Proceedings of
SDSS’Rio 2010
International Colloquium
Stability and Ductility of Steel Structures
First published in Brazil in 2010 by Federal University of Rio de Janeiro and State University of Rio de Janeiro
All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording or by any information storage and retrieval
system, without permission in writing from the publisher.
ISBN: 9788528501377
Distribution:
Federal University of Rio de Janeiro and State University of Rio de Janeiro
Telephone: (+55 21) 25628474 and (+55 21) 23340469
Email: batista@coc.ufrj.br ; pvellasco@globo.com and luciano@eng.uerj.br
ORGANISATION
Federal University of Rio de Janeiro, UFRJ
State University of Rio de Janeiro, UERJ
Structural Stability Research Council, SSRC
EDITORS
Eduardo de M. Batista
Pedro C. G. da S. Vellasco
Luciano R. O. de Lima
SPONSORS
SCIENTIFIC COMMITTEE
ORGANIZING COMMITTEE
VOLUME 1
PREFACE v
2. KEYNOTE LECTURES 7
Some issues for column stability criteria 9
Reidar Bjorhovde
The effect of edge support on tensile membrane action of composite slabs in fire 21
Anthony K. Abu and Ian W. Burgess
Latest developments in the GBT analysis of thinwalled steel structures 33
Dinar Camotim, Cilmar Basaglia, Rui Bebiano, Rodrigo Gonçalves and Nuno Silvestre
Dualsteel frames for multistory buildings in seismic areas 59
D. Dubina
Advanced stability analysis of regular stiffened plates and complex plated elements 81
László G. Vigh and László Dunai
Direct strength design of coldformed sections for shear and combined actions 101
Cao Hung Pham and Gregory J Hancock
Manufacturing specifications for hollow sections in 2010 115
Jeffrey A. Packer
Numerical study on stainless steel beamcolumns with transverse loading 123
N. Lopes, P. Vila Real and L. Simões da Silva
v
Contents
Elegance and economy  a new viaduct over the river Llobregat 157
Peter Tanner, Juan L. Bellod and David Sanz
Design of beamtobeam buttplate joints in composite bridges 165
A. Lachal, S.S. Kaing and S. Guezouli
4. CONNECTIONS 173
Experimental analysis of composite connections using slab made by precast joist
with lattice and bricks 175
William Bessa, Roberto M. Gonçalves, Carlo A. Castiglioni and Luis Calado
Response of endplate joints under combined forces 183
N. Baldassino, A. Bignardi and R. Zandonini
Influence of member components on the structural performance of beamtocolumn
joints of pitched roof portal frames with class 3 and 4 sections 191
I. Mircea Cristutiu, Dan Dubina
Application of Eurocode 3 to steel connections with four bolts per horizontal row 199
J.F. Demonceau, K. Weynand, J.P. Jaspart and C. Müller
A new hybrid testing procedure of the low cycle fatigue behaviour
for structural elements and connections 207
Carlo Andrea Castiglioni, Alberto Drei and Roberto Goncalves
Proposal of a threedimensional semirigid composite joint: tests and finite
element models 215
Beatriz Gil, Rufino Goñi and Eduardo Bayo
Strength and ductility of bolted Tstub macrocomponents under monotonic
and cyclic loading 223
Nicolae Muntean, Daniel Grecea, Adrian Dogariu and Dan Dubina
Prediction of the cyclic behaviour of moment resistant beamtocolumn
joints of composite structural elements 231
Nauzika Kovács, László Dunai and Luís Calado
Numerical modeling of joint ductility in steel and steelconcrete
composite frames 239
Leslaw Kwasniewski and Marian Gizejowski
Influence of chord axial loading on the stiffness and resistance of welded “T”
joints of SHS members 247
Rui M. M. P. de Matos, Luís F. CostaNeves and Luciano R. O. de Lima
Experimental studies of behaviour of composite beamcolumn flush end plate
connections subjected seismic loading 255
Olivia Mirza and Brian Uy
Modelling connections of moment resisting steel frames for seismic analysis 263
L. Mota, A. T. da Silva, C. Rebelo, L. Simões da Silva and Luciano R. O. de Lima
Influence of local plastic buckling of joint on carrying capacity of a thinwalled truss 271
H. Pasternak, G. Kubieniec and V. Bachmann
vi
Contents
Numerical analysis of endplate beamtocolumn joints under bending and axial force 279
Monique C. Rodrigues, Luciano R. O. de Lima, Sebastião A. L. de Andrade,
Pedro C. G. da S. Vellasco and José G. S. da Silva
Loss of preload in bolted connections due to embedding and self loosening 287
Roland Friede and Jörg Lange
Plastic resistance of Lstubs joints subjected to tensile forces 295
M.Couchaux, I.Ryan and M.Hjiaj
Composite beam modelling at real scale including beamtobeam joint 303
S. Guezouli and A. Lachal
Resistance of laser made t RHS joints under compression load 311
Jerzy K. Szlendak
Coldformed steel and concrete composite beams: study of beamtocolumn
connection and region of hogging bending 319
Mairal R. and Malite M.
Shear bolted connections: numerical model for a ductile component,
the platebolt in bearing 327
J. Henriques, L. Ly, J.P. Jaspart and L. Simões da Silva
vii
Contents
viii
Contents
Numerical analysis of steel columns considering the walls on fire condition 521
Jonas B. Dorr, Jorge M. Neto and Maximiliano Malite
On the strength and DSM design of coldformed steel columns failing
distortionally under fire conditions 529
Alexandre Landesmann, Dinar Camotim
VOLUME 2
ix
Contents
x
Contents
The fatigue and serviceability limit states of the webs of steel girders subjected
to repeated loading 771
M. Škaloud and M. Zörnerová
Imperfections in steel plated structures and their impact on ultimate strength 779
J.Kala, M.Škaloud, J.Melcher and Z.Kala
Imperfection sensitivity analysis of longitudinally stiffened plated girders subjected
to bendingshear interaction 787
F. Sinur and D. Beg
9. SHELLS 795
Influence of geometry on the dynamic buckling and bifurcations of cylindrical shells 797
Zenon N. del Prado and Paulo B. Gonçalves
Buckling of a shallow rectangular bimetallic shell subjected to outer loads
and temperature 805
M. Jakomin and F. Kosel
A geometry based method for the stability analysis of plates 813
Hesham Ahmed, John Durodola and Robert G. Beale
Critical loads and stability of an open elasticplastic cylindrical shell with the core
of variable stiffness 821
Jerzy Zielnica
xi
Contents
xii
Contents
xiii
Contents
AUTHOR INDEX
Volume 1 537
Volume 2 1201
xiv
7. MEMBERS’ BEHAVIOR:
TENSION, COMPRESSION, BEAMS,
BEAMCOLUMNS
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
NOTATION
539
Luciano M. Bezerra and Ramon S. Y. C. Silva
1. INTRODUCTION
Submarine pipelines are often laid on relatively rough seabottom terrains and, consequently, may be
supported by soil only intermittently, without intermediate support. Such spans are identified as
“freespans”. The scope of this paper is to predict the behavior of freespan pipelines under compressive
loads originating from effects such as temperature differentials. The paper deals exclusively with free
span pipelines under compressive load combined with internal pressure. The compressive load, P, is
assumed to be applied at the ends of the pipe segment and to be collinear with a line through the end
supports of the pipe segment. Consequently, the load is considered to act along the chord connecting the
two ends of the freespan segment, without change in the load direction. The collapse mechanics of a
segment of a free span pipeline (FSP) under compressive load is not necessarily the same as for a buried
pipeline (BP). Adequate support around a BP may prevent it from buckling globally. Assuming a FSP
under compression deforms as shown in Fig.1, the collapse mode of a FSP under compression, depends
upon the length of the free span, and will be different than for local wrinkle formation typically observed
in short segments of BPs. For short free span lengths, the collapse mode of the FSP might be similar to
the local wrinkle formation mode observed in BPs. For long free spans, the collapse mode might be
comparable to the global buckling collapse mode observed in a structural column.
S 2 EI
PE (1)
L2
To determine a more realistic behavior of FSPs under compressive load it is necessary to admit the
existence of initial imperfections, the possibility of inelastic behavior, and the mobilization of fully plastic
moment capacity of the pipe section. Let us consider the effect of initial imperfection and plastic
deformations, using the mechanical model shown in Fig.2 [6]. Subsequently, it is possible to examine for
the effects of initial imperfection and inelastic material behavior on the buckling behavior. The
mathematical model consists of two rigid arms pinned together at the span centerline at C. On the ends
(A and B) they are pinned too, as in Fig.3. A vertical spring, with stiffness K, is attached at C. Applying
an increasing horizontal axial force P at point A through the centroidal crosssectional axis, with G = 0,
will make P reaches its critical load, P cr.
At this critical load, when buckling takes oplace, the model forms a mechanism in which point C
displaces laterally through a distance G, and the arm rotates D  see Fig.2b. Prior to instability, the force
in the spring, Ps = 0. As soon as the instability takes place, P s=Kδ  with K being the spring constant.
From moment equilibrium of the arm from A to C, about C, for Fig.2b, we can write for small angles
540
Luciano M. Bezerra and Ramon S. Y. C. Silva
§ PS ·§ L · Pcr L
Pcr G ¨ 2 ¸¨ 2 ¸ or (2)
© ¹© ¹ Ps 4G
In Fig.2a, to simulate the Euler’s buckling load, the lateral defection G will take place abruptly when the
critical load reaches the Euler’s load. Substituting KG for Ps in Eq. (2) and equating Pcr = PE yelds
KL 4PE
Pcr PE or K (3)
4 L
Figure2: (a) An elementary buckling model and Figure3: The mechanical model with initial
(b) freebody diagram. imperfection.
Up to now, the prebuckling shape is a straight line, however, by now let’s consider the existence of an
initial imperfection Go z 0 in Fig.3. Note that small initial imperfections will be amplified by the axial
force. The model of the FSP with an initial imperfection Go is in Fig.3a. Go exists initially, for P = 0 and
Ps = 0. For P z 0, the incremental displacement at the centerline increases by the amount G due to the
rotation of the arms – see Fig.3a. The total displacement, due to the arm rotation, E, becomes Gtot = G0 + G
 Fig.3b. The moment equilibrium of arm AC, about C, for Fig.3b may be expressed by Eq.(4).
§ PS ·§ L · § Ps L ·
PG tot ¨ 2 ¸ ¨ 2 ¸ or G tot ¨ 4P ¸ and Gtot = Go + G (4)
© ¹© ¹ © ¹
As PS = KG and δ = δtot – δ0, and using Eq.(3) for K; Eq.(4) can be transformed into the following
equation
§ Ps L · § KLG · § 4PE LG · § PE · § 1 G0 ·
G tot ¨ 4P ¸ ¨ ¸ ¨ 4P ¸ ¨ P ¸ G tot G o or P ¨ ¸ PE (5)
© ¹ © 4P ¹ © ¹ © ¹ © G tot ¹
The model for the FSP with an initial Go will result in an increase in bending moment at the center
of the span as P and Gtot increase. However, P in Eq. (5) indicates that the compressive load for the model
of the imperfect column (or FSP) will never reach the Euler’s load, PE, but approaches PE asymptotically.
In addition, the maximum bending moment that can arise at the central section cannot exceed that
associated with the fully plastic condition for the pipe. At the central section M = PGtot and the collapse
load for can be determined by the load P that produces the moment which, when combined with the axial
effects, mobilizes the fully plastic capacity of the pipe section ( M Tpc ). To compute the full plastic
capacity of the pipe it is necessary a yield criterion.
Pressurized pipes are subjected to hoop and longitudinal stresses due to axial forces and transverse
bending moments acting on the pipe cross section. For a thinwalled pipe, the hoop stress is considered
constant and stresses other than hoop and longitudinal may be neglected. The longitudinal stress Vl and
the hoop stress VT are identified as the principal stresses V1 and V2, respectively. Using the VonMises
541
Luciano M. Bezerra and Ramon S. Y. C. Silva
Hencky yield criterion (with Vy as the uniaxial yield strength) the maximum (and minimum) longitudinal
stresses that the fullyplastic pipe cross section can sustain on the cross section may be calculated as
2
§ Vl · § 2VT · § 3 ·§ V ·
¨ ¸ ¨ ¸ r 1 ¨ ¸¨ T ¸ (6)
¨V y ¸ ¨ Vy ¸ © 4 ¹ ¨© V y ¸
© ¹ © ¹ ¹
Eq.(6) is also valid for the ultimate stress (Vu) in the place of the yielding stress (Vy). Eq.(6) also
identifies the two values of Vl that produce yielding for a specified VT. One corresponds to a compressive
stress (Vl = Vc), and the other, to a tensile stress (Vl = Vt)  Fig 4. These values represent the maximum
longitudinal compressive and maximum longitudinal tensile stresses that can be developed on the extreme
fibers of the pipe crosssection for the given VT. If VT = 0, then, for yielding, Vl=Vc=Vt=Vy. If VT z 0, then,
the longitudinal stress Vl required to origin yield in tension is Vl = Vt which is different than that required
in compression (Vl = Vc) (See Fig.4). Naming [ = VT /Vl and K = Vl /Vy, the VonMisesHencky yield
criterion is shown in Fig.4 [4, 5]. From Fig.4 and Eq.(6), the extreme values for Vl and VT. For the
determination of the fullyplastic capacity of the pipe section, we will assume that the stressstrain curve
shows a well defined yieldstress plateau. The yield stress is an important engineering property in order to
establish limits on the longitudinal and hoop stresses. The hoop stress VT is given by
pr
VT (7)
(Rr)
The longitudinal stress acting on the pipe crosssection will depend on the axial force P and the
bending moment. The limiting combinations of axial force and bending moment that develop the fully
plastic capacity of the pipe section can be presented on an interaction diagram due to [2, 3]. In the
following Section, the equations for the fully plastic moment capacity of the FSP pipe section will be
derived.
P P
V
ªS R r 2 º
(8)
A0 2
¬ ¼
The stresses on the pipe section at the point of maximum moment are in Fig.5 which is a fully plastic
condition. At such a point, at the center of the span, we have a combination of stress from bending
moment plus stress from axial loading. However, at the ends of the FSP (see Fig.1 and 3); the force P
acts in concert with the transverse force of Ps/2, and the combination of these loads must be equilibrated
by the stress distribution of Fig.5 at the centerline of the span. Therefore, at the point of maximum
moment, the resultant longitudinal force given by the difference between the tensile force F t = σtAt and
compressive force Fc = σcAc, in Fig.5, must be in equilibrium with the external applied force P at the
ends of the FSP. The areas Ao, At, and Ac in Fig.5 can be expressed as
§\ · 2 2 ª§ 2S \ · º
S R 2 r 2 , At ¨ ¸ R r and Ac R r 2 «¨ ¸»
2 (9)
Ao
©2¹ ¬«© 2 ¹ ¼»
542
Luciano M. Bezerra and Ramon S. Y. C. Silva
From Eqs. (8) and (9) we can write the following longitudinal equilibrium equation
ª 2S \ º 2 §\ ·
P VS R 2 r 2 Fc Ft V c R2 r 2 « » V t ( R r )¨ ¸
2
(10)
¬ 2 ¼ ©2¹
The angle \ can be calculated as a function of the stresses V, Vt, and Vc of Eq.(10).
ª § V V · º
\ 2S «¨¨ c ¸¸ » (11)
«¬© V c V t ¹ »¼
A search for P (that causes the stress distribution depicted in Fig.5) is the same as a search for the
equivalent stress V = P/Ao at the end of the FSP. The arms yt & yc of the respective forces Ft and Fc,
such forces are at the centroids of the areas At and Ac in Fig.5  can be calculated as
ª§ § \ · ·º ª§ § \ · ·º
«¨ 4 R r sin ¨ ¸ ¸ » «¨ 4 R r sin ¨ ¸ ¸ »
3 3 3 3
yt «¨ © 2 ¹ ¸ » and y «¨ © 2 ¹ ¸»
«¨ 3\ R r ¸ » «¨ 3 2S \ R r 2 ¸ »
2 2 c 2
(12)
«¨ ¸» «¨ ¸»
¬© ¹¼ ¬© ¹¼
Knowing At, Ac in Eq.(9); yt , yc in Eq.(12); and Vc and Vt in Eq.(6); the maximum plastic resisting
moment MTpc can be determined due to the load P (or stress σ) at the ends of the FSP. MTpc is in
equilibrium with the moment caused by the external force P and the eccentricity Gtot of Fig.3 and Eq.(4),
therefore
M T pc Fc yc Ft yt V c Ac yc V t At yt PG tot V Ao G tot (13)
Using Eqs.(9), and (12) in Eq.(13), the expression for the maximum plastic bending moment is
§ 2S \ · 2 4 R3 r 3 §\ ·
M T pc Vc ¨
© 2 ¹
¸ R r 2
3 2S \ R 2 r 2
sin ¨ ¸
©2¹
(14)
§\ · 4 R3 r 3 §\ ·
V t ¨ ¸ R2 r 2 sin ¨ ¸
© ¹
2 3\ R r
2 2
©2¹
Substituting into Eq.(14) the expression for the angle< from Eq.(11), we arrive at the following
simplified version of Eq.(14) which is an expression for the maximum moment capacity for the FSP
§2· ª§ S V c V · º
M T pc ¨ ¸ V t V c R r sin «¨¨ ¸¸ »
3 3 (15)
©3¹ ¬«© V c V t ¹ ¼»
543
Luciano M. Bezerra and Ramon S. Y. C. Silva
capacity in Eq.(15) shows that for an increase in V, there will be a decrease in MTpc at center span. The
formulation contained, herein, is based upon the argument that, to find the compressive collapse stress V
of a FSP, the effect of outofstraightness must be taken into account. In reality, every structure has
imperfections in geometry; but long structures like FSP laid on rough terrains, are more susceptible. The
initial imperfection G0 is taken into account in Eq.(5). Such equation represents the behavior of the FSP in
the elastic range until the fully plastic stress distribution of Fig.5 is developed giving rise to Eq.(13).
Note that Eq.(5), expressed in terms of Euler’s critical stress VE = PE/A0, can give an expression for Gtot as
P ª § G0 · º § PE · ª § Go ·º
V «1 ¨ ¸ » ¨ ¸ or V «1 ¨ ¸ » V E and, G tot G o ª¬1 V V E º¼ (16)
Ao ¬« © G tot ¹ ¼» © Ao ¹ ¬« © G tot ¹ ¼»
Once yielding has fully developed, put MTpc from Eq. (15) into Eq.(13) to get an expression for Gtot as
§ M T pc · ° ª 2 V t V c R r º ½
° ª S V c V º
3 3
V ¨¨ ¸¸ or G tot ®« » ¾ sin « »
(17)
© AoG tot ° «¬ 3S R r V »¼ ¿ ° «¬ V t V c »¼
2 2
¹ ¯
Finally, by equating the right hand sides of Eq.(16) and Eq.(17), we arrive at the following
transcendental equation for the determination of the collapse stress V, which will be designated as V
° ª 2 V t V c R r º ½
· ª S V c V º
3 3
°§ V
« » ¾¨ 1 » VG o
(18)
® ¸ sin « 0
° «¬
¯
3 S R 2
r 2
»¼ ¿° © E V ¹ ¬« V c V t ¼»
Developing Eq.(18) into a Taylor series and keeping only two terms of this series, one obtains:
ª R2 r 2 º
V ² CV D 0 where C V E V c 1,5G oV E « 3 3 » and D V EV c (19)
«¬ R r »¼
Figure 4: VonMisesHencky Yelding Criterion Figure 5: Idealized fully plastic stress distribution
The solution for the collapse stress V in Eq.(19) takes into consideration: (a) the geometric
properties of the pipe section; (b) the initial imperfection for the particular FSP; (c) an upper bound limit
represented by the Euler’s buckling load; (d) the fullyplastic stress distribution, (Fig.5); (e) the fully
plastic capacity depends on both the plasticity criterion and the hoop stress, which is a function of the
applied internal pressure; (f) long structures, with initial imperfection, never reach the Euler’s Load
(which is an upper bound limit); (g) the Euler’s load (or stress) which is a function of the modulus of
544
Luciano M. Bezerra and Ramon S. Y. C. Silva
elasticity, pipe crosssection properties, and pipe length; and (h) the consideration of initial imperfection
that is essential as it triggers the limiting fullyplastic moment capacity mechanism.
0 0.00 f f 7871.194 NA NA
4 2.902 2.50E+05 1.37E+04 6694.619 8041.080 16.74
6 6.529 1.11E+05 6.09E+03 6546.097 7324.820 10.63
8 11.607 6.25E+04 3.43E+03 6348.670 6828.900 7.03
12 26.119 2.78E+04 1.52E+03 5844.567 5913.680 1.17
15 40.815 1.78E+04 9.75E+02 5408.596 NA NA
20 72.577 1.00E+04 5.48E+02 4660.695 4027.00 15.74
6. CONCLUSIONS
This paper presented a mathematical formulation regarding the investigation of compressive
collapse loads of pressurized FSPs. A strategy for obtaining collapse loads as a function of the span
length, initial imperfection, and fully plastic stress capacity has been presented and discussed. Examples
of collapse loads, for pressurized FSPs with a variety of lengths and initial imperfections, were compared
to the sophisticated FE results from the ABAQUS program. The numerical tests show that the proposed
analytical formulation represents a good approximation to freespan solutions. Instead of yield stress, the
analytical solutions were almost coincident with the collapse results generated by ABAQUS FE analyses.
Each complex nonlinear FE run in ABAQUS took approximately 5 hours of CPU on a SUN workstation.
Finally, it is noted that the scope of the present formulation is not to propose a method to substitute
precise FEM modeling and analyses, but to provide an easy, faster and practical way for a first assessment
of compressive collapse loads of pressurized FSPs for the petroleum industry.
545
Luciano M. Bezerra and Ramon S. Y. C. Silva
REFERENCES
[1] ABAQUS. (2000). Standard User’s Manual. Version 6.3. Hibbitt, Karlsson and Sorense, USA.
[2] DOREY, A.B.. (2001). Critical Buckling Strains for Energy Pipelines. PhD Thesis, University of
Alberta, Edmonton, AB, Canada.
[3] MOHAREB, M., D. W. MURRAY. (1999). Mobilization of Fully Plastic Moment Capacity for
Pressurized Pipes. Journal of Offshore Mechanics and Arctic Eng., ASME. vol. 121. p. 237241.
[4] POPOV, E. (1998), Eng. Mechanics of Solids. Prentice Hall. Englewood Cliffs, New Jersey, USA.
[5] HOFFMAN AND SACHS, (1953), Theory of Plasticity, McGrawHill Book Inc. New York, USA.
[6] SHANLEY, F. R. (1957). Strength of Materials. McGrawHill Book Company Inc. New York,
USA.
546
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
* Structural Engineering Research Group/LGCGM, INSA de Rennes, 20 avenue des Buttes de Coësmes
35043 Rennes Cedex France
emails: Rabe.Alsafadie@insarennes.fr, Mohammed.Hjiaj@insarennes.fr
** Department of Civil and Architectural Engineering, KTH, Royal Institute of Technology, SE10044
Stockholm, Sweden
email: JeanMarc.Battini@sth.kth.se
Keywords: Geometrically nonlinear, 3D beams, corotational formulation, mixed finite element analysis,
arbitrary crosssections, elastoplastic material behavior, HellingerReissner functional.
Abstract. The corotational technique is adopted for the analysis of 3D beams. The technique applies to a
twonoded element a coordinate system which continuously translates and rotates with the element. In
this way, the rigid body motion is separated out from the deformational motion. Then, a mixed
formulation is adopted for the derivation of the local element tangent stiffness matrix and nodal forces.
The mixed finite element formulation is based on an incremental form of the twofield HellingerReissner
variational principle to permit elastoplastic material behavior. The proposed element can be used to
analyze the nonlinear buckling and postbuckling of 3D beams. The mixed formulation solution is
compared against the results obtained from a corotational displacementbased formulation having the
same beam kinematics. The superiority of the mixed formulation is clearly demonstrated.
1 INTRODUCTION
In recent literature, there have been notable contributions to improve the accuracy and efficiency of
displacementbased finite elements. This approach has the limitation in elastoplasticity since the
approximations of the axial strains and curvatures are constrained by the element's assumed displacement
fields. Nonetheless, these curvatures can vary in a highly nonlinear fashion along the length of an elasto
plastic structural member. For example, Izzuddin and Smith [1] found that a large number of
displacementbased beam finite elements are typically required to represent elastoplasticity behavior
accurately. In the mixed formulation, both internal forces and displacements are interpolated
independently. This formulation addresses the fundamental limitation of conventional displacement
based elements: the inability of simple displacement polynomials to represent the highly nonlinear
distribution of the curvatures along the member lengths due to general distributed yielding.
The corotational approach has been recently adopted by several authors to handle the geometric
nonlinearity in 3D displacementbased beam models (Alsafadie et al. [2], [3], Battini and Pacoste [4], [5],
Crisfield and Moita [6]). This paper extends the works of Battini on corotational beam elements by
applying the twofield HellingerReissner variational principle for the development of a mixed local
formulation. The corotational approach is employed to handle the geometric nonlinearity, where, in the
corotational frame, the element rigid body motion has been removed and the formulations focus solely on
the element deformational degrees of freedom.
