end moment can be transmitted form the column because the axial
compression may act at a small eccentricity from the centre-line of the column.)
9.2 Column design model

9.2.1 The new column design model
The design model allows the beams to be designed independently from the columns. Each
floor level can be analysed individually assuming the beams are on knife edge supports.
166
In this Chapter, the design model is written assuming that it is being applied using
EN 1993-1-1 [BSI2005a, CEN 2005a], but the design model can be used with any design
code by using the appropriate value of initial imperfection es derived from the strut buckling
resistance of the code.
9.2.2 Breath to thickness ratio of wall

The wall breadth to thickness ratio must be sufficient to avoid local deformations of the
walls. The required values are derived in Section 7.3 and the design limits are repeated in
Figure 9.2. The curve is shown in terms of the overall breadth, B, which is not used in
EN 1993-1-1. This is easier for designers to use because the designer is not required to find
the internal radius of the section, which has been incorporated into B. The equation of a line
that fits the data closely and may be taken as the design limit is:
Bit = -14Log lO(9/1000) +38.5 Eq 9.1
where 9 is in radians
design limits: L/8 ~ 10.7

30
25
20
4::. -+- B~limit
all 15
10 -e-B~= 37.5 -14\oglO(rotnj
5
o
o 10 20 30 40 50 60 70
End rotation
(milliradians)
Figure 9.2 Wall slenderness limitations, overall breadth/thickness, Bit
9.2.3 The design process

The design method is applied as follows:
1. Analyse the beam, as if on knife edge supports, carrying the factored floor loading
2. Calculate the slope of the beams carrying the factored floor loading at each column
167
3. Taking the end slope, 6, of the column equal to the greater of the slopes of the beams
above and below the column plus the frame sway, calculate the design deflection at
mid-height of the column caused by the end slopes, as shown in Figure 9.3 (b).
Eq9.2
where ed is the design deflection along one rectangular axis of the SHS
6 is the end-rotation applied to the column (the root of the sum of the
squares of the end-rotations in the two rectangular planes)
ej is the imperfection along a rectangular axis of the SHS. In this case, the
basic design model, ej = es which is the imperfection along a rectangular axis
of the SHS to make the resistance to axial compression equal to the pin-
ended strut resistance
h is the storey height of the column.
6(hl2 I
6 hl2 e·
,
,,
,," hl2
column
hl2
beam
(b)
Design eccentricity ed
(a)
Load and shape diagram including ej
Figure 9.3 Design eccentricity including imperfection
168
4. Calculate the axial compression resistance Na,Rd using the plastic resistance model
in Figure 9.4 and (where there are no additional moments applied) the applied
bending moment is
Eq 9.3
Where there is an additional moment Mext from external forces, such as wind load or
explosion load as in design for key elements, see Sections 9.6.3 and 9.8.1, the
applied bending moment is
y y
I I
I I
Y Y
~ = area resisting axial compression NO,Rd
Figure 9.4 Plastic cross-section to calculate No from My only
This design method is similar to design using ENI993-1-1 [BSI2005a, CEN 2005a]
Section 5.3.2 applied to an individual column using the initial bow imperfection with plastic
analysis as in (3)b.
169
9.2.4 Defining the strut imperfection
The imperfection is defined in terms of the imperfection, es, that must be used with the
plastic cross-section model in Figure 9.4 to give a resistance equal to that of a pin-ended
strut. For a strut with a gross cross-sectional area, A, the imperfection, es, may be calculated
as follows:
I. Calculate the buckling load of the column, Nb,Rd, as if it were a pin-ended strut
using the specified design code.
2. Calculate the area, Ab, stressed to yield stress, fy, required to resist the pin-ended
strut load, Nb,Rd,
Eq9.S
3. Assume the remainder of the area of the section, (A - Ab) is available to resist
bending working at yield stress in tension and compression, and calculate the
reduced plastic moment of resistance, MN,Rd, of the section
4. Calculate the imperfection, es, at axial load Nb,Rd from the assumption that
Eq 9.6
Eq 9.7
This value of es is the value of eo that would be derived from ENI993-1-1 [BSI200Sa,
CEN 200Sa] Section S.3.2(3)b using the UK National Annex which allows eo to be
calculated by calibration with the buckling resistance calculated using EN 1993-1-1.
9.3 Beam design

9.3.1 General
Beams can be designed as simple beams (statically determinate beams), comprising both
simply supported beams and cantilever-and-suspended-spans, or as continuous beams
(statically indeterminate beams). Modem design codes allow beams to be designed to resist
bending moments equal to the plastic moment of resistance. However, design to the plastic
moment of resistance would result in very large deflections if the yield stress were no greater
than the design value and the loading reached the factored design value. These large
deflections would be accompanied by large slopes of the beams which would cause large
170
end-rotations of the columns. Therefore the beams must be designed to be elastic except at
internal supports of statically indeterminate continuous beams, as described in Section 9.3.3.
These limits do not prevent the use of plastic design in the hogging regions, provided the
sagging region moments are limited to the first yield moment and the end-rotations are
calculated as described below for continuous beams.
A similar problem arises in the design of conventional rigid-jointed multi-storey frames. In

BS 5950-1 [BSI2000a] Annex E, Effective length of compression members in continuous
structures, beams are considered as pinned where the moments applied to the beams exceed
90% of the reduced plastic moment capacity. The limit of 90% of the reduced plastic
moment capacity is used because it is approximately equal to the first yield moment of bi-
symmetric hot-rolled I-beams. Where bi-symmetric beams are used, the 90% plastic
moment limit may be applied as in Annex E to ensure that the beams remain essentially
elastic. Asymmetric beams have a much greater difference between the plastic moment of
resistance and the first yield moment, so it is proposed that, where asymmetric beams are
used, the sagging moments should be limited to the first yield moment of the section,
= yield x lowest elastic modulus, to ensure that the beams remain elastic.
In the type of floor construction expected in buildings using the new design method, the
floor slab is concrete (precast or composite), at least as deep as the steel beam and fully
grouted. In certain cases, as described in SCI P175 [SCI 1997], the slab acts compositely
with the beam, increasing the sagging moment of resistance. Where this type of construction
is used, it is proposed that should be applied to the plastic moment of resistance of the
composite beam. The use of the 90% limit is justified because there is no essential
difference between a composite beam made with an asymmetric section and one made with a
symmetric section, so what is acceptable for Annex E should equally be acceptable for this
design method to ensure that the beams remain essentially elastic.
In this report, the term "essentially elastic" is used to refer to

1. the 90% limit from BS 5950-1 Annex E for bi-symmetric steel I-beams and
composite beams, or
2. elastic design for asymmetric steel beams,
whichever is appropriate as discussed above in this Section.
171
9.3.2 Simple beams
Simply supported beams are shown in Figure 9.5 in which the beam splice resists shear alone
and is depicted as a circle. For bi-symmetric beams and composite beams, the beams should
be designed to be essentially elastic. The column end rotation is the same as the beam end
rotation. For elastic behaviour, the end rotation for a simply supported beam under
uniformly distributed loading is (1I24)(WeIEI) where W is the total load on the beam, Lis
the span, E is the young's modulus and I is the second moment of area.
Beams showing deflections and connection
Bending Moment Diagram
Figure 9.5 Simply supported beams (connections at colunul face)
Cantilever-and-suspended-span beams are shown in Figure 9.6 in which the beam splice
resists shear alone and is depicted as a circle. The beams should be designed to be
essentially elastic in both the sagging region and in the hogging region because the beams
are statically determinate. The column end rotation is the same as the beam slope at the
colunm. The beam slope at the colunm is reduced by the cantilever.
172
Beams showing deflectiqns and connection
I
Bending Moment Diagram
Figure 9.6 Cantilever and suspended span
9.3.3 Continuous beams
Continuous beams may be designed as either elastic throughout or as plastic in the hogging
regions of the bending moment diagram (over interior supports) and elastic in the sagging
regions (to control the deflections). Figure 9.7 shows the latter, with the plastic hinge
location depicted as a black circle. In the sagging region, the bending moments in the beams
should be limited so that the beams remain essentially elastic as described in Sections 9.3.1
and 9.3.2. In the hogging region, the beams should be designed so that the applied moments
do not exceed the plastic moment capacity reduced by shear (and also reduced by axial if
there is axial force in the beam, as would oCCill' in a braced bay).
173
Beams showing deflections and plastic hinge
Bending Moment Diagram,

equal spans, equally loaded
Bending Moment Diagram,

equal pans, unequally loaded
Figure 9.7 Continuous beams
If the beams are designed as elastic, the column end rotation is the same as the beam slope at
the column. If the beam are designed a plastic in the hogging region, the beam slopes will
be different on either side of the colwnn because of the rotation of the plastic hinge. The
end-rotation of the column will be somewhere between these two beam slopes, but the end-
rotation is not easily determined because of the range of behaviour from initial plasticity to
strain-hardening as de cribed by Home [Home 1960] and Davies [Davies 1966]. Initially
the plastic hinge rotations will start in the span in which the reduced plastic moment of
resistance is lowest, which depends on the coexistent shear force and axial force in the beam.
Initially, the end-rotation of the column is governed by the opposite span, the span with the
higher value of reduced plastic moment of resistance. However, as the rotation continues,
strain hardening may occur and this will increase the moment of resistance at the plastic
hinge so much that the other span starts to form a plastic hinge and the end-rotation of the
column is now governed by the span that ftrst formed a plastic hinge. TIllS makes it difficult
to be sure what the end-rotations of the column will be, so it is recommended that the end-
rotation is taken as the greatest end-rotation in the beam on either side of the column. The
174
end slopes can be calculated easily from the bending moment of the adjacent spans as shown
in Figure 9.8.
Where pattern loading is a design case, the column end-rotation should be calculated not
only for maximum load on the adjacent spans but also for alternate spans without live load.
Bending Moment Diagrams of adjacent spans,

equal spans, unequally loaded
Figure 9.8 Bending moment diagrams used to calculate slopes at columns
9.4 Frame effects

9.4.1 Effects of sway deformations of the frame on column design
All beam and colunm frames will sway to some extent, even if they are braced frames . The
sway deflections have three principal causes:
1. Externally applied horizontal forces , the most common being wind loads
2. The effect of out-of-vertical of the columns, causing a horizontal component of load
3. In rigidly jointed frames , asymmetry of loading or asymmetry of structural stiffness
In designing any steel frame, the effect of sway on the stability of the frame should be
considered. Both BS 5950-1 [BSI 2000a], in Section 2.4.2, and EN 1993-1-1 [CEN 2005a,
BSI 2005a], in Section 5.3.2, require that the sway effects of frame imperfections should be
considered in design. The detailed requirements of these two codes are different, but the
broad principles are the same, except that EN 1993-1-1 requires that the sway imperfection is
included in all load combinations whereas BS 5950-1 only requires that the sway
imperfection is included in Combination 1, which comprises factored dead and vertical
imposed loads. Both codes assume a basic value of frame imperfection of 1:200.
EN 1993-1-1 allows this to be reduced for increasing colwll11 height and greater numbers of
175
columns. In both codes, the sway effects will be increased in frames with significant second-
order effects, but the methods and the limits differ.
Wind loads cause sway of frames even when they are braced frames. For initial design, the
effect of sway can be estimated from the Serviceability limit on slope which is set by the UK
National Annex to EN 1993-1-1 at hJ300 similar to traditional UK practice. This is at SLS,
so the deflection at ULS can be expected to be (hJ300) times the partial safety factor for wind
load that would produce the greatest lateral deflection from wind. This factor is 1.5, so the
deflection at ULS would be approximately 1.5(hJ300) = hJ200. Therefore the sway angle
would be not greater than (hJ200)/h = 11200 = 0.005 radians = 5 milliradians. For final
design in a structure in which the effect of wind is expected to be small, it might be worth the
effort to use more accurate calculations of deflection from wind.
The simple design model in a frame with no sway deformation is to calculate the end-
rotation of the beam at the top of the column and the end-rotation of the beam at the bottom
of the column and design the column assuming that both the top and the bottom of the
column have the same rotation as shown in Figure 9.9.
Figure 9.9 End-rotations assumed where there is no sway
Sway deformations affect the angle between the beams and the end-to-end lines of the
columns, increasing the angle at one end of the column and decreasing it by an equal amount
at the other end. If the column end-rotation were 9 without sway, as shown in Figure 9.9,
and then the frame were to sway by the angle <1>, as shown in Figure 9.10, the column end-
rotation relative to the line through the column end points (the "end-to-end line") would be
9 + at the bottom and 9 - at the top as shown in Figure 9.10. Following the simple design
model in which the design is made for the greatest end-rotation at top or bottom, the design
value of the column end-rotation should be increased to the sum of the beam rotation at the
176
column plus the sway angle. Therefore, it is recommended that the sway effect should be
applied as an additional rotation to the column, in addition to the beam end rotation. The
sway effect should be amplified if required according to the code being used, as explained
below.
Angle between column-end and

end-to-end line of column
decreased at top to a - cI>
~i
I '
i.jway angIe 'I'tI\
I ,
I ,
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.~ .-.-.-.-.-.-.-.-.-.-.-.-.~
I '1 ------
• ~=----~
-"' ......... . -------
.
beam slope a at top / !
and bottom of column
Angle between column-end and end-

to-end line of column increased at
bottom to a+ cI>
Figure 9.10 End-rotations modified by sway
When using EN 1993-1-1, the amplification factor may be calculated from Section 5.2.2(5)
& (6) and the sway angle due to the frame imperfection may be calculated directly from
Section 5.3.2, or taken as 1:200 = 5 milliradians for simplicity. In EN 1993-1-1, the frame
imperfection must be considered in all load cases. In addition to the frame imperfection, the
wind sway angle must be included in load cases. It may be taken as an additional
5 milliradians for frames with wind deflections limited to height/300 at SLS.
177
When using BS 5950-1, the amplification factor may be calculated from Section 2.4.2.7.2(b)
and the imperfection sway angle may be taken as 1:200 = 5 milliradians. In BS 5950-1, the
sway due to imperfections (represented by the Notional Horizontal Forces of BS 5950-1) is
only applied in load cases without wind. In load cases with wind, the sway angle may be
taken as 5 milliradians for frames with wind deflections limited to heightl300 at SLS.
9.4.2 Additional sway forces on bracing

Bracing needs to be designed to resist
1. Imperfection effects (Sway + Splice)
2. Externally applied lateral loads
3. Additional sway forces from column end-moments as described in Section 9.4.3
The design for imperfection effects and externally applied lateral loads is the same as for the
design of conventional simple-construction frames as SCI P334 [SCI 2004], the new SCI
guide to the design of braced frames to Eurocode 3.
The addition of sway forces from column end-moments is not part of conventional simple-
construction frame design but it is an integral part of the design of a frame that is both braced
and rigid jointed.
9.4.3 Additional sway forces from column end-moments

The design method assumes there are no end-moments transferred from the beams to the
columns and, therefore, there are no column end-moments. However, because the beams
will often have greater strength and stiffness than the minimum required, there might be
some end-moments in the columns and these might apply additional shear to the vertical
bracing. There are two fundamentally different mechanisms for creating the moments.
The first mechanism is where the column ends do not rotate as much as the beams. This is
shown in Figure 9.11. This would occur with small column loads. Assuming the load is
applied to the beam at the face of the column, the moment on the column at each end is
given by Mend = NEd(B/2). The sway shear caused by each column is equal to the sum of the
end-moments on the column at the top and bottom divided by the column height. Therefore,
for 140x 140SHS columns 3.0 metres high, the shear per column
178
(~) B
VH = 2N Ed - - =- NEd
140
=- - N Ed = 0.0467NEd Eq9.8
h h 3000
There would be some small additional shear from the moment caused by the stiffnes of the
end-plate, but this would be small if the end-plate were thin. Therefore a reasonable estimate
of the additional shear on the bracing is 0.05 = 5% of the column loads. This is a
considerable load compared with the Equivalent Horizontal Forces of EN 1993-1-1 Section
5.3, which are of the order of 0.5% of the vertical load, or the Notional Horizontal Forces of
BS 5950-1 Section 2.4.2.4, which is 0.5% of the vertical load. However, this force would
only occur where the columns are loaded to a low proportion of their compression resistance,
so the shear applied to the bracing would not be very large. With high compression in the
column, the column would defonn, reducing the eccentricity of the load on the column ends
and so reduce the sway shear. The two end columns would give opposite directions of shear
force which would cancel out and, depending on the positions of the columns with respect to
the end of the beam segments, it is possible that some of the other columns would give
shears that cancel. The bracing could be designed for 5% of the force in the columns, but it
would be more economical to arrange the beam ends and column positions so that as many
column shear as possible cancel.
.-.-.-.-.-.-.-.-.-.-.-.-.-.-~.~~~
---
Figure 9.11 Columns not deflected to slope of beams
The second mechanism is when the column is forced to take the slope of the beam by the
axial compression. In this case, the column end-moment is dictated by the column stiffness
and the beam slope, as shown in Figure 9.12 for simply supported beams and in Figure 9.13
for continuous beams.
179
Figure 9.12: Column end-rotation equal to beam slope, simply supported beams
Figure 9.13 Column end-rotation equal to beam slope, continuous beams
Because continuity reduces the beam slope at the interior columns, it is clear that the case of
simply supported beams gives the worst case. As an example the beam end-slope of 0.034
radians calculated in Section 3.6 is considered. For a 140x140x1OSHS column 3.0 metres
high and assuming equal end-rotations at the top and bottom of the column, the end moment
induced assuming elastic behaviour would be
M = 3EI8/(heightl2) = 6EI8/(height) Eq 9.9
= 6x210000x 14160000xO.034/3000 N-mm = 202 kN-m
However, the characteristic value of the plastic moment resistance of the column is
fyWpl = 355x246000 N-mm = 87.3 kN-m Eq 9.10
Therefore, the end moment will be much less than the moment calculated by elastic analysis.
For a column carrying half of its yield force, the reduced plastic moment of resistance is
equal to the moment of resistance of the two opposite sides. Therefore, for 140x 140x 10 SHS
columns, the characteristic value is
(fyAl4)(B-t) = (355*5090/4)(140-10) N-mm = 58.7 kN-m Eq 9.11
180
Therefore, for a 140x140xlO SHS column 3.0 metres high with equal end-moments of
58.7 kN-m, the characteristic value of sway shear applied from each column is
VH = 2x58.7/3.0 = 39.1 kN Eq 9.12
The characteristic value of half of the yield force is
355x5090/2 = 903 kN Eq 9.13
Therefore the sway shear expressed as a proportion of column load is
VH = 39.11903 = 0.0433 Eq 9.14
This may be expressed more generally as
2[fyA (B - t)]~
4 h (B-t)
VH = fA - Eq 9.15
y h
2
Note that for axial compressions that are higher proportions of the squash load, the available
moment of resistance is reduced so the sway shear is reduced.
This is close to the value BIh for the case where the column ends do not rotate as much as
the beams. As in that mechanism, the bracing could be designed for 5% of the force in the
columns, but it would be more economical to arrange the beam ends and column positions so
that as many column shear as possible cancel.
In conclusion, for rapid design, a sway shear of 5% vertical could be assumed, but this could
be reduced by appropriate arrangement of the beam ends relative to the columns. 5%
vertical might be uneconomic, but it might serve to give a frame that is sufficiently stiff to
avoid calculation of second-order sway effects for the design of the bracing.
9.5 Connection design fundamentals

9.5.1 General
For discontinuous columns, the connections are different from the beam-column connections
in conventional simple construction because the columns are discontinuous and the beams
lSI
are continuous. Typical structural arrangements have been published in AD 281 and AD 283
[SCI 200Sa, SCI 200Sb].
This section develops typical details that could be used to integrate discontinuous columns
into a complete frame design. It is not intended to be exhaustive, but to show the range of
sizes required for normal buildings
There are five common requirements for the design of beam-column connections for frames
with discontinuous columns.
1. The connection must be made so that any moments induced in the column at the
ends do not rupture the connection.
2. In frames in which plasticity of the beam is to be used, the top flange of the beam is
not adversely affected by the beam-column connection.
3. The building regulation requirement to avoid disproportionate collapse should be
observed.
4. The connections should minimise the storey-shears (or sway shears) that might arise
from the rigidity of the beam-column connections and will be applied to the vertical
bracing of the building.
S. The connections should not be expensive to fabricate.
Design to Eurocode 3 [BSI 200Sa, CEN 200Sb, BSI 200Sa, BSI 200Sb] using the UK
National Annex gives broadly the same results as design to BS S9S0-1 [BSI2000a]. The
calculations in this section are made using Eurocode 3.
9.5.2 Avoiding connection rupture from column end moments