547
R. Alsafadie et al.
d l = d l (d g ) (1)
Then, d l is used to compute the internal force vector fl and tangent stiffness matrix K l in the local
frame. The transformation matrix B between the local and global displacements is defined by
δ dl = Bδ d g (2)
and is obtained by differentiation of (1). The expression of the internal force vector in global coordinates
f g and the tangent stiffness matrix K g in global coordinates can be obtained by equating the internal
virtual work in both the global and local systems, i.e.
f g = BT f l , K g = BT K l B + ∂ (BT fl ) / ∂d g (3)
fl
Relations (1), (2) and transformations (3) are explained in details in [4].
ay
ax
ry R u
az P
rx G II
I
rz x
Let x0P ( x, y, z ) denote the position vector of an arbitrary point P in the initial configuration and let
x P ( x, y, z ) denote the position vector of P in the current configuration (see figure 1).
x0P ( x, y, z ) = x G0 ( x) + y ry + z rz
(4)
x P ( x, y, z ) = x G ( x) + y a y ( x) + z a z ( x) + α ( x) ω ( y , z ) a x ( x)
where x0G and xG denotes the position vectors of the centroid G in the initial and current onfigurations,
respectively. In the case of thinwalled open crosssections, the normalized warping displacement is
expressed as the product of the warping parameter α ( x ) and the warping function ω ( y , z ) . To handle in
a convenient way nonsymmetric crosssections with distinct shear center and centroid, the warping
function ω is defined according to SaintVenant torsion theory and refers to the centroid G, [7]:
548
R. Alsafadie et al.
ω ( y, z ) = ω − yc z + zc y (5)
and ω refers to the shear center defined by its coordinates yc , zc . The orthonormal triad a i , i = ( x, y , z )
which specifies the orientation of the current crosssection, is given by
a i = R ri , i = ( x, y , z )
(6)
The rotation defined by the matrix R can be considered as the sum of two bending rotations and a
twist rotation, and given by (cf. [8]).
where v , w and ϑx are the transverse displacements and the twist rotation of the crosssection centroid
relative to the local coordinates system, respectively. Introducing the local rotation matrix defined in (7)
into(4), the displacement vector can be evaluated as
U = u − y (v, x + w, x ϑx ) − z ( w, x − v, x ϑx ) + ωα
V = v − z ϑx (8)
W = w + y ϑx
To obtain the strain vector the following assumptions are adopted: the nonlinear shear strain
components generated by warping are omitted since warping effects are rationally taken into account in a
linearized way only, the warping deformations are proportional to the variation of the torsional angle
(Vlasov assumption), an average value of the axial strain is taken in order to avoid membrane locking and
finally the nonlinear terms in the expressions of the bending curvatures and are neglected. With these
modifications, the following strain expressions are obtained:
1 Io 2
ε xx = ε av − y κ z + z κ y + (r 2 − ) κ x + ω κ x, x
2 A
2 ε xy = (ω, y − z ) κ x (9)
2 ε xz = (ω, y − z ) κ x
Since the strain field in (9) is obtained from the local displacement field d l , therefore, all the
components of the strain vector deduced from d l will be designated with a superimposed hat and
combined as İˆ = (εˆxx 2 εˆxy 2 εˆxz ) . Based on the above expression for strain vector idealization, the strain at
any point in the crosssection of the beam element can be related to the crosssectional generalized strain
1
vector eˆ = (ε av κ y − κ z κ x2 κ x , x κ x ) as
2 İˆ = A( y, z ) eˆ ( x) (10)
In the present formulation, the axial rotation ϑx is interpolated with shape functions based on the
closedform solution of the torsional equilibrium equation for an elastic prismatic and geometrically
linear beam. Cubic Hermitian shape functions are chosen for the transverse displacement v and w of the
549
R. Alsafadie et al.
centroid of the crosssection relative to the local element axes. And finally, linear interpolation is adopted
for the axial elongation u of the local element. Thus, the variation in the crosssection deformation
ê can be written as δ eˆ = N eˆ δ dl . Hence, an infinitesimal change in strain vector can be written as
δ İˆ = A N eˆ δ d l (11)
S = N S1 fS (12)
where fS =( N MIy MIz BI T ȍ MIIy MIIz BII) the corotational force degrees of freedom of the mixed
formulated element (where I: first node, II: second node and T a constant torque) and NS1 is the force
shape functions matrix satisfying the equilibrium equations.
ª1 0 0 0 0 0 0 0 0º
«w 1 − x / L 0 0 0 −x / L 0 0 0 »»
«
«v 0 x / L −1 0 0 0 x/ L 0 0»
NS1 = « » (13)
« 0 0 0 0 0 0 0 0 1»
«0 0 0 f BI 0 0 0 f BII 0»
« »
«¬ 0 0 0 f BI, x 1 0 0 f BII, x 0 »¼
sinh [ k ( L − x) ] sinh(k x) GJ
where f BI = − , f BII = and k =
sinh( k L) sinh( k L) E Iω
The resulting element will subsequently be termed as bmw3d element. It should be mentioned that
relation (12) includes Pį effects in the internal moment fields, based on the interpolated transverse
displacements. The variation of the generalized stress resultant internal force vector, may be expressed as
δ S = N S 2 δ dl + NS1 δ fS (14)
where EQ and EC are the element equilibrium and element straindisplacement compatibility equations,
respectively. A third equation, the crosssection equilibrium, may be expressed as
SQ = S Σ − S = 0 (17)
550
R. Alsafadie et al.
where SȈ is given by the nonlinear crosssection constitutive relation and represents a general function
that permits the computation of crosssection stress resultants for given crosssection deformations. The
linearization of the crosssection constitutive relation SȈ= SȈ(e) is obtained using the crosssection tangent
stiffness matrix k = ∂S Σ / ∂e . The crosssection tangent exibility matrix q is obtained by inverting the
crosssection tangent stiffness matrix: q=k1. Furthermore, S is the interpolated generalized stresses acting
over a crosssection and defined by (12).
3.4 Linearization of the HellingerReissner functional
The nonlinear system of equations EQ=0, EC=0 and SQ=0 may be solved using various combinations
of Newton iteration at the element, and crosssection levels. Since interelement compatibility is not
enforced for the generalized stress variables interpolation, the nonlinear discretized straindisplacement
compatibility equation EC=0 can be solved iteratively at the element level for every global equilibrium
iteration. Similarly, the nonlinear constitutive equation SQ=0 can be solved iteratively at the crosssection
level for every element level iteration. In the following Subsections, the consistent linearization of the
above nonlinear equations is presented. In the process of consistent linearization, it is important to
recognize the arguments of any given function.
3.4.1 Linearization of the Crosssection Constitutive Equation
By expanding SQ=0 about the current crosssection state while holding S constant, we can write
∂ SQ
S Qj +1 ≈ S Qj + Δe j Δe j = −qS Qj = q (S − S Σj ) (18)
∂e
∂EQ ∂EQ
EQn +1 ≈ EQn + Δd ln + ΔfSn − ΔFlext , n = 0 (21)
∂d l ∂fS
This equation can be rewritten by substituting ΔfSn from (20), and solving for Δd ln . Hence,
K l Δd ln = Flext , n +1 − fl (22)
551
R. Alsafadie et al.
5. Repeat the above step until S i +1 − S Σj +1 ≤ tolerance , then consider ei +1 =: e j +1 upon convergence.
6. Repeat the above steps from 2 to 5 until EiC+1 ≤ tolerance , then consider fSn =: fSi +1 , e n =: ei +1 and
ECn =: ECi +1 upon convergence.
7. Calculate local element forces fl and tangent stiffness matrix K l
4 EXAMPLES
552
R. Alsafadie et al.
5 CONCLUSION
This paper proposed an efficient local mixed finite element formulation for the analysis of 3D
Bernoulli beams with small strains and large displacements and rotations. The corotational technique
proposed in [4] is employed here. The local strains are derived based on a consistent secondorder
linearization of the fully geometrically nonlinear Bernoulli beam theory. A 3D, geometricnonlinear,
elastoplastic local beam element based on the incremental form of the twofield HellingerReissner
553
R. Alsafadie et al.
functional has been presented. This element is targeted particularly for the analysis of thinwalled beams
with generic open crosssection where the centroid and shear center of the crosssection are not
necessarily coincident. Several numerical examples have demonstrated the superiority of the mixed
formulation over displacementbased one: the use of mixed formulation leads to a considerable reduction
in the number of elements needed to perform the analysis with the same accuracy.
REFERENCES
[1] Izzuddin B.A., Smith D.L. “Largedisplacement analysis of elastoplastic thinwalled frames. I:
Formulation and implementation”. Journal of Structural Engineering (ASCE), 122(8), 905914,
1996.
[2] Alsafadie R., Battini J.M., Somja H., Hjiaj M. “Local formulation for elastoplastic corotational
thinwalled beams based on higherorder curvature terms”. Finite Elements in Analysis and
Design, submitted.
[3] Alsafadie R., Battini J.M., Hjiaj M. “Efficient local formulation for elastoplastic corotational
thinwalled beams”. Communications in Numerical Methods in Engineering, in press.
[4] Battini J.M., Pacoste C. “Corotational beam elements with warping effects in instability
problems”. Computer Methods in Applied Mechanics and Engineering, 191(17), 17551789,
2002.
[5] Battini J.M., Pacoste C. “Plastic instability of beam structures using corotational elements”.
Computer Methods in Applied Mechanics and Engineering, 191(51), 58115831, 2002.
[6] Crisfield M.A., Moita G.F. “A unified corotational framework for solids, shells and beams”.
International journal of Solids and Structures, 33(2022), 29692992, 1996.
[7] Gruttmann F., Sauer R., Wagner W. “Theory and numerics of threedimensional beams with
elastoplastic material behavior”. International Journal for Numerical Methods in Engineering,
48(12), 16751702, 2000.
[8] Van Erp G.M., Menken C.M., Veldpaus F.E. “The nonlinear flexuraltorsional behavior of
straight slender elastic beams with arbitrary crosssections”. ThinWalled Structures, 6(5), 385
404, 1988.
[9] FineLg User's Manual. V9.0. Greisch Info S.A.  Department ArGEnCo  Liege University
(ULg), 2005.
554
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
* Department of Civil and Environmental Engineering, Imperial College London, London SW7 2AZ
emails: a.insausti@imperial.ac.uk, leroy.gardner@imperial.ac.uk
Keywords: Compression; Elliptical hollow sections; Local buckling; Plastic mechanism; Steel structures.
Abstract. The plastic collapse response of structural steel elliptical hollow section (EHS) profiles in
compression is examined in this paper. As an initial step, a parametric study to identify the factors that
determine which plastic mechanisms would arise has been carried out using finite element (FE) results
from the current work and experimental data from the literature. Following this, an analytical model to
describe the “split flip disc” plastic collapse mechanism in compressed EHS is derived. The parameters
controlling the shape and size of the plastic hinges have been investigated and found to be of key
importance; hence, special care has been taken in their definition. Finally, the analytically derived load–
displacement curves have been compared with FE results. The comparisons have revealed good
agreement, confirming the ability of the developed analytical models to predict the plastic collapse
response of elliptical tubes.
1 INTRODUCTION
Hotfinished structural steel elliptical hollow sections (EHS) have recently been introduced to the
construction sector. These structural elements can offer greater structural efficiency than circular hollow
sections (CHS) when subjected to bending or combined loading, or when used as columns with
intermediate restraint about the weaker axis, since they posses different major and minor axis flexural
properties. Despite recent investigations involving the testing, numerical modelling and development of
design rules for EHS, a number of aspects of their structural response remain unexplored. In particular,
the behaviour of elliptical profiles in the post ultimate region has not yet been examined.
The aim of the present work is to develop an analytical model to predict the load–deformation
response of EHS under pure compression. To this end, rigidplastic theory has been applied to EHS based
on the method presented by Murray for plates [1]. As an initial step, a parametric study was carried out to
identify the most common local plastic collapse mechanisms arising in EHS. The study involved finite
element (FE) modelling and the analysis of existing test data [2]. Whilst a number of failure modes were
identified, an inward plastic collapse mechanism of the form shown in Fig. 1 was the most prominent. An
analytical description of this collapse mechanism is therefore the focus of the present study. The key
parameters controlling the size and shape of the hinge lines in the plastic mechanism were carefully
examined and their influence on the overall load–deformation response was assessed. Finally,
comparisons between the analytical model and the results obtained from the FE models are presented.
2 LITERATURE REVIEW
The intermediate response of elliptical tubes between that of flat plates and circular shells has been
previously identified in terms of elastic buckling [3]. In anticipation of an analogous scenario for plastic
collapse, previous studies on rigidplastic failure mechanisms in flat plates and circular shells are initially
reviewed. For rigidplastic mechanisms in flat plates, pioneering work was carried out by Murray [1],
who introduced a number of different failure modes and derived corresponding load–deformation curves.
Among the common platelike patterns identified, the so called “flip disc” mechanism was presented in
555
Aimar Insausti and Leroy Gardner
detail. This mechanism is formed by two curved hinge lines, one of them folding outwards and the other
inwards, creating a disc shape plate within the hinge lines that flips around the horizontal mid axis.
Research on compressed CHS has identified two main local plastic failure modes  the “elephant
foot” and “Yoshimura” mechanisms. The elephant foot is an axisymmetrical mode with outwards
deformation that creates a concertina shape in the surface, while the Yoshimura pattern includes several
sequential folding lines that deform the crosssection in a non axisymetrical way. The boundaries that
define the occurrence of each plastic failure mode were studied experimentally by Andrews et al. [4].
Both failure modes as well as a mixed mode were identified in the experimental work and a classification
chart to predict their occurrence was developed. Later, the available experimental data on compressed
CHS was expanded by Guillow et al. [5] carrying out more tests over a wider crosssection slenderness
range. As a consequence, a revised classification chart was developed. For the elephant foot mode, load
displacement curves were derived by Grzebieta [6] using the plastic mechanism approach. Further
experimental results on circular tubes specifically proportioned to develop axisymmetric failure modes
were presented by Gupta and Velmurugan [7] while Johnson et al. [8] studied the Yoshimura type
mechanism by means of tests on circular PVC tubes.
The key difference between the elements previously studied (i.e. flat plates and CHS) and elliptical
hollow sections lies in the continuously varying curvature brought about by the following geometrical
definition where the symbols are defined in Fig. 1.
z 2 y2
1 (1)
a 2 b2
With the recent introduction of hotfinished EHS into the construction sector, heightened interest in
the structural behaviour of elliptical profiles, as well as the need to develop design guidance, have
emerged. Structural performance data have been generated on elliptical sections in compression [2, 9]
and bending [10]. The result of the 25 compression tests (stub column tests) given in [2], together with
numerically generated results, have been used in the present study for the development and validation of
the analytical model. Based on the results of compression and bending tests [2, 10] and an analysis of the
elastic buckling of EHS, a crosssection slenderness parameter was derived for the purpose of cross
section classification [11]. The slenderness parameter was based on an equivalent diameter De, which
allowed the classification of EHS to be made on the basis of the CHS slenderness limits. For the pure
compression case, De = 2a2/b, which corresponds to the point in the section with maximum radius of
curvature, 2a and 2b being the larger and smaller outer dimensions respectively of the EHS, as shown in
Fig. 1. This point of the section was identified by Kempner [12] as suitable for use with the classical
formula for CHS in determining elastic buckling stresses for EHS. Later, more precise expressions for
determining the equivalent diameter were proposed [3, 13], and an alternative approach to EHS
classification, based on an equivalent rectangular hollow section has also been investigated [9].
556
Aimar Insausti and Leroy Gardner
experimentally. The member length was fixed at 300 mm, which was sufficiently short to ensure no
global buckling, and all sections were subjected to concentric compression. All models were assigned
rigidplastic material properties without strainhardening or residual stresses to allow direct comparison
with the analytical model developed herein.
Fixed boundary conditions were applied to the ends of the columns, with all degrees of freedom,
other than vertical displacement at the loaded end, restrained. Initial geometric imperfections were
introduced in the shape of elastic buckling modes obtained from eigenvalue analyses. Three modes were
considered, in which the number of half sine waves along the stub column length was either odd or even,
the imperfection was either positive or negative (positive being inwards at the midheight of the stub
column) and the imperfection was either symmetrical or asymmetrical about the midheight. Three
imperfection amplitudes were also considered: t, t/10, t/100, where t is the section thickness.
Figure 1. (a) Split flip disc failure pattern observed in tests and (b) illustration of the mechanism.
Variation of the initial geometrical imperfections described in the previous section trigger four
different plastic failure patterns, two of them akin to platelike behaviour and the other two akin to shell
like (CHS) behaviour. Within the plate failure modes, one of them is similar to the flip disc mechanism
proposed by Murray [1] for flat plates, while the other is a variation of it. This variation includes an extra
straight hinge line in the middle of the mechanism splitting the disc into two half parts as shown in Fig. 1.
This plastic mechanism is referred to herein as the split flip disc (SFD) mechanism. The shelllike plastic
557
Aimar Insausti and Leroy Gardner
collapse mechanisms observed in the EHS are the elephant foot and the Yoshimura patterns. All four
mechanisms are shown in Fig. 2.
The specific failure mode that a given EHS profile would succumb to was found to be influenced by
the shape and amplitude of the initial geometric imperfection, as well as the slenderness (De/t, where
De = 2a2/b) of the crosssection. The FE models showed that platelike failure modes dominate the
profile’s behaviour for smaller (and more practical) levels of initial imperfection, and that, overall, the
split flip disc mechanism was the most common.
The experimental work presented in [2] has also been used to identify the plastic failure mechanisms
in compressed EHS. The tests exhibited three of the four modes revealed in the numerical study: the flip
disc, the split flip disc and the elephant foot modes. However, the elephant foot was present only in four
out of 25 tests. The flip disc and the split flip disc modes appeared in the remaining 21 tests in no clear
pattern; overall, the split flip disc mechanism appeared more frequently. Hence, the most commonly
arising mode in both the experimental and numerical studies was the split flip disc mode; consequently,
development of an analytical description of this plastic failure mechanism is the focus of this paper.
(a) Flip disc (b) Split flip disc (c) Elephant foot (d) Yoshimura
Figure 2. Plastic collapse mechanisms obtained from FE models for EHS.
5 ANALYTICAL MODELLING
In this section, an analytical model for the split flip disc failure mode arising in compressed EHS is
developed based on the plastic theory for thinwalled structures presented by Murray [1]. As stated by
Murray, assuming that the material stressstrain curve is a step function with a step height between tensile
and compressive yielding of 2fy, where fy is the material yield strength, a crosssection’s loadcarrying
capacity can be derived as a function of the displacement from equilibrium, based on an assumed plastic
collapse mechanism. Rigidplastic material behaviour is therefore used, which neglects strain hardening,
and assumes that all deformation is localised along the hinge lines with no deformation elsewhere.
The moment capacity of a plastic hinge, based on a rectangular element of width b and thickness t is:
fy b t2
M pl (2)
4
And the reduced plastic moment M'pl in the presence of an axial load N may be shown to be:
§ N2 ·
Mcpl M pl ¨1 2 ¸ (3)
¨ Ny ¸
© ¹
where N is the axial load and Ny = fybt is the yield load in the element. Eq. 3 applies when the hinge
line is perpendicular to the direction of the thrust. However, for an inclined hinge line, the reduced plastic
moment is given by Eq. 4, where ȕ is the angle between the line perpendicular to the thrust and the
studied inclined hinge line.
Mcplc Mcpl sec2 E (4)
558
Aimar Insausti and Leroy Gardner
In the present work, both parabolic and elliptical functions were considered for the description of the
curved hinge lines in the plastic mechanism. However, the parabolic hinge lines were found to more
accurately replicate the actual loaddisplacement response of EHS, thus, this shape is used throughout the
present derivation. A parabola can be defined using two parameters that fix the height Xh and the width
Sh of the curve on the surface of the EHS. Hence, the values of Xh and Sh define the mechanism along the
development of the plastic hinge (Eq. 5). Fig. 1 summarises the notation used in the present section to
define the modelled failure mechanism, as well as the angle ȕ used in Eq. 4.
§ s2 ·
x (s) X h ¨¨1 2 ¸¸ (5)
© Sh ¹
With reference to Fig. 1, the load transmitted by the complete crosssection can be obtained in terms
of the load inside the plastic hinges (Nin) plus the load outside the plastic hinges (Nout). Furthermore, by
using symmetry, only one quarter of the crosssection needs to be analysed with the result for the full
crosssection being factored accordingly. The contribution to the loadcarrying capacity of the cross
section from within the hinge lines and outside hinge lines are derived in the following two subsections.
5.1 Loadcarrying contribution within the plastic hinges
Since the inclination of the hinge lines is variable around the crosssection, a differential strip of
material is analysed, as depicted in Fig. 1, with the following reduced plastic moment:
2
f y t 2 §¨ § dN · ·¸
M cplc ¨ ¸ 2
(6)
¨1 ¸ sec E dS
4 ¨ ¨ f y t ds ¸ ¸
© © ¹ ¹
in which dN is the load in the strip and dS is the strip width. Fig. 3(a) presents the free body diagram
of the material strip, that relates the lateral displacement of the strip ǻds to the applied load dN and the
reduced plastic moment. The bending moment diagram in the material strip is presented in Fig. 3(b). The
load borne by the strip dN can be related to the lateral displacement by considering equilibrium at a null
bending moment point. For the split flip disc mechanism, the point of zero bending moment lies between
the curved hinge line and the straight hinge line that splits the flip disc into two equal parts. Denoting ǻ'
the straight distance between the null bending moment point and the undeformed position (see Fig. 3(a)),
this distance can be related to ǻdS through the magnitudes of M'pl and M''pl, and consequently as a
function of the plastic hinge inclination ȕ:
M cplc sec 2 E
'c ' dS ' dS (7)
M cpl M cplc 1 sec 2 E
Figure 3. (a) Free body diagram and (b) bending moment diagram of the material strip.
Having obtained ǻ', the load borne by the strip dN may be related to the reduced plastic moment:
dN 'c M cplc M cpl sec 2 E (8)
Merging Eq. 7 and Eq. 8, introducing the reduced plastic moment in the material strip (Eq. 6) and
rearranging, we obtain.
559
Aimar Insausti and Leroy Gardner
§ 2 ·
¨ § 2 ' dS · 2 ' dS ¸
dN f y t ¨ ¨¨ 2
¸
¸ 1 2 ¸dS (9)
¨ © (1 sec E ) t ¹ (1 sec E ) t ¸
© ¹
The lateral displacement in the strip ǻdS can be related to the maximum lateral displacement in the
hinge ǻ through Eq. 10, where x(s) defines the parabolic shape of the curved hinge line (Eq. 5).
x (s ) § s2 ·
' dS ' ' ¨¨1 2 ¸¸ (10)
Xh © Sh ¹
From Eq. 9 and 10, and following some manipulation, we obtain the load carried by each strip as:
§ 2 ·
¨ § S2h (S2h s 2 ) ' · S2h (S2h s 2 ) ' ¸
dN f y t ¨ ¨ 4
¨
2 2 ¸ 1 (S4 2 X 2 s 2 ) t ¸dS
¸ (11)
¨ © (Sh 2 X h s ) t ¹ h h ¸
© ¹
Eq. 11 can not be integrated explicitly; hence, in the present work Simpson’s rule has been employed
to obtain the load–lateral displacement curve, as advised in [1]. The load in the strip was evaluated at
s = 0, s = Sh/2 and s = Sh, leading to the following relationship between load within the plastic hinges Nin
(for one quarter of the section) and lateral displacement ǻ for the split flip disc mechanism:
f y t Sh § ' 6S2h ' '2 9S4h '2 ·
N in ¨1 1 4 1¸ (12)
¨ 2 2 2 2 2 2 2 ¸
6 t ( 2Sh X h ) t t 4( 2Sh X h ) t
© ¹
5.2 Loadcarrying contribution outside the plastic hinges
The area of the crosssection outside the plastic hinges is considered to be working at the yield stress.
Hence, the loadcarrying contribution from outside the plastic hinges is proportional to the arc length
outside the hinge Sout, and is given for one quarter of the section as:
N out f y t Sout (13)
The total load carried by the full crosssection is obtained from Eq. 12 and Eq. 13 as:
N ( ' ) 4N out N in (14)
560
Aimar Insausti and Leroy Gardner
140
120
100
Xh (mm)
80
60 ABAQUS data
40 Tests data
20 Eq. 15
0
0 20 40 60 80 100 120
(Det)0.5 (mm)
Figure 4. Xh values from the FE models and the tests.
The FE models showed that the Sh parameter increased more rapidly at the beginning of the plastic
deformation than when the mechanism was fully developed. Hence, it was chosen to represent Sh with a
rational expression of the form given by Eq. 16 that tends to Sh,a, and where C is a constant.
'
Sh Sh,a (16)
'C
Based on observations of the test failure patterns, it was found that the maximum extent of the plastic
mechanism was approximately 75% of the way around the quarter perimeter of the section – i.e.
Sh,a = 0.75P/4 = 3P/16, where P is the perimeter of the ellipse. Furthermore, from the FE results, it was
found that C = t provided a good approximation of the progression of Sh towards its asymptotic value.
Hence, Eq. 17 was established:
3P § ' ·
Sh ¨ ¸ (17)
16 © ' t ¹
0.6 0.6
N/Afy
N/Afy
0.4 0.4
0.2 0.2
0.0 0.0
0 10 20 30 40 0 10 20 30 40
ǻ (mm) ǻ (mm)
(a) t = 4 mm (b) t = 8 mm
Figure 5. Loadlateral displacement comparisons for 150×75 EHS with (a) t = 4 mm and (b) t = 8 mm.
561
Aimar Insausti and Leroy Gardner
7 CONCLUSIONS
Local plastic collapse mechanisms in compressed EHS have been examined in this study. Four
collapse mechanisms were identified, two of which were akin to platelike behaviour and two to shell
like behaviour. A numerical study, coupled with examination of existing test data, revealed that the so
called split flip disc mechanism arose most frequently. Hence an analytical model to describe the load–
lateral displacement response of EHS under pure compression following this failure pattern was derived.
Simple expressions, in terms of section geometry, to determine the key parameters required to fully
describe the shape of the collapse mechanism were developed. Comparisons between the analytical
model and FE model revealed good agreement over a range of crosssection slenderness, with a
maximum discrepancy of 6.7%. It is concluded that the derived analytical model provides an accurate
means of predicting the loadlateral displacement response of a compressed EHS undergoing local plastic
collapse in the split flip disc mechanism.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the Basque Government (Department of Education,
Universities and Research) for the financial support given under the overseas postdoctoral development
scheme in 2009 and 2010.
REFERENCES
[1] Murray NW. Introduction to the theory of thinwalled structures. Oxford University Press 1984.
ISBN 0198561865.
[2] Chan TM and Gardner L. Compressive resistance of hotrolled elliptical hollow sections.
Engineering Structures 2008;30(2),522–532.
[3] RuizTeran A and Gardner L. Elastic buckling of elliptical tubes. ThinWalled Structures
2008;46(11):13041318.
[4] Andrews KRF, England GL and Ghani E. Classification of the axial collapse of cylindrical tubes
under quasistatic loading. International Journal of Mechanics and Science 1983;25(9):687696.
[5] Guillow SR, Lu G and Grzebieta RH. Quasistatic axial compression of thinwalled circular
aluminium tubes. International Journal of Mechanical Sciences 2001;43(9):2103–2123.
[6] Grzebieta RH. An alternative method for determining the behaviour of round stocky tubes
subjected to an axial crush load. ThinWalled Structures 1990;9(14):6189.
[7] Gupta NK and Velmurugan R. An analysis of axisymmetric axial collapse of round tubes. Thin
Walled Structures 1995;22(4):261274.
[8] Johnson W, Soden PD and AlHassani STS. Inextensional collapse of thinwalled tubes under axial
compression. Journal of Strain Analysis 1977;12(4):317330.
[9] Zhao XL and Packer JA 2009. Tests and design of concretefilled elliptical hollow section stub
columns. ThinWalled Structures, 47(67): 617628.
[10] Chan TM and Gardner L. Bending strength of hotrolled elliptical hollow sections. Journal of
Constructional Steel Research 2008;64(9):971986.
[11] Gardner L and Chan TM. Crosssection classification of elliptical hollow sections. Steel and
Composites Structures 2007;7(3):185200.
[12] Kemper J. Some results on buckling and postbuckling of cylindrical shells. Collected papers on
instability of shell structures. NASA TND1510, Dec. 1962:173186. Polytechnic Inst. Brooklyn.
[13] Silvestre N. Buckling behaviour of elliptical cylindrical shells and tubes under compression.
International Journal of Solids and Structures 2008;45(16):44274447.
562
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Institut for Steel Structures and Material Mechanics, Technische Universität Darmstadt, Germany
merle@stahlbau.tudarmstadt.de, lange@stahlbau.tudarmstadt.de
Abstract. The collapse of a steel structure required a detailed analysis of the reasons for the catastrophe.
Several parts of the whole process of the design and construction had to be assessed. The results of this
analysis are presented. Furthermore it will be shown how the design process, the autonomous checking
procedure of the structural design, the workmanship, the construction on the erection side, the material
characteristics as well as human failure affect the failure.
1 INTRODUCTION
Trusses with hinge joints experience only tension and compression forces. Compression forces in
combination with hinge joints lead to a structural design using the well known Euler’s cases especially
the Euler case II. By designing the hinge joints the mounting of the joints has to be considered. This asks
for a long connection area. Due to the compression, the joints have to be as compact as possible. These
two oppositional requirements have to be considered in the design and the construction. The analysis of a
collapse of a truss within the construction of a coal power plant showed that the failure of noncompact
joints led to a catastrophe. The noncompact joints increased the buckling length severely beyond the
Euler case II. The collapse of the joints led to a global collapse of the truss.
563
Heiko Merle, Jörg Lange
was stiffened with many braces. The trusses outer diagonals were hollow sections with a height and width
of 350 mm and a thickness of 30 mm with a steel grade of S 355.
Level + 82.740m
4740
Level + 78.000m
36500
A portal frame transferred the loads from the sections standing above into the truss. This structure
may be reduced to the single girder with vertical loads from the portal frame. A static analysis of the
upper auxiliary steel structure as a twodimensional framework led to four main vertical loads for the
girder. The loading points were the outer joints of the upper chord and the outer diagonal. These loads
were design loads with specific partial safety factors. Figure 2 shows the loads and loading points.
Level + 82.740m
Level + 78.000m
3 INTERNAL FORCES
The inner diagonals as well as the vertical struts were designed for small forces with max
Nd = 130 kN. The bottom chord got a tension force of Nd = 3300 kN. The upper chord received a
compression force of Nd = 2500 kN. Due to its large dimension buckling was eliminated. At last the outer
diagonals were analyzed. They receive a compression load of Nd = 4160 kN. A comparison of the original
structural design and a new and autonomous structural design after the collapse led to the same
compression and tension forces. All members were strong enough to carry the applied loads.
564
Heiko Merle, Jörg Lange
π2 π2 (1)
N ki ,z = ⋅ EI = ⋅ 21000kN / cm 2 ⋅ 66112cm 4 = 22442kN
sk
2
(781cm)2
By using the German code DIN 18800 [2] the effective slenderness can be calculated with the help of
the buckling curve c.
The ultimate load is twice as big as the compression force of the outer diagonal. Therefore a collapse
of the outer diagonal according to Euler’s case II is not expected.
A
0
40
5
70
400
1 0
78
t=20
35
M27  10.9
0
M27  10.9
4
70
Plate 30 x 295; S355 AA
1
58 Plate 30 x 350; S355
t=20
HEB 700; S355 0
40
730
300
5
70
A
20
54
4
70
1
58
565
Heiko Merle, Jörg Lange
By analyzing the workshop drawings it can be assessed that the support conditions are not as assumed
in Euler’s case II. The top and bottom flanges of the hollow section were cut out at the end of the beam to
enable the boltability. Figure 4 shows an overview of the outer diagonal and its joints. Special attention
shall be given to the 704 resp. 705 mm of the compression strut that are without top and bottom flange.
The hollow section is connected with the joint plate by bolts. The joint plate is connected to the chord
by a welding seam. Therefore the modeling of the outer diagonal to the Euler’s case is incorrect. The
slenderness of the joint plate and its length lead to the model of a segmented beam with different
stiffnesses. Rotation springs as well as translation springs have to be used at the end of the segments.
Whereas the Euler’ case II has an elastic buckling mode the segmented beam has a combination of an
elastic buckling and a rigid body motion (fig. 8). The stiffness of the springs as well as the relative
stiffness of the equivalent end segments built by the joint plates are difficult to define. Therefore the
whole truss was modeled as a threedimensional finite element system.
5.1 Finite element method – elastic eigenvalue analysis
By using the finite element software SOFiSTiK the top as well as the bottom chord, the outer
diagonals, and the joints of the girder were modeled with shell elements. The stiffeners of the upper chord
were modeled too, to avoid local buckling effects. The inner diagonals and the vertical compression struts
were modeled using beam elements. A very fine discretization of the model was used to get results as
close to reality as possible. The model was supported as shown in figure 1. To avoid a horizontal
displacement of the system a support in the point of symmetry was modeled.
The material characteristics were ideal elastic and ideal plastic functions with a yield strength of
fy = 240 N/mm² (S235) resp. fy = 360 N/mm² (S355) and a modulus of elasticity of E = 210000 N/mm².
The loads were as shown in figure 2. Figure 5 gives a view of the finite element model.
Figure 6 (left) shows the connection between the outer diagonal and the bottom chord whereas figure
6 (right) shows the connection between the diagonal and the upper chord.
Upper joint
plate
Bottom joint
plate
By using the finite element model and the loads an elastic eigenvalue analysis was done. 10% of the
design loads were applied to the system. Thirty eigenvalues with the associated buckling modes were
566
Heiko Merle, Jörg Lange
calculated. The eigenvalues can be divided into two major groups. Firstly there are many values with
local buckling figures of the slender web of the upper chord. These values will not be pursued. The other
group is made up of eigenvalues with buckling modes containing both outer diagonals.
The result of this analysis is shown in figure 7. Eigenvalues 1 to 3 are buckling modes of the outer
diagonals. The first and second mode is a rigid body motion of the diagonal with a sideways buckling of
the bottom joint plates. There is a symmetric rigid body motion of both diagonals in eigenvalue 1 whereas
in eigenvalue 2 an asymmetric rigid body motion can be observed. The third eigenvalue describes a rigid
body motion of the diagonal with a sideways buckling of the top joint plates. The first eigenvalues are
close together between the load factor 13.1 and 14.1. By using 10% of the design load, the ideal buckling
load can be calculated.
Figure 7 shows the 27th eigenvalue and the buckling mode analogous to Euler’s case II. This buckling
mode has an ideal buckling load of Nki,Euler = 24200 kN with the load factor of 59.1. Compared with
formula (1) there is an explicit accordance between the two ideal buckling loads. This shows the usability
of the finite element model. Altogether the first three eigenvalues show a rigid body motion. The failure
mode is not the buckling of the outer diagonal but rather the buckling of the joint plates. The joint plates
act like a spring at the end of the diagonal. Schmidt et al. [3] showed an equivalent system that is given in
fig. 8.
Joint plates
Outer diagonal
Euler's case two Rigid body motion
5.2 Finite element method –plastic and nonlinear ultimate load analysis
To get the ultimate load of the finite element model a plastic and nonlinear load analysis was used.
Therefore the ideal elastic and ideal plastic function of the material characteristics was used. To include
the nonlinear secondary order theory the first eigenvalues buckling mode was fit in the model as its
imperfection. To take a realistic value the imperfection was set to 1/500 of the length of the joint. Then an
567
Heiko Merle, Jörg Lange
ultimate load iteration was started. Within load steps the design load was raised to the ultimate load of the
system.
Design load factor
1
0,8
Ultimate load factor
Load factor
0,6
0,4
0
0 2 4 6 8 10 12 14 16 18 20
Sidways deformation [mm]
Figure 9: Sideways deformation of the joints in the plastic and nonlinear load analysis
Figure 9 shows the joint plates sideways deformation of the connection between the chords and the
diagonals which are shown in figure 6. The plates get a clear nonlinear sideways deformation. At the load
factor 0.69 the deformations will be infinite and the outer diagonals as well as the joint plates fail. At this
load factor the system analysis is not convergent and the whole system collapse at 69% of the design
loads. The deformation figure in the last convergent load step is affine to the first buckling mode of the
first eigenvalue.
568
Heiko Merle, Jörg Lange
As the finite element system showed, the connections between the outer diagonals and the chords fail
first. In this moment the whole system is still stable. The inner forces of the truss without the diagonals
create bending moments in the bottom chord. These bending moments create the extreme bending
deformation shown in figure 10. Due to the large deformations the whole girder slips off the support and
the system collapses.
7 CONCLUSION
In the process of design and construction of steelwork structures there is a straight progression.
Firstly the structural analysis gives the loads and forces as well as profiles and defined materials.
Subsequently the detail design with workshop drawings and detail drawings follows.
The truss was firstly designed as a framework. All structural elements and their material definitions
were determined by using the theory of stability respectively the elastic second order theory. This
structural analysis was done by the structural engineer. After this the design engineer determined the
structural details. He has to be aware of problems like the boltability and the erection. The design
engineer receives the forces and the stresses from the structural analysis. The load flow in the details had
to be analyzed. The discrepancy between a compact design of the connection and montage aspects as well
as the boltability had to solved. Therefore the structure had to fulfill structural restraints as well as the
static requirements. After the detail design was finished the structure was fabricated.
There was no feedback loop between the engineers. Figure 11 shows that after the construction of the
details the structural engineer should have checked the design. The outer diagonals in coherence with the
new designed details had to be analyzed again. There the engineer had to notice that an important
alteration of the support conditions of the outer diagonal happened. An updated structural analysis should
have been done. This might have led to the recognition of the structures ultimate load and failure
criterion. The design engineer would have had to redesign the details to fulfill the requirements of the
structural design. This kind of feedback loop will take more time but maybe the collapse of the girder
could have been avoided.
569
Heiko Merle, Jörg Lange
REFERENCES
[1] Petersen C., “Stahlbau”. Vieweg , Braunschweig , 1993
[2] DIN 188002 “Steel structures – Part 2: Stability – Buckling of bars and skeletal structures”,
2008/11
[3] Schmidt H. et al., “Ein ungewöhnliches Stabilitätsproblem verursacht Schadensfall – An
uncommon stability problem causes failure”, Stahlbau 77 (12), 862869, 2008
[4] Unterweger, H.; Ofner, R, “Traglast von Verbandsstäben aus Hohlprofilen mit quasizentrischem
Knotenblechanschluss – Load bearing capacity of bracing members with almost centric joints“,
Stahlbau 78 (6), 425436, 2009
570
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Department of Civil Engineering, McMaster University, Hamilton, ON, CANADA, L8S 4L7
email: siva@mcmaster.ca
Key words: Experimental, Steel beams, Flange holes, Ductility, Steel standards
Abstract: This paper presents an experimental investigation, involving twenty five steel beam specimens,
on the effects of flange holes on the flexural behaviour of steel Ibeams. This study used ASTM A992
grade steel beams. Circular holes of various diameters, ranging form 0% to 48% of the gross flange area
are under consideration Based on the experimental results, this research study recommends a design
approach analogous to the axial tension member provisions as per the current CAN/CSAS16.01
standard [1]. The comparison of the proposed procedure with the 15% exemption rule as per current
steel standard S16.01 [1] demonstrated that the current code provision is unnecessarily conservative for
steel grades such as A992 steel. On the other hand, the current code provision may not be adequate for
higher strength steels such as HSLA 80 steel, ASTM A913 Gr: 60 and HPS485W having the minimum
yieldtoultimate strength ratio of more than 0.85.
1 INTRODUCTION
Flange holes are frequently made in structural steel construction, primarily for bolting purposes. The
influence of flange holes on the flexural behaviour of beam members has been the focus of debate for
many years. Early North American design codes allowed a designer to place holes in flanges up to 15% of
the gross area of the tension flange without penalty. If more than 15% of the gross flange area is removed,
the amount of area exceeding 15% would be deducted in calculating the section properties and typically,
only the yield moment could be used rather than the plastic moment. This provision was based on the
study by Lilly and Carpenter [2] on riveted plate girders made of ASTM A7 steel having the yieldto
ultimate strength (Fy/Fu) ratio of about 0.5. However, in 1989, the Allowable Stress Design version of
specification [3] adopted a new provision that altered the use of the 15% exemption rule in this subject
matter. This specification introduced for the first time a mathematical formula based on the ratio of the
fracture strength of net area (AfnFu) and the yield strength of gross area (AfgFy) of the tension flange to
ignore the effects of holes. The present AISCLoad and Resistance Factor Design version [4] of the
specification also follows the same procedure as specified in the 1989AISC specification [3] to ignore
the effect of flange holes.
The present trend in steel construction industry is to use higher strength steels with better structural
performance over traditionally used ASTM A36 steel. These high strength steels have the specified yield
toultimate strength ratio ranging from 0.75 to a code permitted maximum of 0.85. In some instances,
steels such as HPS485W, HSLA 80 steel and ASTM A913 Gr: 60 exhibit yieldtoultimate strength
values of more than 0.85[5]. Nevertheless, the comparisons of corresponding various international code
provisions indicate that the 15% exemption rule which is currently in use as per the clause 14.1 of the
current Canadian Steel Design Code [1] is more restrictive for steel grades having the yieldtoultimate
strength of less than 0.85, whereas it is inadequate and inappropriate for the high strength steels with the
minimum yieldtoultimate strength of more than 0.85.
571
K.S.Sivakumaran et al.
572
K.S.Sivakumaran et al.
573
K.S.Sivakumaran et al.
3 TEST RESULTS
Some of the results reported herein include normalized quantities of (1) load versus midspan
deflection, (2) moment versus load point rotation and (3) moment versus beam end rotation. The rotation
is the average of the rotations measured underneath the two load points. Table 1 presents the measured
peak moments Mm associated with each test and the theoretical grosssection plastic moment MP of each
test specimen. Theoretical MP considers the openings and the resulting neutral axis shift. Table 1 also
provides the percentage reduction in strength as compared to the solid beams (see column 7).
Table 1: Comparison of Experimental Peak Moments with Theoretical Plastic GrossSection Moment
Ave
Type [Afn Mm Mave %difference
Beam [AfnFu/ MPave rage
of /Afg] (Test) (Test) compared to
ID AfgFy] (kNm) Mm/
Test (%) (kNm) (kNm) solid section
(2) (4) (8) MP
(1) (3) (5) (6) (7)
(9)
A1001 100 1.30 215
Series1 A1002 100 1.30 214
214 0.0 176 1.22
A1003 100 1.30 214
A1004 100 1.30 214
A901 91 1.18 214 214 0.0 176 1.22
A851 85 1.10 216 216 0.9 176 1.23
A801 79 1.03 212 212 0.9 174 1.22
Series2 A751 74 0.96 210 209 2.3 175 1.19
A75
A7012 74
71 00.92
96 206
204 204 4.7 176 1.16
A70
A6012 71
62 00.81
92 205
197 195 8.9 174 1.12
A60 2 62 0 81 194
A501 52 0.67 178 178 16.8 172 1.03
A85B1 86 1.17 210 210 1.8 178 1.18
Series3 A75B1 74 0.96 200 200 6.5 176 1.14
A70B1 70 0.91 197 197 7.9 176 1.12
A60B1 63 0.82 192 192 10.3 179 1.07
A85F1 85 1.10 212 212 0.9 175 1.21
A75F1 74 0.96 210 210 1.9 174 1.21
Series4 A70F1 70 0.91 207 207 3.3 177 1.17
Series1: Solid Beam Tests: The maximum moment carrying capacity of solid beams A1001, A100
2, A1003 and A1004 were 215 kN.m, 214 kN.m, 214 kN.m and 214 kN.m, respectively. However, the
corresponding measured average load point rotations corresponding to peak moment were of 0.0938,
0.0972, 0.0949 and 0.0878 radians, respectively, resulting in the maximum deviation from the average
measured rotation (0.0934 radians) of approximately 6%. The normalized moment (M/MP) versus the
normalized load point rotation (Ө/ӨP) relationship for each solid beam was established. The moment
versus load point rotation relationship was in close agreement up to the peak moment, even though slight
variations were observed perhaps due to the inherent variability associated with the presence of residual
stresses and initial geometric imperfections. Two different rotation capacities such as Ry (a measure of
574
K.S.Sivakumaran et al.
available rotation capacity corresponding to the plastic moment MP obtained on the unloading branch)
and Rm (rotation capacity at peak moment) were established in this research program. The average
measured Ry and Rm of the solid beam specimens were 23.5 and 13.1, respectively. The failure of the
solid beam was due to local flange buckling of the compression flange which was followed by lateral
torsional buckling in the critical span region.
Series2: Beams Having Holes in Tension Flange Only: Figure 4 shows the normalized moment,
M/MP versus the normalized average load point rotation, Ԧ/ԦP for the beam specimens with holes in
tension flanges only. In order to illustrate how the flexural behaviour of steel member could be influenced
due to the presence of holes in the tension flanges, the momentrotation response of a solid beam (A100
3) is also shown in the same figure. Figure 5 shows a close up view of failure pattern of the beam
specimen (A603) failed as a result of netsection fracture through the holes in tension flange.
Figure 4: Normalized Moment Versus Figure 5: Failure Pattern of beam with holes in
Normalized Load Point Rotation Tension Flange
From figure 4, it can be noted that the rotation capacities of the flexural members were reduced even
when the holes removed was small, say approximately 10% (A901). However, it can be observed that the
strength of the flexural members was not significantly impacted provided the nominal netsection fracture
strength was greater than nominal grosssection yield strength (AfnFuAfgFy). This ratio is given in Table
1 Column 4. The percentage reduction in strength (Table 1 Column 7) increased as the AfnFu/AfgFy ratio
became lower than 1.0. Thus, for beam specimens having the AfnFu/AfgFy ratio of 0.96 (26% flange holes
of gross flange area) and 0.92 (29% flange holes of gross flange area), the percentage reductions in the
average maximum load were of 2.3% and 4.7%, respectively, compared to that of solid beams. These
specimens having the AfnFu/AfgFy of 0.96 and 0.92 eventually failed by net section fracture which
occurred after visible local bulking of the compression flanges in the uniform moment region which can
be seen in Figure 5. However, for beam specimens having the AfnFu/AfgFy ratio of 0.81 (38% holes of
gross flange area) and 0.68 (48% holes of gross flange area), which were well below 1.0, the percentage
reductions in the average maximum applied load were of 8.9% and 16.8%, respectively. Such beam
specimens failed by netsection fracture, prior to local bucking of compression flange. The reduction in
the moment capacity of beam specimens, having the AfnFu/AfgFy>1.0, was not substantial, although a
slight reduction did occur with increasing hole size. Based on these results, suppose it is presumed that
any strength reductions within ±5% range can be ignored from a design stand point, then the tension
flange holes of up to 29% of the gross flange area can be safely ignored in beams made of ASTM A992
steel having yieldtoultimate strength ratio of 0.77. Table 1 Column 9 gives the ratio of test moment to
theoretical moment resistance. For series2 specimens, since the Mm/MP for all specimens were more than
1.0, it can be concluded that the tension flange rupture did not occur prior to the attainment of the gross
section plastic moment, when the holes removed was from 9% to 48%.
575
K.S.Sivakumaran et al.
Series3: Beam Having Holes in Both Flanges: As presented in Table 1, the percentage decrease in
the moment capacity of beam specimens with holes in both flanges (Series3) having the AfnFu/AfgFy
ratio of 1.17, 0.96, 0.91 and 0.82, compared to the solid beams, were 1.8%, 6.5%, 7.9% and 10.3%,
respectively. As expected, the flexural behaviour of beam specimens in terms of strength and rotation
capacity was considerably influenced as holes were present in both flanges. The percentage decrease in
the maximum moment capacity of beam specimens having the A fnFu/AfgFy ratio of 1.17 (A85B1), 0.96
(A75B1), 0.91 (A70B1) and 0.82 (A60B1), compared to the corresponding beam specimens having
holes in the tension flanges only (Series2) were of 2.7%, 4.2%, 3.2% and 1.4%, respectively. The beam
specimen having the AfnFu/AfgFy ratio of 0.96 (26% holes of gross flange area), which is closer to 1.0,
failed due to local buckling of the compression flange whereas, similar beam specimen (A fnFu/AfgFy
=0.96) having holes in the tension flange only failed due to netsection fracture of the tension flange,
which occurred after noticeable local buckling of the compression flange. This can be attributed to the
fact that the compression flange was weakened locally due to the presence of holes which resulted in
early yielding of the locally buckled compression flange. However, the beam specimens having the
AfnFu/AfgFy ratio of 0.91 and 0.82 failed due to tension fracture.
Series4: Beam Tests Having Holes With Fasteners in Both Flanges: These tests were somewhat
similar to Series3 tests, in that both set of beams had holes in both flanges, except that fasteners were
present in the holes in the current set of beams. For beam specimens having the AfnFu/AfgFy ratio of 1.10
(A85F1), 0.96 (A75F1), 0.91(A70F1) and 0.82 (A60F1), the percentage reduction in the maximum
moment capacity in compared to similar solid beam specimens, were of 0.9%, 1.9%, 3.3% and 10.3%,
respectively. The moment capacities of beam specimens were greatly improved when the holes in the
compression flanges were filled with the standard size of fasteners. The percentage improvement in the
maximum moment capacity of beam specimens having the AfnFu/AfgFy ratio of 1.10, 0.96, 0.91 and 0.81
when compared to the similar beam sections having holes in both flanges were of approximately 50%,
71%, 58%, and 10%, respectively. This clearly indicated that the presence of fasteners within the holes in
the compression flanges improved the moment resistance of beams with flange holes.
576
K.S.Sivakumaran et al.
ratio was less than 1.0. Also, the failure of beam specimens in such cases was mainly due to local
buckling of the compression flange preceded by lateral torsional buckling in the critical span. On the
other hand, as the AfnFu/AfgFy ratio became less than 0.95 the MP/Mfn ratio was increased to more than
1.0. The failure of beam specimens in such cases was mainly due to netsection fracture of the tension
flange. By considering all the scenarios tested in this research program, a factor of 0.85 can be considered
as an optimum upper bound that should be used to multiply the theoretical netsection fracture moment
(Mfn). Therefore, this research study suggests a design check, which is analogous to the tension member
provision as per the current CAN/CSAS16.01 [1] standard,; (a) The grosssection shall be designed for
the grosssection plastic moment capacity, MP (ZgFy) (or lower if compression flange or web limit states
control) (b) Calculate the factored netsection fracture moment, Mfnf =0.85ZnFu. If MP ≤ Mfnf, the effects
of holes (or fastener holes) shall be ignored and the flexural member shall be designed for its gross
section plastic moment as usually followed in the design solid beams. Otherwise, design the member to
carry the factored netsection fracture moment.
Overall, the design moments calculated as per the proposed design approach resulted in higher design
moments than that permitted by the current code provisions for flexural members having either flange
holes or flange fastener holes. The 15% exemption rule, which is still in use as per the current CAN/CSA
S16.01 (Clause 14.1) code provision [1], is conservative for currently used structural steels which often
possess a yieldtoultimate strength ratio of less than 0.85. Note that the design moment as per the
proposed design procedure in this investigation has a reduction factor of approximately 0.85 as compared
to the maximum measured moments associated with the netsection observed in this investigation.
Moreover, the suggested design method was analogous to the tension member provision as per the current
CAN/CSAS16.01 (Clause 13.2) code provision [1] eliminating unnecessary ambiguity in regards to the
design of flexural members having holes (or fastener holes) in tension flanges. That is, the clause: 14.1 of
the current CAN/CSAS16.01 standard [1] treats the effects of holes and the effects fastener holes in
different manner, in which when holes occur in flanges a theoretical netsection calculations shall be
followed whereas, when fastener holes in beams is considered, the 15% exemption rule would be applied.
However, the proposed method in this investigation follows a unified approach, in which the effects of
holes or fastener holes that may present in flanges of a flexural member or a tension member would be
treated in an identical manner. In addition, the proposed method as opposed to the current CSA code
provision [1] takes into account the material characteristics in terms of yieldtoultimate strength ratio.