In cases where the column has greater capacity than required, it is likely that the column end-
rotations will be less than the beam rotations. This will be a common case when using the
design method proposed because it assumes single curvature of the column so that the design
method can remain simple. Where the column end-rotation is less than the beam rotation,
forces and moments will be induced in the connection. To keep the design method simple,
analysis of this effect should be avoided by appropriate detailing of the column-beam
connection. This can be achieved by ensuring that the welds and bolts in the connection are
so strong that they remain intact while the plates in the connection yield. In simple end plate
connections, which are the first choice of connection for hollow section columns, it is easy to
select a thickness of plate to yield along plastic hinges at loads lower than the loads that
would rupture the bolts and welds. A typical end plate connection is shown in Figure 9.14.
182
--- - -- - - -- - - -----.
: • 1"====:;1 • I
II II I
i:
I. II II. I
I LL._-_-_-_-JI I
L.. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ I
Figure 9.14 Typical thin end-plate connection
If the beam rotation is different from the column end rotation, the end plate will deform with
plastic hinges along the lines shown in Figure 9.15, provided that the bolts and the weld
between the end-plate and the column do not rupture. The forces on the weld and on the
bolts are easily calculated from the plastic moment of resistance of the plate along the hinge
lines shown.
From simple statics:
the resistance of the end-plate to vertical shear is given by
v = (Mpl,Rd 1 + Mpl,Rd2)lLa Eq 9.16
There is a prying force, Q, generated by the moment Mpl,Rdl such that
Eq 9.17
The total tensile load in the two bolts is
2P=V+Q Eq 9.18
The values ofMpl,Rdl and Mpl,Rd2 are commonly calculated as
Eq 9.19
183
This ignores issues of reduction of effective yield stress by shear stress, increase in effective
yield away from the edges of the plate by Poisson effects, reduction of the section at the bolts
by the bolt holes and increase in resistance by strain hardening. However, Eurocode 3
EN 1993-1-8 Table 6.2 [BSI 2005b, CEN 2005b] uses the above simplifications in the
formula described as Method 1.
,,,---- - --------
,--
-------
------,
, (D , -- ---,, 0
,, ,, ,,
,, ( D '-------,
------- 0
L ___
-------- -------
Mpl,Rdl
M pl,Rd2
Figure 9.15 Bending moments in end-plate
If a connection has a 120x120 SHS and a 160mm wide by 10mm thick end plate in S275
steel, with 6mm fillet welds and 20mm diameter Grade 8.8 bolts at placed 70mm from the
face of the column with 50mm edge distance, the calculations to EN 1993-1-8 [BSI 2005b,
CEN 2005b] are as follows:
Weld resistance stress on S275 plate is
f d = f).J3 = 430/.J3 = 199 N/mm2 Eq 9.20

wv ~w 'YM2 0.85x1.25
Weld resistance along one side of the column (length of weld is 120 mm) is
184
6
199x 120x .fi = 101 kN Eq 9.21
Plate plastic moment of resistance of the plate 160 mm wide and 10 mm thick is
be
MplRd= - - =
fy 160x1Q2
x-
275
= 1l00kN-mm Eq 9.22
, 4 'YMO 4 1.0
La = 70 - 0.8x6 = 65.2 mm Eq 9.23
v = 2Mpl,Rd/La = 2x 1100/65.2 = 33.7 kN Eq 9.24
Q = Mpl,Rd/ep = 1100/50 = 22.0 kN Eq 9.25
Forcelbolt = (33.7 + 22.0)/2 = 28 kN Eq 9.26
Bolt resistance FtRd = k2fui>Asi'YM2 = 0.9x800x245/1.25 = 141 kN Eq 9.27
The bolt resistance and the weld resistance far exceed the plate resistance, so the connection
is ductile, allowing end rotation of the column without rupture of the bolts, welds or end-
plate. Therefore it is safe to use the proposed column design method without checking the
implications of the actual end rotation of the column being less than the rotation of the beam.
9.5.3 Allowing plastic rotations of the beams at the columns

One of the structural benefits of discontinuous columns is that it is relatively inexpensive to
have continuity of the beams to increase the load bearing capacity of the beams and/or to
reduce the deflections. Therefore it is not desirable to drill large holes through the tension
flange of the beams. Because the top flange of a continuous beam at a columns is in tension,
the vertical tying needs to be designed to pass either side of the top flange. This is especially
true for designs in which plastic redistribution of the beam moments is desired.
EN 1993-1-1 Section 5.6, Cross-section requirements for plastic global analysis, subsection
(4) requires that any fastener holes in tension in the area of plastic hinges should satisfy
Section 6.2.5(4). This section is to avoid rupture at the holes. It requires that the nett cross-
sectional area of the tension flange at the fasteners is no smaller than (fylO.9fu)x(YMlYMo)
times the gross cross-sectional area of the flange. For S355 steel and the UK. National Annex
value of 1.15 for 'YM2 at holes in tension flanges, (fylO.9fu) x('YM2/'YMO) =
(355/0.9 x510)x(1.15/1.0) = 0.89, so only very small holes are acceptable. The total width
of holes at a section is 11 % of the flange width. Given that the top flanges of ASBs vary
from 175 mm to 203 mm, the allowable diameter if there were two holes would be from 9 to
11 mm which is impractically small. Therefore the bolts at the level of the top flange need to
185
pass either side of the top flange. There is no limitation on holes in the compression flange
at plastic hinges. A simple detail is shown in Figure 9.16.
1 1 1
Q 1
1 1 1
1
1
1
1 b
--,
1 1
1 1 1
--+-~
---I--r
1
---I
1----
1---- I!--~---
~__ L ___
1 1 1 1
1 1
A Q 1
1 1 1
i Pit
t
1 1 1 1
1 1 1 1
SectionAA
Sectional plan
Figure 9.16 Typical beam-column connection for plastic rotations in the beam
186
9.6 Avoiding disproportionate collapse
9.6.1 Overview of Building Regulation requirements in the UK
Building regulations in the UK. require that buildings should not be so sensitive to accidents
so that damage is disproportionate to the cause. There are three different domains of
regulation in the United Kingdom alone which are:
England and Wales
Scotland
Northern Ireland
The following discussion is in terms of the England and Wales Building Regulations for
simplicity. The England and Wales regulations cover the greatest proportion of construction
in the UK. and the other regulations are similar in intent.
In the England and Wales Building Regulations, the Regulation A3 requirement is:
"The building shall be constructed so that in the event of an accident the building will not
suffer collapse to an extent disproportionate to the cause." Approved Document A gives
ways that are deemed to satisfy the regulations. Section A3 divides structures into a variety
of Classes. These Classes relate to both the size and the use of the building. The different
Classes have different requirements in Approved Document A. Steel frame structures will
usually be used for Classes 2A, 2B and 3. The Building Regulations Classes are given in
Table 9.1:
187
Table 9.1 Classes of buildings
Classes Building type and occupancy

1 Houses not exceeding 4 storeys
Agricultural buildings
Buildings into which people rarely go (with limitations)
2A 5 storey single occupancy houses
Hotels not exceeding 4 storeys
Flats, apartments and other residential buildings not exceeding 4 storeys
Offices not exceeding 4 storeys
Industrial buildings not exceeding 3 storeys
Retailing premises not exceeding 3 storeys of less than 2000m2 floor area in
each storey
Single-storey educational buildings
All buildings not exceeding 2 storeys to which members of the public are
admitted and which contain floor areas not exceeding 2000m2 at each storey
2B Hotels, flats, apartments and other residential buildings greater than 4 storeys
but not exceeding 15 storeys
Educational buildings greater than 1 storey but not exceeding 15 storeys
Retailing premises greater than 3 storeys but not exceeding 15 storeys
Hospitals not exceeding 3 storeys
Offices greater than 4 storeys but not exceeding 15 storeys
All buildings to which members of the public are admitted which contain
floor areas exceeding 2000m2 but less than 5000m2 at each storey
Car parking not exceeding 6 storeys
3 All buildings defmed above as Class 2A and 2B that exceed the limits on
area and/or number of storeys
Grandstands accommodating more than 5000 spectators
Buildings containing hazardous substances and/or processes
The requirements in Approved Document A for hot-rolled steel framed buildings may be
summarised as follows.
1. For Class 2A: to provide effective horizontal ties as described in BS 5950-1:2000.
2. For Class 2B: to provide effective horizontal ties together with effective vertical ties
in all supporting walls or columns as described in BS 5950-1 :2000. Alternatively,
188
limited areas of collapse following notional removal of individual columns may be
used or design of members as "key elements".
3. For Class 3: following a systematic risk assessment of the building accounting for
normal and abnormal hazards, to design the structure to reflect the conditions that
can reasonably be foreseen as possible during the life of the building.
9.6.2 Class 2A buildings

For Class 2A buildings, BS 5950-1 requires that the horizontal ties comply with Section
2.4.5.2 Minimum requirements. In brief, this requires lines of horizontal ties with a
resistance of75 kN at the perimeter and on the column lines in two directions, approximately
at right-angles. The ties can be the steel beams. This requirement is easily achieved with
almost any beam-beam and beam-column connections with not less than two 20 mm
diameter bolts. With discontinuous columns, tying along the line of main beams is simple
because the beams are not interrupted by the beam-column connections and it is easy to
provide simple end-plate or web-cleat connections to provide the necessary continuity.
Equally, the tying at right-angles is simple to provide with the lateral tie-members
recommended in AD 281, AD 283 and AD 285 [SCI2005a, SCI2005b and SCI2005c].
Therefore it is clear that steel frames with discontinuous columns can easily satisfy the
requirements for Class 2A. This means that satisfying the building regulation for
disproportionate collapse for 4-storey residential or office buildings is simple.
9.6.3 Class 28 buildings

For Class 2B buildings, BS 5950-1 requires that the horizontal ties comply with Section
2.4.5.2 Minimum requirements and 2.4.5.3 Limiting effects oj accidental removal oJsupports.
Following the Approved Document A,Section 2.4.5.3 presents three different approaches to
limit the effects.
Horizontal and vertical tying
Notional removal of columns
Design of members as "key elements"
In the aftermath of the Ronan Point collapse, designers had to demonstrate that new designs
were not sensitive to accidents by one of two approaches. One was to show that removal of
anyone vertical element did not cause extensive collapse, which is the philosophy of
notional removal of columns. The other was to show that the important support members
would not be blown away by the pressure of a gas explosion, which is the philosophy of
189
"key elements". The tying requirements are a derivative of the methods used to justify a
frame by notional removal of columns, served up as a recipe that is comparatively easy to
apply to the majority of conventional building frames. Designers of conventional steel
frames should try to use the requirements for horizontal and vertical tying because these
were developed by the steelwork industry to make it simpler to show that a frame is
sufficiently robust. The application of horizontal and vertical tying to frames with
discontinuous columns is investigated in 9.7.
9.7 Connection design to avoid disproportionate

collapse
9.7.1 Horizontal and vertical tying
For frames of hot-rolled steelwork, the relevant BS remains BS 5950-1 because Approved
Document A of the Building Regulations refers to this for the design forces for horizontal
and vertical tying. Eurocode 1 Part 7 contains a similar Annex, but the factors applied to the
loads give less onerous requirements than BS 5950-1, which is not thought to have been
intended. Therefore the requirements of BS 5950-1 will be applied in this study.
The clauses in BS 5950-1 are written on the assumption that steel frames follow the common
practice of continuous columns, so does not consider discontinuous columns explicitly.
However, initially it is seems reasonable to design frames with discontinuous columns to
resist the loads specified for continuity of columns at column splices. For this, BS 5950-1
Clause 2.4.5.3(c) requires that the splice should be capable of resisting a tensile force equal
to the largest total factored vertical dead and imposed load applied to the column at a single
floor level. In the case of discontinuous column construction this might be interpreted
simply as the factored load applied to the column by the floor at that floor level.
The tying clauses in BS 5950-1 have been developed from the option of "notional removal".
The philosophy of the tying clauses is to find a structure that will limit the deformations
throughout the structure when a column is removed or damaged so that it has almost no
capacity. The structural behaviour assumed is as shown in Figure 9.17.
190
,-------,..:;..:- - - - - - - - - - -.- - - - - - - - - - --:.:>1'"-----,
I
I
I
t------k.::-- - - - - - - - - - ---------- -~-+_---__I
Figure 9.17 Model assumed in BS 5950-1 tying
The beams above the damaged column act as inclined ties to support the loads on the
columns above the damaged columns. The vertical tying requirement is to share unequal
reactions on the damaged column line amongst all the beam-ties. The unequal reactions may
arise either from different design load intensities or from the different amount of load
actually applied on each floor. This is important in the majority of steel-frame buildings
because the design values of the tie forces are small to ensure economy in normal frames of
"simple construction".
Vertical tying should not be brittle so it is wise to use bolts of no higher grade than Grade
8.8. These have a minimum ultimate tensile strength of 800 N/mm2 and a minimum 0.2%
proof stress of 640 N/mm2. The tension capacity of Grade 8.8 bolts is
Eq 9.28
where fu = 800 N/mm2 and 1M2 = 1.25
0.9xfu 11M2 = 0.9x800/1.25 = 576 N/mm2 Eq 9.29
The tying forces depend on the column grid and the floor loads. To demonstrate a range of
details, the tying forces required for a larger grid commercial building and a smaller grid
residential building are calculated. The calculations use loads derived from SCI P 342
Design of Asymmetric Slimflor Beams with Precast Concrete Slabs.
191
9.7.2 Vertical tying for a larger grid commercial building
Taking a 7.5m by 9.0m grid, the reactions based on grid area are:
PC units (1.25 x 3.3) x 9.0 x 7.5 278 kN
Concrete (1.25 x 1.23) x 9.0 x 7.5 104
Beam weight (1.25 x 0.2) x 9.0 x 7.5 17
Ceilings and services (1.25 x 0.1) x 9.0 x 7.5 8
Partitions (1.5 x 1.0) x 9.0 x 7.5 101
Imposed loading (1.5 x 2.5) x 9.0 x 7.5 253
TOTAL 762 kN
A typical column for a four storey building with this floor load and 2 kN/m2 for the roof load
would carry 2x9x7.5 + 3x762 = 135 + 2286 = 2421 kN. At 0.7 firM" the area required is
9741mm2.
200x200x 16 SHS has A = 11500mm2 and 200x200x 12.5 has A = 921Omm2, so the column
would probably not be less than a 200x200 SHS.
The tying force must be resisted by the end plate. If the column is assumed to be a 200 SHS,
a reasonable width for the end-plate is 220mm or more. An example beam-column
connection is shown in Figure 9.18. Assuming that the bolts are 70mm from the face of the
column and that the end-plate is S275 plate 25mm thick by 260mm wide, the vertical tying
resistance is 797kN. Therefore, at plastic deformation of the end-plate, the tying force per
bolt is 199kN. If the end-distance of the bolts is 7Omm, the prying force at full plasticity
increases the total bolt tension to 276kNlbolt. The tensile resistance of an M30 Grade 8.8
bolt is 323kN. The tensile resistance of an M24 Grade 8.8 is only 203kN which is
insufficient. However, this end-plate arrangement gives a load-path of the tying force from
the column into the end-plate concentrated through two sides of the column only. This
would necessitate 10mm leg welds along these two sides because the weld strength is
governed by the yield of the S275 plate. The other two sides could use 6mm leg fillets.
192
, ,,
,, ,,,
Q ,, ,, ,,,---- ,b
,,--1---
--+-~ --:j, 1----
__ L ___ A
---'--r ,
+ rl ,
, ,,
'-": ,
,, ,,
, '
,,' h
,1'-.../
+
Sectional plan
Section through beam
Figure 9.18 Example beam-column connection
Nonnallya lOmm weld will require 3 passes of weld so it is more expensive than nonnal.
To reduce the weld leg size, either a greater length of weld must be mobilised or a higher
grade of plate must be used. Raising the grade of plate to S355 gives a higher weld
resistance, but does not allow the use of an 8mm leg weld that only requires 1 run of weld, so
it does not help reduce the cost of the weld. Mobilising a greater length of weld would mean
using a much wider plate, approximately square on plan. This would also need the bolts to
be spaced out further, necessitating two vertical stiffeners either side of the web, instead of
the one shown. While this is possible for the connection above the beam, it is less attractive
for the connection below the beam because asymmetric beams have wide bottom flanges and
the spacing of the bolts clashes with the edges of the flanges making the details even more
difficult. The larger end-plates and the extra web stiffeners add to the costs making this a
less attractive detail for economy.
The ductility of the end connection must also be considered. The column-beam connections
are quite stiff if they are to resist vertical tying forces, so any column end-moments will
cause significant end moments. These might be so large as to fracture the bolts and welds,
so either the bolts and welds must be clearly stronger than the plate, or the end moments
must be calculated so that the welds and bolts can be checked for strength. The welds could
be increased to 12mm without increasing the number of weld passes above 3. For rotation
193
of the column in the plane of the column and the beam, the bolts in tension will develop
individual yield-line patterns, as shown in Figure 9.19.
~:j
1
1
1
1 :b.
~
1 1
r :::] 1----
~---
1 1
d
1
1 1 1
1 1 1
1 1 1 1 1
1 1 1 1
1
Figure 9.19 End-plate yield line patterns
The plate resistance per bolt is now so great that the bolts and the welds need to be even
larger than those required to resist the vertical tying force, making the connection even larger
and more expensive.
9.7.3 Vertical tying for a smaller grid residential building

Taking a 7.5m by 6.0m grid, the reactions for the 1.5 kN/m2 areas based on grid area are:
PC units (1.25 x 3.3) x 6.0 x 7.5 186 kN

Concrete (1.25 x 1.23) x 6.0 x 7.5 69
Beam weight (1.25 x 0.2) x 6.0 x 7.5 = 11
Ceilings and services (1.25 x 0.1) x 6.0 x 7.5 6
Partitions (1.5 x 1.0) x 6.0 x 7.5 68
Imposed loading (1.5 x 1.5) x 6.0 x 7.5 101
TOTAL = 440 kN
This load is resisted by 4 bolts through plastic end-plates detailed to give prying forces of
50%, then the force per bolt = 1.5 x 44014 = 165 kN
194
Therefore minimum area of Grade 8.8 bolts = 165x 103/576 = 287 mm2
M24 bolts to BS 4190 have a tensile area of353 mm2, M20 bolts have only 245 mm2•
9.7.4 Design options

If conventional vertical tying is used to satisfy the Building Regulation, it can be seen from
Sections 9.7.2 and 9.7.3 that many building types and floor grids will need bolts of 24 mm or
30 mm diameter when 4 bolts are used to make the connection.
The options are:
1. use very big bolts to allow the end moments to be ignored
2. calculate the end moments to check that the bolts will not fracture
3. find alternative methods of satisfying the disproportionate collapse regulation
4. do not use this type of construction for Class 2B buildings
9.8 Alternative methods to satisfy the regulation

9.8.1 Design as "key elements"
A standard method of design to avoid disproportionate collapse is to design the members to
resist a lateral pressure defined by the design code. This method, called "key element"
design in BS 5950-1, could be used for structures with discontinuous columns.
9.8.2 Notional removal

One option for demonstrating resistance to disproportionate collapse is the approach often
called "notional removal". The philosophy of notional removal is to find a structural system
in the damaged structure that will limit the deformations throughout the structure when either
a column or a beam is removed (or damaged so that it has almost no capacity) such that
equilibrium is maintained. It is probable that many structures could be shown to satisfy this
requirement if the horizontal tying resistance were sufficiently high.
If the horizontal tie resistances were high, there would be no need to share the floor loads
because each floor could support its own loads. This is particularly appropriate for the form
of construction foreseen in this study because the beams are well restrained by the floor in
the depth of the beams, the beams can form plastic hinges if there are no holes in the top
flanges at the columns and it is easy to provide beam splices that will develop large tying
resistances. Therefore, for this type of construction it is appropriate to design for horizontal
tying alone, but with larger forces than in BS 5950-1.
195
10 WORKED EXAMPLE
10.1 Introduction
The purpose of this worked example is to illustrate the calculations used in the design
method. It does not include all possible load combinations and all possible structural
arrangements. Designers must consider all load combinations even when they are not shown
below.
The design method comprises two parts:
1. Design of the floor slab and main beams.
2. Design of the columns to be compatible with the slopes of the slab and beams at
Ultimate Limit State.
The design process is shown with pattern loading for the slab, because this is the more
complicated case. It is shown both with and without pattern loading for the beams to show
how the end slopes are calculated when plastic redistribution occurs due to plasticity in the
hogging region.
The use of pattern loading or uniform loading will depend on the design code specified for a
particular structure and the judgement of the designer.
10.2 Structural arrangement

The structural arrangement is shown in Figure 10.1. The beam span is 6.0 metres, so, if
these are continuous over two spans, the piece length is 12 metres which is reasonable for
transport. The slab is assumed to be made of precast concrete units with a concrete overlay
with some continuity steel across the tops of the main beams.
196
6.0m I 6.0m I 6.0m I 6.0m ~ I
I I I I
.,. : "! "!
I I I
of i
7.5m iI'i PslabM I
I
I
I
I
I
I
I
Main
I
I I I I
I
I beams
I I
I I
7.5m
I I
I I
I I ~
T T T T
Plan of floor
Beams continuous Beam splice - pinned
SectionAA
Figure 10.1 Structural arrangement
197
10.3 Loading
The loading used in the example is given in Table 10.1
Table 10.1 Loading
Item Gamma Specified Factored Load

load
(kN/m2)
PC units 1.25 3.3
Concrete 1.25 1.23
Beam weight 1.25 0.2
Ceilings and services 1.25 0.1
Total permanent load 1.25 4.83 6.04 kN/m2
Partitions 1.5 1.0

Imposed loading 1.5 1.5
Total variable load 1.5 2.5 3.75 kN/m2
10.4 Slab
In this example it is assumed that the slab is constructed so that the self-weight of the floor is
carried as if the slab were simply supported along each main beam, but that the continuity
steel across the main beams is sufficient for· the slab to act as continuous for support of the
variable loads (imposed floor load and partition load).
10.4.1 Column end-rotation due to slab deflections

The deflection of the slab can cause end-rotation of the column. For this particular structure,
the self-weight is assumed to be supported as if the slab were simply supported and that there
is no end-rotation induced in the column due to self-weight when the structure is complete.
Therefore the only cause of end-rotation in the columns arising from the slab in the
completed structure is from the variable load effects and these effects will be worst with
alternate spans loaded. This condition is shown for an infmitely long structure in Figure
10.2.
198
Variable load
~ I I 1 I
~ t ~ ~
I I
,~:~__-----.~~~'______~~~~~5~m~sp~a=n=s~~~:~~----~~~:~~----~~:
I
1I~-~4-===~K---r~~I=-=-=--S----..:.; ~~:
----=---~
y
w ?124
V I, I~ /1 I~ /~ "
I~I I~I I
I I I I I I
,I ,I
Figure 10.2 Slab variable load and bending moment diagram
The EI value of the concrete floor is taken as 18.3 xl0 12 N-mm2/metre-width.