5 CONCLUSIONS
The following points summarize the main observations of this research program:
[a] Experiments considered ASTM A992 steel with the measured yieldtoultimate strength ratio of
0.77 beams having flange holes as high as 48% of the gross area of the tension flange. Though tension
flange rupture was observed in some cases, the peak moments in all of the beams were higher than the
grosssection plastic moment (MP) for the beam [b]The strain measurements indicated that no great
deviation occurred with regards to the position of the neutral axis of the gross crosssection when holes
when holes were made in the tension flange only (or fastener holes occurred in both flanges) [c]The strain
measurements made in the vicinity of hole region of beam specimens A753 and A752, in which holes
existed in tension flanges only, were about 1.2% and 2%, respectively when the beam members reached
the grosssection plastic moment, MP. This yielded a conclusion that the flexural members with holes in
tension flanges only require a strain in the range of 610 times the yield strain (0.2%) for the ASTM A992
steel as has been already verified by Dexter et al.[5] who performed flexural tests made of HPS 480W
steel grade. [d]When holes were present in the tension flange only, and for the cases of fastener holes in
both the tension flange and the compression flange, the failure of flexural members having the
AfnFu/AfgFy≥1.0 was primarily due to lateral torsional buckling which was eventually followed by local
buckling in the critical span (midspan) region. It was noted in such cases that the grosssection plastic
momenttothe netsection fracture moment (MP/Mfn) ratio was less than 0.85. [e] The design moment
calculation as per the proposed design method was quite comparable with the present AISC code
577
K.S.Sivakumaran et al.
provision [4]. However, beyond a threshold value, depending on the yieldtoultimate strength ratio, the
proposed method allowed higher moments on netsections than the presently used design code provisions.
It should also be noted that the proposed design moments are lower than the experimentally measured
maximum moments on the netsection. Therefore, the design moments as per the proposed design method
would be safe. [f] The ratio of the nominal netsection fracture strength (AfnFu)tothe grosssection yield
strength (AfgFy) did not seem to be as of a significant parameter for flexural members as it is for the
tension members in determining the required strength since the flexural member (A501) having the
AfnFu/AfgFu ratio as low as 0.67 attained the maximum netsection moment which is more than the gross
section plastic moment. However, this parameter seemed to significantly influence the available total
rotational capacity of flexural members having flange holes and fastener holes. [g] All beam specimens
tested in this investigation attained more than the required rotation capacity of 3 before the onset of local
buckling However, the required rotation capacity for nonseismic applications as per the current AISC
specification [4] is greater than or equal to 79. In this investigation, beam specimens having the
AfnFu/AfgFu ≥ 1.0 exhibited substantial inelastic rotation capacity beyond the maximum load and were
able to reach the grosssection plastic moment on the unloading branch. Thus, beam specimens with holes
in the tension flanges only and fastener holes in both flanges satisfying A fnFu/AfgFu≥1.0 exhibited a total
available rotation capacity, Ry of more than 9. If the condition was violated, the beam specimens failed
primarily due to a rupture of tension flange through the flange holes which occurred before the flexural
members reached the grosssection plastic moment again on the unloading branch. That is, for flexural
members having the AfnFu/AfgFu<1.0 in the tension flanges, the inelastic deformation beyond the ultimate
load was substantially reduced. However, the beam specimens with holes in both flanges satisfying the
AfnFu/AfgFu≥0.95 exhibited a total available rotation capacity, Ry of more than 9. It should be noted that
the available rotation capacities would substantially vary depending on many parameters, such as the
crosssectional geometry of the beam specimens, bracing locations (closer bracing will result in higher
rotation ductility), material strain hardening, local instabilities associated with flange and/or web
buckling, presence of initial geometric imperfections, etc. Thus, the generalization of available rotation
ductility from a certain type of flexural test is not reasonable. Further details of this investigation are
available in the thesis by Arasaratnam [6].
REFERENCES
[1] CSA (2003), Limit States Design of Steel Structures, CAN/CSAS1601, Canadian Standard
Association, ON, Canada.
[2] Lilly, S.B., and Carpenter, S. T., (1940), Effective Moment of Inertia of a Riveted Plate Girder,
Transactions, American Society of Civil Engineers, Paper No. 2089, pp.14621517.
[3] AISC (1989). Allowable Stress Design for Structural Steel Buildings,9th Edition, American
Institute of Steel Construction, Inc., Chicago, Illinois, USA.
[4] AISC (2005). Load and Resistance Factor Design Specification for Structural Steel Buildings, 4th
Edition, American Institute of Steel Construction, Inc., Chicago, Illinois, USA.
[5] Dexter, R.J. Alttstadt, A. and Gardner, C.A. (2002). Strength and Ductility of HPS70W Tension
Members and Tension Flanges With Holes, Research Report, University of Minnesota,
Minneapolis, MN, 554550116, USA.
[6] Arasaratnam, P. (2008), “Effects of Flange Holes on Flexural Behavior of Steel Beams”, Ph.D.
Thesis, McMaster University, Hamilton, Ontario, Canada, p.xxv, p. 350.
578
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. This paper presents residual stress distributions in roller bent wide flange HE 100B sections
obtained from experiments. The wide flange steel beams are curved at ambient temperature by means of
the roller bending process, which alters the initial residual stress pattern. Curved wide flange steel
sections are frequently used in large span structures like roofings and bridges. Their geometry and
loading often make these arches susceptible to instability phenomena’s. An accurate knowledge of the
residual stresses is therefore necessary. An experimental program was set up to investigate the residual
stresses in roller bent sections. Measurements were carried out on straight and curved sections. It was
found that the residual stresses in roller bent sections differ significantly from those in their straight
counterparts.
1 INTRODUCTION
A structural steel section contains residual stresses which are a result of the manufacturing process.
Residual stresses are in general of primary importance for structures susceptible to loss of stability, since
the presence of these stresses causes early yielding and thereby a reduction in load carrying capacity.
Extensive research has been carried out on the measurements of residual stresses in straight hot rolled
sections, [1], [2], [3] and [4] amongst others. Results were summarized and published in various forms
[5], [6] and [7]. It was found that residual stresses in straight hotrolled sections are characterized by
compressive stresses in the flange tips and tension in the webflange junction. The web can be either
under compressive or tensile residual stress depending on the shape of the crosssection.
Cold forming structural steel at ambient temperature changes the residual stress pattern due to plastic
strains. This characteristic of cold forming was first observed in [8] where the residual stress distribution
in coldformed circular hollow sections was investigated. Residual stresses in pressbraked plates were
measured in [9], [10]. The residual stresses in bent sheet metals were reported in [11].
Theoretical models are available to obtain a residual stress distribution in coldbent sections. A first
solution was proposed in [12] for bars under uniform bending based on elastic perfectlyplastic material
behavior. The basic theory is illustrated in Figure 1, where Į=ratio between the plastic and elastic section
modulus or shape factor, fy=yield stress, h=height of the cross section and R=radius of the circular arch.
When a stress free bar or plate is plastically bent into a specific radius, a plastic stress distribution
emerges (Figure 1(b)). After releasing the acting moments, an elastic release or springback of the member
takes place, thereby imposing an elastic stress distribution on the already present loading stresses (Figure
1(c)). The result is a stress distribution, which is a summation of the loading and unloading stresses
(Figure 1(d)) thereby assuming a uniaxial stress condition. The theoretical model has found widespread
application in structural analysis of curved steel, as reported in [13].
579
R.C. Spoorenberg et al.
This theoretical model has been used to find residual stresses in roller bent sections. However the
model is questionable since the true bending process features a complex interaction between rollers and
beam which cannot be represented by a uniaxial loadingunloading procedure. At Eindhoven University
of Technology an experimental and numerical investigation has been started to study the roller bending
process and its effects on wide flange steel beams. This project includes a variety of wide flange sections
and different steel grades. Some of the experimental results have been published in [14].
A short description of the roller bending process is presented in section 2. The experimental program
and test setup are given in section 3. In section 4 the results of the residual stress measurements as taken
from both straight reference sections and roller bent sections are presented. Section 5 discusses the results
of the measurements and the paper ends with conclusions in section 6.
A straight member is placed in the machine (Figure 2(a)) and successive rolling and movement of the
rolls induces permanent curvature of the sections (Figures 2(b) and (c)). The top flange and bottom flange
are elongated and shortened respectively. The roller bending process is mainly featured by continuously
changing bendingtype deformations as the section moves through the rollers. During the bending process
all three rollers are driven and automatic speed compensation is implemented for the difference in surface
speeds between the inner and outer circumference of the section being rolled. Depending on the machine
type, sections with a height of up to 1 m can be bent. Several additional passes are required until the
designated radius is attained. Steel members can be bent the easy way or the hard way which involves
bending about its weak axis or about its strong axis respectively. When bent the hard way, a small roller
580
R.C. Spoorenberg et al.
is placed at the inside of the top flange to prevent web crippling. This investigation is limited to wide
flange sections bent about the strong axis. Although complex curvatures are possible, this study is
confined to beams bent into a circular shape. Due to placement requirements within the bending machine,
it is impossible to impose a permanent curvature along the complete length of the beam. The ends remain
straight and have to be considered as waste material (Figure 2(c)). A more extensive description of the
roller bending process is reported in [15].
The initial residual stress distribution was determined from straight reference sections from which the
curved ones were made. Tensile tests were carried out on coupons taken from the flange of the straight
and curved members in order to assess the increase in yield stress and ultimate tensile stress as a result of
the roller bending process (Figure 3).
600 600
500 500
stress [N/mm ]
400
stress [N/mm ]
400
2
300 300
200 200
100 100
S235 S355 S235 S355
0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
strain [] strain []
Figure 3. Stress strain relationship for HE 100B with steel grade S235 and S355, straight reference
section (left) and roller bent section with a radius of 2546 mm (right).
581
R.C. Spoorenberg et al.
3.2 Preparation
The sectioning method was used to measure residual stresses in roller bent steel arches. The test
specimen was saw cut from larger steel arches. Electrical strain gauges were applied to the surface of the
roller bent wide flange sections. For this investigation small (2 x 6 mm) electric strain gauges
manufactured by Tokyo Sokki Kankyujo Co. Ltd. were used. The arrangement of the strain gauges is
shown in Figure 4(a). To reduce end effects, the test area was a distance of 2.5 times the height of the
beam from the ends (Figure 4(b)). Only longitudinal stresses were measured.
A transverse saw cut and subsequent longitudinal saw cuts were made with an electrical band saw and
hand saw respectively. The influence of heat release from the electrical band saw cuts was suppressed by
supplying fluid coolant. Shortcircuiting of the electrical strain gauges was prevented by covering the
gauges with a protective layer of paraffin. Strain release was recorded during the entire saw cutting
procedure. Strain measurements were recorded until approximately 30 minutes after the end of the
cutting. Strain measurements were converted into stress values by multiplying the strain by the Young’s
modulus as obtained from the tensile tests on the straight reference sections, thereby assuming elastic
release of the strains. Stress values on opposite sides of the flanges and webs were averaged.
582
R.C. Spoorenberg et al.
4 RESULTS
100
50
25
0
25
50
100 25
0 0
100 50 0 50 25 50 0 50
200 50
50 50
25 25
0 0
25 50 0 50 25 50 0 50
50 50
50 50
Top fl. Top fl.
4 56
S355 S355
0 0
50
100
50
0
50
100
25 25
0 0
25 50 0 50 25 50 0 50
50 50
Figure 5 Hot rolled residual stresses in straight HE 100B reference sections in N/mm2
583
R.C. Spoorenberg et al.
300
150
0
150
300
400
300
150
0
150
300
0 0 0
50 0 50 50 0 50 50 0 50
200 200 200
400
200
0
200
400
400
200
0
200
400
Figure 6 Residual stress distributions after roller bending of HE 100B sections in N/mm2.
584
R.C. Spoorenberg et al.
5 DISCUSSION
The residual stresses as shown in Figure 6 display the following characteristics. High tensile residual
stresses were observed at the webtoflange junction in the bottom flange. The web of the roller bent
specimens is primarily under compression. Small residual stresses, either in compression or tension, were
found in the top flange. The results show a symmetrical stress pattern with respect to the minor axis of
bending.
The net effect of the roller bending process on the residual stresses can be observed by comparing
Figure 5 with Figure 6. It can be seen that the roller bending process results in a residual stress pattern
that is entirely different from the hotrolled pattern prior to bending. In particular, the maximum residual
stress of the roller bent sections is much larger compared to the maximum residual stress of the straight
hotrolled sections.
Bending a beam into a smaller radius requires an increase in cold work and hence an expected
increase in residual stress. However, it can be observed that the bending radius has small influence on the
residual stress distribution.
The effect of the steel grade on the residual stress distributions in roller bent sections for the HE 100B
series can be observed by comparing the results of specimens 1 to 3 (Figure 6) with those of specimens 4
to 6. The effect of the steel grade is significant since the residual stresses of the specimens 1 to 3 (S235)
are much smaller than those of specimens 4 to 6 (S355).
The measured residual stresses of the roller bent specimens are generally below the yield stress of the
straight material, although the yield stress was exceeded by the residual stress values in the bottom flange.
A higher residual stress than the yield stress can be expected as a result of cold working the material
during the roller bending process. The additional tensile tests on coupons taken from curved sections
support this theory as they showed an increased yield stress (Figure 3). Recent findings as reported in
[17] have shown a similar phenomenon for residual stresses in stainless steel sections.
The experimental results are quite different from the theoretically obtained residual stresses in cold
bent sections as proposed in [12]. The theoretical residual stress distribution does not have a stress
gradient along the flange width, which was clearly observed in all experimental results. An antisymmetric
stress pattern with respect to the major axis of bending postulated by theory has not been measured.
6 CONCLUSIONS
This paper presents experimental results of residual stress measurements carried out on straight and
roller bent HE 100B sections. The results are part of a larger experimental and numerical study on
residual stresses due to roller bending of wide flange steel sections. The sectioning method in conjunction
with electrical strain gauges was employed to measure the residual stresses. The experimental results for
the roller bent specimens showed a significantly different residual stress distribution when compared to
the residual stress distributions in their straight counterparts. In the roller bent specimen, high
compressive and high tensile stresses were observed in the web and in the bottom flange respectively. The
observations clearly indicate that a hot rolled residual stress pattern is not applicable to a roller bent
specimen. Also it can be stated that the theoretical solution for residual stresses in coldbent members as
proposed by Timoshenko is not suitable for roller bent wide flange steel sections. The large deviations
between the theory and the experiments show that the roller bending process of wide flange steel sections
cannot be simplified by a beam subjected to loading and unloading in uniform bending.
ACKNOWLEDGEMENTS
This research was carried out under the project number MC1.06262 in the framework of the Research
Program of the Materials innovation institute M2i (www.m2i.nl). The majority of the experiments were
585
R.C. Spoorenberg et al.
carried out by H.L.M Wijen and T.J. van de Loo in the Pieter van Musschenbroek Laboratory and their
help is greatly acknowledged. The specimens were supplied and bent by Deltastaal BV and Kersten
Europe BV respectively.
REFERENCES
[1] Beedle L.S. and Tall L. "Basic Column Strength". Transactions of the ASCE, 127, 138179, 1962.
[2] Mas E. and Massonet Ch. "Part prise par la belgique dans les recherches experimentales de la
convention europeenne des associations de la construction metallique sur le flambement centriques
des barres en acier doux". AcierStahlSteel, 9, 393400, 1966.
[3] Lay M.G. and Ward R. "Residual stresses in steel sections". Journal of the Australian Institute of
Steel Construction, 3(3), 221, 1969.
[4] Young B.W. "Residual stresses in hot rolled members". IABSE reports of the working commissions,
23, 2538, 1975.
[5] ECCS. Manual on stability of steel structures  ECCS Committee 8 Stability. 1976.
[6] Ballio G. and Mazzolani F.M. Theory and Design of Steel Structures. Chapman & Hall, Oxford,
1983.
[7] Galambos T.V. Guide to Stability Design Criteria for Metal Structures. John Wiley & Sons inc.,
New York, 1998.
[8] Kato B. and Aoki H. "Residual Stresses in Coldformed Tubes". Journal of Strain Analysis, 13(4),
193204, 1978.
[9] Weng C.C. and White R.N. "Residual Stresses in Coldbent Thick Steel Plates". Journal of
Structural Engineering ASCE, 116(1), 2439, 1990.
[10] Weng C.C. and Pekoz T. "Residual Stresses in ColdFormed Steel Members". Journal of Structural
Engineering ASCE, 116(6), 16111625, 1990.
[11] Tan Z., Li B., Persson B. "On Analysis and Measurement of Residual Stresses in the Bending of
Sheet Metals". International Journal of Mechanical Sciences, 36(5), 483491, 1994.
[12] Timoshenko S.P. Strength of Materials. Part II Advanced Theory and Problems. D. Van Nostrand
Company, Inc., New York, 1940.
[13] King C. and Brown D. Design of Curved Steel. The Steel Construction Institute, Berkshire, 2001.
[14] Spoorenberg R.C., Snijder H.H., Hoenderkamp J.C.D. "Experimental investigation of residual
stresses in roller bent wide flange steel sections". Journal of Constructional Steel Research, 2010
doi: 10.1016/j.jcsr.2010.01.017.
[15] Bjorhovde R. "Cold Bending of WideFlange Shapes for Construction". Engineering Journal,
43(4), 271286, 2006.
[16] Daddi I. and Mazzolani F.M. Determinazione sperimentale delle imperfezioni strutturali nei
profilati di acciaio. Universita degli studi di Napoli istituto di tecnica delle costruzioni, Bari, 1971.
[17] Cruise R.B. and Gardner L. "Residual stress analysis of structural stainless steel sections". Journal
of Constructional Steel Research, 64(3), 352366, 2008.
586
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Richard Stroetmann
Richard.Stroetmann@TUDresden.de
Keywords: Lateral torsional buckling, space structures, couple effects, calculation methods.
Abstract. The structural behavior of space structures with ,beams subject of lateral torsional buckling
can often be described only insufficiently by plane subsystems. This is caused e. g. by couple effects of
stabilizing and destabilizing beams or transmissions of rotations and displacements from cross girders at
the connecting points. By means of special finite elements, which are designed for ,beams in space
structures, an efficient calculation will be possible. Beam systems with crossconnected structural mem
bers, like cross girders or trapezoid sheets will allow a derivation of approximate solutions for standard
applications. The intended paper deals with calculation methods for space structures with ,beams and
demonstrates structural behavior as well as applications on the basis of typical examples.
1 INTRODUCTION
Methods for analyzing space structures with beams subject to lateral torsional buckling are mainly re
stricted to plane systems. Usually, the effect of adjacent structural elements is taken into account by
definition of bearings, discrete and continuous translational or rotational restraints. In many cases these
simplifications describe the structural behavior of such beams with sufficient accuracy.
Problems may occur when adjacent members do not have a stabilizing, but a destabilizing effect. Of
ten deformations (displacements, angular rotations) are transferred to the beams to be stabilized. In con
sequence, bearings, discrete and continuous restraints cannot describe the stiffening effect entirely. The
following examples will document the difficulties mentioned above.
Figure 1 shows two transversely loaded beams that are connected by a lattice bar at the midspan. In
case of lateral torsional buckling with different loading conditions qz,1 and qz,2 stabilizing forces will be
transmitted by the joining member. The more loaded beam will be restrained by the less loaded beam.
The interaction can only be recorded considering the entire global system.
qz,1 qz,2
y y
x z
587
Richard Stroetmann
qz
cross girder
y y
x z
The system in Figure 2 shows a cross girder with uniformly distributed load that is supported by a
longitudinal girder. The deformation of both girders at the connecting point can be assumed to be identi
cal. The angular rotation at the end of the cross girder and the vertical deflection of the longitudinal
girder result in a threedimensional deformation shape. In the longitudinal girder bending moments Mz
and torsional action effects will occur. They depend on the load level, the stiffness ratio and the dimen
sions of the system. Connections without stiffeners lead to additional distortions of the crosssection in
the area of the connecting point.
Figure 3 shows a roof structure consisting of purlins, trusses and a roof bracing. Lateral displace
ments of the trusses and the roof bracing are linked by means of the purlins. On the one hand the bracing
has a stabilizing effect on the trusses. On the other hand deformation, caused e. g. by wind loads, is trans
ferred into the trusses. This causes deflecting forces in the trusses that lead to additional loadings in the
bracing. The assumption that purlins connected to the roof bracing act as rigid supports for the top flange
of the trusses is inaccurate. Especially when diagonal bracings are realized as round steel tension bars and
the bracings have to span large distances the influence of the deformation may become important.
The preceding examples demonstrate that the structural behavior of space structures with Ibeams
subject to lateral torsional buckling can often only insufficiently be described by a division into plane
subsystems. Although known there is a lack of practical calculation tools that consider the interaction of
the involved components with reasonable effort.
plan view
section
q
purlin
w
truss
bracing
trusses
588
Richard Stroetmann
as membrane, solid or shell elements. Decisive disadvantages of the analysis of ,beam structures subject
to lateral torsional buckling are:
x The input of the structural systems including the definition of load and boundary conditions as well
as geometrical imperfections such as initial bow or sway imperfections is very complex and time con
suming.
x Generating the numerical model requires a comparatively large number of elements. The resulting
meshes have a large number of nodes. To calculate the response of the structure a large system of
equations needs to be solved.
x The analysis of the huge data output requires a graphical post processor. The transparency of results
is partially lost since action effects are only presented as stresses, bending moments and inplane
forces per unit length.
In practice such calculations remain restricted to special cases, e. g. when manufacturing a large num
ber of identical girders justifies these efforts. However, it is more efficient to use “macroelements” with
modified mechanical properties that represent the essential properties of the components to be modeled.
The effort needed for modeling, calculation and analysis can be minimized by using such elements, so
that FEM calculations can more widely be used for applications as described here.
Within the framework of the research project [1] tools for the calculation of structural systems with ,
beams subject to lateral torsional buckling were developed based on the finite element method. Among
other aspects the research aimed at providing a design tool for space structures and thus to remove limita
tions to plane systems such as single span or continuous Ibeams and Icolumns. It was intended to find a
possibility to describe the impact of adjacent structural members directly by means of suitable elements
and not only indirectly through the definition of discrete or continuous restraints and bearings. In addi
tion, it was important to eliminate the assumption that there no change in crosssection shape during
buckling, because some section designs with thin webs and support conditions of beam flanges require
that web distortions need to be taken into account. The concept was to limit the mechanical properties of
the elements that are essential for the modeling of members and connections. Effects such as transverse
shear or membrane strains of ,beams were neglected. In a first step, the development of finite elements
was limited to linearelastic material properties and linearized distortiontranslation relationship (second
order analysis).
For the modeling of beams and columns with doublesymmetric ,sections a “macro“ or “super ele
ment” composed of a group of element types (figure 4) was developed. The structural behavior in the
plane of web is idealized by a beam element. Perpendicular to the plane of web the ,section consists of
two flange members and a web plate.
x,
ˆ uˆ x,
ˆ uˆ
internal forces and moments
he
y,
ˆ vˆ ˆ wˆ
z,
le
ˆ wˆ
z,
element loads continuous restraints
qz,o cxx,o
qy,o cy,o tfl czz,o
mx,o s s
he/2
qz,m r, f r tw
S
s, f s cxx,u
he/2
Figure 4: ,ProfileElement – reference system, loads, bedding, internal forces and moments
589
Richard Stroetmann
The splitting of ,sections into different elements has the advantage that web distortions can be de
scribed with twodimensional buckling shapes. The separation of the cross section eased the application
of eccentric forces and moments, the modeling of practical supporting conditions and the attachment of
further structural elements to the flanges of the beams and columns. A frequently discussed issue in the
beam theory is how to model geometric and/or static coupling conditions (keyword transmission of warp
ing deformations) especially in cases when the beams are attached perpendicularly or in a random angle
to each other or if they consist of stepped cross sections. The modeling with this type of element elimi
nates such kind of problems.
Compared to the modeling of ,profiles with shell elements the advantage of the proposed concept is
that a relatively small number of elements is required for discretization. This leads to a significant reduc
tion in computing time. Generating the model is very practicable and the numerical results are easier to
analyze and interpret. The influence of crosssection fillets in rolled sections that significantly increase
torsional stiffness can easily be defined by means of modifying the stiffness values. Modeling with solid
elements requires a large number of elements, thus further increasing the effort for calculation and
evaluation of the results.
Besides the ,profile element additional elements were developed. A stiffening element can be used
for the modeling of stiffeners at load applications and of end plates; a trapezoidal profile element, a beam
element for the modeling of bracings and lattice bars as well as different spring elements for the consid
eration of connection flexibility.
mx
ˆ uˆ
x, ˆ uˆ
x,
ˆ vˆ
y, ˆ vˆ
y, cxx
ˆ wˆ
z, ˆ wˆ
z, q
ˆ uˆ
x,
ˆ uˆ
x,
ˆ vˆ
y, ˆ wˆ
z,
ˆ wˆ
z,
Figure 5: Assembly of the finite elements for trapezoidal sheeting, connection springs and ,beams
590
Richard Stroetmann
pling of beams or columns. Due to the structural detailing diaphragm actions cannot always be taken into
account.
The following description will briefly present different methods of calculating buckling loads of lat
eral coupled beam systems. Knowing the ideal buckling load a simplified verification of buckling resis
tance, e.g. according to EN 199311 [3], is possible.
One option to perform the structural analysis of coupled beam systems is to derive stiffness matrices
of the beams separately (based on second order analysis) and use suitable coupling conditions to obtain
the matrix formulation for the overall structure. Rigid couplings can be described by kinematic con
straints and semirigid ones by coupling matrices. Connections can be at discrete points or in closely
spaced intervals along the beams. The influence of constraining effects against twist rotations at the
connecting areas can be considered by discrete or continuous torsional restraints.
For simple structures approximation formulas and diagrams can be derived to determine the buckling
load of coupled beams. Moreover, making use of programs for the calculation of singlespan and con
tinuous beams the lateral torsional buckling load of rigid coupled beam systems may by determined
according to second order analysis of lateral torsional buckling and nonlinear spring characteristics of
the beams. A detailed description of the methods is given in [4].
3.2 Approximation formulas for rigid coupled ,beams
In case of simply supported beams with uniform distributed and single loads the assumption of buck
ling shapes in the form of half sinus waves for twist rotations  and lateral displacements vM of shear
centre axis leads to an acceptable approximation for the buckling load. When the beams are rigidly cou
pled, the kinematic constraints provide the transformation rules of the stiffness matrices. In case of two
coupled beams the degrees of freedom will be reduced from four to three. The assembly of beam stiffness
matrices to the global stiffness matrix and the derivation of the determinantal equation result in a charac
teristic cubic polynomial with three eigenvalue. A closedform solution is possible.