The end slope of the slab if simply supported is
1 we3 1 3.7Sx7.S 3 xl0 9

ess = - -- =
24 EI 24 18.3xlO ' 2
0.0036 radians Eq 10.1
The end slope is reduced by any hogging moments at SUppOltS
eend = eSS _(Mnea/ + Mrar.e J = e - (Mnear + MrarJ ~ Eq 10.2

3EI 6EI ss 3 6 EI
Taking the hogging moment at the support as
= we = 3.7Sx7.s
2
.
M _v_ =8.79 kN-m/metre-wldth Eq 10.3
24 24
e =0.0036_(8.79 + 8.79JX10 6 X 7S00 Eq lOA

end 3 6 18.3xl0 ' 2
= 0.0036 - 0.0018 = 0.0018 radians
In this case the column end-rotation in the plane transverse to the main beam is very small.
However this cannot be assumed in all cases, especially if a beam carrying significant floor
load frames into the column in this plane.
10.4.2 Slab reaction on main beams

The slab reaction along the main beams due to variable load should not be expected to be
equal to {variable load x slab span}. Where pattern loading is applied, as shown in Figure
10.3 , the load along the most heavily loaded beam will be increased significantly. Even
199
where pattern loading is not applied, the load along the beam next to the edge beam will be
increased . In the case of a two-span slab on beams that have equal deflections, as shown in
Figure 10.4, the load along the central main beam is l.25 {variable load x slab span} .
For this example, the load will be taken as l.20 {variable load x slab span}.
Variable load
I I I I
is
I is Zi is
I
:l1li
I
~iIIII
J..5m spans ~:1111 I
~'
I I
I
I
I I I I I
~
I c::::::::::::::
/If't---- . .~.
I I <::::>'
&
I <::::::>
/". ;----...,.1\
I I "J
I I I I
I I I I
I I I I
I I I I
Figure 10.3 Slab variable load for max reaction on main beams
Figure 10.4 Two-span slab
10.5 Main beams

10.5.1 Initial analysis of main beam
The main beams are two-span continuous beams with pinned splices to one side of a column.
This will give high moment and shear in the beam where it is continuous across the column
line. However, the highest column end-rotation can be expected in the columns at the pinned
splice in the beam.
200
Permanent load
,1111111111111111111111111111111111111111
, ,111111111111111111111111111111
, 1111111111
,
1$ L$ t:,, 1$ ~ i,\,
,6.0m spans ' , :6.0m spans :
,:~ ~:--
, ~:
, ,,~ ~,~
, ~'
,
:, :,, k j W ,,, L'/16
~
,': ,,
,, ,, , ,
,~, c::::::=::::::::, :~, ,
;~
,:.WW:i
, , ,
, ' ,
(a)
(b)
Max hogging moment
Max sagging moment and
max column end-rotation
Figure 10.5 Main beam load and bending moment
Load/m along main beam:

Factored permanent loads/m,
Wp = 1.00 x 6.04 kN/m2 x 7.5m = 45.3 kN/m Eq 10.5
Factored variable loads/m,
Wv = 1.20 x 3.75 kN/m2 x 7.5m = 33.8 kN/m Eq 10.6
Assuming elastic behaviour:

The maximum hogging moment is
Eq 10.7
2 2
45.3x6.0 +33.8X6.0 = 204+152 = 356kN-m
8 8
The coincident shear force is
Eq 10.8
= 1.25x6.04x6.0 + 1.25 x3.75 x6.0 = 169.8 + 126.6 = 296 kN
201
Maximum sagging moment:
The maximum sagging moment occurs at zero shear
Shear at end of span is
M M
V d = V, + ~ _-----.f!L Eq 10.9
en ss L L
where
Vss = (w p +wJL = (45.3+33.8}6.0 =237kN Eq 10.10

2 2
Mnear = zero for the end of the beam which is simply supported, so
Mnear = 0 Eq 10.11
L
W p L2
Mfar = - - +
W
_v_
e Eq 10.12
8 16
2 2
= 45.3x 6.0 + 33.8x 6.0 = 203.8 + 75.9 = 280kN-m,
8 16
M 280
=>-----.f!L = - = 46.6kN-m Eq 10.13
L 6.0
=> Vend = 237 + 0 - 46.6 = 190.5 kN Eq 10.14
so the shear force is zero at a distance x from the end of the beam is
x= Vend = 190.5 = 2 410m Eq 10.15

(w p +wJ (45.3+33.8) .
(W p +WJX2
Msag = Vend X - 2 Eq 10.16
= 190.5x2.410- (45.3 + 33.8}x 2.4e = 459 _ 229.5 = 230kN-m

2
10.5.2 Choice of beam size

Max hogging moment = 356 kN-m with coincident shear force = 296 kN
Max sagging moment = 230 kN-m from pattern loading
Try 280 ASB 74, yield stress = 355 MPa
202
WeI = 776 cm3,
Me,eI = 355x776x1000 = 275 kN-m Eq10.17
Msag = 230 = 0.833 => OK Eq 10.18

Me,el 275
Check classification to see if plastic cross-sectional resistance can be used

Bottom flange:
c= Bb -tw -2r = 285-10-2x24 =113.5 Eq 10.19

2 2
so ~ = 113.5 = 8.107 Eq 10.20

tf 14
Class2 limit is
.!!.. =10. =10

tf
t fy
35
=IO~ 235 =8.136
355
Eq 10.21
Therefore 280 ASB 74 is Class 2, so the plastic moment of resistance can be used.
Check if the plastic moment of resistance is reduced by shear:
Eq 10.22
V 296
.·.~=-=0.386<0.5 Eq 10.23
Vpl,Rd 767
so the moment of resistance need not be reduced for the effects of coexistent shear force.
Wpl = 979 cm3,
Me,pl = 355x979x 1000 = 348 kN-m Eq10.24
M hOg = 356 = 1.082 Eq 10.25

Me,pl 348
EN 1993-1-1 Section 5.4. 1(4)B allows 15% redistribution from elastic analysis for Class 2
cross-sections provided that the conditions of this Section are satisfied, which they are for
this particular structure. For this beam, the required redistribution is (356 - 348)/356 x l00%
= 2.3%, so the redistribution is acceptable.
203
The sagging moment resulting from the redistribution must also be checked, but the pattern
load case used to check sagging above is more onerous, so the sagging resistance is
acceptable.
Therefore the 280 ASB 74 is satisfactory.
10.5.3 Calculation of beam slopes

E = 210000 MPa,
I of280 ASB 74 = 12190 cm4 ,
The end slope of the beam if simply supported:
1 (w +wJO
ess = 24 p
EI Eq 10.26
3 9
=_1 (45.3+33.8)x6.0 xl0 =0.0278 radians
24 21Ox10 3 x121.9x10 6
The end slope is reduced by any hogging moments at supports
L
eend = e SS
_(Mnear
3EI
+ MfarL)
6EI Eq 10.27
At end of two-span beam, Mnear = 0,
so
eend = eSS - Mfar L
6EI Eq 10.28
At centre of two-span beam, M far = 0,
e = e _ Mnear L Eq 10.29
ss 3EI
With pattern loading:

The hogging moment at mid-length of the two-span beam = 280 kN-m
At end of two-span beam,
e - e _ Mfar L Eql0.30
end - SS 6EI
6
=e - 280x10 x 6000 = 0.0278-0.0109 = 0.0169 radians
ss 6x21Ox10 3 xl21.9xl0 6
At centre of two-span beam,
204
e - e _ MnearL Eq 10.31
- ss 3EI
6
=e _ 280 x 10 X 6000 =0.0278 _ 0.0219 =0.0059 radians
ss 6x21Ox10 3 x121.9x10 6
Without pattern loading:

The hogging moment at mid-length of the two-span beam is limited to the cross-sectional
moment of resistance of the beam, reduced by shear if appropriate. In this case is no
reduction from shear because the applied shear is less than 50% of the plastic shear
resistance. Therefore the hogging moment = 348 kN-m
At end of two-span beam,
e = e _ Mnear L
Eql0.32
end SS 3EI
6
=e -0- 348x10 x 6000 =0.0278-0.0136 =0.0142 radians
ss 6x21Ox10 3 x121.9xl0 6
At centre of two-span beam,
e =e _ Mnear L _ MfarL Eql0.33

end SS 3EI 6EI
6
=e - 348xl0 x 6000 -0 =0.0278-0.0272 =0.0006 radians
ss 3x21Oxl0 3 x121.9x10 6
10.6 Column
10.6.1 Design values of end-rotations
The design values of end rotations are taken from the beam analysis without pattern loading
to show the values calculated to the Eurocodes.
The greatest end-rotation in Section 10.5.3 is at the end of the two-span beam, so this case
will be used to show the design method. It should be noted that the axial load in this column
will be less than in the column at mid-length of the two-span beam because of the hogging
moments in the beams at mid-span. Therefore, the columns at mid-length might be the most
heavily loaded and must be checked.
205
Table 10.2 Design end-rotation (end of two-span beam)
End of two-span beam End-rotation

(radians)
In plane of beam and column Beam slope 0.0142
Frame imperfection 0.005
Horiz deflection from wind 0.005
TOTAL 0.0242
Perpendicular to main beam Slab, see Note 1 0.0018
Design end-rotation 0.0243
End-rotation =9 = ~9~ + 9~
Note 1: The end-rotation was calculated with a different pattern of floor
loading (alternate spans of slab loaded) from the pattern that gives the
highest beam slopes, but the value is so small that it is not worth re-
calculating the slope for the coexistent case.
10.6.2 Wall thickness of column
The design limits are given in Figure 9.2. From this it can be seen that the limiting Bit is
approximately 18 for the design end-rotation of 0.0243 radians = 24.3 milliradians.
Using the formula for the limiting BIt:
Bit = 37.5 -1410g lO(end-rotation in milliradians) Eq 10.34
Bit = 37.5 -19.4 = 18.1

If the section is a 140x140 SHS, then B = 140, so the minimum wall thicknes of the column
is 140/18.1 = 7.7 mm.
10.6.3 Design equations
Where there is no externally applied moment from lateral loads, the design equation is
Eq 10.35
Where there is an externally applied moment from lateral loads, the design equation is
Eq 10.36
206
but Ne,Rd ~ flexural buckling resistance of a pin-ended strut with the the same lateral
loading.
The resistance equation for the column depends on the axial compression applied.
If the axial compression resistance does not exceed 0.5 NeRd, the resistance may be calculated
from Appendix A.2.4:
Eql0.37
If the axial compression resistance is not less than 0.5 NeRd, the resistance may be calculated
. _ N •. Rdm-M~,
.. N e.Rd - ( b)
e+-
Eql0.38
where b = B - t = the distance between mid-planes of the SHS walls
10.6.4 Resisting vertical loads only
Design end-rotation, e= 0.0269 radians and storey height, h = 3.0 metres.

Find the resistance of a 140x 140x 10 SHS with yield stress of 355 N/mm2,
Find the design imperfection ed
The design imperfection ed = es, the imperfection calculated from the flexural buckling
resistance of the column if acting as a pin-ended strut.
Calculate the pin-ended strut resistance to EN1993-1-1:
Ler) (3000)
(
~=~= i = 52.7 = 56.9 =0.745 Eql0.39
A, [ xff,) (X 2~~500 J 76.4
<l> = 0·4 + (l(~- 0.2)+ ~2 J= 0.5[1 +0.21(0.745 - 0.2)+ 0.745 2 ] Eq 10.40
207
=0.5[1 + 0.114 + 0.555] =0.835
1 1
X= =~ ]= 0826
- r l]
Eql0.41
[ <P + ~(<p' '0.835 + ~(0.835' - 0.745' ) .
:. Nb Rd = AXfy = 5090xO.826x355 =1492 kN Eql0.42

''YMI l.0
Calculate the squash load:
N = Afy = 5090x355 =1807kN Eql0.43

pl,Rd 10
'YMO •
Calculate strut action moment Ms assuming rectangular stress blocks:

In this case, the area required to resist axial compression alone is more than the area of the
two sides acting as webs. Therefore the area resisting bending is confined to the two sides
acting as flanges as shown in Figure 10.6 as the two unshaded areas. The force resisted by
. (N pI Rd - N b Rd)
each unshaded area IS equal to ' '.
2
y
I
I
Y
~ = area resisting axial compression Na,Rd
Figure 10.6 Plastic stress blocks in column
208
Therefore the strut action moment resistance available is
Ms= (Np),Rd -Nb,Rd)b = (1807-1492)0.130 =20.47kN-m Eq 10.44

2 2
Calculate imperfection es:
M 20.47
es = _ s _ = - - = 0.01372 metres = 13.72 mm Eq10.45
Nb Rd 1492
Calculate total eccentricity on column:
-t
e=e s + ( h ) 8=13.72+ (3000)
-2- xO.0243 Eq 10.46
= 13.72 + 36.45 = 50.18 mm
Calculate the axial compression that can be resisted at the end of the two-span beam from
either Appendix A.2.4 for Na,Rd not more than 0.5Nc ,Rd or Appendix A.3.4 for Na,Rd not less
than 0.5Npl,Rd.
Initially try Na,Rd not less than 0.5Npl,Rd'
When there is no externally applied bending moment:
Na,Rd = Np),Rd
(*)(b) Eq 10.47
e+ -
2
= 1807 0.065 = 1807 X 0.564 = 1020 kN

0.05018 + 0.065
From Eq 10.43, O.5Np l,Rd = 0.5x 1807 = 904 kN => Na,Rd > 0.5Npl,Rd
NaRd 1020 .
- - '- = - - = 0.684 even for smgle curvature of the column Eq10.48
Nb,Rd 1492
For the column at mid-span of the two-span beam, the design end-rotation is 0.0106 radians,
so Na,Rd = 1241 = 0.832 Eq 10.49

Nb,Rd 1492
even for single curvature of the column
209
10.6.5 Resisting reduced vertical and lateral loads
Therefore the design equation derived from moment equilibrium at mid-height becomes:
Eq 10.50
but Na,Rd ::: resistance of a pin-ended strut with the the same lateral loading.
Take the factored horizontal load for this load combination WH = 15 kN/m acting on the
storey height of 3.0 metres as a uniformly distributed load,

Therefore the bending moment applied to the column is:
2 2
M - w H h = 15x3.0 = 16.88kN-m Eq 10.51
ext- 8 8
Using EN 1993-1-1 Section 6.3.3 and Annex B, the maximum axial compression coincident
with this moment is 1107kN.
Using the new design model:
Eq 10.52
The resistance function for the column depends on the axial compression applied.
If the axial compression resistance is not less than 0.5 NeRd, the resistance may be calculated
N pl,Rd (~ ) - M ext 1807( 0.130) -16.88

N ---~~~---
Q,Rd - ( b)
e+-
2
(0.05018 + O. ~30)
= 873 kN Eq 10.53
If the axial compression resistance does not exceed 0.5 NeRd, the resistance may be calculated
Eq 10.54
Afy
0.5Np l,Rd = 0.5 - = 903kN Eq 10.55
'YMI
so N does not exceed 0.5Npl,Rd,
210
Therefore with externally applied moment = 16.88 kN-m, max compression = 871 kN.
This is less than the compression resistance of 1107 kN for the pin-ended strut resisting the
same applied bending moment of 16.88 kN-m because the design end-rotations, caused by
the beam slopes, give a lower resistance than the pin-ended condition.
211
11 CONCLUSIONS
11.1 Summary of what has been achieved

11.1.1 Design philosophy
A design model has been developed for discontinuous square hollow section (SHS) steel
columns with flexible end-connections. It uses the non-linear behaviour of the column
arising from extensive plasticity. This has necessitated a thorough investigation of the
behaviour of columns with zones of plasticity that extend beyond the zones of plasticity that
occur in columns designed using current structural design codes such as BS 5950-1 and
EN 1993-1-l.
The model allows beams to be designed as beams independently of the columns and the
columns designed for end-rotations equal to the rotation of the beam. The design model
assumes that the column is not required to resist externally applied moments to maintain
static equilibrium, eg moments from the beams, so the resistance to axial loads for a given
end-rotation is higher than allowed by normal codified design methods.
11.1.2 Column design model

The column design model uses the classic plastic cross-sectional resistance model about a
rectangular axis together with a member model that is rigid-plastic with an initial
imperfection. The initial imperfection is calibrated from whatever design code is specified
so that at zero end-rotation, the column model gives the same resistance to axial compression
as the code gives for a pin-ended strut. The design model has been demonstrated to be
conservative so it is safe for use in the design of structures.
11.1.3 Design of a complete frame

Beams may be designed as simple or continuous, either entirely elastic or elastic in the spans
and plastic at the interior supports. Appropriate design methods are given for the beams,
connections and to ensure overall frame stability and compatibility of the frame sway
deflections with the column design. A worked example including beam and column design
is included.
212
11.1.4 Economy of structural material
The design method requires the beams to carry all the bending moments from the loads on
the beams, leaving the column to resist axial load alone. Thus the reduction of resistance of
the column below the pin-ended strut resistance is not as severe as in normal codified
methods. Therefore this design method offers economy of material relative to current frame
and column design methods in appropriate frames.
11.1.5 Economy of design effort
The beams are designed independently from the columns, either as simply supported beams
or continuous beams on knife edge supports. Therefore the analysis is very simple. The
column design model has simple "closed solutions" to calculate the resistance, so this also is
simple.
11.2 Suggestions for further developments

Further work would be valuable in the following areas:
11.2.1 Circular hollow sections
Verifying that the model is applicable to circular hollow sections (CHS).
11.2.2 Rectangular hollow sections

Checking the application to rectangular (not square) hollow sections (RHS) and finding how
to benefit from the greater inertia about one rectangular axis.
11.2.3 I-sections
Investigating the application to I-sections. It is expected that application to end-rotation
about the minor axis will be similar to the behaviour of SHS sections, but with restrictions on
the slenderness of the flange outstands to be defined. Rotation about the major axis forces
shortening in one flange more than in the other, so the behaviour is likely to be similar to two
SHS sections side-by-side. In single curvature bending about the major axis of the column,
one flange can be expected to deflect out-of-plane much more than the other flange, resulting
in significant twisting along the length of the member. This is likely to be a complicated
investigation.
213
11.2.4 Avoiding disproportionate collapse in Class 28 and Class 3
buildings
As discussed in Section 9.8.2, the verification strategy of "notional removal" appears ideally
suited to demonstrating the robustness of frames with discontinuous columns. Reference to
older design documents from the period after Ronan Point would show the methods used by
designers to employ the notional removal philosophy.
214
REFERENCES
Access-steel Determination of moments on columns in simple construction,

SN005a 2005 SN005a-EN-EU; 2005
Access-steel website: www.access-steel.com
AISC 2005 ANSI!AISC 360-05

Specification for Structural Steel Buildings
American Institute of Steel Construction, 2005
AISC 2005a ANSI!AISC 341-05 + ANSI!AISC 341s1-05

Seismic Provisions for Structural Steel Buildings
Including Supplement No.1
American Institute of Steel Construction, 2005
ASCE 1971 Plastic Design in Steel; A Guide and Commentary

ASCE - Manuals and Reports on Engineering Practice - No. 41
American Society of Civil Engineers, 1971
BCSA2002 National Structural Steelwork Specification for Building Construction

(commonly known as the NSSS)
The British Constructional Steelwork Association
Fourth Edition, 2002
BSI1969 BS 449-2:1969
Specification for the use of structural steel in building,
Part 2, Metric units
British Standards Institution 1969
BSI2000a BS 5950-1:2000
Structural use of steelwork in building -
Part 1: Code of practice for design - Rolled and welded sections
215
BSI2000b BS 5400-3:2000
Steel, concrete and composite bridges
Part 3: Code of practice for design of steel bridges
BSI2001 BS 5950-2:2001
Structural use of steelwork in building -
Part 2: Specification for material, fabrication and erection
- Rolled and welded sections
BSI2002a BS EN 1990:2002
Eurocode 0, Basis of structural design.
BSI2002b BS EN 1991-1-1:2002
Eurocode 1, Actions on structures. General actions.
Densities, self-weight, imposed loads for buildings.
BSI2005a BS EN 1993-1-1 :2005