The approximation formulas given in figure 6 provide buckling load values that in most cases deviate
by less than 5 % compared to calculations using more significant buckling shapes. Systems with a high
stiffness of discrete and continuous torsional restraints show larger differences. If four, six or more
a0 2 c1 c2 ( f 2 c1 c2 )
a0 a1 K a2 K2 a3 K3 0 o KKi
a1 2 c1 c2 [ f ( g1 g3 ) g2 g4 ] ( f c1 )2 ( g2 g4 )
qz a2 ( f 2 c1 c2 ) ( g12 g32 ) 2 c1 [ f ( g1 g4 g2 g3 ) g2 g4 ]
a3 2 f g1 g3 ( g1 g3 ) g12 g4 g2 g32
C
qz ,i L2 Fz ,i L 1 2 2
M q ,i M F ,i rM y ³ ( y z ) z dA 2 z M
8 4 Iy A
Fz
x
EIz S 4 ECM S 4 GIT S 2 L
L/2 L/2 c1 c2 c C f ZK ZM
2 L3 2 L3 2 L 2
z Mq,1 S2 MF ,1 S2
g1 (1 ) (1 )
qz,1 qz,2 L 3 L 4
c c
Mq,1 rM y S2 qz ,1 L MF ,1 rM y S2
g2 ( 1 ) ( zq,1 zM ) (1 ) Fz ,1 ( zF ,1 zM )
L 3 2 L 4
y S S Mq,2 S2 MF ,2 S2
ZK ZM g3 ( 1 ) ( 1 )
M M L 3 L 4
f
EA=f Mq,2 rM y S2 qz ,2 L MF ,2 rM y S2
Fz,1 Fz,2 g4 (1 ) ( zq,2 zM ) ( 1 ) Fz ,2 ( zF ,2 zM )
L 3 2 L 4
591
Richard Stroetmann
beams are rigidly coupled and respectively half of them have the same load intensity the buckling load of
these systems can obviously be calculated as well using the formulas given in figure 6.
3.3 Diagrams for calculation of lateral torsional buckling moments
Using the program PROFIL [2] diagrams were created for the determination of lateral torsional buck
ling moments of girder systems with discrete and continuous rigid couplings (see figure 7). The diagrams
are valid for systems of simply supported rolled ,beams. The doublesymmetric cross sections of the
respective Ibeams are identical. The application points of the uniformly distributed loads and the level of
couplings are placed to the centroid from the top flanges of the beams. At the determination of the dia
grams web distortions were excluded. The application is as follows:
The number of coupled beams determines the type of diagram to be used. Curve parameters are given by
the moment distribution My and the load relation q2/q1. Depending on torsion coefficient F (see equation
(1)) the diagram provides the coefficient k to determine the lateral torsional buckling moment of the
whole system (equation (2)). The value My,Ki refers to the total maximum moment My, that means either
the span or support moment. Lateral torsional buckling moments of single beams j will be determined
with equation (3) by the relation of beam load and system load.
ǼI W (1)
Ȥ
Gǿ ȉ L2
k (2)
M y ,Ki GIT EI z
L
18 q2 26 q2
q1 q1
q1 q2
k q2 q1 q2
k M q 1

24 +
M q
 1
16 + 5
0,5
L 22 L f
0,2
Mq = q·L² / 8 5 M q = q·L² / 8 0,2
0,1
14
20
0 0,1
1
1
18 0
12 0,5 1
5
1 5
0,5 f
0,25 0,1
0,1 f
0,25
15,7
16 0
0
10,5
10,1 0,1 0,2
10 14,5
9,3
0
13,7
14
0,1
13,1
8,7 12,8
8,2
8,1 12
8 11,7
11,5
0
7,4 10,9
0,67· M q
 
6,8 + 10
9,7
6,5
6,3 9,1
6 8,6 0,67· M q
 
5,5 +
8
5,1 7,5
Mq + 7,3 Mq +
6,4
4 6
0 0,1 0,2 0,3 0,4 0,5 0 0,1 0,2 0,3 0,4 0,5
a) F b) F
592
Richard Stroetmann
qj (3)
M y , j ,Ki M y ,Ki
¦q
The coefficient k is nearly independent on section series and size. The diagrams in figure 7 are based
on minimum values of the IPE series. In good approximation the application is possible for other rolled
beams with doubly symmetric ,section. In the diagrams for three coupled beams it has to be considered
that the coefficient k increases from load relationships of zero to one and thereafter k decreases again.
Under certain conditions the diagrams may also be used for determining lateral torsional buckling mo
ments of systems of more than three beams. This requires that only two different values of transverse
load are present and that the relation between the number of beams with the same load is n1:n2 = 1:1 or
1:2.
4 EXAMPLES
In order to estimate the buckling loads of discrete coupled beam systems an iterative determination
can be performed. The aim of such determination is to find the load level KKi at which the sum of stabiliz
ing and destabilizing forces that are transferred to the coupling beams reach equilibrium. Figure 8 dem
onstrates the principle by an example. Three differently loaded beams are rigidly connected at midspan to
the top flange. If the spring characteristic Cy for lateral displacement of the Ibeams at the connecting
point is known from a previous calculation performed according to second order analysis the iteration can
be carried out by using the diagram presented.
Figure 9 shows a girder grillage consisting of purlins and trusses, respectively with same cross
section, on which a uniformly distributed load of q=5.00 kN/m2 acts. Lateral torsional buckling of the
purlins is prevented by restraints, e. g. due to trapezoidal sheeting. Caused by the structural system, the
load transfer of the inner trusses is approximately 2.75 times higher than that of the outer trusses. By the
connection with the purlins the trusses are torsionally restrained and coupled in transverse direction at the
top flanges.
qz
250
130,3
Cy [kN/m]
31,6
0
3000 3000
161,9
250
qz,1 = 2,0 qz,2 = 5,0 qz,3 = 10,0
500
q [kN/m]
(1) (2) 290 x 7 (3) 0 10 20 30 z 40
0
150 x 10
1
Cy C y ,B 1
vk
qz qz
Cy,B 1 2
K Ki = 2,44
1 y,v Beam 3
S 3
Cy,B Beam 2
K Beam 1
3000 3000
4
z
Figure 8: Determination of buckling load with spring characteristic (loads [kN/m], dimensions [mm])
593
Richard Stroetmann
The stability of the structural system was calculated by the program PROFIL [2] and various ap
proximation methods as documented in [4]. Moreover, various effects were investigated by means of
different calculations. Disregarding the coupling effect to the outer trusses and the torsional restraints
from the purlins the buckling load factor of the inner trusses is KKi=0.39. With torsional restraints this
factor increases to KKi=1.24. Additionally, the consideration of the coupling effect to the outer trusses
results in a buckling load factor of KKi=1.71. In this particular case it is necessary to consider both effects,
to verify the structural safety of the inner trusses.
IPE220
IPE600 IPE600
8 1,71
purlins IPE220
5x2.500=12.500
1,71
trusses IPE600
4 without 1,24
1,24
2 torsional
restraints with with
1
and couple torsional torsional
8
effects restraints restraints
6 but and
0,39
0,39 without couple
4
couple effects
2
effects
0
1 2 3
safety factor of the ideal buckling load
Figure 9: Girder grillage with trusses and purlins – influence of different stabilization effects
5 CONCLUSION
The stability of space structures is often only insufficiently assessed by the analysis of plane subsys
tems. Specific finite element formulations allow system analyses that include essential effects with regard
to the overall structural behavior. Besides the possibility to consider the transmission of deformation of
adjacent structural members, stabilizing forces of bracing systems can be directly determined. By restric
tion to the effects that are essential for the structural behavior the efforts for modeling, calculation and
interpretation are minimized. In this way, the finite element method can economically be applied to space
structures in practice.
REFERENCES
[1] Stroetmann, R., “Zur Stabilitätsberechnung räumlicher Tragsysteme mit IProfilen nach der Me
thode der finiten Elemente“, Veröffentlichung des Instituts für Stahlbau und Werkstoffmechanik
der Technischen Universität Darmstadt, Heft 61, 1999.
[2] Stroetmann, R., “PROFIL FEMProgram for the structural analysis of space structures with I
Profiles“, Fachgebiet Stahlbau und Werkstoffmechanik, TUDarmstadt, 1999.
[3] EN 199311: “Eurocode 3 – Design of steel structures, Part 11 – General rules and rules for
buildings“, 2005.
[4] Stroetmann R., “Zur Stabilität von in Querrichtung gekoppelten Biegeträgern“, Der Stahlbau
69, Verlag Ernst & Sohn, Berlin, 391408, 2000.
594
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
* Graz University of Technology – Institute for Steel Structures and Shell Structures
emails: taras@TUGraz.at, r.greiner@TUGraz.at
Abstract. This paper presents a new formulation for the design of beamcolumns against inplane
buckling that makes use of an increasingly popular generalized slenderness definition and an “overall”
formulation of the buckling reduction factor for combined load cases. Thereby, great care is placed on
accurately describing the specific behavior of each studied crosssectional type. The result is a
“generalized slenderness” formulation that is as accurate, safe and mechanically consistent as the
familiar and thoroughly studied interactionconcept formulae.
1 INTRODUCTION
Beamcolumns are characterized by the presence of compressive axial forces N and bending
moments M. The resistance of a steel member against either N or M is commonly determined in design
codes by the use of buckling reduction factors F=f( O ), whereby the plot of the function F over the
normalized slenderness O is a socalled buckling curve. The (usually) detrimental effect of the combined
action of N and M is taken into account in design codes by formulae that are based on one of the
following two concepts, see figure 1:
 The interaction concept, found e.g. in clause 6.3.3 of Eurocode 3 [1], makes use of the
information contained in the utilizations nFB and mLT of the buckling checks for flexural
buckling under N alone and LTbuckling under M alone. The combined effect of N and M is
then taken into account by an interaction factor k.
M INTERACTION CONCEPT GENERALIZED SLENDERNESS CONCEPTS
N ("overall slenderness"; "general method"; ...)
My N N
My
My N N
My
My My LEA/MNA GMNIA
N N geometric residual
imperfections stresses
LPFMNA
OGS h/b>1.2
Npl Mpl LPFLBA
O oF OLT o FLT 0.3 fy  + 
Ncr Mcr Rpl e0
=
Rcr h/b<1.2
NRk Npl F MRk Mpl FLT
0.5 fy
595
Andreas Taras and Richard Greiner
 In the second group of concepts, a generalized definition of the (normalized) slenderness is used;
they are therefore called generalized slenderness concepts in the following. Specifically, these
concepts encompass the “overall load case method” used for the design of plates and shells (see
e.g. [2]) and the socalled “general method” for the design of beamcolumns of clause 6.3.4 in
Eurocode 3. These methods have in common that they consider total utilizations for the
combined case as basis for the calculation of the normalized slenderness and of the (overall)
buckling reduction factor. As is illustrated in figure 1, the slenderness O GS is defined in a
generalized form as the square root of the total load proportionality factor LPF for the plastic
collapse load (LPFMNA=Rpl) over the pertinent buckling eigenvalue LPFLBA=Rcr. The buckling
strength is then defined as follows:
FGS R pl
Rb,d t 1.0 (1)
JM1
Thereby, Rb,d is the design buckling resistance (in terms of maximum LPF) of the component or
structure against the studied buckling mode for a given load combination.
Even though the current debate over these two concepts might seem to indicate otherwise, the
concepts are best thought of as two different forms of representation of the same information, with no
basis for attributing an (inexistent) higher degree of mechanical consistency to any of the two. In the case
of the interaction concept as found in the Eurocode, mechanical accuracy and safety/reliability have been
ensured by extensive theoretical, numerical and statistical studies, summarized in [3]. On the other hand,
one could argue that the “generalized slenderness” formulation according to (1) is more “consistent” with
the design checks for the single load cases, in the sense that it also implicitly contains the buckling
checks used for only N or only M. (In the case of the “general method” this is only true for M, since for
N it is based on Fy.Npl instead of on Npl alone). However, the reduction factor FGS must account for the
exact same effects as the interaction factor k. As is indicated by the question mark in figure 1, the values
to be adopted for FGS are not clear and still up for debate, with a common opinion being that they must be
studied and calibrated by means of GMNIA calculations, see e.g. [4].
The following figure 2 illustrates factors FGS obtained from such GMNIA calculations for the in
plane buckling behavior of a beamcolumn under N+M with uniform moment diagram and different
values of the ratio K0=(M/Mpl)/(N/Npl)=m0/n0. Two different definitions of F and O are used. Figure 2a
makes use of Fip and O ip, which are based on the definitions of Rpl and Rcr valid for the overall load case
N+M and inplane buckling behavior.
K0=0.00
1.0 1.0
IPE 500 K0=m0/n0 GMNIA
1 1
2 M K0=0.25 2
Oip My/N=const. Oy
0.8 0.8
K0=0.50
0.3 fy
=Nb/Npl []
 +  0.6 K =1.00
F =R /R []
0.6
0
b,ip pl
K0=0.00
0.4 0.4 K0=2.00
y, K 0
K0=0.25
ip
K0=0.50
K0=1.00 K0=4.00
K0=2.00
0.2 K0=4.00 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
O ip a) Oy b)
Figure 2: GMNIA buckling reduction factors Fip (a) and Fy,K0 (b) for inplane buckling of an IPE 500.
596
Andreas Taras and Richard Greiner
In figure 2b, the familiar definition of the inplane flexural buckling slenderness O y is used, as well
as a representation of the buckling strength for the combined, proportional load case solely in terms of
the achievable axial load Nb. Both types of representation yield the same curve for the case of K0=0.0,
representing the (imperfect) column, and result in distinctly different curves for varying values of K0.
As was mentioned above, any application of a generalized slenderness concept requires a
definition/formulation of the buckling reduction factor FGS that reproduces the same type of information
contained in an interaction factor k, thereby achieving high accuracy when compared to more
sophisticated GMNIA calculations. This paper presents an analytical formulation for the buckling
reduction factors Fip and Fy,K0 (i.e. FGS for the inplane buckling phenomenon of beamcolumns) that
fulfils this requirement.
1
Oip Oy
kn1 K0 km1
1.6 1.0
failure criterion:
1.4 kni * F y,K0 k mi * mII 1.0
0.8
1.2 1/kn1*
strongaxis buckling =1.0
1.0 N+My
K0=m0/n0 0.6
n=N/Npl []
[]
1/kn2 1/kn2*
pl
0.8 mII
n=N/N
0.6
0.4 m0 Oip
m0 1/kn1=1
0.4 F y,K0
1/km1 0.2
K0 n0 1/km2*
0.2 Ksc1
n0 1/km2=1
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0
m=M/M pl [] a) m=M/Mpl [] b)
1/km1*
Figure 3: Linearization of the crosssectional interaction and generalized slenderness definition (a);
definition of a secondorder failure criterion (b).
597
Andreas Taras and Richard Greiner
Table 1: Ccoefficients used for the description of the crosssectional N+M interaction behaviour.
Type of section, loading, underlying
# Parameters of the NM interaction linearization J
residual stress distributions
Isection, strong
1 axis buckling 0.8 0.8
 +  kn1=1.0; km1= 1 0.5 a t 0.75 ; KSC1
N+My 1 0.8 k m1
0.30.5 fy
+
0.2
Rectangular  0.2
kn2= ; km2=1.0; KSC2 f
fy
2 hollow section 1 0.8 k m1 0.4
+
RHS, N+My 
+ + +0.5 fy
Isection, weak
4 axis buckling kn2=0.8/(1.81a) ; km2=(1a)/(10.55a); KSC2=1/a 0.6
0.30.5 fy
N+Mz kn3=0.0; km3=1.0; KSC3 f
A 2 b tf Aw
a
A A
Since the “generalized slenderness” concept operates with load amplification factors R, a reference
load level m0/n0 is proportionally increased in the following, meaning that the ratio K0=m0/n0=K0 is kept
constant. Taking this into account, the following expressions can be found for the crosssectional (plastic)
amplification factor Rpl and the buckling eigenvalue Rcr pertaining to the inplane mode:
1 K0
R pl (3)
n0 c0 m0 c0
1
Rcr 2
(4)
n0 O j
2
With c0 k ni K0 k mi ; O j N pl / N cr, j ; j = axis y or z, depending on the case. The
generalized inplane buckling slenderness can now be written as
O ip O j / c0 (5)
The next step consists of a definition of a (secondorder) failure criterion. The basic concept behind
the adopted criterion is illustrated in figure 3b: the buckling load of the member is reached when at one
crosssection the following condition is fulfilled:
N M II
kni * kmi * 1.0 (5)
N pl M pl
Thereby, MII is the total, secondorder bending moment in the critical crosssection at failure, while
kni* and kmi* are factors derived –once again from a linearization of the crosssectional interaction
diagram (see kni and kmi), but taking into account the transition from the applicability of the plastic and
elastic crosssectional interaction curve with increasing slenderness. This transitional behavior, discussed
e.g. in [3] and [5], is caused by the detrimental effect of extremefiber (e.g. flange) yielding on the
obtainable buckling strength observed in tests or realistic GMNIA calculations. Accordingly, the
following expressions (6) and (7) for kni* and kmi* reproduce a transition from the values kni and kmi valid
for the plastic crosssectional capacity to the values of kni*=1.0 and kmi*=Wpl/Wel=w applicable for the
elastic crosssectional resistance at higher slenderness.
598
Andreas Taras and Richard Greiner
kni *
N
kmi *
CmS M N e0
1
1.0 (9)
N pl M pl 1 N / N cr
The next step, explained in detail in [5], consists of replacing the geometric imperfection amplitude
e0 by the generalized, AyrtonPerry imperfection amplitude that leads to the EC3 column buckling
curves for the studied crosssection and the case where m0 = K0 = 0.0.
M pl
e0
N pl kmi *
Kimp ; Kimp D O 0.2 (10)
599
Andreas Taras and Richard Greiner
0.15 fy
=Nb/Npl []
0.6 K0=1.00
F =R /R []
0.6
b,ip pl
K0=0.00 K0=2.00
0.4 0.4
y, K 0
K0=0.50
ip
K0=1.00
K0=2.00 K0=4.00
K0=4.00
0.2 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
a) Oy d) b)
O ip
Figure 4: Comparison of newly developed analytical and numerical GMNIA buckling reduction factors
Fip (a) and Fy,K0 (b) for a CHS section.
A different type of comparison is shown in figure 5. In this figure, a spectrum of obtainable
combinations of the external loads N+M is plotted for two sections loaded by N + My or Mz, whereby
three different member lengths and a uniform moment diagram are considered. Figures a) and b) compare
the GMNIA results with the current Eurocode “interaction concept” formulae of clause 6.3.3, see [1] and
[3]. Figures c) and d) compare the numerical results with the evaluation of the newly developed “overall”
expression (15). The accuracy of both the two interactionconcept formulae found in the Eurocode
(Annex A and B) and the new formula can generally be said to be excellent.
Some advantages of the new formula can be mentioned here: the fact that expression (15) builds upon
an accurate (linearized) description of the actual, sectionspecific crosssectional N+M interaction
diagram causes it to be “automatically” accurate at very low slenderness, and to lead to a consistent and
“correct” transition from the buckling to the (plastic) crosssectional design check. The latter point is not
the case for the two sets of Eurocode interaction factors, which make use of approximations of the cross
sectional interaction for the buckling check and usually don’t lead to the exact interaction at zero
slenderness. Another advantage of (15) appears in cases with nonequal endmoments, where failure can
be dominated by the crosssectional check at one of the ends instead of by proper buckling. In the
interaction concept, this must be checked specifically, while (15) simply “includes” this check by being
limited by 1.0.
600
Andreas Taras and Richard Greiner
 + 
in Oz=1.5
0.8 Oy=0.5 0.8
t er
ac Oy=1.0
tio
n Oy=1.5 0.3 fy
 +  0.3 fy
0.6 0.6
n=N/Npl []
n=N/Npl []
0.4 0.4
0.2 0.2
GMNIA
EC3A
EC3B
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
m=M/Mpl [] a) m=M/Mpl [] b)
1.0 1.0 N/M
IPE 240 z C Oz=0.5 HEA 500z
N/ line Sin M
M M ariz tera Oz=1.0
lin C atio ctio
 + 
ea S n n Oz=1.5
0.8 riz inte Oy=0.5 0.8
at
ion rac Oy=1.0
tio
n Oy=1.5 0.3 fy
 +  0.3 fy
0.6 0.6
n=N/Npl []
n=N/Npl []
0.4 0.4
0.2 0.2
GMNIA
EQU
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
m=M/Mpl [] c) m=M/Mpl [] d)
Figure 5: N+M buckling interaction curves according to Annex A & B of Eurocode 3 (ab) and the
analytical formulation (cd), compared with GMNIA results
In the final figure 6, the numerical GMNIA results, code formulae and the evaluation results of (15)
are compared on the level of interaction factors kj, computed so that they fulfill the following equation:
M N
nj kj 1.0 ; nj (16)
M pl F j N pl
The top three diagrams in figure 6 compare the values of kj (kyy or kzz) as defined in Annex A and B
of the Eurocode with GMNIA results, while in the bottom three diagrams GMNIA results are compared
with (iteratively determined) results of (15) that fulfill (16). Again, the proposed formulation follows the
GMNIA values of kj quite well, especially qualitatively. The curves obtained from (15) appear to have a
similar course as the ones of the EC3 Annex A formulae, but with some advantages in accuracy
particularly in the case of the circular crosssection, for which the crosssectional interaction is poorly
represented by the Eurocode formulae. It should be noted that the accuracy of the kj values gives a rather
misleading representation of the accuracy of the formulation itself, particularly for higher values of ny or
nz. Even errors of some 2030% in terms of kj only lead to total errors of only a few percentage points at
values of nj beyond 0.5. In this sense, it is a welcome observation that the accuracy of the proposed
formulation in terms of kj is highest for low values of nj, and mostly conservative in all other cases.
601
Andreas Taras and Richard Greiner
 + 
2.0 2.0
kzz []
ky []
1.5 0.3 fy
 + 
ky []
0.15 fy
0.3 fy 1.5 1.5
1.0
GMNIA 1.0 1.0
ny=0.2, 0.4, 0.6, 0.8 EC3A ny=0.2, 0.4, 0.6, 0.8 nz=0.2, 0.4, 0.6, 0.8
EC3B
0.5 0.5 0.5
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Oy a) Oy b) Oz c)
2.0 2.5 2.5
IPE 240 CHS178/10 HEA 500z
M M M
 + 
2.0 2.0
ky []
1.5
kz []
0.3 fy
 + 
ky []
0.15 fy
0.3 fy 1.5 1.5
kz []
1.0
1.0 1.0
GMNIA
ny=0.2, 0.4, 0.6, 0.8 EQU ny=0.2, 0.4, 0.6, 0.8 nz=0.2, 0.4, 0.6, 0.8
Figure 6: Comparison of interaction factors kj; top: Eurocode Annex A & B; bottom: new proposal.
4 CONCLUSION
The design proposal in this paper combines the advantages of the “interaction” and “generalized
slenderness” concepts for the case of inplane beamcolumn buckling. The comparison with numerical
results and current Eurocode rules (with many more examples given in [5]) showed the new proposal to
have a consistent level of accuracy and safety. The proposal could serve as a procedural blueprint for the
expansion of “generalized slenderness” concepts to other memberbuckling cases.
REFERENCES
[1] EN 199311, Eurocode 3. Design of steel structures. General rules and rules for buildings, CEN,
Brussels, 2005.
[2] Rotter, J.M., Shell Buckling and Collapse Analysis for Structural Design: The New Framework of
the European Standard, Festschrift for Prof. Calladine, Cambridge, 2002.
[3] Boissonade, N., Greiner, R., Jaspart, J.P., Lindner, J., Rules for Member Stability in EN 199311,
Background documentation and design guidelines, ECCS TC 8 – Stability, Brussels, 2006.
[4] Greiner, R., Taras, A., On the variety of buckling curves, Proc. of “Stability and Ductility of Steel
Structures”, Lisbon (PT), Sept. 69, 11011108, 2006.
[5] Taras, A., Contributions to the Development of Consistent Stability Design Rules for Steel
Members, PhD Thesis, Graz University of Technology, 2010.
[6] Rondal, J., Maquoi, R., Formulations d’AyrtonPerry pour le Flambement des Barres Métalliques,
Construction Metallique, 4, 4153, 1979
[7] Beer, H., Schulz, G., Bases Théoriques des Courbes Européennes de Flambement, Construction
Métallique, 3, 3757, 1970.
602
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Harald Unterweger*
Abstract. In building constructions for bracing members often hollow sections are used with slotted
gusset plates at the ends. These plates are attached to nonstiffened plates of the adjacent construction. In
practice sometimes a nearly centered joint is designed, by arranging the slotted gusset plate with an
eccentricity of half the plate thickness to the member axis, so that the member eccentricity is minimised.
In the paper the load bearing behaviour of such members under compression and tension is discussed
based on numerical analysis with nonlinear FE – models including imperfections. The geometric
parameters and boundary conditions are varied in such a way that practical cases are covered and that
the typical load bearing behaviour can be seen. Based on these results an engineering model for the
design in practice is represented.
1 INTRODUCTION
Bracing members and truss members are often designed with hollow sections and slotted gusset plates
at the ends. These plates are attached to nonstiffened plates of the adjacent construction, using welds or
bolts. In Fig. 1a some typical joints of this type are represented, including also joints with concrete
foundation. In Fig 1b the representative and idealized joint configuration for these joints  limited to
rectangular hollow sections (RHS)  is shown. The vertical plate (KB2) is restrained in axis I. Here, the
two different border cases related to boundary conditions are considered:  pinned (BC1) or, – fixed
(BC2, e.g. “concrete joint” in Fig. 1a). The slotted gusset plate of the member is joined by a fillet weld
(a1), passed around. An alternative solution with two bolt rows is possible, in accordance with the as
sumptions of the numerical analysis (fixed connection between plate KB1 and KB2 along their borders).
A special feature of the studied joint is the eccentric position of the slotted gusset plate with an
eccentricity of half the plate thickness tKB1, as shown in Fig. 1b. In doing so, the eccentricity of the RHS 
member for the buckling check – relevant is member buckling out of plane (about the z – axis)  is
minimised (e* = 0,5·tKB2). The bending moment along the memberlength is constant.
Based on the minimised eccentricity for the RHS  member, the opinion of practitioners is that only a
buckling check for the member under axial load is necessary to get the load carrying capacity of the
member. The results of this study will show that this approach would lead to high overestimations of the
load carrying capacity, especially for low slenderness ratios of the member. The reason for this, are high
bending moments in the gusset plate out of plane.
The loading of the RHS  member in this study is restricted to axial forces with bending moments
only due to eccentricities of the joints. This paper summarizes the results in [1].
603
Harald Unterweger
First of all the finite element (FE) – model and the executed nonlinear numerical analysis are
presented. Afterwards the studied band width of the varied geometric parameters is summed up. At the
end the results of the numerical nonlinear calculations for an example of a rectangular hollow section are
presented, for different member slenderness and boundary condition.
Figure 1: Studied RHS  member joints: a.) different types in practice, b.) geometry and restraint
conditions of the studied representative joint.
a.) b.)
Figure 2: a.) FE – model of the joint, b.) relevant first eigenmode, L0 = 2 m, for a pinned (above) and
fixed gusset plate (below).
604
Harald Unterweger
In the study a squared, hot finished, RHS  profile with 100 / 100 / 5 mm was used, leading to gusset
plate dimensions of 250 / 130 mm (KB 1) and 100 / 330 mm (KB 2). The calculations were done for total
member lengths L0 = 2, 3, 4, 5, 6, 7 and 8 m. The corner radii of the hollow section were omitted, leading
to an area A0 = 1900 mm2 and a radius of gyration iz,0 = 38,84 mm.
In the calculations an ideal elastic – perfectly plastic material behaviour was considered with a
characteristic yield strength of fy = 235 N / mm2. A modulus of elasticity E = 210000 N/mm2 and a
Poisson ratio of Q = 0,3 were used.
First of all an LBA – analysis (linear buckling analysis) was made, leading to the capacity NLBA of
the member. Based on these results, on the one hand the “real” buckling lengths of the members were
determined (using the formula for the Euler buckling load for the RHS  member section). Due to the
limited bending stiffness of the gusset plates, the buckling length Lcr,0 of the idealized RHS  member
within the end – restraint in axis I (Lcr,0 = L0 for BC 1, Lcr,0 = 0,5 L0 for BC 2) is too small.
Afterwards the results are either based on the idealised slenderness തതതതതߣ௭ǡ (Equ.1), with Lcr = L0 , or on
the slenderness based on the LBA – analyses തതതതതതߣ (Equ. 2).
f y L0 1 L0 1
O z ,0 (1)
V cr ,0 iz ,0 S E f y iz ,0 93,9
N pl ,0 A0 f y
O LBA (2)
N LBA N LBA
On the other hand the eigenmodes of the LBA – analyses, scaled to a maximum value of wmax= L0 /
1000 were used for a GMNIA – analyses (geometric and material nonlinear analyses with imperfections).
This was done with care, considering different eigenmodes (not only the one for the minimum ideal
buckling load), to get a minimum for the load carrying capacity NR of the member. In Fig. 2b the relevant
eigenmodes for a member with small length L0 is shown for the two different boundary conditions.
Additional GMNA – calculations were used to check the GMNIA – results. Residual stresses were
ignored, because they affect the buckling capacity of RHS – members not significantly (e.g. [4]).
Also for cold formed RHS  members the presented results mainly are valid, only for high slenderness
ratios – where the overall buckling of the member is relevant – the appropriate buckling curve should be
used (e.g. curve c instead of a, using Eurocode 3 [3]).
2.2 Studied joint parameters
The numerical study was limited to rolled RHS – members. The joint geometry is restricted to the
dimensions of Fig. 1b. Very important is the slotted length Ls 1,5·h in the RHS – member. Otherwise
sometimes significant smaller load bearing capacities would occur. The overlapping length of the two
gusset plates was fixed with 0,75·h. The distance between member end and restraint axis I is limited to
L1 = 1,25·h.
The thickness of the two gusset plates was varied in such a way that the area ratio AKB1 / A0 = 0,8 to
1,4 and tKB2 = (0,5 to 1,0 )·tKB1.
2.3 Results of the nonlinear calculations
The results of the nonlinear calculations for different member length (i.e. different member
slenderness) and gusset plate thicknesses are presented in Fig. 3, based on an effective width of beff =
3,3·h = 330 mm of gusset plate KB2. The load carrying capacity NR is related to the section capacity of
the RHS  member Npl,0 = A0·fy =19,0·23,5 = 446,5 kN. The slenderness ratio തതതതതത
ߣ (Equ. 2) is based on
the results of the LBA–analysis  that means based on the “real” buckling length.
In Fig. 3a the overall carrying behaviour is shown, based on the GMNA – results, without geometric
imperfections. The effect of these geometric imperfections is quantified in Fig. 3b, where the results with
and without imperfections are visible. It can be seen that the reduction of load carrying capacity is
comparatively small.
605
Haraldd Unterweger
Inn Fig. 3 also thhe buckling currve a, relevant for the bucklinng load capacity y of a hot finishhed, RHS –
membber under onlyy axial load in Eurocode
E 3 [3]] is plotted (dotted line). It caan be seen that the special
featurre of the studieed joint – minimmising of the member
m eccentriicity – is only usable
u for high slenderness
ratioss and thick gussset plates. For example
e with a gusset plate thiickness of tKB1 = tKB2 = 12 mm m, leading to
a ratiio AKB1 / A0 = 0,82 a dramatiic reduction of the load carryiing capacity occcurs in case off the pinned
gusseet plate KB2, also
a for very low w slenderness ratios
r (NR § 0,16·Npl,0 ). A fix xed gusset platee, however,
increases the load carrying capacitty significantly,, but also for veery low slenderrness ratios thee capacity is
far beelow the sectioon capacity of thhe RHS  mem mber (NR § 0,544·Npl,0 ). Not beefore the plate thickness
t is
increased significanntly (t KB1 = tKB22 = 20 mm, leaading to a ratio AKB1 / A0 = 1,37) and the gussset plate is
fixedd (BC2 in Fig. 1b) nearly about 80 % of thhe section capacity Npl,0 for small slendernness ratio is
availaable.
curve a
curve a
slenderneess slendern
ness
Fiigure 3: a.) GM
MNA  results deepending on thee slenderness raatio, b.) GMNA  results in com
mparison
with GMMNIA  results, for pinned (RB1=BC1) and fixxed (RB2=BC2 2) gusset plates.
T
The significant reduction of thhe load carryingg capacity – alsso related to the buckling capacity of the
RHS––member (see Fig.F 3) – is caussed by the locall bending momeents in the gussset plate, particuularly at the
end oof the RHS – member.
m This caan be seen in Fiig. 4 for a very short member with the thin gusset
g plates
mentiioned before. In
I Fig 4a the pinned
p and in Fig.
F 4b the fixeed ended gusseet plate can be seen at the
ultim
mate limit state. The gusset plaate section at thhe member endd reaches its seection capacity under axial
forcee and bending moment.
m
606
Harald Unterweger
Summing up, the load carrying capacity of the RHS  member with the specific joint configuration of
Fig. 1b has an upper limit – also for very low slenderness ratios – which primarily is influenced by the
gusset plate thickness tKB1 and the boundary condition of the gusset plate (pinned or fixed). The influence
of the effective width beff of the gusset plate is significant smaller.
These correlations are presented in Table 1. The tabulated load carrying capacities are calculated with
the engineering model presented in chapter 4.
607
Harald Unterweger
For pinned gusset plates (BC1) simplified LLBA = L0 , as long as L1 < 1,5·h. In addition the constant
moment M = N2,Rd·e* = N2,Rd·(tKB2 / 2) (see Fig. 1b) should be used in the buckling check. For higher
slenderness ratios the influence of the moment decreases and it can be omitted.
For fixed (BC2) gusset plates the eccentricity e* can be omitted, if LLBA = L0 is used. This
simplification leads to conservative results, mainly for high thicknesses tKB1 (see Table 2). Otherwise the
moment should be included (LLBA§ 0,85·L0 , as long as N2,Rd is relevant for design).
4.2 Upper limit for the member capacity – N1,Rd
The engineering model for the upper limit capacity N1,Rd of the member is defined by the load
carrying capacity of the gusset plate – section at the end of the member (axis II), considering 2nd order
effects. The model is summed up in Fig. 5.
The load bearing capacity N1,Rd is based on the full utilization of the plastic section capacity, due to
axial force and bending moment. As defined in Eurocode 3 [3], the acting axial force is considered by a
reduced moment capacity MN,Rd, given in Equ. 3 (fyd is the design yield strength).
2
§ § N · · hKB t 2 KB1 § N 21, Rd ·
¨
M N , Rd M pl , KB1 1 ¨ 1, Rd
¸ ¸ f yd ¨1 2 ¸ (3)
¨ ¨ N pl , KB1 ¸ ¸ 4 ¨ h KB t 2 KB1 f 2 yd ¸
© © ¹ ¹ © ¹
To calculate N1,Rd an iterative approach is necessary, until Equ. 4 is fulfilled.
M II d M N , Rd (4)
The acting bending moment MII in the gusset plate, depends on the axial force N1,Rd and the actual
boundary condition. For pinned gusset plates (BC1) the full eccentricity e0 is relevant in section II (see
Fig. 5), leading to Equ. 5. The 2nd order effect, is covered by the factor fII in form of a so called
“Dischingerfaktor”, including the Euler buckling load Ncr,BC1 for the gusset plate (Equ. 6), based on the
relevant buckling length lcr = 2·L1.
§t t · 1
M II N1, Rd e0 f II N1, Rd ¨ KB1 KB 2 ¸ (5)
© 2 N
¹ 1 1, Rd
N cr , BC1
1 §t t · 1
M II N1, Rd f M e0 f II N1, Rd ¨ KB1 KB 2 ¸ (8a)
I z , KB 2 © 2 ¹ 1 N1, Rd
1
I z , KB1 N cr , BC 2
1 §t t · 1
M II N1, Rd ¨ KB1 KB 2 ¸ (8b)
beff t 3 KB 2 © 2 ¹ 1 N1, Rd
1
hKB t 3 KB1 N cr , BC 2
608
Harald Unterweger
Figure 5: Engineering model for the determination of the compression load capacity N1,Rd,;
a.) actual situation, b.) determination of N1,Rd for BC1 (above) and BC2 (below).
609
Harald Unterweger
3
t KB1 § t KB1 ·
M II Nt , Rd et N t , Rd ¨ ¸ (10)
5,14 © 20 ¹
Table 2: Compression load capacity ratios NRd / Npl,0 (RHS  profile 100/100/5, beff = 330mm), with
imperfections as well as based on an engineering model for 3 variants:
 tKB1 = tKB2 = 12 mm (V1),  tKB1 = tKB2 = 20 mm (V2),  tKB1 = 20 / tKB2 = 12 mm (V3).
boun തതതത
slenderness ߣ ௭
dary calcu 0,548 1,097 1,645 2,194
cond. lation
V1 V2 V3 V1 V2 V3 V1 V2 V3 V1 V2 V3
BC 1 GMNIA 0,161 0,309 0,367 0,163 0,302 0,357 0,160 0,272 0,295 0,148 0,177 0,183
Eng. 0,172 0,309 0,373 0,172 0,309 0,373 0,172 0,276 0,291 0,172 0,172 0,178
model +6,8% +0,0% +1,6% +5,5% +2,3% +4,5% +7,5% +1,5% 1,4% +16% 2,8% 2,7%
BC 2 GMNIA 0,526 0,825 0,643 0,525 0,654 0,643 0,361 0,425 0,424 0,219 0,286 0,275
Eng. 0,458 0,786 0,545 0,458  0,598 0,545  0,318  0,318 0,318 0,188 0,188 0,188
model 13% 4,7% 15% 13% 8,6% 15% 12% 25% 25% 14% 34% 31%
6 CONCLUSION
The almost centric joint in Fig. 1 only for RHS  members in tension gives a load bearing capacity
comparable with the section capacity Npl,0 (about 90 % of Npl,0 ). However, thick gusset plates are
necessary, leading to an area ratio of AKB1 / A0 > 1,25.
Also for such thick gusset plates with pinned ends the load bearing capacity in compression is limited
to about 50 % of the section capacity Npl,0 , independent of the member slenderness. The reason for that
is the high additional bending moment in the gusset plate at the member end (axis II in Fig. 5).
In case of a fixed gusset plate a maximum of about 70÷80 % of the section capacity Npl,0 is available.
Only for very high member slenderness (ߣ തതതതത
௭ǡ > 1,5 / 1,0 for pinned / fixed gusset plates) the buckling
check of the member is relevant for design.
It should be noted that the studied joint configuration of Fig. 1 is nearly not usable, if fatigue loads are
relevant (stress cycles due to axial force). The reason for this, are very high stress peaks near the welds,
relevant for the fatigue check.
REFERENCES
[1] Unterweger, H. and Ofner, R., “Traglast von Verbandsstäben aus Hohlprofilen mit quasi
zentrischem Knotenblechanschluss“, Stahlbau, 78(6), 425436, 2009.
[2] ABAQUS, Software package, Version 6.7, 2007.
[3] Eurocode 3, European Standard, Design of steel structures – Part 11: General rules and rules for
buildings, 2006.
[4] Wilkinson, T., “The plastic behaviour of cold – formed rectangular hollow sections”, Doctoral
Thesis, University of Sydney, 1999.
610
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Steel Frame, Angle Bracing, Bolted Connection, Welded Connection, Numerical Modeling.
Abstract. This paper summarizes experimental and numerical investigations on the behavior of steel sub
frame specimens braced with use of a diagonal angle member. The laboratory tests were designed to
investigate the effect of bracing member end connections (bolted asymmetric and welded symmetric) and
the effect of bracing member slenderness on the frame behavior in the whole range of frame load
displacement characteristics. An advanced finite element numerical model is developed with use of
commercial ABAQUS code. Elastic buckling modes are evaluated. Displacement controlled Riks
geometrically and materially nonlinear analysis is carried out for the reproduction of the behavior of
specimens tested in laboratory. Numerical frame loaddisplacement characteristics are compared with
experimental ones. A simple analytical model of the compression member behavior developed elsewhere
is compared with experimental angle brace characteristics. Practical recommendations are formulated.
1 INTRODUCTION
Steel truss bracing systems are commonly used for the enhancement of sway performance of
structural frames in multistorey buildings. Vertical truss bracings are composed of diagonals made of
rolled profiles, single or compound. In typical braced frames of moderate height, diagonals are made of
single angles connected through bolted asymmetric joints or welded symmetric joints to continuous
columns, or columns and beams, of the primary load bearing frame.
The behavior of structural members is usually examined experimentally as isolated elements; see
Gizejowski et al. [1]. This type of experiments is helpful in predicting the effect of member slenderness
on the buckling strength but is not adequate to model the member performance as an element of structural
systems, especially with reference to the influence of real end conditions of bracing members on their
buckling strength and forcedeformation characteristic. The behavior of members acting integrally as
elements of braced frame structures has been less investigated.
Analytical models, one based on tangent modulus theory and the second  on the evaluation of overall
member forcedeformation characteristic, have been developed by Barszcz and Gizejowski in [2] for the
prediction of buckling strength according to Eurocode 3. Model based on the evaluation of overall
member forcedeformation characteristic for the assessment of loaddisplacement characteristic of more
complex structures was presented in [3]. The above mentioned analytical model of the member behavior
is further verified in this paper for angle bracing members tested as components of subframe specimens.
A summary of experimental work concerned with the behavior of angle member as an element of
braced subframe portal specimens is presented hereafter. The experimental loaddisplacement behavior
of tested specimens is compared with the FE results obtained with use of commercial ABAQUS code.
Experimental forcedeformation characteristics of the brace angle are compared with the analytical model
611
Anna M. Barszcz et al.
dealt with in [2] and [3]. Conclusions directed towards the practical application of developed analytical
model for modeling of the behavior of bolted and welded angle braces are drawn.
2 EXPERIMENTAL INVESTIGATIONS
Tests were designed to examine experimentally the behavior of two sets of braced subframes that are
presented in figure 1. Figure 1a illustrates the general layout of subframe specimens BL with the angle
brace jointed to gusset plates with use of highstrength bolts though close tolerance holes, and figure 1b –
specimens WL with the angle brace welded to gusset plates. Details of both types of the connection are
given in figure 2a and b, respectively.
a) b)
a) b)
Each set of tested specimens consists of three subsets characterized by different beam length L and
the diagonal distance Ld. Notation for each subset of specimens is given in table 1. In each subset of BL
and WL specimens, three specimens were tested.
Tests were conducted for subframes mounted in an upsidedown position. In the upper left node, the
bracing member axis coincided with the left column axis and the node was held in position but allowed
for the inplane rotation. The same boundary conditions were applied to the upper right node of the right
column end. The upper left node was loaded with a horizontal load F and the upper right node was
subjected to a reactive force. The beam of subframe specimens was restrained in the outofplane
direction in order to ensure that the frame deflects primarily inplane. The incremental displacement
612
Anna M. Barszcz et al.
controlled loading program, corresponding to the horizontal load F, was applied. Details of test rig,
testing procedure and measurement devices were described in [4].
Figure 3: Frame critical loads and buckling modes from numerical simulations.
The multi point constraints option (MPCs) available in ABAQUS code is used to model the bolts in
the braced subframe specimens BL. For each bolt MPC, rigid beam like constrains are created at
matching nodes of two midsurfaces of the angle brace leg and the gusset plate around the bolt shank
613
Anna M. Barszcz et al.
circumference approximated by a square in order to simplify capturing the behavior of two plate elements
jointed with highstrength close tolerance bolts. This helps also to avoid early analysis termination in case
of excessive plastic deformations in concentrated bearing zones, between the bolt shank and connecting
plate elements. Such a modeling technique seems to be accurate from engineering point of view.
The angle brace welded to gusset plates is modeled with connecting the angle brace to gusset plate
directly via common nodes along the intersection line between the angle brace and the gusset plate. It is
strongly believed that this modeling technique is quite accurate for the welded joint as long as no fracture
is developed in the area of weld material or its neighborhood.
In order to estimate the elastic failure loads and to find the sensitivity of tested specimens to buckling
effects, eigenproblems are solved. Critical loads Fcr and buckling modes are evaluated. Figure 3 illustrates
the buckling modes and the values of critical loads. It is observed that for the shortbeam specimens a
distortional form of bracing member buckling governs while for the longest beam specimens  an overall
buckling mode is detected. It is clear that buckling is associated primarily with the bracing member
deformations and rather a negligible contribution of the frame beam and column deformations to the
buckling profile is observed.
a) WL 1320: Fult=124,6 kN
b) WL 1520: Fult=124,4 kN
c) WL 1925: Fult=115,3 kN
Figure 4: Frame deformation of welded brace specimens from tests and numerical simulations.
Finally, ultimate loads and deflected profiles are evaluated from fully nonlinear Riks analysis using
ABAQUS code. Figure 4 gives the values of the frame ultimate loads Fult and illustrates the deformed
profiles of inelastic WL subframes at failure, comparing them with those recorded during tests. One can
614
Anna M. Barszcz et al.
observe that the deflected frame profile is characterized by localized plastic deformations of the lower
node of the right column. Since bracing members in these specimen subsets are connected concentrically,
their deflected profile is associated with buckling deformations. The maximum coordinate of buckling
profile shifts from the lower joint for a shorter brace length to the midlength for a longer brace length.
Figure 5 gives the values of the frame ultimate loads Fult and illustrates deformed profiles of inelastic
BL subframes at failure, comparing them with those recorded during tests. One can observe that the
deflected frame profile of BL specimens is similar to that of WL subframes. Since bracing members in
all BL subsets of tested specimens are connected eccentrically, their deflected profile is associated with
bending and torsion, without distinguished buckling inplane deformations.
a) BL 1320: Fult=113,1 kN
b) BL 1520: Fult=109,3 kN
c) BL 1925: Fult=105,8 kN
Figure 5: Frame deformation of bolted brace specimens from tests and numerical simulations.
The most stressed joint in all the tests appeared to be the right lower frame joint at which three
elements are connected – the frame beam and column, and the bracing member. The detailed deformation
of this joint obtained from laboratory tests and from numerical modeling is shown in figure 6.
Figure 6: Details of WL 1320 columnbeambracing joint deformations from tests and simulations.
615
Anna M. Barszcz et al.
Frame loaddisplacements characteristics from laboratory tests and from numerical modeling are
presented in figure 7. All six frame experimental characteristics are evaluated as average curves from
three tests of each subset. Since numerical modeling is carried out for initially perfect specimens, the
accuracy of computer simulations is of a different degree depending on the sensitivity to imperfections
and type of bracing member connections.
a) BL 1320 d) WL 1320
b) BL 1520 e) WL 1520
c) BL 1925 f) WL 1925
616
Anna M. Barszcz et al.
postlimit branch of equilibrium path if compared with those obtained experimentally. Effect of geometric
imperfections of the bracing member does not play important role in this case since the member is under
combined bending, torsion and compression from the beginning of loading process.
b) For frame specimens WL with concentric connections of the bracing member, computer
simulations lead to a very close estimation of the initial stiffness but a higher ultimate loads and a lower
placement of the postlimit branch of equilibrium path if compared with those obtained experimentally.
Higher values of the ultimate load from computer simulations can be attributed to the effect of geometric
imperfections of the bracing member that is not accounted for in analysis. Since the bracing member is
concentrically connected with regard to outofplane deformations, it is predominantly axially loaded
almost up to the attainment of the frame buckling strength. A sharp drop in the value of applied load is
observed in the postlimit range. A more close estimation could be expected if the effect of geometric
imperfections is accounted for in computer simulations.
b) BL 1520 e) WL 1520
c) BL 1925 f) WL 1925
Figure 8: Bracing member forcedeformation characteristics from tests and analytical model.
617
Anna M. Barszcz et al.
Finally, the behavior of bracing member assessed on the basis of test results is compared with simple
modeling of forcedeformation characteristic developed in [2]. Forcedeformation characteristics
presented in figure 8 are constructed in terms of the member axial force and member generalized strain
(shortening due to compression and bending divided by the length). Results of three experimental curves
are plotted for each subset of tested frame specimens. The upper bound of analytical solution is
constructed for the effective length factor ȝ=0.5 while the lower bound it corresponds to the effective
length factor ȝ=1.0.
The following observations are made:
a) The evaluation of welded connection bracing member behavior is reasonable enough using the
upper bound analytical solution. It indicates that that the forcedeformation characteristic of the welded
type of member connection may be evaluated with the effective length factor equal 0.5.
b) The bolted connection bracing member behavior can not be evaluated so accurately using the cited
analytical formulation. The values of experimental initial stiffness are lower than those from analytical
solution for both values of the effective length factor. The experimental buckling strength is placed
between those corresponding to two extreme values of the effective length factor, i.e. for two different
types of boundary conditions assumed for the connection of brace member. This clearly indicates that the
strut model developed in [2] has to be refined in case of bolted connections in order to account for the
effect of connection eccentricity.
4 CONCLUSIONS
This paper presents experimental, numerical and analytical results of investigations aiming at the
development of a simple and yet reliable model for the angle member behavior in bracing systems of
structural frame structures. Welded brace angle forcedeformation characteristics may be evaluated using
directly the analytical formulation presented in [2] for the effective length factor equal 0.5. Bolted brace
angle forcedeformation characteristics can not be evaluated so accurately using the formulation
presented in [2]. The refinement is needed with regard to the effect of connection eccentricity and
inclusion of corresponding bending deformations in the evaluation of member generalized strain. This
would affect predominantly the region of prebuckling deformations and the level of buckling strength,
resulting in lowering of the member forcedeformation curve and bringing the analytical curves closer to
the experimental ones. A more accurate model would therefore be developed and the effective length
factor suggested accordingly.
The development of a refined model of the angle strut behavior in case of bolted connections is being
underway, and better prediction of the bolted brace forcedeformation characteristic is expected.
REFERENCES
[1] Gizejowski M.A., Barszcz A.M., Foster J.D.G., Uziak J., Kanyeto O.J., “Experimental
investigations of the behaviour of angle struts”, Proc. of ICMS2006 XIth International Conference
on Metal Structures, M. A. Gizejowski, A. Kozlowski, L. Sleczka and J. Ziolko (eds.), Taylor &
Francis, London / Leiden / New York / Philadelphia / Singapore, 152153, 2006.
[2] Barszcz, A.M., Gizejowski, M.A. “An equivalent stiffness approach for modelling the behaviour of
compression members according to Eurocode 3”. Journal of Constructional Steel Research, 63(1),
5570, 2007.
[3] Gizejowski M.A., Barszcz A.M., “Advanced analysis of inelastic steel truss and frame structures: a
unified approach”, Proc. of SDSS2006 International Colloquium on Stability and Ductility of Steel
Structures, D. Camotim, N. Silvestre and P.B. Dinis (eds.), IST Press, Lisbon, 431438, 2006.
[4] Barszcz A.M., “Modelling and experimental investigations of the behaviour of angle bracing strut
in steel frames”, Proc. of Local Seminar of IASS Polish Chapter on Lightweight Structures in Civil
Engineering: Contemporary Problems, J. Obrebski (ed), MicroPublisher, Warsaw, 106113, 2007.
618
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Pedro D. Simão*, **, Ana M. Girão Coelho*, *** and Frans S. K. Bijlaard***
Keywords: Column Stability, GBT, Lagrange Multipliers, RayleighRitz method, Spliced Columns.
Abstract. The paper presents a study on the influence of splice connections on the stability behaviour of
compressed steel columns. The column is modelled as two independent prismatic parts connected by a
rotational spring at the splice location and rotational and extensional springs at the column ends to
represent the effect of the adjacent structure. The general behaviour is characterized using a polynomial
RayleighRitz approximation substituted into the potential energy function, in combination with the La
grange’s method of undetermined multipliers, and based on this model the critical load is found. The
loadcarrying capacity is analysed with respect to the following variables: (i) location and rotational
stiffness of the splice, (ii) change in the column section serial size and (iii) column endrestraints stiffness
coefficients. A nonlinear regression model is developed to predict simple relationships between the criti
cal load and the relevant column characteristics.