Eurocode 3: Design of steel structures - general rules and rules for
buildings
BSI2005b BS EN 1993-1-8:2005
Eurocode 3: Design of steel structures: Part 1.8: design of joints
BSI2006a BS EN 10210-1:2006
Hot finished structural hollow sections of non-alloy and fine grain
steels
Part 1: Technical delivery conditions
216
BSI2006b BS EN 10210-2:2006
steels
Part 2: Tolerances, dimensions and sectional properties
BSI2008 BS EN 1090-2:2008
Execution of steel structures and aluminium structures.
Part 2: Technichal requirements for the execution of steel structures.
Calladine 1969 Calladine, c.R.,

Engineering Plasticity,
Pergamon Press 1969
Carr 1993 Carr, J.F.,

A simplified approach to the design of semi-rigidly connected columns
in multi-storey non-sway steel framed buildings
M. Phil Thesis, University of Sheffield, 1993
CEN2002a EN 1990:2002
Eurocode 0, Basis of structural design.
CEN (European Committee for Standardization), 2002
CEN 2002b EN 1991-1-1:2002

Eurocode 1, Actions on structures. General actions.
Densities, self-weight, imposed loads for buildings.
CEN 2005a EN 1993-1-1:2005

Eurocode 3: design of steel structures - Part 1.1: General rules and
rules for buildings
CEN2005b BS EN 1993-1-8:2005
Eurocode 3: Design of steel structure -: Part 1.8: Design of joints
217
CEN 2006a EN 10210-1 :2006
steels
Part 1: Technical delivery conditions
CEN 2006b EN 10210-2:2006

steels
Part 2: Tolerances, dimensions and sectional properties
CEN2008 EN 1090-2:2008
Execution of steel structures and aluminium structures.
Part 2: Technichal requirements for the execution of steel structures.
Davies 1966 Frame instabilty and strain hardening in plastic theory

Journal of the Structural Division,
Proceedings of the American Society of Civil Engineers, June 1966
Dowrick 1977 Dowrick, D J,

Earthquake Resistant Design,
John Wiley & Sons, 1977
ECCS 1976 European Convention for Constructional Steelwork

Second International Colloquium on Stability
Introductory Report
(includes the ECCS Manual on Stability of Steel Structures)
Tokyo, September 1976
Liege, April 1977
Washington, May 1977
ECCS 2006 Boissonnade N., Greiner R., Jaspart J.P. and Lindner J.
Rules for member Stability in EN 1993-1-1, Background
documentation and design guidelines
ECCS Technical Committee 8 - Stability;
ECCS No 119,2006
218
Fewster et al Fewster M., Giradier v., Owens G,
1993 Economic Design and the Importance of Standardised Connections,
New Steel Construction, February 1993
Galambos Galambos T. V.
1998 Guide to Stability Design Criteria for Metal Structures, fifth edition,
John Wiley & Sons, 1998
Gent 1966 GentAR.,

Elastic-plastic column stability and the design of no-sway frames,
Proceedings of the Institution of Civil Engineers, 34, June 1966 p129-
152
Gent & Gent, AR. and Milner, H.R.,

Milner1968 The ultimate load capacity of elastically restrained H-columns under
biaxial bending,
Proceedings of the Institution of Civil Engineers, 43, December 1968
p685-704
Gibbons 1990 Gibbons, C

The strength of biaxially loaded beam-columns in flexibly connected
steel frames,
Ph.D. Thesis, Department of Civil and Structural Engineering,
University of Sheffield, 1990
Gibbons et al Gibbons, c., Nethercot, D.A, Kirby, P.A, and Wang, Y.C.,
1993 An appraisal of partially restrained columns behaviour in non-sway
steel frames,
Proceedings of the Institution of Civil Engineers,
Structures and Buildings, 99, Feb 1993, p15-28
Gibbons et al Gibbons, C., Kirby, P.A., and Nethercot, D.A.,

1993a Experimental behaviour of partially restrained steel columns,
Proceedings of the Institution of Civil Engineers,
Structures and Buildings, 99, Feb 1993, p29-42
219
Greiner et al Greiner R, Ofner Rand Salzgeber G,
1998 Review of rules for members in bending and axial compression,
Report 3, Flexural buckling of beam-columns,
ECCS Validation Group, Working Item 8
Institut fur Stahlbau, Holzbau und Flachentragwerke,
Technishe Universitat Graz,
Home 1960 Home,M.R,

Instability and the plastic theory of structures
Transactions of the Engineering Institute of Canada, Vol 4, No.2, 1960
Home 1963 Home,M.R,

Elastic-plastic failure loads of plane frames
Proceedings of the Royal Society, Vol A, Part 274, pp 343-64, 1963
King 2006 King, C.M.,

Column design for axial compression and end rotation
Report to the University of Sheffield and the SCI
SCI RT 1111, 2006
King 2007 King,C.M.,

Design of discontinuous columns for axial load and end-rotations in
rigid jointed braced frames
SSRC, 2007 Annual Stability Conference, April 18 - 21,
New Orleans, Louisiana
Kirby et al Kirby, P.A., Bitar, S. and Gibbons, C.,

1992 Design of columns in non-sway, semi-rigidly connected frames,
Proceedings of First World Conference on Constructional Steelwork
Design, ed. Bowlins, P., J., et aI.,
Elsevier Applied Science, London, 1992, pp 54-63.
ODPM2004 Approved Document A - Structure

The Building Regulations 2000, 2004 edition,
Office of the Deputy Prime Minister
220
SCI 1997 Lawson, R.M., Mullett, D.L., Rackham, lW.,
P175, Design of asymmetric Slimfor beams using deep composite
decking.
The Steel Construction Institute, 1997
SCI 2001 P 202, Steelwork Design Guide to BS 5950-1 :2000, Volume 1

Section Properties and Member Capacities
(commonly referred to as the "Blue Book")
SCI 2004 Brown, D.G, King, C.M., Rackham, lW., Way, A.,
P334, Design of multi-storey braced frames.
SCI2005a AD 281: The use of discontinuous columns in simple construction

New Steel Construction, February 2005
SCI2005b AD 283: The use of discontinuous columns in simple construction

(Part 2)
New Steel Construction, March 2005
SCI2005c AD 285: Floor systems for simple construction using discontinuous

columns and continuous beams.
New Steel Construction, April 2005
SCI2005d AD 288: Discontinuous columns in simple construction: Beam-column

connection.
New Steel Construction, July/Aug 2005
SCI2005e AD 292: The use of discontinuous columns and shallow deck

composite slabs on floor beams.
New Steel Construction, Oct 2005
221
SCI 2009 Brown, D.G., !les, D.C, Yandzio, E.,
Steel Building Design: Medium Rise Braced Frames
Wood 1973 Wood,RH.,

A new approach to column design
Building Research Establishment Report
Department of the Environment, HMSO, 1974
222
APPENDIX A SHS SECTION PROPERTIES
A.1 Introduction
A.1.1 General
There are two basic cases of resistance to axial and bending. One is where the axial
compression is sufficiently small to be resisted only by the "webs". In a square section, this
occurs when the axial compression is not greater than 0.5 Npl • The other is where the area of
the "webs" alone is not sufficient to resist the axial compression. In a square section, this
occurs when the axial compression is greater than 0.5 Npl •
The calculations are made using an "ideal square" section. This is a perfectly square
centreline model, to avoid the complexities of rounded corners with varying thickness. The
finite slice model used an equivalent thickness and distance from mid-plane to mid-plane of
the walls calculated to give the same area and inertia (second moment of area) as the real
SHS section. In the Abaqus model, the distance from mid-plane to mid-plane of the walls is
the same as in the SHS being analysed. It used the thickness of the SHS being analysed
because the local stability of the walls is automatically included in the Abaqus model.
Therefore the comparison of resistance of the Abaqus model with actual SHS sections, as in
the analysis of the laboratory tests, requires that the Abaqus prediction is reduced according
to the ratio of cross-sectional area of the SHS to the Abaqus model.
The formulae are written to be compatible with EN 1993-1-1 and are based on Np\ and Mp\
to avoid the difference between an "ideal square" section and the actual SHS section with
rounded corners.
A.1.2 Partial safety factors for resistance
The calculation of the resistance uses the strut buckling resistance to establish the design
value of initial imperfection and uses plastic cross-sectional resistance calculations to find
the resistance with applied end-rotations. The calculation of resistance could be done either
by using characteristic values of resistance and then dividing the characteristic value by the
appropriate partial safety factor for resistance or it could be done by using the design values
of both strut resistance and cross-sectional resistance.
Most designers will not be accustomed to working with characteristic values, so this report
recommends the use of design values. In EN 1993-1-1, different partial safety factors are
223
used for cross-sectional resistance (YMO) and for buckling (YM!)' In many countries, including
the UK, the National Annex for EN 1993-1-1 will use the same numerical value for both YMO
and for YM! so there is no difference in the partial safety factor used. In countries using a
higher value of YM!. than for YMO, the design value of initial imperfection will be higher, so
even though YMO is used in the cross-sectional resistance calculations, the resistance will be
correctly calculated.
A.2 Axial not exceeding 50% squash load

A.2.1 Plastic stress blocks
An idealised SHS section is shown in Figure AI. It is subject to axial compression and
bending about the YY axis. The area required to resist axial compression at yield stress is
denoted by the length of wall within the shaded area which extends to the distance "ca "
either side of the YY axis.
b
14
1
'-'-1-'-' b
~~-- ._._.
I
Y
~ = area resisting axial compression Ne,Rd
Figure A.1 Plastic stress distribution, low axial load
224
A.2.2 Derivation of depth of block resisting axial, dimension "ca"
In an ideal square cross-section. the plastic moment of resistance with no axial compression
is given by:
EqA 1
For compression resistance not exceeding 50% Npb ie Na,Rd ~ 2thfyl '¥Mo.
the axial compression resisted is
EqA2
therefore the plastic moment of resistance of the section is reduced by the plastic moment of
resistance of the wall lengths "ca" either side of the YY axis
Therefore the reduced plastic moment of resistance is
EqA3
If there is no externally applied bending moment such as from wind
EqA4
then
3 2 2) tfy 4ctfy
( -b -2c - = - - e
EqA5
2 'YMO 'YMO
:. -b 2 2 = 4c a e
3
-2c a
EqA6
2
EqA 7
EqA8
- b± .Jb 2 - 4ac
Then from ax + bx + c = 0 ~ x = - - - - - -
2
2a
EqA9
EqA 10
225
A.2.3 Derivation of "e" from failure load No
To calculate e so that the design model gives the correct failure load No as calculated by
Abaqus:
EqA 12
where M ext is an externally applied moment on column, for example from wind load,
Therefore e = (MN - M ext)/ No EqA 13
EqA 14
and No = 4tcafy => c a = No EqA 15

4tfy
A.2.4 Derivation of failure load Ne,Rd from "e"

To calculate N 9,Rd from "e":
EqA 16
where Mext is an externally applied moment on column, for example from wind load,
For N~O.5Np" Ne,Rd = 4tcfyl 'YMo EqA 17
EqA 18
EqA 19
EqA20
4c a tfy N 9,Rd
Na,Rd = => Ca = EqA21
'YMO 4tfYd
226
•
(N )(N
.. N 9,Rd e + - - - -
2
9,Rd
4tfYd
9,Rd
J
=Mpl-Mext EqA22
2
N 9Rd
. ' + N 9,Rd e = Mpl,Rd - M ext EqA23
8tfyd
.'. N 9,Rd 2 + 8tfyd N 9,Rde - 8tfyd (Mpl,Rd - M ext ) = 0 EqA24
EqA25
in which "+" gives the relevant root
EqA26
A.3 Axial compression exceeding 0.5 Npl •

A.3.1 Plastic stress blocks
An idealised SHS section is shown in Figure A2. It is subject to axial compression and
bending about the YY axis. The area required to resist axial compression at yield stress is
denoted by the length of wall within the shaded area which extends beyond the ''webs''. This
idealisation uses a "centre-line" model, so the area of the flange is taken to extend to the
distance "Cb" either side of the :xx axis.
227
I.. b ~I
1 1
Y 1
1 1
1 1
;
._._.
I
Y
~ = area resisting axial compression Na,Rd
Figure A.2 Plastic stress distribution, high axial load
A.3.2 Derivation of depth of block resisting bending, dimension

"Cb", assuming no externally applied moments
In an ideal square cross-section, for axial compression exceeding 50% N pl•Rd ,
The axial compression resisted is
EqA27
The reduced plastic moment of resistance is
EqA28
If there is no externally applied bending moment such as from wind
MN,Rd = Na,Rde, then EqA29
EqA30
EqA31
228
EqA32
EqA33
4be 2be
:,C b = =-- EqA34
2b+4e b+2e
A.3.3 Derivation of "e" from failure load N assuming no externally

applied moments
To calculate e so that the design model gives the correct failure load Ne as calculated by
Abaqus:
EqA35
where MN,Rd is the plastic moment of resistance of the plastic stress block NOT resisting
the axial force,
Therefore MN = (Npl- Ne)(b/2) EqA36
Therefore e = [(Npl - Ne)/Ne](b/2) EqA37
A.3.4 Derivation of failure load Ne,Rd from "e"

To calculate Ne ,Rd from "e":
For N~O.5NpI. the moment resitance is
M = (N pl,Rd -N e,Rd JU
h...
EqA38
N,Rd 2.
. N e + M = (N pl,Rd -N e,Rd JU
k..
.. e,Rd ext 2 EqA39
:. N e+M = (N pl,Rd -N e,Rd JU

\.... = N pl,Rd b _ N e,Rd b
EqA40
e,Rd ext 2 2 2
NeRdb Npl,Rd b
:. Ne,Rd e + i = 2 - M ext EqA41
EqA42
229
Npl,Rd ("2b)_M ext
:. NO,Rd = ( b)
e+-
EqA43
230
APPENDIX B BS 5950 CALCULATIONS
B.1 Buckling resistance of Square Hollow Sections

This Appendix gives the calculations for the axial compression resistance according to
BS 5950-1:2000 [BSI 2000] and conventional analysis.
BS 5950-1:2000 clause 4.8.3.3.3, More exact method for CHS, RHS or box sections with
equal flanges gives:
For square hollow sections (so no lateral-torsional buckling) with moments about one axis
only:
Eq B. 1
EqB. 2
By inspection, the first equation is seen to be critical
EqB. 3
EqB.4
EqB. 5
EqB.6
EqB.7
Hot finished 140xl40xl0 SHS in S355

Sx = Sy = 246 cm3, so Mp = 355x246x10-3 = 87.3 kN/m
231
B.2 Single curvature
B.2.1 3 metre column height
General
Pcx = 1550 kN from Blue Book
6
M = 2EI8 = 2x205000x14.2x10 8xlO-6 = 19418 kN-m EqB. 8
L 3000
m = 1.0 for single curvature

40 mllll-radians end slope
End slope = 0.040 radians at the top and -0.040 radians at the bottom for single curvature.
2EI
M = - 8 = 19418 = 1941 x 0.040 = 77.6 kN-m EqB. 9
L
Eq B. 10
Eq B. 11
= 1- 0.889 1550 = 0.111 1550 = 119 kN

1+ 0.5 x 0.889 1.445
30 mllll-radians end slope

2EI
M =-8 = 19418 = 1941xO.030= 58.2 kN-m Eq B. 12
L
Eq B. 13
Eq B. 14
= 1- 0.667 1550 = 0.333 1550 = 387 kN

1+ 0.5 x 0.667 1.334
232
20 mllli-radians end slope
2EI
M =- 8 = 19418 =1941xO.020 = 38.8 kN-m Eq B. 15
L
EqB.16
Eq B. 17
= 1- 0.444 1550 = 0.556 1550 = 705 kN

1+ 0.5 x 0.444 1.222
2EI
M = - 8 = 19418 = 1941xO.01O = 19.4 kN-m EqB.18
L
EqB.19
EqB.20
= 1- 0.222 1550 = 0.778 1550 = 1085 kN

1+ 0.5 x 0.222 1.111
233
APPENDIX C FINITE ELEMENT MODEL
DETAILS
C.1 Introduction
This Appendix contains details of the finite element models numbering systems.
C.2 Full model

The key nodes are shown in Figure C.l and Figure C.2. The wall node numbering scheme is
shown in Figure C.4, Figure C.5, Figure C.6 and Figure C.7. The nodes used to predict the
inclinometer output are shown in Figure C.S. The input lines for the end constraints and
rigid bodies are shown Table C.l.
z
t1
1
__-+1---'::--------- 3910
I ",
1 "
------- ---;~:~---------.
, / 1 3950
"" I
3930 -------'...:........j-~----
__- ; . - - i - - : 7 " ' - - - - _ 110
130
Figure C.l Node numbering of comer nodes of SHS
234
z
3950 ...-
150 .......~:.....--+y
Imperfections and nodes to apply eccentric load
235
z
t1
1
1 _-_~--::::oo
392.0_ ..... __".. . 3910
3930 __--tIt--......
~......._!t_e~ _ ____
130 140
Figure C.3 Node numbering of rigid bodies at ends
236
3910
3940
110
140
Figure C.4 Wall 1, node and element numbering
237
3920 3921 3922 3923 ~924 3925 3910
I""
" " "
220 221 222 ~23 224 225 210

e21 e22 e23 e24 e25 e26
120 121 122 123 124 125 110
238
3920
3930
230
130
239
~930 3941 3942 2943 ~ 944 3945 3940
'" '" '"
.230 ~41 ~42 1243 .244 1245 . 40

."
e41 e42 e43 e44 e45 e46
130 141 142 143 144 145 140
Figure C.7 Wa114, node and element numbering
240
z
t
1
1
1 3910
I ",
I ,,'
-----------~~:~---------.
,,' 1 3950 ...............'---t- 176. mm
" 1
3930 ---------..:....,'-+--...;.1---
--36-1-0--l~265. mm
010
410
mm
--31-0-+ 176.
110
130---------~----w
Figure C.S Node numbering at mid-height and inclinometers
241
Ifable C.l Input numbering for end nodes, end constraints and loading
** nodes on centre-line
**
** at top
** central node
*NODE
~950, 0.0, 0.0, <height>
** added 1st Jan08, eccentricity to simulate stabilising moment from friction
*NODE
3951,0.0, <eccytop>, <height>
**
** at bottom
** central node
*NODE
150,0.0,0.0,0.0
** added 1st Jan08, eccentricity to simulate stabilising moment from friction
*NODE
151,0.0, <eccybottom>, 0.0
**
** node sets for rigid bodies at top and bottom
*NSE~NSET=TOPNODES
3910,3911,3912,3913,3914,3915,
~920,3921,3922,3923,3924,3925,
~930,3931,3932,3933,3934,3935,
~940,3941,3942,3943,3944,3945
**
*NSE~NSET=BOTTO~ODES
110, 111, 112, 113, 114, 115,

120, 121, 122, 123, 124, 125,
130, 131, 132, 133, 134, 135,
140, 141, 142, 143, 144, 145
**
** v2: define sets for friction eccentricity
*NSE~NSET=TOPREFNODE
242
3950
*NSET, NSET=BOTTOMREFNODE
150
**
*NSE~NSET=TOPNODESPLUS
3910,3911,3912,3913,3914,3915,
3920,3921,3922,3923,3924,3925,
3930,3931,3932,3933,3934,3935,
3940,3941,3942,3943,3944,3945,3950,
3951
**
*NSE~NSET=BOTTOMNODESPLUS
110,111,112,113,114,115,
120, 121, 122, 123, 124, 125,
130, 131, 132, 133, 134, 135,
140, 141, 142, 143, 144, 145, 150,
151
**
*********************************************
** Boundary conditions
**
*BOUNDARY, TYPE=VELOCITY
151,1
151,2
**150,3,,10
151,3,,10
151,5
151,6
~951,1
~951,2
**3950,3,,-10
~951,3,,-10
~951,5
~951,6
**
243
C.3 Half model
The key nodes of the finite element model are shown in Figure C.9 and Figure C.1 O. The
nodes on line 123 to 7623 and 143 to 7643 are constrained against displacement in the X
direction and against rotations about the Y and Z axes using the Abaqus XSYMM boundary
constraint command. The nodes used to predict the inclinometer output are shown in
Figure C.11.
Z
tI
I
: 7623 ___- - -__7610
I ",
I "tI'
7650,~:-----
" I
"", I
7643 __--+-----t...
110
143
Figure C.9 Node numbering of SHS - half model
244
7650 - - -
ISO .......-=-
Figure C.l 0 Imperfections and eccentric load application - half model
245
z
t
I
I
I 7623 7610
I
1 ___ /
I "
7650~-::____ ~ 7210
"" I 176: mm
,-' I ./'
7643 " I
I
76 0 265'< mm
7010
=
4~351O
=
610 265.1 mm
~
176. mm
410
~y
V
123
17' k:::
143
_/1~~*-
/
V -·x
JlO
140
Figure C.11 Node numbering at mid-height and inclinometers - half model
246
APPENDIX D STRUT RESISTANCE FROM
INITIAL IMPERFECTIONS
0.1 Introduction
This appendix shows how the strut imperfection is used in BS 5400-3 [BSI 2000b] Annex G
in the Perry strut approach. It is included as an example of a strut buckling fonnula in which
there is magnification of the design imperfection.
0.2 Equilibrium
Calculate the equilibrium stresses in the simple strut with an initial imperfection in the shape
of a half-sine wave
Using the Perry type offonnulation for strut stability:
EqD.I
where
cr max is the maximum stress in the strut assuming elasticity
cr a is the mean axial stress in the strut
cr b is the bending stress in the strut at the extreme fibre assuming elasticity
N is the axial compressive force on the strut