1 INTRODUCTION
In structural engineering practice and due to manufacturing, transportation and/or handling restraints,
individual steel elements are usually fabricated with a maximum length of 12 meters. During erection of
a steel frame and where the element length is insufficient, splices are provided to form a single and
longer element. Designers often use the splices for changing crosssections, in view of a more economical
and rational design.
In steelwork construction, column splices are located at a convenient distance for erection and con
struction above floor beam level and have to be designed (i) to join lengths in line, (ii) to transmit forces
and moments between the connected member parts and (iii) to maintain continuity of strength and stiff
ness through the splice to safeguard the robustness of the structure [1]. Column splices are usually disre
garded in determining the distribution of moments and forces in the structure and when the design of the
columns itself is being considered, assuming that the splice is providing full continuity in stiffness and
strength of the column. This practice is questionable as the splices most times do not provide this conti
nuity. So, the splices may adversely affect the overall frame behaviour, from a stiffness and strength
point of view.
Previous research pertaining to the loadcarrying capacity of spliced columns includes investigations
by Lindner [2], Snijder and Hoenderkamp [3] and Girão Coelho et al. [4,5,6]. Lindner [2] carried out
experimental and numerical tests on different column splice types and highlighted the existence of eccen
tricities at the splice. An adjusted buckling curve for columns having contact splices at column mid
height was later proposed Lindner [2]. Snijder and Hoenderkamp [3] conducted a series of experimental
tests to analyse the influence of end plate splices on the loadcarrying capacity of slender columns. These
619
Pedro D. Simão et al.
tests were used to make design recommendations for column splices. Girão Coelho and coauthors [4,5]
further extended this work to produce a relatively simple yet reasonably accurate engineering method for
predicting the critical behaviour of spliced columns in steel frames. This paper is a followup study to
this research.
The current work presents a generalized energy formulation of a framed spliced column in sway and
nonsway frames (Figure 1). For analytical modelling, a framed column is represented by means of ex
tensional and rotational restraints at the ends A and B. The splice is modelled as a rotational spring at
point C. The potential energy functional of this system uses a RayleighRitz approximation of the rele
vant deformation modes of the column. This formulation is presented together with the method of La
grange multipliers to deal with the constraints at the column splice. Elastic buckling analysis is carried
out to find the critical load of the system. The concept of end fixity factor C [7] is successfully applied
and the significance of this factor in simplifying the analysis of results is emphasized. Simple relation
ships between C and the relevant characteristics of the column and splice are derived to a point where the
critical load can be readily determined by hand or by computer.
x,u
NEd
B
LII Member II:
E, III
C
L
LI Member I:
E, II
z, w
A
2 ENERGY FORMULATION
620
Pedro D. Simão et al.
whereby zj is the distance between point j and the neutral axis, iu(y, z) and iw(y, z) are preestablished
modal displacement patterns defined along the member crosssection, and ifu(x) and ifw(x) are modal
amplitude functions for warping and transverse displacements, respectively. For any mode of deforma
tion the amplitude functions for axial and transverse displacements are related in the form [8,9]:
d ª¬ i f w x º¼ i
i
fu x f wc x i f c (3)
dx
The extensional strain of a column segment of length dx is readily defined as:
2 2
§ wu · § ww ·
Hx¨1 wx ¸ ¨ wx ¸ 1 (4)
© ¹ © ¹
Expansion in Taylor series, neglecting higherorder terms, yields the following kinematic relation:
22
wu 1 § ww · 1 2 2 k l k l
Hx  ¨
wx 2 © wx ¹ k 1
¸ ¦ u f cc ¦¦ w w f c f c
k k
2k 1l 1
(5)
From Hooke’s law (constitutive relation), the longitudinal stress is then given by:
ª 2 1 2 2 º
V x EH x E «¦ i u i f cc ¦¦ i w j w i f c j f c »
2i1j1
(6)
¬i 1 ¼
The internal strain energy of the member, Um, is then equal to [5]:
1 EA 1 2 EI 2 2 EA 2 21 EA 2 4
Um ³ V x H x d:
2 2 ³
f cc d x
2 ³
f cc d x
2 ³ f c f cc d x
8 ³L
fc dx (7)
: L L L
where ȍ denotes the member’s volume, A the cross sectional area and I the moment of inertia.
2.2 Strain energy stored in the springs
The energy stored in the linear springs (Figure 1) is given by the following expressions:
1 E I II 2
Rotational spring at end A: U șa
2
K șa T a2
2L
kșa 2 f Ic
x 0
(8)
2
1 E I II
Rotational spring at end B: U șb K șb T b2 kșb 2 f IIc (9)
2 2L x LII
2
1 E I II
Rotational spring at splice: U șc K șc T c2 kșc 2 f IIc 2 f Ic (10)
2 2L x 0 x L I
2
1 E I II
Extensional spring at end B: U 'b K 'b ' b2 k 'b
2
f II (11)
2 2 L3 x LII
where KT and K' are rotational and extensional spring constants, respectively, and k are spring coeffi
cients that are defined in nondimensional form.
2.3 Work done by load
The final component of energy to be identified is the work done by the load. For a centrally loaded
column, the potential energy of the external loading is given by:
3 N Ed u 1 f IIc (12)
x LII
621
Pedro D. Simão et al.
§ ·
uI,C uII,C 0 1 i
f Ic
¨¨ G1 a j , N Ed 0 ¸¸
x LI
1 f IIc
x 0
0
© ¹
(14)
§ ·
wI,C wII,C 0 2
fI
x LI
2
f II
x 0
i
0 ¨ G2 a j , N Ed 0 ¸
¨ ¸
© ¹
We now wish to find a stationary value of a functional subjected to some subsidiary conditions or con
straints Gk(iaj,NEd). The problem is easily tackled by using the approach proposed by Lagrange [10]. The
technique is to form a modified potential energy expression:
m
V i a j , Ok , N Ed
V i a j , N Ed ¦ Ok Gk i a j , N Ed
k 1
(15)
where Ok are the Lagrange multipliers. For the spliced column, the modified potential energy functional
may be written as:
V U m U șa U șb U șc U 'b 3 O1 fc
1
I x L
I
1 f IIc
x 0 O f 2
2
I x L
I
1 f II
x 0 (16)
These polynomials form a set of coordinate functions that satisfy the kinetic boundary conditions of the
problem and are orthonormal functions that enable fast convergence of the method [9]. The coordinate
functions are given by:
Member I:
1
M1,I 5 LI x 2 LI
2
M1,I 3 LI x
2
M 2,I
5 LI 3 x 4 x 2 LI
2
M3,I
7 LI 6 x 20 x LI 15 x3 LI 2 2
(18)
2
M4,I
9 LI 10 x 60 x LI 105 x LI 56 x 4 LI3
2 3 2
2
M5,I
11 LI 15 x 140 x 2 LI 420 x3 LI 2 504 x 4 LI3 210 x5 LI 4
2
M6,I
13 LI 21x 280 x 2 LI 1260 x3 LI 2 2520 x 4 LI 3 2310 x5 LI 4 792 x 6 LI5
2
M7,I
15 LI 28 x 504 x LI 3150 x LI 9240 x LI 13860 x LI 10296 x LI 3003 x 7 LI 6
2 3 2 4 3 5 4 6 5
622
Pedro D. Simão et al.
Member II:
1
M1,II x LII
1
M 2,II 3 LII 1 x LII
2
M1,II LII
2
M2,II 3 LII LII 2 x
(19)
2
M3,II
5 LII LII 6 x 6 x 2 LII
2
M4,II
7 LII LII 12 x 30 x 2 LII 20 x 3 LII 2
2
M5,II
9 LII LII 20 x 90 x LII 140 x LII 70 x 4 LII 3
2 3 2
2
M6,II 2
11 LII LII 30 x 210 x LII 560 x LII 630 x LII 252 x 5 LII 4 3 2 4 3
2
M7,II
15 LII LII 42 x 420 x LII 1680 x LII 3150 x LII 2772 x LII 924 x 6 LII5
2 3 2 4 3 5 4
Equilibrium of the system is obtained by rendering stationary the total potential function with respect
to the generalized coordinates iaj and Ok. The solution that emerges from the unloaded state, the funda
mental path (FP) is a function of NEd. A sliding set of incremental coordinates iqj and qk is then defined by
the following equations [11]:
i
a j i a j FP N Ed i q j and Ok Ok FP N Ed qk (20)
A new energy function W is now introduced [11]:
W i q j , qk , N Ed
V i a j FP i q j , Ok FP qk , N Ed (21)
i
A global numbering for coordinates qj and qk (ql) can now be adopted. The equilibrium and stability
conditions hold good for this transformed energy function W. In this new NEd – ql space, the fundamental
path is defined trivially by ql = 0.The critical points along the fundamental path are now those points that
render zero the determinant of the total potential energy Hessian matrix along the fundamental path:
H FP H FP,0 N Ed H FP,1 (22)
The relevant states of critical equilibrium are identified via a local linear eigenvalue equation
HFP q = 0, q representing the local eigenvector [11]. By substituting the forms in Eq. (22), we thus obtain
the critical state identity:
H FPq H FP,0 N Ed H FP,1 q 0 (23)
This analysis yields the critical buckling load of the spliced column, Ncr that can be expressed in terms of
an end fixity factor C [7]:
N cr CS 2 EI II L2 (24)
3 NUMERICAL RESULTS
The purpose of this numerical study is to ascertain the effect of the following variables on the general
equilibrium response: (i) splice location (LI = DL), (ii) ratio between second moment of area of lower and
upper column members (E = II/III), (iii) splice rotational stiffness (kTc) and (iv) endrestraints stiffness
coefficients (kTa, kTb and k'b). Results are independent from the column length L. These properties are
varied parametrically as shown in Table 1.
623
Pedro D. Simão et al.
These results form a comprehensive analytical database. We can now generate a continuous function
that approximates the values of the end fixity factor within the domain of analyses and with a minimum
error. The analysis results are then used to develop a multiple regression model to approximate the end
fixity factor C { Cfit(Į, ȕ, kșa, kșb, kșc, kǻb) from the data in the database, by means of piecewise approxi
mations. In developing the regression model, the relationship between the dependent variable and each
independent variable is studied separately, while all other independent variables are kept constant. The
dependent variable is approximated by a continuous function that is linear in terms of a set of regression
coefficients, which are determined by enforcing the method of least squares that minimizes the sum of
the squares of the residuals. Approximating (or coordinate) functions are then selected for each inde
pendent variable. The multiple regression model is formed as the product of the individual coordinate
functions.
kTa = 0.01, kTb = 100, kǻb = 5 kTa = 100, kTb = 100, kǻb = 0.01
kTc = 0.01, 1, 5, 10, 20, 100 kTc = 0.01, 1, 5, 10, 20, 100
624
Pedro D. Simão et al.
Forming the regression model as a product allows the effect of each independent variable to be exam
ined separately and facilitates the process of selecting suitable coordinate functions for the individual
independent variables. Some key results are illustrated graphically in Figure 2. The graphics suggest that:
1. The response C vs. D can be approximated by a quadratic function.
2. Typical C vs. Ebehaviour is characterized by a monotonic increasing function that can be gener
ally approximated by a simple linear relationship.
3. The degree of rotational and extensional end restraint is an essential parameter for the computation
of C. The shape of the curves C vs. k (k { kTa, kTb), C vs. kTc and C vs. k'b is best described by an arctan
gent function.
The end fixity factor is then predicted by means of an expression in the form:
ª § k ·º
Cfit D , E , k'b , kșa , kșb , kșc C1 1 C2D C3D 2 1 C4 E «1 C5 arctan ¨¨ 'b ¸¸ » u
¬« © C6 ¹ ¼»
(25)
ª § kșa · º ª § kșb · º ª § kșc · º
u «1 C7 arctan ¨¨ ¸¸ » «1 C9 arctan ¨¨ C ¸¸ » «1 C11 arctan ¨ C ¸ »
¬« © C8 ¹ ¼» ¬« © 10 ¹ ¼» ¬« © 12 ¹ ¼»
where Ci are regression coefficients.
As expected in developing a predictive regression model, many models were tried, analysed and as
sessed for accuracy and effectiveness. The final model presented here evolved out of several attempts to
develop conventional (nonlinear) regression models by means of simple mathematical functions. The
overall character of the response is well captured and the number of regression coefficients is kept small
in order to provide a compact procedure for the simplified method. The accuracy of the model is meas
ured by means of the RSquared value (R2). The RSquared gives the fraction of the variation of the
response that is predicted by the model. A good model fit yields values of Rsquared close to unity.
Nonlinear regression analysis is performed with the Mathematica software [12]. Regression coeffi
cients are determined for the spliced column using piecewise approximations depending on the nature of
the segments that comprise the above relationship. The domain of analyses of the spring coefficients is
divided into three intervals: G1 for k ]0,3], G2 for k ]3,15] and G3 for k ]15,100]. Table 2 sets out
the computed regression coefficients and values for the RSquared factor are also given.
625
Pedro D. Simão et al.
REFERENCES
[1] CEN (European Committee for Standardization), EN 199318 – Eurocode 3: Design of steel struc
tures – Part 18: Design of joints, Brussels, 2005.
[2] Lindner, J., “Old and new solutions for contact splices in columns”. Journal of Constructional Steel
Research, 64, 833844, 2008.
[3] Snijder, H.H., Hoenderkamp, J.C.D., “Influence of end plate splices on the load carrying capacity
of columns”. Journal of Constructional Steel Research, 64, 845853, 2008.
[4] Girão Coelho A.M., Bijlaard F.S.K., “Requirements for the design of column splices”, Stevin Re
port 6083, Delft University of Technology, 2008.
[5] Girão Coelho A.M., Simão P.D., Bijlaard F.S.K., “Stability design criteria for steel column
splices”. Journal of Constructional Steel Research, 66, 12611277, 2010.
[6] Girão Coelho A.M., Bijlaard F.S.K., Simão P.D., “Stability design criteria for steel column splices
in nonsway frames”. Proceedings of the fourth international conference on structural engineering,
mechanics and computation SEMC 2010, (accepted for publication).
[7] Simitses, G.J., An introduction to the elastic stability of structures, Krieger Publishing Company,
Malabar, 1986.
[8] Schardt, R., Verallgemeinerte Technische Biegetheorie, Springer, BerlinHeidelberg, Germany,
1989.
[9] Simão, P.D., Postbuckling bifurcational analysis of thinwalled prismatic members in the context
of the Generalized Beam Theory, Ph.D. thesis, University of Coimbra, Portugal, 2007
(www.inescc.pt/documentos/dissertations/PDSimao_PhDthesis.pdf).
[10] Richards, T.H., Energy methods in stress analysis, Ellis Horwood, Chichester, UK, 1977.
[11] Thompson J.M.T., Hunt G.W., A General Theory of Elastic Stability, John Wiley & Sons, London,
UK, 1973.
[12] Mathematica 6, Wolfram Corp., Champaign, USA, 2007.
626
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Tadeh Zirakian
Abstract. The experimental determination of critical buckling load of structures undergoing lateral
buckling has usually been accompanied by the application of certain and just a few deformation
characteristics such as lateral displacement and/or twist. This paper explores the possibility of
application of various deformation variables such as web transverse and longitudinal strains, vertical
deflection, and angles of twist of top and bottom flanges for experimental determination of the critical
buckling load of Ibeams with different initial geometrical imperfections undergoing elastic lateral
distortional buckling. After demonstrating the linear relationship between lateral displacement and the
various aforementioned deformation variables, the four Southwell, Massey, Modified, and Meck
extrapolation techniques are applied on these various deformation variables, and consequently
satisfactory estimates are acquired for the critical buckling loads.
1 INTRODUCTION
The extrapolation or plotting techniques are experimental methods developed for determining
experimentally the critical buckling load of structures, without having to test them to failure. By plotting
the results of a structure test in a certain manner, it would be possible to determine the structure’s
buckling load.
Southwell [1] initially proposed a plotting method for a concentrically loaded pinended column with
a sinusoidal initial imperfection. Later on, Massey [2], Trahair [3], and Meck [4] successfully applied this
method and variations of it to predict buckling loads for beams.
A search of the literature shows that the extrapolation techniques have mostly been used on certain
and just a few deformation characteristics such as lateral displacement and/or twist, and also Mandal and
Calladine [5] demonstrated that lateral displacement tends to be proportional to rotation as deformations
increase in lateraltorsional buckling mode. In a recent research work reported by the author [6], it was
demonstrated that lateral displacement in Ibeams undergoing lateraldistortional mode of buckling tends
to be directly coupled with the web transverse strains developed as a result of occurrence of web
distortion, and accordingly the application of the extrapolation techniques on the web transverse strains
yielded good predictions for the critical buckling load.
Based on findings of the previous studies, the possibility of application of various deformation
variables for experimental determination of the critical buckling load is explored in this paper. Initially,
the relationship between lateral displacement and the various considered deformation variables including
web transverse and longitudinal strains, vertical deflection, and angles of twist of top and bottom flanges
of Ibeams with different initial geometrical imperfections is investigated, and subsequently the
applicability of the Southwell, Massey, Modified, and Meck Plots on the aforementioned deformation
variables is studied.
627
Tadeh Zirakian
Four finite element lateraldistortional buckling solutions are developed using the ABAQUS software
system [7]. All of the solutions are of simplysupported steel Ibeams subjected to uniform bending
moment with identical crosssection dimensions hw=900 mm, bf=240 mm, tw=5 mm, tf=20 mm, and
L=7000 mm. The material properties adopted for the beams are E=200 GPa, v=0.3, Fy=345 MPa, and
G=0.385E. The beam crosssection components, i.e. flanges and the web, were modeled using a four
node shell element S4R5. Finally, the details of various geometrical imperfection forms of the beams are
provided in Table 1.
Lateral displacement vs. web longitudinal strain: In addition to the web transverse strain, the
relationship between lateral displacement and web longitudinal strain is investigated as well. As seen in
Figure 2, lateral displacement is plotted against web longitudinal strain (measured at midspan and
midheight), and the obtained linear equations as well as the Rsquared values are displayed on the chart.
It is clearly observed that after the initial loading stages, the data points align with the linear portion near
the latter loading stages and the direct coupling of the two deformation characteristics becomes evident.
628
Tadeh Zirakian
400 400
350 350
įL (mm)
200 200
150 150
100 100
50 50
0 0
0.0003 0.0002 0.0001 0 0.0001 0.0002 0.0003 0.0004 0.0002 0 0.0002 0.0004
İT İT
350 350
300 300
į L = 139884İ T + 282.27 į L = 127840İ T + 284.66
250 2 250
2
R = 0.9977 R = 0.9971
įL (mm)
įL (mm)
200 200
150 150
100 100
50 50
0 0
0.0006 0.0004 0.0002 0 0.0002 0.0004 0.0006 0.0004 0.0002 0 0.0002 0.0004
İT İT
350 350
300 300
į L = 87525İ L + 253.42
į L = 92964İ L + 247.77
250 250 2
2 R = 0.9979
R = 0.9978
įL (mm)
įL (mm)
200 200
150 150
100 100
50 50
0 0
0.0005 0 0.0005 0.001 0.0015 0.0005 0 0.0005 0.001 0.0015
İL İL
350 350
300 300
2
200 2 200 R = 0.9983
R = 0.9987
150 150
100 100
50 50
0 0
0.0005 0 0.0005 0.001 0.0015 0.0005 0 0.0005 0.001 0.0015
İL İL
629
Tadeh Zirakian
Lateral displacement vs. vertical deflection: The proportionality between the lateral displacement
and vertical or inplane deflection at midspan and midheight of the analyzed Ibeams undergoing lateral
distortional buckling is investigated in here. The plots of lateral displacement against vertical deflection
are shown in Figure 3. The linear equations and Rsquared values are displayed in the figures as well. To
a fair approximation, lateral displacement and vertical deflection seem to be proportional to each other
and the linearity range is comparatively large in this case.
400 400
350 350
300 300
250 250
į L = 3.0863į V + 113.9
įL (mm)
įL (mm)
200 2 200
R = 0.9979
į L = 4.1982į V + 15.085
150 150 2
R = 0.9807
100 100
50 50
0 0
0 20 40 60 80 0 20 40 60 80 100
į V (mm) į V (mm)
400 400
350 350
300 300
250 250
įL (mm)
įL (mm)
200 200
į L = 4.41į V + 6.6499
150 į L = 4.8986į V  3.5458 150
2
R = 0.9922
R 2 = 0.992
100 100
50 50
0 0
0 20 40 60 80 0 20 40 60 80 100
į V (mm) į V (mm)
Lateral displacement vs. angle of twist: Lastly, the relationship between the lateral displacement and
angles of twist of top and bottom flanges of the Ibeams is examined. Unlike the lateraltorsional mode of
buckling, in lateraldistortional buckling mode top and bottom flanges have different angles of twist,
hence the two angles of twist are taken into consideration in this study. Plots of lateral displacement
against angles of twist of top and bottom flanges are made and shown in Figure 4. It is evident that lateral
displacement and angles of twist of the two flanges are directly coupled.
630
Tadeh Zirakian
400 400
250 250
įL (mm)
įL (mm)
200 200
50 50
0 0
0.1 0 0.1 0.2 0.3 0.1 0 0.1 0.2 0.3
ș (rad.) ș (rad.)
400 400
350 350
300 300
į L = 410.34ș TF + 278.89 į L = 361.52ș TF + 281.38
2 250 250
R = 0.9981 2
R = 0.9965
įL (mm)
200
įL (mm)
200
150 150
100 į L = 1287ș BF  3.0457 100
R 2 = 0.9999 į L = 1292.1ș BF + 4.0579
50
50 2
R = 0.9996
0
0
0.2 0.1 0 0.1 0.2 0.3
50 0.2 0.1 0 0.1 0.2 0.3
ș (rad.) ș (rad.)
3 APPLICATION
Based on the linear relationship between lateral displacement and the other considered deformation
variables, it seems logical to obtain straight lines by applying the extrapolation techniques on the various
deformation variables, and also acquire favorable predictions for the buckling load. To prove this, the
following five case studies are performed.
Lateral displacement: In this case, Southwell, Massey, and Modified Plots are applied on the lateral
displacement of the top flange at midspan. As an example, the Southwell, Massey, and Modified Plots for
beam 4 are shown in Figure 5. The extrapolatedtoultimate failure moment ratios for the three methods
are also given in Table 2.
631
Tadeh Zirakian
6.00E04 300000
0.25
5.00E04 250000
0.2
įL / M2
įL * M
įL / M
4.00E04 200000
0.15
3.00E04 150000
0.1
2.00E04 100000
0 0.00E+00 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
įL įL įL
Table 2: Comparison of the ultimate failure moments with the extrapolated moments for the case of
lateral displacement
Beam number MSouthwell/Mu MMassey/Mu MModified/Mu
1 1.038 0.992 1.025
2 1.069 1.352 1.051
3 1.137 1.174 1.189
4 1.165 1.474 1.232
Average Discrepancy (%) 9.07 18.44 10.50
In spite of some scatter in the results, it can be seen from the table that the extrapolated moments are
generally in good agreement with the ultimate failure moments. Moreover, as it is seen, the lowest
average discrepancy is found in the case of the Southwell Plot.
Web transverse strain: The four Southwell, Massey, Modified, and Meck Plot methods are applied
on the web transverse strains captured at the midheight and midspan of the analyzed beams. The
extrapolatedtoultimate failure moment ratios for the plotting methods are given in Table 3.
Table 3: Comparison of the ultimate failure moments with the extrapolated moments for the case of web
transverse strain
Beam number MSouthwell/Mu MMassey/Mu MModified/Mu MMeck/Mu a
1 1.038 0.992 0.995 1.040
2 0.950 1.022 1.003 1.033
3 1.010 1.017 1.002 1.009
4 1.036 1.042 1.008 0.992
Average Discrepancy (%) 3.29 2.15 0.45 2.18
a
MMeck is obtained as a result of the use of Meck Plot method on lateral displacement and web transverse
strain.
632
Tadeh Zirakian
As it is seen in Table 3, some scatter in the results is present in this case as well. In general, the
agreement between the extrapolated and the ultimate failure moments is satisfactory in all cases. Lastly, in
this case, the lowest average discrepancy is found in the case of the Modified Plot.
Web longitudinal strain: In this case, the use of Southwell, Massey, Modified, and Meck Plots on
the web longitudinal strains captured at the midheight and midspan of the considered Ibeams, is
investigated. The extrapolatedtoultimate failure moment ratios for the four considered Plot methods are
presented in Table 4.
Table 4: Comparison of the ultimate failure moments with the extrapolated moments for the case of web
longitudinal strain
Beam number MSouthwell/Mu MMassey/Mu MModified/Mu MMeck/Mu a
1 1.038 0.992 1.044 1.014
2 1.069 1.022 1.051 1.092
3 1.010 1.087 1.044 1.018
4 1.036 1.114 1.059 1.091
Average Discrepancy (%) 3.64 5.28 4.73 4.96
a
MMeck is obtained as a result of the use of Meck Plot method on lateral displacement and web
longitudinal strain.
As can be seen from the results in Table 4, despite some scatter, the agreement between the
extrapolated and the ultimate failure moments is satisfactory. Furthermore, in this case, the lowest
average discrepancy is found in the case of the Southwell Plot.
Vertical deflection: The applicability of the extrapolation techniques on the beam midspan vertical
deflection is studied in this case. The extrapolatedtoultimate failure moment ratios for the four applied
plotting methods are given in Table 5.
Table 5: Comparison of the ultimate failure moments with the extrapolated moments for the case of
vertical deflection
Beam number MSouthwell/Mu MMassey/Mu MModified/Mu MMeck/Mu a
1 1.038 1.072 1.025 1.022
2 1.069 1.209 1.040 1.104
3 1.137 1.174 1.126 1.010
4 1.165 1.318 1.154 1.077
Average Discrepancy (%) 9.07 15.74 7.72 4.92
a
MMeck is obtained as a result of the use of Meck Plot method on lateral displacement and vertical
deflection.
In spite of the scatter in the results, it may be concluded that the extrapolated and the ultimate failure
moments are generally in good agreement in all cases. Also, the lowest average discrepancy in the present
case study is interestingly found in the case of the modified Meck Plot.