M is the moment caused by strut buckling
A is the cross-sectional area of the strut
We is the elastic modulus resisting bending in the plane of buckling
eN is the equivalent geometrical imperfection in the plane of buckling under loadN
eo is the initial equivalent geometrical imperfection in the plane of buckling
Ncr is the relevant buckling force
247
eN
I
I eO I N e I
i++I I
I I
I I
I I
:\ I
I I\ I
I : \
I ! \I
I
I i. \\ II
I
,
'I
I
,,
I
I
I
I
I I
: I
!I
Figure D.1 Imperfections; eo at zero load and ~ at load N
D.3 Derivation of strut formula
EqD.2
= 0' + 0' (AeoYJ 1 EqD.3

8 8 I 0' 8
1__
O'er
EqD.4
EqD.5
where
248
(J cr is the elastic critical buckling stress in the strut
I is the inertia in the plane of buckling

r is the radius of gyration in the plane of buckling
EqD.6
When (J max = (J y' where (J y is the yield stress,
EqD.7
EqD.8
EqD.9
EqD.1O
Dividing through by u/
EqD.ll
Eq D. 12
Re-writing with cr a =k and cry =~2

cry cr CT
EqD.13
EqD.14
2 -b±Jb2_4~
From ax +bx+c=O=>x=-----
2a
Eq D. 15
249
The relevant root is
Eq D. 16
EqD.17
Eq D. 18
This is the form of strut buckling equation that appears in BS 5400-3 [BSI 2000b] Annex G
250
APPENDIX E TRANSVERSE BENDING OF
WALLS
E.1 Introduction
This appendix contains the study of wall thickness made in the early phase of this project.
The goal was to establish if the requirements of codes such as BS 5950-1 and EN 1993-1-1
are sufficient for the type of column behaviour considered in this project. The study was
made on columns of 140xl40xi0 SHS, 3 metres long, in S355 steel bending about a
rectangular axis. The study showed clearly that the limits given in these codes are not
sufficient. The application of the method in this appendix was not extended to other
geometries of column because the column analysis method was changed from the finite slice
method of Section 4.2 to the Abaqus finite element analysis of Section 4.3 which
incorporates the effects of wall thickness as seen in Section 7.
The structural mechanics of the effects of curvature are discussed in Section 7.1. In
conventional design procedures, the transverse bending of the walls of members is not
analysed explicitly. Instead, the wall thickness is chosen to satisfy the "classification"
requirements of the design code for the particular use of the member. For example,
designing to BS 5950-1, a beam required to develop a plastic hinge must satisfy the breadth
to thickness limits of Class 1. These are given for flanges and webs of 'open' sections and
for walls of hollow sections.
The limits in BS 5950-1 for Class I, Class 2 and Class 3 for a 140xl40 SHS in S355 steel
are as follows:
£ = ...J(275/355) = 0.880 EqE.l
Class I: bit = 28£ =28xO.880 = 24.64 EqE.2
so Bit = 3 + 24.64 = 27.64 EqE.3
so t ~ 140/27.64 = 5.1 mm EqE.4
Class 2: bit = 32£ =32xO.880 = 28.16 EqE.5
so Bit = 3 + 28.16 = 31.16 EqE.6
so t ~ 140/31.16 = 4.5 mm EqE.7
251
Class 3: bit = 40£ = 40xO.880 = 35.2 EqE.8
so Bit = 3 + 35.2 = 38.2 EqE.9
so t ~ 140/38.2 = 3.7 mm EqE.I0
In certain cases, such as in design for earthquake resistance of some forms of structures,
greater wall thickness is required as shown in textbooks and codes such as the AISC Seismic
Provisions [AISC 2005a]. The dominant phenomenon requiring extra wall thickness is
transverse bending of the wall caused by the curvature of the member. The curvature of the
member produces radial forces in the members across the full width. These radial forces
cause transverse bending moments. If these are large, they will cause distortion of the cross-
section such that the resistance might be significantly reduced. In this Appendix, the issues
involved in defining a suitable breadth to thickness ratio are considered because the normal
classification limits may not be adequate for the high curvatures that may occur in columns
checked to the proposed design method.
Finite element analysis using 3-D elements with 3-D elasto-plastic material properties are
commonly used for detailed studies of this issue. However, it is possible to develop a good
understanding of this behaviour by using the simplified analyses in this Appendix.
All common columns carrying their maximum load will have plastic strains on the most
compressed face, so the behaviour is not entirely elastic. However, it is helpful to develop
understanding of the phenomena involved by starting with a study of behaviour that is
elastic. Therefore, the following studies start with elastic behaviour in Section E.3 and then
develop equations for elastic-plastic behaviour in Section E.4. Equations are developed for
columns bending about a rectangular axis and about a diagonal axis. It was observed in
Section 7 that members with applied end-rotations in the X = Y axis tend to deflect as
expected in that plane, but as they approach maximum load, they tend to deflect towards a
rectangular axis if the initial imperfection was in that rectangular axis. This results in an
end-rotation in the rectangular plane that approaches IISin45° times the end-rotation in the
X = Y axis. Therefore columns with applied end-rotations in the X = Y axis need to have
wall slenderness suitable for IISin45° times the end-rotation if it were to be applied in a
rectangular plane.
252
E.2 Second order elastic behaviour
The deflections of the wall due to the radial pressure vary along the member because the
curvature varies and so the radial pressure varies. The variation of deflection causes an
additional curvature along the centre-line of the wall which will cause additional deflections.
A simple model of this effect can be made by assuming that this curvature is a Sine curve
with a certain effective half-wave length of leff and maximum displacement = a.
The radial pressure then becomes the sum of
1. The radial pressure from the member curvature, PRC = to'clR = to'c(d2y/dx?), as
described in Section 7.1, plus
2. The pressure from the curvature of the local buckling of the wall, PRC = to'c(d2y/dx?).
It is most convenient to represent the curve of the local buckling as a Sine curve, so that
Eq E. 11
EqE.12
At the maximum value of Sine,
EqE.13
Assuming the transverse stiffness depends on transverse bending stiffness alone, the
deflection may be written for the general case as
Eq E. 14
EqE.15
EqE.16
Eq E. 17
253
EqE.18
The term [ 2] is an elastic magnifier.

I
4
- -J
b (1t
I-a td--
C f EIwall eft'
E.3 Elastic behaviour bending about X or Y axes

E.3.1 Radial pressure
As described in Section 3, curvature causes radial pressure on the wall:
PRC = tOclR = PRC = tOcCEq E.19
where C is the curvature = llR
For the case of a 140x140xlO SHS 3 metres long in S355 steel bending about a rectangular
axis, the curvature, C, reaches 2xl0-4 for the case of single curvature with both the initial
imperfection and the imposed end rotations acting in the same plane.
Taking
the wall thickness is 10mm and
the axial stress in the wall is at yield stress of 355N/mm2
:.Axial force/mm, txoc = 10x355 = 3550 N/mm EqE.20
:.Radial pressure, PRC = txocC = 3550x2xlO-4 = 0.71 N/mm2 EqE.21
E.3.2 Transverse bending moments

Assuming the stress along the tension face of the section is also at yield, there is symmetry
of the bending moment diagram about the axis of curvature of the section.
The free bending moment on curved wall = WL2/8 with free end slope = wL 3/24EI where
W == PRC because member under consideration is bending about a rectangular axis so the
radial pressure is perpendicular to the wall of the SHS.
The symmetry of the bending moment diagram gives a uniform moment along the walls in
the plane of curvature. Writing this uniform moment as M, the end slope =MLl2EI
For strain compatibility,
the end slope at end of wall in the plane of curvature = End slope at end of curved wall
254
EqE.22
2EI 24EI 2EI
. 2(Mb)_ wb
.. 2EI - 24EI
3
EqE.23
wb
:.M=-
2
EqE.24
24EI
Therefore the moment along walls in the plane of curvature is constant and is:
Eq E. 25
Therefore the moment at mid-span of curved wall is:
EqE.26
Figure E.I BMD with equal stresses in the compression and tension
flanges
E.3.3 Stresses
The span of the walls is the width of the section minus half the wall thickness at each end
Therefore maximum moment is
2
/12 __ O.71x130
2
wb - - - - = 1000 N-mm EqE.27
12
If t = 10 mm, the elastic modulus of a 1 mm wide stip is
BD2 lx10 2
Z=--= = 16.7mm3/mm EqE.28
6 6
255
Therefore, bending stress is
0= 1000/16.7 = 60 N/mm2 EqE.29
Shear force (at inside face of wall at right-angles) is
v = 0.71x120/2 = 42.6 N Eq E. 30
Average shear stress is
'tave = 42.6/10 = 4.3 N/mm2 Eq E. 31
At extreme fibre, shear stress = 0

At mid-plane, shear stress is
't = 1.5x42.6/1O = 6.4 N/mm2 EqE.32
E.3.4 First order deflections

The deflection of wall at mid-span of the wall is:
5 wb 4 1Mb 2 5 wb 4 1 wb 2 b 2
8=-----=------ EqE.33
384 EIwall 8 EIwall 384 EIwall 8 24 EIwall
3 wb 4
:.8=-- EqE.34
384 EIwall
3
where 1= BD3 = Ix10 = 83.3mm 4 Imm EqE.35
12 12
4
:.8= 3 0.71 x 130 = 0.091mm Eq E. 36
38421OxlOOOx83.3
E.4 Effects of plastic flow

E.4.1 Introduction
Where material reaches yield and is strained beyond the onset of yield, there is deformation
due to plastic flow of the material in addition to deformations due to the elastic strains. This
means that the deflections arisng from transverse bending will be different from that
calculated by assuming pure elastic behaviour.
Plastic flow is explained with elegant simplicity by Calladine, Engineering Plasticity,

(Calladine 1969) in section 2.8 Yield Surface for Three-dimensional Stress, especially page
36, and in section 2.13 The Mises Yield Condition and Associated Flow Rule, especially page
49. From this we have that the plastic flow vector, (&t. &2, &3) is parallel to the deviatoric
stress vector, (0' I, 0' 2, 0' 3) in which
256
Eq E. 37
EqE.38
Eq E. 39
where ao = (al + a2 + (3)/3 EqE.40

and at. a2 and a3 are the principal stresses.
At the extreme fibre, the transverse direct stress, a2, is at its maximum and the shear stress is
zero. In this study, the maximum shear stress is small compared with the maximum
transverse direct stresses, so the study will focus on the zones of maximum direct stresses,
which are at the extreme fibres.
Clearly there is an interaction between the longitudinal stress at which yield will occur and
the transverse bending stresses. In a wall with longitudinal compressive stresses, on the face
with tensile transverse bending stresses, the stress at yield is less than the uni-axial
compressive yield whereas on the face with compressive transverse bending stresses, the
stress at yield is greater than the uni-axial compressive yield.
Strain hardening will affect the actual behaviour. So also will the unloading of longitudinal
stress as the longitudinal "slices" deflect towards the centre of the section due to the
transverse bending deflections. Both of these effects will be ignored to simplify the analysis.
The plastic strains are calculated from the plastic plastic flow vector, (&10 &2, &3). The
plastic flow vector is found by calculating the deviatoric stress vector, (a' t. a' 2, a' 3), which
is found from the principal stresses, at. a2 and a3.
E.4.2 Calculating the principal stresses at the surface

The von Mises solution for stresses in 2 directions only is used, because there is no through-
thickness stress along the flat regions of the walls. This is solved to find the value of
longitudinal direct stress, at. that causes yield in the presence of the bending stress that was
calculated from the member curvature.
Eq E. 41
EqE.42
EqE.43
257
a 2-V2
+ Ia 2 + ia
~y
2 _ a 2 _ 3't 2 )
2
:.a, = - - 0 - - - - -- - - - - EqE.44
2
:.a, =~±
2
a/ +(a 2 -a
4 ~ y 2
2 -3't2) EqE.45
EqE.46
EqE.47
At the surface, 't = 0, so substituting 't = 0 into Eq E. 47
EqE.48
E.4.3 Calculating the deviatoric stress vector at the surface
At the surface, a3 = 0, and a, is defined by Eq E. 48, so Eq E. 40 becomes
ao = 3 =~--------3----~--- EqE.49
EqE.50
3 2 3
Calculating a' ,
.,
.. a , =a, - a o = [a2-+[
2
G,' -{~),)J-[G; ±~[ G,' -{G;),)J EqE.51
G,' -{ G; J] EqE.52
Calculating a' 2
258
Eq E. 53
Calculating 0' 2
EqE.54
Therefore the deviatoric stress vector (0' I, 0'2, 0' 3) =
E.4.4 Calculating the normalised plastic flow vector at the surface

The plastic flow vector is parallel to the deviatoric stress vector, so the plastic flow vector,
(&1. &2, &3), normalised to a longitudinal plastic strain of 1.0 =
EqE.56
~ 7+~[ ":-\7)'] -7+~[ ":-{7)']

±~[ ",' _{ "; )')' ±~[ ",' _\ ";
I, n Eq E. 57
EqE.58
259
The vector is non-linear in terms of 0'2.
The effect of plastic flow is shown schematically (ie not to scale) in Figure E.2 in which the
elastic stress-strain is compared with stress-strain when there is plastic flow form a
longitudinal strain of 2 yield strain and 3 yield strain.
Elastic strain
~ Elastic strain + plastic flow
i at longitudinal strain
,i / of 2 yield strain
, ,
,i/ / " , Elastic strain + plastic flow

/ " at longitudinal strain
,i / of 3 yield strain
,i/'
i'
"
~--------------~--------------~E2
,
/,?
,
, /'/
,
/' ,/
/ " "
"
/ ,
Figure E.2 Schematic comparison of (elastic strain) v (elastic strain + plastic flow)
For the purposes of this present report, the relationships will be assumed to be linear to
estimate the magnitude of the effects of transverse bending and the associated deflections.
From Eq E. 58, it is seen that
30'2
& ±
2~ [O,'_'~2
4
rfz _1 &
I
EqE.59
The term {=+= ~} is the component of plastic strain that expands the total width of the wall,
but does not affect the deflection of the wall.
260
The term is the component of plastic strain that increases the
curvature of the wall. It is proportional to the bending stresses and inversely proportional to
the yield stress modified by the transverse bending stress. This means that the deflections
from transverse bending are magnified according to both the transverse bending stresses and
the longitudinal plastic strain. The term is not linear in 02, but for smaller values of 02 it is
not far from linear. To make a simple calculation possible, for small values of 02, this term
can be taken as:
30 2
4 EqE.60
Calling the plastic strain that increases the curvature of the wall E2b, then Eq E. 59 becomes
30 2
1.':_
ut:.2b =ut:.) x--
1.':_
40
Eq E. 61
y
E.4.5 Second-order effects where plasticity occurs

It is shown in Section E.2 that, for elastic behaviour, the deflections are magnified by:
I
( J2]
b4 1t
I-a t d - - - -
[ C E1wall f elf
where d~ is the deflection of the wall calculated by first-order elastic analysis.

E1wall
The bending moment diagram will be modified by plastic flow. The parts of the cross-
section experiencing plastic flow deflect as if they have reduced stiffness, so the restraint
from the walls in the plane of the curvature will be somewhat increased. However, a
simplified study of the wall behaviour can be made by assuming that the elastic amplifier can
be modified to represent behaviour in the presence of plasticity by increasing the deflections
by an amplification factor, A. Therefore, the magnification factor would become:
261
1
( J2]
b4 1t
1-0' tAd-- - -
[ Y EIwall f elf
The amplification factor, A, may be calculated approximately without iterations. The

amplification is due to the plastic strains that increase the curvature of the wall. These are
the strains in direction '2' of the plastic flow vector (& .. &2, &3) assuming that the plastic
strain along the column is in direction '1'. Integrating Eq E. 61, in which E2b is the plastic
strain that increases the curvature of the wall,
EqE.62
Both the transverse bending stress, 02, and the longitudinal strain in the column, E.. are
functions of the overall curvature of the column. In Section 7.1 it was shown that the radial
pressure is a function of the curvature and in Section E.3.3 the transverse stresses were
calculated from the radial pressure.
2
Writing the overall curvature ofthe column, d ;, as 'C' for convenience, and writing
dx
EqE.63
and EI = FxC, EqE.64

where D and F may be considered to be independent of C, then
Ern
= Emf ~1 =_3_ J(DC}i(FC)= 3DFErnf~c= 3DF[c ]Ernax
2
E2b EqE.65
40 y 40 y 40 y 40 y 2
lOy lOy lOy £y
. 3DF r 2-C 2]
=--lC EqE.66
•. E2b em y
SOy
where CErn is the curvature at a cross-section

and C y is the curvature at a cross-section when the longitudinal strain reaches
1.0 yield strain.
The total bending strain is the sum of the elastic bending strain and the plastic bending strain.
Writing the elastic bending strain as E2be, the total bending strain = E2be + E2b.
262
The amplification factor, A, is the ratio of the total bending strain, including plasticity, to the
elastic bending strain, therefore
EqE.67
DC
where £2be =-02
E
=- E
EqE.68
EqE.69
where CErn is the maximum curvature at the cross-section under consideration.
The factor F was defined in Eq E. 64 as £, = FxC, therefore
F=~ EqE.70
C
Writing the number of yield strains at maximum strain as 'n',
EqE.71
EqE.72
Substituting into Eq E. 69
EqE.73
EqE.74
E.4.6 Minimum thickness of wall to avoid local instability

The wall thickness needs to be sufficient to prevent gross deformation of the wall of the
column, because gross deformation would reduce the resistance of the column and
263
precipitate failure. The minimum thickness of the wall cannot be less than the thickness at
which the magnifier would become infinite. From the formula for the magnifier above, we
see that the magnifier becomes infinite when:
EqE.75
EqE.76
:.1 = 12cr y Ad- ( b4

Et
2
-
1t
'-elf
2
J
EqE.77
:. t = 12cr Ad~(~J2 = l2AdG, (~Jb2 EqE.78

y E '-elf E '-elf
E.5 Elastic-plastic behaviour bending about X or Y

axes
E.S.1 Example of strains
For the case of a 140x140xlO SHS 3 metres long in S355 steel, for the case of single
curvature with end rotations equal and opposite of 0.040 radians and with both the initial
imperfection and the imposed end rotations acting in the same plane, the curvature reaches
2xlO-4.
On the face with compressive stresses of 02 = 60 N/mm2 from transverse bending, as

calculated in Eq E. 29, and the shear stress of zero at an extreme fibre,
EqE.79
Therefore the "hydrostatic" stress is
00, = (01 + 02 + (3)/3 = (381.3 + 60 + 0)/3 = 147.1 N/mm2 EqE.80
Therefore
264
0"] = 0'] - 0'0 = 381.2 -147.1 = 234.1 N/mm2 Eq E. 81
EqE.82
0"3 = 0'3 - 0'0 = 0.0 - 147.1 = - 147.1 N/mm2 Eq E. 83
Therefore the deviatoric stress vector is
(0"], 0"2, 0"3) = (234.1, - 87.1, -147.1) EqE.84
The plastic flow vector, (0£], 002, 003) is parallel. Dividing through by 0"], so that oo] = 1.0,
the plastic flow vector = (1.0, -0.372, -0.628) EqE.85
On the face with tensile stresses of 0'2 = 60 N/mm2 from transverse bending, as calculated in
Eq E. 29, and the shear stress of zero at an extreme fibre,
0'
]
=~±
2
0' 2
y
_3(~)2
2
-3't 2
=-30±~3552 -330 2 -3X02 =-30+351.2=321.2 N/mm2 EqE.86
Therefore the "hydrostatic" stress is
0'0, = (0'] + 0'2 + 0'3)/3 = (321.2 - 60 + 0)/3 = 87.1 N/mm2 EqE.87
Therefore
0"] = 0'] - 0'0 = 321.2 - 87.1 = 234.1 N/mm2 EqE.88
0"2 = 0'2 - 0'0 = -60.0 - 87.1 = - 147.1 N/mm2 EqE.89
0"3 = 0'3 - 0'0 = 0.0 - 87.1 = - -87.1 N/mm2 EqE.90
Therefore, the deviatoric stress vector is
(0"], 0"2, 0"3) = (234.1, -147.1, -87.1) EqE.91
The plastic flow vector, (0£], 002, 003) is parallel. Dividing through by 0" ], so that oo] = 1.0,
the plastic flow vector = (1.0, -0.628, -0.372) EqE.92
Therefore, under longitudinal plastic straining, the compression face plastic flow (or plastic
strain) is -0.372 and the tension face plastic flow (or plastic strain) is -0.628. The mean
value is -0.500, so the difference from mean to maximum is +1-0.128. This means that for an
increment of longitudinal strain of 1.0, there is a plastic flow equivalent to a transverse
bending strain of 0.128. For each increment of longitudinal strain of one yield strain,
265
£y == a/E, the increase in the bending strain at the surface of the wall will be equivalent to an
elastic bending stress of o.1 28><yield stress == 0.128x355 == 45 N/mm2 due to plastic strain.
For the same member in single curvature with an end rotation of 0.020 radians, the most
severe curvature is 5.3xlO-5 and the maximum strain is less than 5 yield strains. Therefore
the radial pressure and the transverse bending stress are reduced by a factor of about 4,
giving commensurately reduced plastic flows.
E.S.2 Example of minimum thickness for high end-rotation

The elasto-plastic second-order finite slice model of Section 4.2 of a 140x140x1O SHS
3 metres long in S355 steel bending about a rectangular axis gave the following results for
the case of single curvature with both the initial imperfection and the imposed end rotations
of 40 milliradians acting in the same plane. Assuming no strain hardening:
1. at mid-height (the point of maximum curvature), the curvature reaches 2xlO-4 and
the strain reaches 14 yield strain,
2. at 100 mm from mid-height of the member, the curvature is IxIO-4 and the strain is 7
yield strain.
At mid-height of the member, the maximum strain is 14£y, so taking
C/CErn == 1414 £y == 0.07 Eq E. 93
:. A = 1 + 3n
8
[1_(~)2]
CErn
= 1 + 3x14 [1- (0.07)2 ]=6.2
8
EqE.94
At 100 mm from mid-height of the member, the maximum strain is 7Ey and Cmax == 1 X 10 4,
so taking
C/CErn == 147 Ey == 0.14 EqE.95
8 CErn
y
:.A=I+ 3n[I_( C )2]=1+ 3x7
8
~-(0.14)2 ]=3.6 EqE.96
To find the magnification factor, the effective length lefT needs to be found. Assuming that
the effective length remains the same, this can be calculated from the ratio of the deflections
at mid-length of the column and at the points 100mm either side.
266
The deformation arising from the second order effects described in Section E.2 is a half Sine
curve of length [elf. The deformation can equally be defined as a half Cosine curve with the
maximum displacement at mid-length of the column and the curve form:
EqE.97
Therefore, the deflection at 100mm from mid-length is
I:: _I:: C 7txlOO EqE.98

UIOO - Umax os
I elf
. C 7tX 100 _ 0100

.. os --1::- EqE.99
I elf Umax
from which [elf can be calculated.

b4
The elastic deflection of the wall was calculated in E.3.4 as 0 =_3_ PRc
384 EIwall
Therefore the total deflection, Bt, including the effects of plasticity is
I:: _ 3A PRc b4
Ut - EqE.100
384 E1wall
Using Eq E. 100 to calculate :100

max
Eq E. 101
AIOO= 3.6 from Eq E. 96

Amax = 6.2 from Eq E. 94
C 100 = 1xlO-4 from finite slice model

Cmax = 2x 10-4 from finite slice model
EqE.102
Cos
1txlOO = --.!9Q..
<>
= 0.290 = Cos( )
1.277rads EqE.103
1elf <>max
267
:. 1tX 100 = 1.277 Eq E. 104
leff
1tx100
:. leff = = 246 mm EqE.105
1.277
Substituting into Eq E. 78, t = ( 12Acr

E Yd J( £ J
1t b
eff
2
= (12X6.2X355
205000
3)(~)1302
384 246
=6.8mm Eq E. 106
Therefore the estimate of the thickness for wall stability is clearly greater than the thickness
required for Class 1 to BS 5950-1 which is 140/27.6 = 5.1 mm for a 140x140 SHS. The
thickness for wall stability is 6.8/5.1 = 1.3xCIass 1 thickness to BS 5950-1.
E.5.3 Example of lower end-rotation and lower curvature

To show the influence of the end-rotation, the calculation is repeated for the case of equal
end rotations (point of contraflexure at mid-height of column) with the end-slope equal to
0.020 radians bending about a rectangular axis.
Assuming no strain hardening:
1. at mid-height (the point of maximum curvature), the curvature reaches 0.58x10-4 and
the strain reaches 4.7 yield strain,
2. at 100 mm from mid-height of the member, the curvature is 0.41xlO-4 and
the strain is 3.5 yield strain.
At mid-height of the member, the maximum strain is 4.7ty, so taking
CyfC max = 1ty/4.7Ey = 0.21 EqE.107
EqE.108
At 100 mm from the point of maximum strain, the greatest strain is 3.5ty.
CyfC max = 1Eyf3.5ty = 0.29 Eq E. 109
-
268
J]
:. A=1+ 3; [1-( CC,:. =1+ 3X:.5 ~-(0.29)' 1= 2.20 EqE.110
AlOO = 2.2 from Eq E. 110

Amax = 2.7 from Eq E. 108
CloO = 0.41xlO-4
C max = 0.58xlO-4
• Bt,lOO = AlOOC 1OO 2.2x0.41xlO-4 = 0.576 Eq E. 111

Bt,max AmaxCmax 2.7xO.58xlO-4
1txlOO B
Cos = ---.!QQ... = 0.576 = Cos(0.957rads) EqE.112
f elf Bmax
:. 1tXlOO = 0.957 Eq E. 113

felf
:. f = 1tXlOO =328mm Eq E. 114

elf 0.957
Substituting into the Eq E. 78, t =

~ 2Adcr Y ( -1t-Jb 2
E Reff
t=
(
12Acr y d
E
J( Reff J
- 1t b2
= (12X2.7X355 3 )(~)1302 =3.4mm Eq E. 115

205000 384 ·328
This is less than 6.7mm required for single curvature with end-slopes of 0.040 radians. It is
also much less than the BS 5950-1 Class 1 limit of 5.lmm. Therefore, the simplified
analysis above suggests that it is sufficient for the columns to be Class 3 for equal end slopes
and 0.020 radians end slope.
E.6 Summary
From the analysis shown above it is clear that some columns can be Class 3 or Class 2 if the
end-rotations are low, but guidance is needed to say under what conditions they need to be or
Class 1 of even thicker walled than Class 1. Wall slenderness is investigated in Sections 5, 6
and 7 and design limits are given in Section 7.4 and repeated in Section 9.2.2. Generally this
will not be an issue because designers will almost always choose the thicker walled sections
269
available to allow them to use the smallest overall section size to be hidden in the walls of
the building with the greatest ease. There is also an incentive to use the thickest walls so that
the columns designed for least load, ie for the higher floors, are easily distinguished on site
by having smaller breadth and width than the sections with higher design loads.
270
APPENDIX F FINITE SLICE SPREADSHEET
F.1 Outline of spreadsheet calculations

F.1.1 Introduction
This Appendix describes the CMK. elasto-plastic analysis spreadsheet used to calculate the
axial compression resistance of columns for particular end rotations and end moments.
The spreadsheet allows the calculation of the resistance for any end rotations both in the
region between zero load and maximum resistance and in the region beyond maximum load,
often referred to as the "falling branch".
At October 2005, the calculation considers only flexural buckling. It does not consider
torsional or lateral-torsional effects. It is intended to add the calculation steps to consider
lateral-torsional effects in the future.
F.1.2 Overall method

The solution is iterative.
The core of the method is to find the axial compression that can be resisted at any cross-
section along the column by the moment induced by the curvature of the column, as shown
in Figure F.l. The axial load resisted, Nr (kN), at any section is equal to the moment, Mx or
My (kN/m), divided by the eccentricity from the line of thrust at that section, Da_mod (mm),
so that:
Nr
Deflected form
of column
Nr
Figure F.l Bending moment in column
271
The curvature of the column is the principal variable in the analysis. It is the variable that is
modified to start each new cycle of iteration. At the end of each cycle of the iteration, a
modified curvature is calculated from the difference between the moment induced by
curvature and the moment from axial load at the eccentricity between the line of thrust and
the centre line of the column at each section.
Cycles of iteration are repeated until the equilibrium at each section along the length of the
column is within acceptable limits of convergence. In early September 2005, a new method
was developed for improving the curvature and the convergence achievable is now closer
than 0.1%.
Variables at October 2005 are:
1. curvature
2. initial geometrical imperfection
3. residual stress (to date considered to be parabolic)
4. length
5. yield stress
6. elastic modulus
7. section properties
Strain hardening is not yet included. Strain hardening is an extremely complicated issue
because local buckling phenomena often reduce or cancel the benefits that would be
expected.
F.2 Calculation of the 'line of thrust'

The line of thrust is defined as the straight line off-set by the distance DN (mm) from the
undeformed column centre-line and passing through points at the ends of the member
defmed by the end-moment M_end (kN/m) divided by the applied axial load N (kN), so that
the ends:
DN = Mend xlOOO
N
The eccentricity for the deformed column from the line of thrust, Da_mod (mm) is:
Do_mod = Do - DN
For pin-ended columns with zero end moments, the line of thrust passes through the ends of
the column, because the end moments are zero.
272
The end moment is defined on the Sheet "Data" - at October 2005 it is the [joint+beam
stiffness] *[the end rotation of the column in excess of the end rotation of the beam if the
beam where simply supported].
The axial load is defined on "Sheet 1"
The eccentricity of the line of thrust from the column at the ends is calculated
The line of thrust at intermediate points is calculated by interpolation
The eccentricity of the column centre line from the line of thrust is found from the deformed
shape of the column.
F.3 Generation of the curvature and the deflected

shape
F.3.1 Outline
The initial geometrical imperfection DO, the initial curvature (ie the curvature from the initial
geometrical imperfection) KO, the curvature due to bending KflexF and the deflections due
to bending DflexF are calculated on Sheets "Mode Shape Y" and "Mode Shape X" and
exported to Sheet 1. On Sheet 1, the total deflection of the member Da is calculated from Da
= DO + DflexF.
F.3.2 Generation of initial imperfection and imperfection curvature

The initial geometrical imperfection and the imperfection curvature are generated on sheet
"Mode Shape Y" in columns E to K (I, J and K are not used at October 2005). At October
2005, it is assumed that the initial imperfection is a sine curve. The maximum displacement
is defmed on the sheet as a i in row 5.
F.3.3 Generation of curvature due to bending

Starting curvature
The starting curvature is most conveniently generated by using the imperfection curvature
plus a number of multiples of the imperfection curvature by setting U9 to the multiple
required. End moments are most conveniently included in the starting curvature by copying
them into column S. At October 2005, when end moments are being taken as equal and
opposite, the values are copied by CopylPaste Special- Values from S8 = M_end/(dMldK) =
M_end*1000000INI0 where NIO = dMldK copied from Sheet 1. For defined end slopes, the
value of U9 must be adjusted until the defined values of end slope, 9d, is generated. This can
be found by finding the end slope with curvatures from end moments alone, 9Mend, then
finding the end slope from imperfection curvature alone, 9;, and then finding the necessary
value of U9 from S45
273
U9=( 0. -:,- )-1
Calculation of Improved curvature
The subsequent values of curvature are generated with U9 = O.
On Sheet 1, after calculating the axial load that can be resisted, a curvature correction is
calculated for each section. This is done by finding the difference between the moment at
the section and the moment required at the section to resist the applied axial load.
The curvature correction is calculated from [curvature correction] = [difference in
moment]/[dMldK], in which [dMldK] can either be the secant value or a value nearer to the
tangent stiffness as the user chooses. The ratio dMldK is calculated on Sheet 1 in column 00
as (My - old My)/(KflexF - old K) in which 'old My' and 'old K' are in column OM and
column ON. They are imported manually from column OV and column OW into which they
were imported from column OK and column OL. For the secant value of dMldK, OM and
ON are set to zero. For a value nearer the tangent dMldK, OM and ON are generated from a
cycle in which the curvature KflexF is reduced on sheet Mode Shape Y by using a value in
Z9 less than 1.0 so that the K values in column Z (used to generate the output from Mode
Shape Y) are reduced.
The curvature corrections are exported to sheet "Mode Shape Y" into columns P and Q.
When ready to iterate on Mode Shape Y, these values in P and Q are copied into R and S. T9
must be set to 1.0 if 1.0 times the correction is to be used in the iteration. T41 shows the sum
of the curvature corrections from Sheet 1. If this is not zero, the end slopes will change
unless a further correction equal and opposite to T41 is also added. This further correction is
applied in column W, which adds column V times -T41 at each section. The sum of column
V should be 1.0 and the sum of column W should be equal to -T41. The values of V are
entered manually, generally at the central section alone or shared by the central three
sections.
F.3.4 Generation of the deflected shape due to bending

Outline
The deflected shape due to bending is calculated on sheet Mode Shape Y, columns AA to BL.
The method is, on each section in turn, to
1. calculate the change of angle of the member, N j , at each section from the curvature
times segment length,
2. impose the change of angle on the member at that section, as Figure F.2.
3. calculate the displacements at all the sections from the change of angle at this section
-
274
4. sum the displacements from all the changes of angle
5. correct at each section for the difference between change of angle at a point and
curvature, as Figure F.3.
275
Internal sections, sections 2 to 30 inclusive:
~
~.
Figure F.2 Displacements of nodes from curvature at node "i"
¢; = K; X s; where 's/ is the segment length at section 'i'
Zt; = (L - xl';)¢;
z, =zfi )=[(L-XY;ltPJ (Xi )=(L-XY;{ xn~, =(L-XY;{ xi )K,S,

In this model, the segments are all of equal length s = L , where 'n' is the number of
n
segments
The deflection at other sections, 'j', due to the curvature Ki at section 'i' is given by
..
J<z-+z.=z.
J I.
- (j)
Z
j=i-+z.J =z.I
..
J>z-+z.=z.
J I
- -• (n-n-zj)
where
n = the number of segments of the member = 30,
in the spreadsheet, n = nsegments,
i = the number of segments from the bottom of the column at which the change of angle
occurs
-
276
in the spreadsheet =
nofsr -1 = number of section (for rows), from 1 to 31, -1.
j = the number of segments from the bottom of the column at which the deflection is
calculated
nofsc - 1 = number of section (for columns), from 1 to 31, -1.
The formula in each cell for all the internal sections (sections 2 to 30) and the end sections
(sections 1 and 31) is:
=KYr*(IF((nofsc-l )«nofsr-l ),zir*(nofsc-l )/(nofsr-l ),IF((nofsc-l )=(nofsr-
1),zir,zir*((nsegments+ I-nofsc)/(nsegments+ I-nofsr»»)
The deflections calculated as above for the change of angle at each section are summed to
find the total deflection at each section. This sum is an over-estimate of the deflection
because it assumes that the change of angle occurs as an abrupt change at a point, whereas it
occurs as a curve over the length of a segment, as shown in Figure F.3:
Figure F.3 Correction of displacements
=-KYc*(Llnsegments )A2/8
277
Ends of member, sections 1 and 31
<Pe sl <Pesl
~
~ ~
·I ze .I .I . ze .
.I I I
"-- :--r--r---!d
I I
· -- . ----
---T--
I
/2
s
~1r
I
--- !
.I
--,--r---
I
. .I
I
r-- i ----
·i .i s/2
---.+---
· --.+-----
. ---.+-- l--'+-----~-L.
·II .II I I
L
I I
Figure FA Member shape at ends
At the ends of the member, the change of angle is shown in Figure FA.
¢>.e = Ke 2 (!...)
therefore the offset of the straight line projected to the end of the member would be Ze:
==-KYc*(Llnsegments)"2/8
It should be noted that the spreadsheet only considers no-sway columns, so 'z' at the end of
the column = 0,
The deflection at other sections, 'j', due to the curvature Ke at the end is given by
Xj < x, -Hi = z{:: ). which gives the deflections due to curvature at the top
x j > Xi ~ Z j = Z i
L-X_) which gives the deflections due to curvature at the bottom
( L-x-, J,
where
n == the number of segments of the member = 30,
in the spreadsheet, n = nsegments,
i == the number of segments from the bottom of the column at which the change of angle
occurs
278
nofsr -1 = number of section (for rows), from 1 to 31, -1.
i = the number of segments from the bottom of the column at which the deflection is
calculated
nofsc - 1 = number of section (for columns), from 1 to 31,-1.
The formula in each cell for all the internal sections (sections 2 to 30) and the end sections
(sections 1 and 31) is:
=KYr*(IF«nofsc-l )«nofsr-l ),zir*(nofsc-l )/(nofsr-l ),IF«nofsc-l )=(nofsr-
1),zir,zir*«nsegments+ I-nofsc)/(nsegments+ I-nofsr»»)
The values of deflection due to the curvature concentrated at each section are summed in
column BI:
31
Sumzi= LZ1
The value of the deflection at section ';' due to flexure allowing for the curvature dz; is given
in column BL:
DiF = Sumzi + dz;
The values of DiF are copied into the column BS via BQ and exported automatically from
BS into Sheet 1 in which they are the values ofDflexF, the deflection due to flexure.
F.3.S Calculation of end-slope

The end slope is calculated from the sum of:
the slope due to the deflection of the internal sections
the slope from the curvature at the end sections.
The slope due to the deflection of the internal sections is given by the slope of a straight line
from the end of the member to the penultimate section
= Zptis, where Zpt = SurnziYat the penultimate section at the top of the member
= ZpJs, where Zpb = SurnziY at the penultimate section at the bottom ofthe member
see Figure F.5.
The slope from the curvature at the end sections = cl>er shown in Figure F.5.
cl>er=cI>e-cI>epwhere ¢e =Ke(~)and ¢ep = z;
279
z~ ~
.--f5\----------
._-+- -------- =ts ~
<Pe S/
I
----,.----
I
------ ~
.I Ze .I .I
I I I
.-. i\.:,-r- T----ld/
2
.--t-\;- -- ' ---- S
i, \., i, s/2
._-+-- -- ---
e
I
I
I
.--~~
--- ------
.--~- -------- ~
.__ I __________ --XS
z~ ~
Figure F,5 End-shape from curvature in end sections
F.3.6 Calculation of next shape

On Sheet "Mode Shape Y", the next shape is generated as follows:
1. the total change of slope at all 31 nodes is found by summing dK
2. for solutions with a defined (equal and opposite) end slope, an equal and opposite total
curvature is assigned to the nodes so that there is no change to the end-slopes. This
assignment is done manually according to the deformed shape. Generally the most
efficient method is to assign all this to the central nodes, often assigning all to the
central node.
3. the new curvature from flexure is generated by adding
the previous values of curvatures from flexure, KflexY from Sheet 1,

the change in curvature, dKY from Sheet 1,
the manually assigned equal and opposite curvatures
4. the new deflections from flexure arising from the new curvature from flexure are
generated
5. the new curvature from flexure and the new deflections from curvature are exported to
Sheet 1
280
F.4 Calculation of the moment induced by curvature
The calculated moment allows for the effects of plasticity and for the effects of residual
stresses and of the axial load.
The calculations, performed for each of 30 sections along the length of the section, are as
follows:
I. Calculate the total stresses from curvature + axial load + residual stresses, assuming the
section remains elastic
2. Calculate the elastic-plastic stress diagram
3. Calculate the moment about the centre-line of the column limiting the compressive
stresses to yield
4. Check that the calculated tensile stresses do not exceed yield.
Step 4 is required because at October 2005, the spreadsheet does not handle tensile strains in
excess of yield strain.
F.4.1 Calculate the total "elastic" stresses

Calculate the res/dual stresses
The residual stresses are defined on Sheet I and a stress distribution is calculated on Sheet
"res stress"
Input data are:
I. fr/fy, the ratio of maximum compressive residual stress to the yield stress
2. Aw, the web area, which appears because the spread sheet is derived from a spreadsheet
for I-sections; for SHS, Aw = 0 but it needs reconsidering for RHS
frmin=-fr/(2+ 3Aw/4tfY)
a=(fr-b')N"2
It is assumed that the stress distribution is parabolic, with compression at the comers of the
section, as shown in Figure F.6.
tfr "1"f r,max
ffr,min
Figure F.6 Residual stress distribution
Assuming that the residual stress distribution is parabolic
IT =ay 2 +IT min
281
At the edge ofthe section, I r max = aY 2 +1 r min
:.ay2 =/rmax -/rmin
:.a= (frmax -/rmin)

y2
The total force on the flange is:
But the total force from residual stresses = 0, so:
: . /r min = Ir max for a Square Hollow Section

2
Calculate the "elastic" stress from axial compression
The "elastic" stress from axial compression I a =I ave = N ,as shown in Figure F. 7.
Ag
where N is the applied axial compression

and Ag is the gross area ofthe section
Figure F.7 Average compression stress
-
282
Calculate the "elastic" stress from curvature
y
-x ....c:c----'--->~x
p2 ...--_-----=f2=-;-_ _ _.p1 x
y
13
-x ----- f1 --- x
---------r--------- --
i
I
I +
I -y
p3 f4 I p4 ~---' }!bX
-y
-y~x
Figure F.8 "Elastic" bending stresses
The "elastic" stress from curvature, as shown in Figure F.8, is given by:
./bx = E x y X K flex Fx for curvature about the x-x axis

./by = E x x X K flex Fy for curvature about the y-y axis
where E is Young's modulus
y is the distance from the elastic neutral axis parallel to the x-x axis
x is the distance from the elastic neutral axis parallel to the y-y axis
KflexFx is the curvature due to flexure about the x-x axis
and KflexFy is the curvature due to flexure about the y-y axis
F.4.2 Calculate the elastic-plastic stress diagram
The elasto-plastic stress diagram is shown in Figure F.9 for one side of a member of total
width = 2X. The procedure followed by the spreadsheet is:
1. Find where the "elastic" stresses exceed the yield stress.
2. Limiting the stresses to yield stress in compression, calculate the axial load resisted
3. Find the axial load error between the design axial load and the axial load resisted
4. Find the increase required of the axial stresses to correct the axial load error
5. Iterate to correct axial compression
Foran SHS
Find where the "elastic" stresses exceed the yield stress
The "elastic" stress is the stress the would occur if there were no plasticity. This is the strain
multiplied by the Young's modulus = tE.
283
The "elastic" stress along any side is the sum of:
the residual stress which along faces 2 and 4 isl r = ay 2 +1 r min
the mean axial stress/a = lave = N

Ag
the bending stress which is I b = E x K flexFy XX + E x K flexFx X Y

along faces 2 and 4, the "elastic" bending stress from curvature about the x-x axis is:
fbx =±/bxY =ExKflexFx(±Y)

:·/ b =E X KflexFy xx +E X KflexFx x(±Y )=E X K flexFy xx ±/bxY
The general case of the "elastic" stresses is:
along faces I and 3,/e = (rr min + ay 2 )+ la ±/byX + E x K flexFx xy
along faces 2 and 4,/e = (rr min + ax 2)+ la ±/bxY + Ex K flexFy XX
The "elastic" stresses at each of the comers pI, p2, p3, p4 are fepl, fep2, fep3, fep4, where
f ep = la ±/bxY ±/byX +1 rmax

eg, fep I =fa+tbxY+tbyX +frmax
the "elastic" stresses at mid-length of each of the sides fl, fl, f3, f4 are fefl YO, feflXO,
fef3YO, fef4XO, where:
eg feflXO=fa+tbx Y +frmin
The coordinates of the point of minimum "elastic" stress in the plane of each side, which
may occur outside the section, are yfminfl, xfminfl, yfminf3, xfminf4.
For the minimum stress along face 2, d if e ) = 0,

dx
dx --~J'f
:.0= d(fJ dx 1\ rmin +ax 2) + f a +f·
- bxY + E X K flexFy xx }
:.0 = {(O+ 2ax)+O±O+ Ex K flexFy }
:. 2ax + E x K flexFy = 0
. Ex K flexFy
•• X min! 2 =- 2a
The minimum "elastic" stress in the plane of each face is fminfl, fminfl, fminf3, fminf4,
Where:
along face 2, I min 12 = {rr min + ax min 12 2 )+1 a +1bxY + E x K flexFy X X min 12
the coordinates of the points at which the "elastic" stress reaches yield in the plane of each
face, which may occur outside the section, are ypfyfl, ynfyfl, xpfyf2, xnfyf2, ypfyf3,
Yllfyf3, xpfyf4, xnfyf4.
284
If fminf is greater than the yield, fy, then the value of are ypfyfl, ynfyfl, xpfyfl, xnfyfl,
ypfyf3, ynfyf3, xpfyf4, xnfyf4 is taken as 0 because this value is used to define the extent of
the plastic zone.
Along face 2, where I min 12 ~ I y the point at which the "elastic" stress reaches yield is
2
calculatedfrom/e =/y = (rrmin +axfyf2 )+/a +/bxY +ExKflexFy xXfyf2
2
:.(rrmin +ax fy12 )+/a +/bxY +E X KflexFy XXfyf2 - / y =0
Re-arranging in the classic solution of a quadratic equation:
2 - b ± ~ b 2 - 4ac
ax +bx +c = 0 ~ x = - - - - ' - - - - -
2a
:.axfyf/ +ExKflexFy xXfyf2 + (-/y +/rmin +/a +/bXY)=O
- (E x K flexFy ) ± ~ (E x K fleXFy)2 - 4a(- I y +1 r min +1 a +1bxY )

:. x fyf2 =
2a
-(EXKfleXFy)+~(EXKfleXFyF -4a(-/ y +/rmin +/a +/bXY)

:. X pfy12 =
2a
- (E x K fleXFY)- ~(E xK flexFy F - 4a(-/y + Irmin + la + IbxY)

:. X nfyt2 = 2a
xpfyfl=IF(fminfl>fy,O,( -(E*KflexFy)+SQRT«E*KflexFy)A2-4*a*(-
fy+frmin+fa+tbxY»)/(2*a»
xnfyfl=IF(fminf2>fy,0,(-(E*KflexFy)-SQRT«E*KflexFy)A2-4*a*(-
fy+frmin+fa+tbxY»)/(2 *a»
The limit of the elastic zone (p = positive, n = negative) along each face is defmed as ypecfl,
ynecfl, xpecfl, xnecfl, ypecf3, ynecf3, xpecf4, xnecf4
If the "elastic" stress does not reach yield within the section, then the values of ypecfl,
ynecfl, xpecfl, xnecfl, ypecf3, ynecf3, xpecf4, xnecf4 are taken as the distance to the edge
of the section from the neutral axis as follows:
ypecfl =Y, ynecfl =-Y, ypecf3=Y, ynecf3=-Y
xpecfl=X, xnecfl=-X" xpecf4=X, xnecf4=-X
285
If the "elastic" stress does reach yield within the section, then the values of ypecfl, ynecfl,
xpecf2, xnecf2, ypecf3, ynecf3, xpecf4, xnecf4 are as follows:
ypecfl=ypfyfl, ynecfl=ynfyfl, ypecf3=ypfyf3, ynecf3=ynfyf3,
xpecf2= xpfyf2, xnecf2 =xnfyf2, xpecf4= xpfyf4, xnecf4 =xnfyf4,
Where the "elastic" stress exceeds yield throughout the length of the side, the IF statements
will use the values for reaching yield, eg xpfyf2 and xnfyf2 which will be =0 if the "elastic"
stress exceeds yield throughout the length of the side.
==1F(xpfyf2> X,X,(IF(xpfyf2<-X,-X,xpfyf2»)
==1F(xnfyf2<-X,-X,(IF(xnfyf2>X,X,xnfyf2»)
r --- --- - --- -- - •.
l
00
00
00
00
00
00
ri------- --
00 0 f
:': : ap1
, 00
, 00
, 00
, 00
I I / I I
: I,':
- ------,--
- - -- - - - -:- - -
I
o fa
Figure F.9 Stress diagram for one wall of width 2X
Calculate the axial load resisted

The axial compression resisted by the plastic zones, if any, are defined as Nefl, Nef2, Nef3,
Nef4.
The force is the integral of the stress over the elastic area (which is the area where the stress
is less than yield).
N ef = Jr e x dA where f e = (r r min + ax 2 ) +f a ±f bxY + E x K flexFy XX

Ae
The extent of the elastic zone depends on the "elastic" stresses along each face.
As shown in Figure F.10(a), the stresses in the section might all be less than yield, so the
extent is from -X to +X or to -Y to +Y. Alternatively, the "elastic" stress may reach yield
within the width of the section, giving a reduced extent of the elastic zone, as shown in
Figure F.lO(b) and (c). The variables ypecfl, ynecfl, xpecf2, xnecf2, ypecf3, ynecf3,
Xpecf4, xnecf4 are so defined as to give the limit as either the edge of the section or the point
- 286
at which the "elastic" stresses reach yield. Writing the positive extent as "xp", eg instead of
xpecf2, and the negative extent as "xn", eg instead ofxnecf2, the integral can be written:
xp xp
N ef = fife} x (I x dx ) = I f~ 2 + (E x K flexFy ~ + if r min + f a + f bxY )}Ix

xn xn
P
lX
=1 [ a~+(E XKfleXFyr2 +ifrmin +fa +fbxY)x xn
3 2
=I [ a X p
3
- X n
3
3
(
+ E x K flexFy
)X P
2
- X n
2
2
+
if r min + f a + f bxY )X p( )]
- X n
eg for face 2, Nef2=tf*(a*«xpecf2}"'3-(xnecf2}"'3)/3+E*KflexFy*«xpecf2}"'2-

(xnecf2}"'2)/2+(frmin+fa+fbxY)*(xpecf2-xnecf2»/IOOO
The axial compression resisted by the plastic zones, if any, are defined as Npp 1fl, Npp 1f2,
Npp2f2, Npp2f3, Npp3f3, Npp3f4, Npp4f4, Npp4fl. Where there is a plastic zone, the force
is the integral from the end of the elastic zone to the edge of the section.
N p = Jr y x dA = Jr y (I x dx )
Ap Ap
This is calculated using the variables ypecfl, ynecfl, xpecf2, xnecf2, ypecf3, ynecf3, xpecf4,
xnecf4 and the distances to the extreme fibres of the section along the side considered, + X, -
X, +Y, -Yo For example, on face 2 and at point 1,
x x
N pplf2 = Jry (I xdx ) =fy XI fdx =fy xt[x ]~Cf2 =fy XI (X -xpecf 2)
xpecJ 2 xpecJ 2
The axial force on all the plastic and elastic zones that is actually resisted by the stress
diagram derived above is defined as Na:
N =LNa
ij
ppifj + LN
j
efj
287
fyl------------~'-------
-i-+-~'!------------;-------------------!/.-
I I " • I
: : :: I :
1 : ~ i l l
:' ---~---~
------------ ..... : :: i 1
I: x;ecf :1: XtCf:1 I: ~necf :1: xXecf il

(a) All (b) Two plastic zones
I
"
n"
,
,/ i
,,':,
I
---------1--,/ i
fy l----·----'::~'-'~-----------
~ .......................... --- --"""
------------------------------------"-"-''''''''''~''''''''''''-------------
(c) One plastic zone
Figure F.IO Possible stress distributions
Find the error In the axial compression actually on the section

dNa = the difference between the design load, N, and the force actually resisted by the stress
diagram above load, Na.
dNa =N -N a
Calculate the correction required for the axial compression
To increase the axial force resisted by the stress diagram, the axial strain must be increased.
This is shown below in terms of a corrective compressive "elastic" stress, fc, which equals
the corrective compressive strain, EC, multiplied the Young's modulus, E. The spreadsheet
Uses fc and does not explicitly use EC x E.
-
288
i
------ - •••• J••• _ •••
I
dri>-. ~ ~f
I1
fy I
i dx dx I
dxL • dxR ! I
i1 I1
I1
I
I
x x I I -xnecf2 xpecf2 I
I 1
I
I.
"I'- ~ I
• ~
I I1
:..
X
~ . X
~
I
I
Figure F.Il Correction of axial force resisted
The increase in the axial force resulting from the "elastic" stress fc is shown by the shaded
area in Figure Ell. The areas either side are at yield stress, so there is no increase in axial
force resisted from the increased strain. Assuming the sides of the shaded area are straight
lines at the slope of the stress diagram at the top of the shaded area, with slopes of -dfL/dxL
and +dfR/dxR, then
dN =lex( 1 [dx
-Ie --let I [dx
\.
peefl - X neefl ' -
2 dl + -Ie
2
--let
dl R
R
)
L
L )
For convenience write in terms of "y" instead of "x" for faces 2 and 4 and "y" for faces 1
and 3
IdN ={- ~2 I[[~)

f I dl + [dxdl )]t}fe + {I(x
f I R
L
L
2
f I
peefl -x }}fe neefl
Writing A, ~ {~ (x ",a - x "~a } } ,

To limit the number of columns in the spreadsheet, which is an important issue for a
manageable spreadsheet, a substitution of an approximation is used as follows:
sumdybydl = {I[[dxdl f I
R )
R
+ [dxdl L
L
J]}, where sumdybydf = 1
mod djbydy
where
289
moddjbydy = sum4fbydy = {~[( f: Hf:)]} but taking mod djbydy =0.01 to
retain mathematical stability if sumdjbydy < 0.01
The substitution amounts to taking f df

/1 dx
= ~1
~ dx
which works well as an approximation
/1 df
because of the shape of the stress diagram. There is no inaccuracy from this approximation
because of the iterative procedure.
/4
Also writing L dN = dNa
/1
Then
-~{sumdybydf}tf/ + Aefc -dNa =0
:. !2 {sumdybydf }tf/ - Aefc + dNa = 0
Re-arranging in the classic solution of a quadratic equation
2 -b±~b2 -4ac
ax +bx+c=O~X=------
2a
Ae ± A/ -4[~(sumdybYdf~ JdNa }
fc=----~~~--------~~-----
2[~(sumdybydf~]
If there is no plasticity, then there is no effect of the slope of the stress diagram and the
change of stress is equal to fcapp, where
dNa
J.capp=A
e
The formula for fc in the spreadsheet is written

==1F(sumdybydf=O,fcapp,(Aea-SQRT(Aea"2-
4*(tf'i'sumdybydf/2)*(dNa*1000»)/(2*(tf'I'sumdybydfl2»)
The values of dy/df and dxldf are found by inverting the values of dfldy and dfldx where the
latter are found for example by
-
290
!
d~) = {Vrmin +ax 2 )+ fa ±fbxY +ExKflexFy xx}={(0+2ax)+0±0+EXKflexF)T
herefore dUe) =2ax + Ex K flexFy and dUe) =2ay + Ex K flexFx calculated at the points
dx dy
where the stress diagram reaches yield, ie at ypecfl, ynecfl, xpecf2, xnecf2, ypecf3, ynecf3,
xpecf4, xnecf4.
Iterate to correct axial compression

The final step in the iterative procedure is to calculate faplusfc = the previous "elastic" axial
stress, fa, plus the correction, fc. For subsequent iterations, the calculations in 4.2.1 to 4.2.4
above are repeated after feeding faplusfc into fa for the next iteration of finding the stress
diagram. The criterion to show conversion is that the axial force resisted, Na, is equal to the
design axial force, N,at all sections. At October 2005, this has always been achieved to less
than IkN in 3 iterations after the first set of calculations.
Check that the tensile stresses do not exceed yield.

Check that the calculated tensile stresses do not exceed yield because, at October 2005, the
spreadsheet does not handle tensile strains in excess of yield strain.
Check is done on the minimum offepl, fep2, fep3, fep4 and feflYO, fef2XO, fef3YO, fef4XO
F.4.3 Calculate the moment resisted by the column

The moment is calculated about the centre-line of the column using the "elastic-plastic"
stress diagram calculated in 4.2 above, in which the compressive stresses do not exceed
yield.
The moments from plastic zones are calculated from the lever-arm multiplied by the force, as
shown in Figure F.12. For the unshaded areas, the contribution to the bending moment is
calculated from the axial force resisted x the lever-arm from the geometrical centres of the
area. For the shaded areas, it is calculated by integration about the axis of bending.
291
Figw·e F.12 Calculation of bending moment resisted
The lever arms from the centre of force of the plastic zones about the neutral axis of the
column are defined as laxNpplfl , layNpplf2, layNpp2f2, laxNpp2f3 , laxNpp3f3 ,
layNpp3f4, layNpp4f4, laxNpp4fl , where lax is about the x-x axis, lay is about the y-y axis.
My = the total moment about the y-y axis
MY=Myfl +Myf2+Myf3 +Myf4
Myfl , Myf2, Myf3 , Myf4 = the moments about the y-y axis from faces 1, 2, 3, 4
Myfl =(Npplfl +Nefl +Npp4fl)*XlI 000
Myf2=Meyf2+layNpp If2*Npp 1f21l 000+layNpp2f2*Npp2f2/1000
MYf3 =(Npp2f3+Nef3+Npp3f3)*( -X)/ 1000
Myf4=Meyf4+layNpp4f4*Npp4f4/1000+layNpp3f4*Npp3f4/ 1000
Meyf2, Meyf4 = the moments about the y_y axis from the elastic zone on each face
Meyf2=tf*(a*((xpecf2),,4-(xnecf2),,4)/4+E*KflexFy*((xpecf2),,3-(xnecf2)"3)/3
+( frmin+fa+fbx y) *((xpecf2)"2-( xnecf2)"2)/2)/ 1000000
Meyf4=tf*(a*((xpecf4),,4-(xnecf4),,4)/4+E*KflexFy*((xpecf4),,3-(xnecf4)"3)/3
+(frmin+fa-fbx Y)*((xpecf4)"2-(xnecf4)"2)/2)/ 1000000
------------------------~-----------------------------
292
Similar expressions are used for the moment about the x-x axix
Mx = the total moment about the x-x axis
Mxfl, Mxf2, Mxf3, Mxf4 = the moment about the x-x axis from faces 1,2,3,4
Mexfl, Mexf3 = the moments about the x-x axis from the elastic zone on each face
The derivation of the moment from each of the elastic zones in the plane of the curvature,
faces 2 and 4 shown shaded above, is as follows:
p p p
Me = JXfe dA = JXfe(txdx) = t Jxfe dx
n n n
2
along faces 2 and 4, fe = Vrmin + ax )+ fa ± fbxY + Ex K flexFy X X
J
n
2
:. Me = t X[Vrmin + ax )+ fa ± fbxY + Ex K flexFy X x}Ix
2
= tJx[ax + ExK flexFy Xx+ (frmin + fa ±fbxY)}Ix
n
3 2
= tJ[ax + Ex K flexFy xx + (frmin + fa ± ibxy)x}Ix
n
4 4 2 2
(x p - Xn ) (x / - x/) ( ) (x P - Xn )]
=t a + Ex K flexFy X + frmin + fa ± fbxY --"----
[ 4 3 2
293
APPENDIX G TABLES OF RESULTS OF
PARAMETRIC STUDY
G.1 Introduction
This Appendix gives the results of the results from the parametric study of the compression
resistance of the columns. All the results are for 140x 140x 10 SHS sections in S355 steel.
G.2 Mid-height displacements

Figure G I shows the mid-height displacements for members with an initial imperfect in the
+Y direction and imposed end rotations in different planes. The figure includes a list of the
Abaqus analysis runs used the generate the mid-height displacements.
-.~ - --_. 40 .--~
1
1- .:"0
i
1- 1--- c-- -3& ---- _.
~J
or ~ -.-SHS3000_X+Y_45(22L2009m09dO
'"
'\ II 2vl
,
I'V
~
I
I 0
_... - .i. .11 /
~, ~<
--_. f- 1&
f-
r :t ;I ~
~
~ ~~~ i
-:~Q. -, ~~,-is.A ~ 0 ~ p
,
~ 3b
~
f-· !
.... J .J
,.
.J
~SHS3000_X+Y_O_2009m09d03(04)
vl
I
J
l
.n
_v
.-.. f----- -------

'"
i I .n
.v
..... sym
X",· ..:::.
.r
mm
SHS3000_X+Y_90_2009m09d04vl
Figure G.I Mid-height displacements - rotations at 22.5 0 intervals
294
G.3 1.5m columns
The results of the Abaqus analysis for columns of 1.5m length are presented in
Table G I for the initial imperfection and end-rotations in the X = 0 plane.
Table G2 for the initial imperfection in the X = 0 plane and end-rotations in the Y = 0 plane.
Table G3 for the initial imperfection in the X = 0 plane and end-rotations in the X = Y plane
by imposing rotation about X = Y axis, rotation forcing displacement in the same direction as
the initial imperfection.
Table G4 for the initial imperfection in the X = 0 plane and end-rotations in the X = Y plane
by imposing equal rotations about X and Y axes, rotation forcing displacement in the
opposite direction as the initial imperfection. This case is not a worst case so is not
especially interesting.
Table G5 lists early analyses performed for the 2007 SSRC paper [King 2007] for the initial
imperfection in the X = 0 plane and end-rotations in the X = Y plane by imposing equal
rotations about X and Y axes. The results of these analyses were somewhat uncertain
because the maximum load was that at the point where the end moment passes through zero
from destabilising to stabilising, but this occurred at different loads for the two axes of
enforced rotation, the X and Y axes. Therefore these results are inferior to those of
Table G3.
295
Table G.1 l.5m column, rotation in x=o plane
Initial imperfection in the X =0 plane in Y +ve direction
End rotations in the plane of the initial imperfection and in the ~ direction as the
initial imperfection
Max from URI = RF3 = end of End

SHS2006m 1Od22_130by 10_by 1500v 1 end rot compn analysis moment
All results other from (radians) (kN) (kN-m)
SHS2006ml0d22_130byl0_byI500v2
Max pin-ended 0.00220 1769.8 buckled o(pinned)
0.00291 1767.1 continued o(pinned)
296
Table G.2 1.5m column, rotn in Y=O plane
Initial imperfection in the X = 0 plane in Y +ve direction
End rotations Qemendicular to the plane of initial imperfection
URI = RF3= end of Endmmt

end rot compn analysis kN-m
(radians) (kN)
SHS2006m 10d24_130by 1Oby 1500 0.00481 1730.7 buckles 2.18
_2planes_vl destabilising
SHS2006m 10d24_130by 1Oby 1500 0.00942 1696.1 buckles 0.45

_2planes_v2 destabilising
SHS2006m 1Od24_130by 1Oby 1500 0.01386 1663.6 continues 0

_2planes_v3
SHS2006m 10d24_130by 1Oby 1500 0.01797 1617.6 ditto 0

_2planes_v4
SHS2006m 1Od24_130by 1Oby 1500 0.03252 1465.2 ditto 0

_2planes_v5b
End moment zero by interpolation
Table G.3 1.5m column, rotation in the X=Y plane (1)
End rotations at 45° to the plane of initial imperfection with the component in the plane
of the initial imperfection in the !!!!!!! direction as the initial imperfection
max end rot RF3= end of End

inX=Y compn analysis moment
plane (kN) (kN-m)
(radians)
SHSI500_X+Y_45_2009m09d27vl 0.00905 1659 end Oby
moment interpola
changing to tion
stabilising
SHS 1500- X+Y-45-2009m08d30v2 0.01695 1559 ditto ditto
SHS 1500_X+Y_45_2009m08d30v4 0.02165 1508 ditto ditto

SHS 1500_X+Y_45_2009m09d28v2 0.02418 1483 ditto ditto
SHS1500- X+Y-45-2009m09d29v3 0.03096 1424 ditto ditto
297
Table GA 1.5m rotation in the X=Y plane (2)
of the initial imperfection in the ol!l!osite direction as the initial imperfection

(radians) compn analysis moment
(kN) (kN-m)
SHS1500- X+Y-45-2009m08d30vl 0.01520 1566 end Oby
moment interpol
changing to ation
stabilising
Table G.5 1.5m column, imperfection in X =0 plane, rotn in X = Y plane (3)
Column end-rotations at 45° to the plane of initial imperfection
Cor_130by 1Oby 1500 max end RF3= end of analysis End

Jotn_in_ 2planes rot compn moment
_ 2009m08d l6.xls/plots (radians) (kN) (kN-m)
SHS- X+Y- 2009m08d13vlb 0.00997 1620 end moment Oby
changing to interpola
stabilising tion
SHS- X+Y- 2009m08d13v2 0.02014 1474 ditto ditto
SHS- X+Y- 2009m08d16vl 0.03034 1326 ditto ditto
SHS- X+Y- 2009m08d16v2 0.04032 1188 ditto ditto
298
G.4 3.0m columns
The results of the Abaqus analysis for columns of3.0m length are presented in:
All the tables, except Table G10 and Table Gll, have the initial imperfection in the X = 0
plane with the imperfection in the +Y direction. Table GlO and Table Gll have the initial
imperfection in the X + Y plane and the imperfection in the X & Y +ve direction.
Table G15 lists early analyses performed for the 2007 SSRC paper [King 2007] for the
initial imperfection in the X = 0 plane and end-rotations in the X = Y plane by imposing
equal rotations about X and Y axes. The results of these analyses were somewhat uncertain
because the maximum load was that at the point where the end moment passes through zero
from destabilising to stabilising, but this occurred at different loads for the two axes of
enforced rotation, the X and Y axes.
Table G.6 3.0m column, rotation in X;:::O plane
End rotations in the plane of the initial imperfection and in the ~ direction as the
Max from URI = RF3= end of End moment

SHS2006m 1Od09_130by 10_ end rot compn analysis (kN-m)
vI (radians) (kN)
All other results from
SHS2006m 1Od 10_130by 10_
v2
Max pin-ended 0.00459 1582.8 buckled o(pinned)
299
Table G.7 3.0m column, rotn in Y=O plane
End rotations Qemendicular to the plane of initial imperfection
UR2 = RF3 = end of End moment

end rot compn analysis (kN-m)
(radians) (kN)
SHS2006m 1Od 13_130by 10_v4 0.00451 1556 buckled destabilising
SHS2006ml0dI3_130byl0_v6 0.00662 1522 buckled destabilising
SHS2006ml0dI3_130byl0_v5 0.00861 1483 buckled destabilising
SHS2006m 1Od 13_130by 10_v3 0.01313 1404 buckled destabilising
SHS2006ml Od 13_130by 10_v2 0.02422 1257 buckled destabilising
SHS3000- X+Y-90- 0.03197 1169 end moment Oby

2009m09d12v4 changing to interpolation
stabilising
SHS2006m 1Od 13 _130by 10_v7 0.04012 1078 ditto ditto
300
Table G.8 3.0m column, rotation in the X=Y plane (I)
of the initial imperfection in the ~ direction as the initial imperfection
max end RF3= end of End

rot inX=Y compn analysis moment
plane (kN) (kN-m)
(radians)
SHS3000- X+Y-45-2010m03d14v2 0.00516 1548
SHS3000_X+Y_45_201Om03d14vl 0.01552 1345
SHS3000_X+Y_45_2009m09d12vl 0.03063 1119 buckled 5.89 kN-m
destabilising
SHS3000- X+Y-45-2010m03d14v3 0.04041 1010
SHS3000- X+Y-45-201Om03d14v4 0.06342 846
of the initial imperfection in the ol!l!osite direction as the initial imperfection

inX=Y compn analysis moment
plane (kN) (kN-m)
(radians)
SHS3000_ X +Y_45_2009m08d22vl 0.02206 1254.8
SHS3000_X+Y_45_2009m08d25vl 0.03092 1129.5
SHS3000_X+Y_ 45_2009m08d27vl 0.04930 946.0
SHS3000- X+Y-45-2009m08d29v2 0.07243 800.1
301
Table G.lO 3.0m column, rotn in the Y=X plane (3)
Initial imperfection in the X = Y plane in X & Y +ve direction
End rotations in the plane of initial imperfection (Y=X plane) and in the ~ direction
as the initial imperfection
Axial force/shortening only, no applied end rotations
initial imperfection in both Y=O max end rot RF3 = end of End
plane and X=O plane = 0.75L11000 about either compn analysis moment
axis (kN) (kN-m)
(radians)
max axial load 1584.2 buckled
SHS3000- X +Y-45-2009m08d30v5
Table G.ll 3.0m column, rotationn in the Y=X plane (4)
Initial imperfection in the X = Y plane in X & Y +ve direction
End rotations in the plane of initial imperfection (Y=X plane) and in the ~ direction
as the initial imperfection
initial imperfection in both Y=O max end RF3= end of End

plane and X=O plane = 0.75L11000 rot about compn analysis moment
either (kN) (kN-m)
axis
(radians)
SHS3000- X+Y-45-2009m09dO 1v6 28.50 1239.2 end Oby
moment interpolation
changing
to
stabilising
SHS3000- X +Y-45-2009m09dO 1v5 46.61 1033.5 ditto ditto
302
Table G.12 3.0m column, rotation in X = Ytan22.5 (I)
End rotations at 22.5° to the plane of initial imperfection (X=Ytan22.5°), with the
component in the plane of the initial imperfection in the ~ direction as the initial
imperfection

about either compn analysis moment
axis (kN) (kN-m)
(radians)
SHS3000- X +Y-22-2009m09d02v2 0.02232 1269.6
SHS3000 X+Y 22 2009m09dl2vl 0.03125 1141.5 end Oby

moment interpola
changing to tion
stabilising
Table G.13 3.0m column, rotation in X = Ytan22.5 (2)
End rotations at 22.5° to the plane of initial imperfection (X=Ytan22.5°) with the
component in the plane of the initial imperfection in the opposite direction as the

axis (kN) (kN-m)
(radians)
SHS3000_X+Y_ 22_2009m09d02vl 0.02348 1356.6
303
Table G.14 3.0m column, rotation in X = Ytan67.5 plane
End rotations at 67.5° to the plane of initial imperfection (X=Ytan67.5°) towards

axis (kN) (kN-m)
(radians)
SHS3000- X+Y-67-2009m09d03vl 21.99 1251.0 buckled destabilis
ing
SHS3000- X+Y-67-2009m09dl2v2 32.51 1187.7 end 0 by

moment interpola
changing to tion
stabilising
of the initial imperfection in the ~ direction as the initial imperfection

about both X compn analysis moment
and Yaxes (kN) (kN-m)
(radians)
SHS2006m II d 14_130by I 0 0.006452 1549.4 end moment Oby
- X+Y v4 changing to interpolatio

stabilising n
SHS2006mlld14_130by10 0.01549 1316.2 ditto ditto
_X+Y_v5b
SHS2006mlld14_130by10 0.02285 1185.4 ditto ditto
- X+Y- vI
SHS2006mlldI4_130byI0 0.03581 976.7 ditto ditto
_X+Y_v6b
304
G.S 6.0m columns
The results of the Abaqus analysis for columns of 6.0m length are presented in
Table G 16, for end-rotations in the plane of the initial imperfection
Table G 17, for end-rotations out of the plane of the initial imperfection
Both the tables have the initial imperfection in the X = 0 plane with the imperfection in the
+Y direction.
Table G.16 6.0m column, rotation in the X=O plane
End rotations in the plane of initial imperfection and in the !!!!!£ direction as the initial
imperfection
Max from URI = RF3 = end of End

SHS2006m 1Od 14_130by 10_by6000v 1 end rot compn analysis moment
All results other from (radians) (kN) (kN-m)
SHS2006m 1Od 14_130by 10_by6000v2
0.0184 671.1 continued o(pin)
Max pin-ended 0.0270 755.7 buckled o(pin)
0.0751 477 continued o(pin)
0.0852 423 continued o(pin)
305
Table G.17 6.0m column, rotation in the Y=O plane
URI = RF3 = End mmt

end rot compn (kN-m)
(radians) (kN)
Max pin-ended 0.0270 755.7
SHS2006mlOd19_130bylOby6000 0.00469 750.5 0.29 destab
_2planes_vi 0
SHS2006m 1Od 19_130by 1Oby6000 0.00924 738.9 0.83 destab
_2planes_v9
_2planes_v8
SHS2006ml0dI9_130bylOby6000 0.01780 711.9 2.23 destab
_2planes_v7
_2planes_v6
_2planes_v5
_2planes_v2
_2planes_v4
_2planes_v3
SHS2006mlOdl9_130by1Oby6000 0.03606 631.1
_2planes_vi
SHS6000- X+Y-90-201 Om05d06v 1 0.07156 447 buckled
306
Table G.18 6.0m column, rotation in the X=Y plane
URI = RF3 = End mmt

end rot compn (kN-m)
(radians) (kN)
SHS6000_X+ Y_ 45_201 Om05d03v2 0.0804 453
307
APPENDIX H IMPROVED ECONOMY WITH
VARIABLE IMPERFECTION
For end-rotations applied in the rectangular planes of the SHS section, the parametric study
showed that use of a constant for eo (eo = es) is unnecessarily conservative and the resistance
could be modelled more accurately by using a value of eo that reduces with increased end-
rotation. This is of practical interest because there might be many structures in which the
slab is very stiff relative to the steel beams, so the end-rotations orthogonal to the end-
rotations caused by the steel beam might be very small.
The values of eo needed for the design model to give the correct failure load for that end-
rotation for rotations in a rectangular plane are shown in Figure H.I together with the dotted
line eO/es that shows a possible tri-linear design line for eoles that could be used for rotations
in a rectangular plane. The equation of the dotted line shown is:
if 9(h/2) < aees,
EqH I
eo = es(1-aJDe-be)( 1-(9(h/2) -aexeJ(mees-aees»+bees, EqH2
if 9(h/2) > mees,
EqH3
where 9 is the end-rotation applied to the column

h is the column height
and the constants have values:
ae = 1.50, be = 0.15, me = 5.0 and De = 3.5.
308
(em-eO)/es, rotation in a rectangular plane
1.0~,-~r--T.--r-r--r-~
, i I I I
I.
I
-r~-'.-l-~~
I Iii
r-l---
I I I I I I I ~ 300m 0 (em-e) /es
,
0.5
~ 300m 90 (em-e) /es
.. i I l' ~ 1.5m 0 (em-e) /es
~ 000 l}~ _l.Sm90 ,om")i"
i I ~600mO(em~e) /es
-005 ··~T~--1 ~ 600m 90 (em-e) /es
- - eO/es
-1.0
Figure H.I eO for failure load for rotations in a rectangular plane.
For rotations that are not entirely in a rectangular plane, the dotted design line in Figure H.I
should not be used unless there is a method to account for the effect of being out of plane.
One method to account for the effect of the end-rotation being out of a rectangular plane uses
the projection of the applied rotation, 9, onto a rectangular axis. The resistance is calculated
by taking the end-rotation, 9r. about the rectangular axis "r" (where the axis "r" is either of
the rectangular axes of the SHS section) as 9r = 9/Coscp, where 9 is the end-rotation applied
to the column and this end rotation is in the plane at an angle cp to the nearest rectangular axis
(ie cp:s 45°).
Another method to account for the effect of the end-rotation being out of a rectangular plane
uses the sum of the end-rotations 9 x and 9 y • The resistance is calculated by taking the end-
rotation, 9r. about the rectangular axis "r" (where the axis "r" is either of the rectangular axes
of the SHS section) as 9r = 9Sincp + 9Coscp, where 9 is the end-rotation applied to the column
and this end rotation is in the plane at an angle cp to the nearest rectangular axis (ie cp:s 45°).
For both of these methods to account for the effect of the end-rotation being out of a
rectangular plane, the imperfection eo needs to be calibrated to ensure it is reliable. The data
available is shown in Figure H.2 compared with the design line shown in Figure H.I.
Data labelled eOp is derived using the projection onto a rectangular plane and data labelled
eOsum is derived using the sum of 9X and 9Y.
309
eOP/es & eO/es, rotation not in a rectangular
1.0
,, ---r
!
- - eO/es
, -o;;!o- eOsum 22.5 /es
..... eOp22.5/es
..
i 0.0 ~eOp45.0/es
~eOsum67.s/es
-0.5
..... eOp67.5/es
~ 1.5m 45 eOp /es
Figure H.2 eO for failure load for rotations not in a rectangular plane
310
APPENDIX I DEFORMED SHAPE OF TEST
SPECIMEN
The correlation factors varies between tests as shown in Figure 6.26 Correlation factors for
rotation. The possibility that these differences arose from a differences in deformed shape of
the members (due to differences in the extent of plasticity) was investigated by studying the
relationship of mid-length displacement to end-rotation for different forms of curve. The
ratio of (mid-height displacement/height) to (end-rotation) in a circular curve is
approximately ~ and for a half-sine curve is lI1t. The plots in Figure 1.1 show ratios from
0.28 to 0.33 at the unloading cycles, the points at which the correlation factors Cr are
calculated, but there is no apparent correlation between these ratios and the correlation
factor, Cr.
ke3 to kelO; Ratio of disp/heightto mean end rotation

04 - 1/4
0.38 -In
c
.g 0 .3 6
- ke3 disp/height/mean
'"
~ 0.34
rotn
-c - ke4 disp/height/mean
c
~ 0 .32 rotn
'"<II - keS disp/height/mean
E 0 .3 rotn
';:;-
..c - keG disp/height/mean
.!!II 0 .28
<II
..c rotn
'-
c:
<II
0 .26 ke7 disp/height/mean
rotn
E
~ 0 .24 ke8 disp/height/mean
'"
Ci
.~ 0 .22
rotn
-c ke9 disp/height/mean
.1: rotn
.!!II 0 .2
<II ke10
..c 0 10 20 30 40
-0 disp/height/mean rotn
~ Mid-height displacement (mm)
Figure 11 Deflected shape of test columns
311
APPENDIXJ CMK 2007 SSRC MODEL
At the 2007 SSRC conference [King 2007], a model was proposed using the data and
understanding at that date. The model proposed is described below. As a result of the full
scale tests reported in Section 5, the analysis reported in Section 6 and further parametric
studies reported in Sections 7 and 8, this model is now superseded by the 2009 basic design
model. The 2007 model is described here for reference purposes only.
The 2007 model was as follows:

In cases where there are end rotations in both planes, the design deflections in both planes
must be included. The appropriate value of imperfection, eO, (see below) is added in only
one plane. Therefore the total design deflections, eX and eYare:
either eX = Oy (h/2) + eO and eY = Ox (h/2)
or eY = Ox (h/2) + eO and eX = Oy (h/2)
where eX is the design deflection along the X axis
eY is the design deflection along the Y axis
Ox is the end-rotation about the X axis
Oy is the end-rotation about the Y axis
Note that for members with rotation about one axis only, there are two checks required.
Using the case of a column with Oy = 0 to illustrate this, the checks are:
(1) eX = (Oy (h/2) + eO =) eO applied together with eY = Ox (h/2)
(2) eY = Ox (h/2) + eO applied together with eX = (Oy (h/2) =) 0
The proposed imperfection was established by calibrating the design model with the finite
element analyses using different imperfections. Analytical studies of square hollow sections
at that date showed that a varying design value of initial imperfection was possible.
The following was proposed for the imperfection, eO:
if eO ~ 2es, eO = es [1 - (eO / 2es)]
but if eO > 2es, eO = 0
where eO = 0 x (h/2)
Following full scale testing and further analytical studies, the imperfection was simplified to
the use of es in the basic design model proposed in Section 8.5.1. This is required because
the formulae above were derived
312
1. before the full scale laboratory testing and calibration reported in Chapters 5 and 6
and
2. before knowing about the difference in the directions of the displacement at
midheight of the column caused by end-rotations not in a rectangular plane, as
described in Section 3.4.5
Both of these show that the resistance may be lower than predicted by the 2007 model
[King 2007], so that model should not be used in design.
313

2 Bolted Connections

Uploaded by

Copyright:

Available Formats

2 Bolted Connections

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2 Bolted Connections

Uploaded by

Copyright:

Available Formats

What is the column design method developed for?

What is the column design method developed for?

What are floor beams designed to do in this method?

What are floor beams designed to do in this method?

UNIVERSITY OF SHEFFIELD ,

Department of Civil and Structural Engineering

by Charles MacIan King

The method incorporates the elasto-plastic behaviour of columns subject to axial

The design method

I would also like to thank following for their contributions:

Dr A R Gent for his original papers on elastic-plastic column stability

Charles MacIan King

2 REVIEW OF DESIGN METHODS 6

3 THE NEW DESIGN APPROACH 33

6 COMPARISON OF TESTS WITH ANALYSIS 100

7 BREADTH TO THICKNESS LIMITS 139

8 VALIDATION OF THE DESIGN MODEL BY A PARAMETRIC STUDY AND

9 DESIGN METHOD 165

10 WORKED EXAMPLE 196

APPENDIX A SHS SECTION PROPERTIES 223

APPENDIX B BS 5950 CALCULATIONS 231

APPENDIX C FINITE ELEMENT MODEL DETAILS 234

APPENDIX 0 STRUT RESISTANCE FROM INITIAL IMPERFECTIONS 247

APPENDIX E TRANSVERSE BENDING OF WALLS 251

APPENDIX F FINITE SLICE SPREADSHEET 271

APPENDIX G TABLES OF RESULTS OF PARAMETRIC STUDY 294

APPENDIX H IMPROVED ECONOMY WITH VARIABLE IMPERFECTION 308

APPENDIX I DEFORMED SHAPE OF TEST SPECIMEN 311

APPENDIX J CMK 2007 SSRC MODEL 312

Figure 1.1 Typical braced frame 2

FigureA.1 Plastic stress distribution, low axial load 224

initial imperfection along one rectangular axis

Ne,Rd resistance from new design model with imposed end-rotation

PRC,max radial pressure on wall of SHS due to curvature at distance the

a non-dimensional imperfection factor (EN 1993-1-1)

normalizing factor for slenderness = 1t fK (EN 1993-1-1)

cro yield stress (Canadine)

cr' I, cr' 2, cr' 3 deviatoric stress vector (Canadine)

crc Longitudinal compressive stress in wan

crmax Maximum elastic stress in column

=0.5l1 + a~ - 0.2)+ ~2 J(EN 1993-1-1)

1.2 Braced frames

1.3 Beam and column arrangements in traditional

1.4 Braced frames with discontinuous columns

(floor and stiffeners to the beam web omitted for clarity)

Design methods comprise the complete process of selecting a fonn of construction,

2.2 Design of traditional braced frames

Figure 2.3 Typical column with beams connected on both sides

v = 0.5(10 x 6.0 x 7.5) = 225 kN Eq2.1

Mx = 225(0.l00+0.l40/2) = 225 x 0.170 = 38.25 kN-m Eq2.2

Mbs = Py X Sx = 355 x 246 x 10-3 = 87.3 kN-m Eq2.4

and re-written again as

= 1550(1-0.219) = 1550xO.781 = 1211kN Eq2.7

2.3 Orthodox design of braced frames with

2. Rigid jointed frame using elastic analysis

3. Flexibly jointed frame using elastic analysis

Figure 2.4 Beam on a thin cap plate on a 140 x 140 x 10 SHS

Substituting the values for this example:

This is a reduction to 69111550 = 45% of the strut buckling capacity.