Angle of twist: In the last case study, the four considered extrapolation techniques are applied on the
angles of twist of both top and bottom flanges captured at midspan of the analyzed beams. The
extrapolatedtoultimate failure moment ratios for the Southwell, Massey, Modified, and Meck Plot
methods are presented in Table 6.
633
Tadeh Zirakian
Table 6: Comparison of the ultimate failure moments with the extrapolated moments for the case of angle
of twist
Beam number Locationa MSouthwell/Mu MMassey/Mu MModified/Mu MMeck/Mu b
TF 1.038 0.992 1.001 0.948
1
BF 1.038 0.992 1.022 1.007
TF 0.950 1.022 1.002 1.037
2
BF 1.069 1.209 1.042 1.046
TF 1.010 1.017 1.007 0.956
3
BF 1.137 1.286 1.181 1.043
TF 1.036 1.042 1.011 1.025
4
BF 1.165 1.474 1.209 1.066
TF 3.29 2.15 0.50 3.92
Average Discrepancy (%)
BF 9.07 18.11 9.72 3.86
a
TF and BF stand for “top flange” and “bottom flange”, respectively.
b
MMeck is obtained as a result of the use of Meck Plot method on lateral displacement and angle of twist.
As it is seen in Table 6, despite some scatter in the results, the agreement between the extrapolated
and the ultimate failure moments is generally satisfactory in both cases. Finally, in this case study, the
lowest average discrepancies for the cases of angles of twist of top and bottom flanges are found in the
cases of Modified and Meck Plots, respectively.
4 CONCLUSION
Based on the key findings regarding the proportionality between lateral displacement and other
considered deformation variables including web transverse and longitudinal strains, vertical deflection,
and angles of twist of top and bottom flanges of the Ibeams undergoing elastic lateraldistortional
buckling, the applicability of the Southwell, Massey, Modified, and Meck extrapolation techniques on the
various deformation variables was investigated in this paper and generally satisfactory and reliable results
were obtained. The results of this study may be considered as an indication of a great extension in the
application of the extrapolation techniques.
REFERENCES
[1] Southwell, R.V. “On the analysis of experimental observations in the problems of elastic stability”.
Proc. of the Royal Philosophical Society of London, 135(A), 601, 1932.
[2] Massey, C. “Elastic and inelastic lateral instability of Ibeams”. The Engineer, 216, 672674, 1963.
[3] Trahair, N.S. “Deformations of geometrically imperfect beams”. Proc. of ASCE, Journal of the
Structural Division, 95(ST7), 14751496, 1969.
[4] Meck, H.R. “Experimental evaluation of lateral buckling loads”. Proc. of ASCE, Journal of the
Engineering Mechanics Division, 103, 331337, 1977.
[5] Mandal, P. and Calladine, C.R. “Lateraltorsional buckling of beams and the Southwell Plot”.
International Journal of Mechanical Sciences, 44, 25572571, 2002.
[6] Zirakian, T. “Lateraldistortional buckling of Ibeams and the extrapolation techniques”. Journal of
Constructional Steel Research, 64(1), 111, 2008.
[7] ABAQUS analysis user’s manual, Version 6.5. ABAQUS, Inc., 2005.
634
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. It is clear from prior research studies that the web distortional flexibility can lead to a
substantial reduction relative to the beam theory lateraltorsional buckling resistance for Isections with
stocky flanges and slender webs. Hence, the 2005 AISC Specification gives specific rules for controlling
the unconservative errors due to the neglect of web distortion effects. The accuracy of the 2005 AISC
code predictions in case of elastic lateraldistortional buckling of singly symmetric Ibeams is
investigated in this paper through comparison with the accurate finite strip analysis distortional buckling
solutions as well as the theoretical predictions of two elastic distortional buckling design equations
proposed by other researchers. The code predictions in case of lateraldistortional buckling of slender
web singly symmetric Ibeams are found to be by and large conservative, and even overconservative in
some cases.
1 INTRODUCTION
For the slenderweb Isections, the 2005 AISC Specification [1] bases the lateraltorsional buckling
resistance on Eq. (1), but the St. Venant torsional constant J is taken equal to zero.
CbS 2 E J (1)
Fcr 1 0.078 Lb rt 2
Lb rt 2 S x ho
In fact, the implicit use of J = 0 in Section F5 of the 2005 AISC Specification is intended to account
for the influence of web distortional flexibility on the lateraltorsional buckling resistance for slenderweb
Isection members [2].
In addition to the destabilizing effect of web distortion in a slenderweb singly symmetric Ibeam
which results in lowering of the torsional rigidity of the beam, this may be coupled with the influence of
the Wagner effect to reduce significantly the buckling strength of the singly symmetric beam [3].
This paper focuses on distortional buckling of singly symmetric Ishaped flexural members with
slender webs, and evaluates the effectiveness of the 2005 AISC code rules by comparing the code
predictions with finite strip analysis (FSA) distortional buckling solutions developed using the finite strip
analysis software CUFSM [4] as well as the theoretical predictions of Bradford’s (Eq. (2)) [5] and Wang
et al.’s (Eq. (3)) [6] proposed distortional buckling design equations in the elastic range of structural
response.
635
Tadeh Zirakian and Jian Zhang
J f<>)D 101 D @ (3)
1 d 1.0
Jo D[) 2
2 CONSIDERED IBEAMS
All of the Ibeams in this study have compact flanges and slender webs in accordance with the
compactflange and noncompactweb limits specified in the AISC Specification [1]. The crosssectional
dimensions, lengths, and yield strengths of the Ibeams considered for each case study are summarized in
Table 1.
636
Tadeh Zirakian and Jian Zhang
As it is seen in the table, the AISC code predictions seem to be remarkably conservative relative to
the FSA solutions particularly in sections as the section monosymmetry increases.
It is generally accepted that the distortional effects are smaller in longer beams. This fact is clearly
demonstrated by the FSA as well as the theoretical predictions of other two design equations, as shown in
637
Tadeh Zirakian and Jian Zhang
Table 3. However, the AISC code predictions demonstrate a distinct trend by providing reductions
increasing from 19% to 53% for sections, and 11% to 41% for sections, as the beam length increases.
It is quite obvious that the 2005 AISC code [1] equations provide remarkably conservative results relative
to the FSA and the other considered theoretical predictions especially in longer beams.
As it is seen in Table 4, the difference between the results of the distortional and lateraltorsional
solutions in both monosymmetry cases tends to increase as the web becomes more slender. However, the
code reductions in case of singly symmetric beams with smaller compression flange are relatively larger
than those of singly symmetric beams with larger compression flange. It is notable that both sets of
reductions are comparatively larger than the respective reductions of the FSA as well as the other
theoretical solutions.
638
Tadeh Zirakian and Jian Zhang
From the table, it is found that the predictions of the AISC code equation are by and large below the
predictions of the FSA as well as the two proposed design equations, and the amount of conservatism of
the code predictions seems to be relatively high in sections with smaller compression flange.
639
Tadeh Zirakian and Jian Zhang
From the table, it is evident that the predictions of the FSA as well as the other proposed equations
are not affected by the variation of the yield strength in both and cases, while the reductions induced
by the AISC code equations are found to increase slightly in case of sections, and also initially decrease
and then increase gradually in case of sections, with the increasing of the yield strength. In any case,
the AISC code equations seem to yield conservative predictions relative to the FSA results, and the
conservatism in case of sections with smaller compression flange is considerably high.
8 CONCLUSION
The evaluation of effectiveness of the 2005 AISC code design rules in case of distortional buckling of
singly symmetric Ibeams demonstrates that the 2005 AISC code equations generally provide
conservative strength estimates for elastic distortional buckling. Even the amount of this conservatism is
found to be relatively high in case of singly symmetric Ibeams with smaller compression flange. This
indicates that the assumption of J = 0, which is used in Section F5 of the 2005 AISC Specification with
the aim of controlling the unconservative errors due to the neglect of web distortion effects, may not be
an appropriate approach to the problem, since it may impose economic burden in some cases.
REFERENCES
[1] American Institute of Steel Construction. Specification for structural steel buildings, Chicago, IL,
AISC, 2005.
[2] American Institute of Steel Construction. Commentary on the specification for structural steel
buildings, Chicago, IL, AISC, 2005.
[3] Bradford, M.A. and Waters, S.W. “Distortional instability of fabricated monosymmetric Ibeams”.
Computers & Structures, 29(4), 715724, 1988.
[4] Schafer, B.W. CUFSM 3.12, Elastic buckling analysis of thinwalled members by finite strip
analysis, 2006. (http://www.ce.jhu.edu/bschafer/cufsm)
[5] Bradford, M.A. “Distortional buckling of monosymmetric Ibeams”. Journal of Constructional
Steel Research, 5(2), 123136, 1985.
[6] Wang, C.M., Chin, C.K., and Kitipornchai, S. “Parametric study on distortional buckling of
monosymmetric beamcolumns”. Journal of Constructional Steel Research, 18(2), 89110, 1991.
[7] Galambos, T.V., Guide to Stability Design Criteria for Metal Structures, 5th Ed., John Wiley &
Sons, New York, 1998.
640
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. The 2010 Aluminum Association Specification for Aluminum Structures has been significantly
revised to include more transparent stability provisions. Secondorder effects, including P and P
moments, and factors known to accentuate these effects, such as geometric imperfections and member
inelasticity, will need to be considered in determining required strengths. This paper provides an
overview of these provisions and describes experimental and analytical studies that investigated their
effectiveness.
1 INTRODUCTION
Widely used in the US since its first publication in 1967, the Aluminum Association’s (AA)
Specification for Aluminum Structures [1] has always addressed the stability of individual structural
members. With regard to beams and columns, the Specification provides equations for determining the
strength of beams and columns that account for local buckling of elements such as flanges or webs, and
flexural, flexuraltorsional, and lateraltorsional buckling of members. Prior to the 2010 Specification, a
momentamplification factor was used to address the P effect, which is the effect of axial load acting on
the deflected shape of a member between its ends, on the stability of beamcolumns.
Although it addressed the stability of individual members, earlier editions of the Specification have
not directly considered the stability of structural systems as a whole. The Specification has never
required engineers to design for the P effect, which is the effect of loads acting on the displaced
location of joints in a structure, and only in more recent editions of the Specification was system response
included through the use of the effective length concept. As a result, the strength of a structural system
designed by previous editions of the Specification can be significantly less than the strength of its weakest
member.
With some collapses of aluminum structures attributed to system instability, the AA decided to
provide more comprehensive and transparent stability provisions in the 2010 edition of the Specification.
Recognizing that accurately determining the effective length of members is complicated by the wide
variety of nonorthogonal structural geometries used in aluminum structures, the AA has abandoned the
use of effective length. In an effort to be more consistent with other US design specifications, the AA
adopted stability provisions similar to those that appear in the 2010 American Institute of Steel
Construction’s (AISC) Specification for Structural Steel Buildings [2]. Because of differences in (1) the
stiffness and strength of steel and aluminum, in particular that the E/y ratio for steel is approximately
twice that of aluminum, and (2) the manufacturing processes of aluminum profiles and hotrolled steel
sections, a study that includes experimental and analytical components was conducted to confirm the
adequacy of adopting the AISC provisions. A summary of this study is presented below.
641
Ronald D. Ziemian and J. Randolph Kissell
Pr
= 1.0 for 0.5
Py
(1)
P P Pr
= 4 r 1 r for 0.5 < 1.0
Py Py Py
in which Pr is the required axial compressive strength (i.e., axial force in member) and Py is the axial
yield or squash load (i.e., Py = Agy).
Because of uncertainty in the stiffness and strength of the structural system, there exists the potential
for larger deflections (due to less than expected lateral stiffness) and correspondingly increased second
order moments. To account for this, a factor of 0.8 must be applied to all axial, flexural, and shear
stiffnesses in the structure. One simple method for achieving this is to employ 80% of the actual modulus
of elasticity in the analysis model.
In addition to the above analysis requirements, the Specification requires that all gravity loads must
be included in the analysis of lateral loadresisting systems. For example, additional lateral load may
result from columns that support only gravity loads but attain their sidesway stability by leaning on the
structure’s lateral system.
642
Ronald D. Ziemian and J. Randolph Kissell
with P/Py < 0.5), the same 0.8 factor can be employed because it is approximately equal to the product of
the AISC resistance factor of 0.9 and the 0.877 reduction factor used in the AISC column curve to
account for member outofstraightness.
In contrast, many aluminum sections are typically extruded and upon cooling are then pulled to
straighten or remove any initial outofstraightness. This stretching process requires axially straining the
material beyond yield and as a result typically relieves the residual stresses that may have developed from
an uneven cooling process. On the other hand, some aluminum sections are fabricated by welding several
profiles together; a fabrication process that may result in significant residual stresses.
Differences in the stressstrain relationships for each material may also be a factor in determining the
appropriateness of adopting the AISC provisions. Hotrolled steels typically have a fairly linear
constitutive relationship with a pronounced yield point. In contrast, the stressstain relationships for most
aluminum alloys are inherently nonlinear and without pronounced yield points [5]; as a result, “yield” for
most aluminum alloys is defined by the stress at a 0.2% strain offset. Hence, the above reasons (e.g.,
absence of residual stresses) for not employing the parabolic form of equation 1 may be offset by the need
to model a nonlinear material.
4 EXPERIMENTAL STUDY
As an initial study, a series of stubcolumn tests were performed according to the Structural Stability
Research Council’s Technical Memorandum No. 3: StubColumn Test Procedure [6]. Using the
experimental setup shown in figure 1(a), three Ishape sections and three hollowrectangular shapes were
compression tested. The dimensions for the Ishape are shown in figure 2 and the rectangular shape was
203.2 101.6 mm with a wall thickness of 12.7 mm. In all tests, the stubcolumns were 6061T6
aluminum of length 0.61 m. In each test, the compression force and longitudinal deformation at two
locations (midflange) over a gauge length of 254 mm were recorded. The force was converted to an
axial stress and the deformations were averaged and then converted to axial strain. Stressstrain plots
were then prepared for each of the specimens for the two shapes. For each shape, the stressstrain data
was averaged to produce a single curve.
An untested fourth specimen for each shape was sectioned and tensile coupons were machined from
material at two locations, including the flange tip and the flangeweb intersection. Tensile tests were
performed on these coupons according to ASTM B557 [7]. The two resulting stressstrain curves for the
tension tests of each shape were averaged using the same procedure as that employed in the above
compression tests.
(a) (b)
Figure 1: Experimental setup and results of stubcolumn axial compressive tests.
643
Ronald D. Ziemian and J. Randolph Kissell
For each shape, the resulting tensile and compressive stressstrain curves were used to determine a
stiffness reduction relationship equivalent to that provided in equation 1. This was done by first
determining the stress 0.2% corresponding to a 0.2% offset strain and the modulus of elasticity Eo at low
stress values (taken as the slope of line “bestfit” to stressstrain curve between 0.150.2% and 0.650.2%).
The stress normalized by 0.2% and the corresponding tangent modulus E (slope of stressstrain curve)
normalized by Eo are then plotted as shown in figure 1(b). Although this figure only includes the
resulting tension and stubcolumn curves for the Ishape, similar results were obtained for the hollow
rectangular section.
A review of the three curves plotted in figure 1(b) indicate:
1. The stiffness relationship for tension confirms that the material has a significantly nonlinear
response. A material with a linear stressstrain relationship and a pronounced yield point
would closely match the grey bilinear curve shown in the upper right of the figure.
2. In comparison to the tension response, the compression relationship indicates that the
material stiffness degrades noticeably faster with increased levels of axial compression.
Unfortunately, it is not clear how much of this difference can be attributed to the loading
direction on material response versus that which may be attributed to the existence of
residual stresses.
3. The stiffness reduction factor defined by equation 1 is conservative (but in the authors’
opinion not necessarily overlyconservative) in estimating the loss of axial stiffness due to
increased levels of compressive force.
5 COMPUTATIONAL STUDY
Using one of the frames appearing in the original AISC calibration studies mentioned above, a
computational study was performed to investigate the impact of adopting the AISC  and 0.8 factors
within the AA provisions. The symmetrical portal frame used in this study is shown in figure 2. Two
ratios of beamtocolumn stiffness were considered, one of which included assuming rigid beams with
(EI/L)c/(EI/L)b = 0, and the other with moderately flexible beams (EI/L)c/(EI/L)b = 3. Using the cross
section geometry of the bisymmetrical Ishape used in the above experimental study, both major and
minoraxis bending behavior of the columns was investigated. In all cases, members were assumed fully
braced outofplane.
For each of the four models investigated, two nonlinear finite element programs were employed to
determine system strengths and obtain interaction curves for a wide range of resulting combinations of
axial force and bending in the columns.
A more refined and commercially available finite element program ADINA [8] was employed to
obtain a theoretically “exact” solution. Threedimensional models of the Ishape were created using fully
integrated, 4node shell elements (MITC4). The cross section was modeled with a mesh density of 10
elements across the flange width and 10 elements through web depth. The number of elements along the
644
Ronald D. Ziemian and J. Randolph Kissell
length of the member was varied to maintain an element aspect ratio of approximately one. A typical
model included approximately 6000 shell elements. All models considered both geometric (large
rotation/small strain) and material (multilinear plasticity) nonlinear effects. A nonlinear stressstrain
response (figure 4) for aluminum was explicitly incorporated. Initial imperfections, including member
outofstraightness and frame outofplumb, were included by distorting the original finite element mesh.
Because the focus of the study was to determine the impact of partially yielded compressive members on
system stability, beam elements were always modeled as elastic and column elements were permitted to
yield by using the constitutive relationship shown in figure 3.
Each ADINA analysis was performed until a strength limit state was detected. Such limit states were
typically defined by an instability resulting from a combination of yielding in the columns and second
order effects.
The frame analysis software MASTAN2 [9] was employed to obtain results that meet the above AA
Specification’s stability requirements. MASTAN2 models secondorder effects through the use of
element geometric stiffness matrices and an incremental solution scheme based on an updated Lagrangian
formulation [10]. Equation 1 is directly included in the analysis, which results in the flexural stiffness
being reduced according to the axial force in each element during every load increment.
Strength limit states for the MASTAN2 analyses were defined by the combination of axial force and
bending moment in the columns that just satisfied the AA Specification’s interaction equation:
Pr M r
+ 1.0 (2)
Pc M c
where, Pr and Mr are the axial force and bending moment from the MASTAN2 analysis, Pc the design
compressive strength determined in accordance with the AA Specification’s column curve with kL = L,
and Mc the design strength determined in accordance with the AA Specification’s requirements for
flexure (which for this fully braced compact section column, Mr = bSy with b = 0.9 and S is the elastic
section modulus). Frame outofplumbness of H/500 was included in these analyses but member outof
straightness was not. The latter is assumed to be included in the AA Specification’s equation for column
strength Pc.
6 RESULTS
Using the same validation approach employed in the AISC studies, the AA Specification stability
requirements can be assessed by comparing PM interaction plots of the limiting strengths from the AA
MASTAN2 approach to the “exact” strength determined from sophisticated geometric and material
nonlinear ADINA analyses.
645
Ronald D. Ziemian and J. Randolph Kissell
Results for majoraxis and minoraxis bending cases are contained in Figures 4 and 5, respectively.
Two plots are presented in each figure, including one for rigid beams with (EI/L)c/(EI/L)b = 0 and one for
a moderate degree of flexibility in the beams with (EI/L)c/(EI/L)b = 3. Two sets of AAMASTAN2 and
ADINA curves are provided in each plot.
The first set allows for a comparison of the ratios of the firstorder moment in the column to its plastic
moment (WLc/Mp, with W and Lc defined in figure 2, and Mp = Zy where Z is the plastic section
modulus). The second set can be used to compare ratios of the total moment (including first and second
order effects) in the column to its plastic moment (Mc/Mp). Each point on the curves represents the results
of an analysis for specific combination of gravity load Q and lateral load W. In total, just under 50
separate ADINA and MASTAN2 analyses were performed in this study.
Based on Figures 4 and 5, several observations can be made:
1. For each analysis type (AAMASTAN2 and ADINA), a comparison of the firstorder moment
ratio WLc/Mp to the total moment ratio Mc/Mp at various values of P/Py indicates that second
order effects are significant for this example. At loads as small as P/Py = 0.1, the secondorder
moments are on the order of 10 to 15 percent larger than the firstorder moments. As expected,
this moment amplification increases significantly for larger values of P/Py. The reason the
ADINA secondorder effects are larger is because this analysis includes material nonlinear
behavior, which tends to reduce lateral stiffness and increase deflections.
2. By comparing the AAMASTAN2 and ADINA total moment ratios Mc/Mp at various values of
P/Py, it is clear that the “exact” bending moment capacity of the column in the presence of any
amount of axial force always exceeds the moment capacity defined by the AA Specification’s
646
Ronald D. Ziemian and J. Randolph Kissell
beamcolumn interaction equation (see equation 2). A significant factor contributing to this is
that the AA Specification moment capacity of the member bSy is less than the actual (ADINA)
upper limit capacity of byZ. Another cause for this is that the AA Specification uses a single
linear interaction equation to represent the strength of beamcolumns; in most other
specifications (e.g., AISC) a bilinear curve is used, which permits larger strengths at low to
intermediate values of axial force, ranging from approximately P/Py = 0.1 to P/Py = 0.5.
3. The WLc/Mp curves also provide a direct indication of the ultimate strength of the frame
predicted by the AAMASTAN2 and ADINA approaches. For example, the coordinate pair
(WLc/Mp, P/Py) = (0.2, 0.4) represents failure at gravity and lateral load combination of Q = 0.4Py
and W = 0.2Mp/Lc. In all majoraxis bending cases, the strength predicted by the AAMASTAN2
approach is less than the “exact” strength predicted by ADINA. This conservatism is repeated
for all minoraxis bending conditions with the exception of the high axial load case (P/Py > 0.5)
in the frame with a moderately flexible beam of (EI/L)c/(EI/L)b = 3. The overpredicted AA
MASTAN2 strength, however, is quite small (see lower plot in figure 4). For a columntobeam
stiffness of (EI/L)c/(EI/L)b = 3, a design method based on effective length would use an effective
length factor of approximately k = 2.5, where as the AA stability provisions permit the use of k =
1.0.
4. The largest P/Py values observed in each AAMASTAN2 case are between 0.6 and 0.7.
Substituting these values into equation 1 results in relatively inconsequential factors of 0.96
and 0.84, respectively. Given that fairly stocky columns (L/r = 20 with r = I A ) were
investigated in this study, it should be noted that larger slenderness L/r values more common to
647
Ronald D. Ziemian and J. Randolph Kissell
design would result in smaller column strengths (i.e. lower P/Py values) and hence, even larger
(closer to 1.0) and less consequential factors.
7 SUMMARY/CONCLUSION
The new stability provisions that appear in the 2010 Aluminum Association’s Specification for
Aluminum Structures have been investigated by a study that contains experimental and analytical
components. Experimental stubcolumn tests were performed on I and hollowrectangular shapes in an
effort to better understand the reduction in axial stiffness in the presence of compressive force.
Computational analyses were then performed on a portal frame that is similar to one used in part to
validate the AISC stability provisions. In this work, various combinations of lateral and gravity loads
were examined as well as varying degrees of frame flexibility.
Insight from this study indicates that use of the AA stability provisions in conjunction with the AA
single linear interaction equation for designing beamcolumns provides moderate to fairly conservative
results. The AA use of the same stiffness reduction factors and 0.8 that appear in the AISC
Specification does not appear to be unreasonable, although it is unclear if the factor is necessary.
The study presents several cases where the AA stability provisions are adequate for allowing the
routine use of an effective length factor of k = 1, even in cases where an effective length design method
requires using two to three times that value. Just as importantly, the research further justifies the need to
consider secondorder effects in the design of aluminum structures.
It is recommended that additional studies be made to determine if the AA could avoid the use of a 
factor in future editions to their specification. Such studies should also explore cases that include builtup
sections, where the effects of welding may result in substantial residual stresses and may justify using the
factor.
8 ACKNOWLEDGEMENT
The authors thank the Aluminum Association for their support of this research under grant number
547.
REFERENCES
[1] Aluminum Association, Specification for Aluminum Structures, Arlington, VA, 2010.
[2] American Institute of Steel Construction, Specification for Structural Steel Buildings, Chicago, IL,
2010.
[3] SurovekMaleck, A., White, D.W. and Ziemian, R.D., Validation of the Direct Analysis Method,
Structural Engineering, Mechanics and Materials Report No. 35, School of Civil and Environmental
Engineering, Georgia Institute of Technology, Atlanta, GA, 2003.
[4] Bleich, F., Buckling Strength of Metal Structures, McGrawHill, New York, 1952.
[5] Kissell, J.R. and Ferry, R.L., Aluminum Structures: A Guide to Their Specifications and Design,
Wiley, Hoboken, NJ, 2002.
[6] Ziemian, R.D. (Ed.), Guide to Stability Design Criteria for Metal Structures, 6th edition, Wiley,
Hoboken, NJ, 2010.
[7] ASTM Standard B557, Standard Test Methods for Tension Testing Wrought and Cast Aluminum
and MagnesiumAlloy Products, ASTM International, West Conshohocken, PA, 2006.
[8] ADINA, Theory Manual, ADINA Research and Development, Inc., Watertown, MA, 2009.
[9] MASTAN2, developed by R.D. Ziemian and W. McGuire, version 3.2, www.mastan2.com, 2009.
[10] McGuire, W., Gallagher, R.H., and Ziemian, R.D., Matrix Structural Analysis, Wiley, Hoboken,
NJ, 2000.
648
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Nonlinear analysis, large span frame, hunched rafter, tapered column.
Abstract. The article describes the applied technological solutions to transform an existing ice rink into
an indoor arena in the city of Târgu Mureú, Romania. The new indoor arena will have a capacity of 1800
fixed seats. Using a large free span (50 m) structure will overcome the in situ technological constraints
due to the position of the existing building. There is limited access due to the fact that the ice rink is
situated between two buildings and the river Mureú which flows alongside the third side. Also, the
existing refrigeration system makes access impossible within the ice pad structure area. Taking into
account the above mentioned restrictions, the article describes the applied structural solutions which will
make the structural steel work erection possible. The structural solution using steel will ensure fast and
easy erection of the structural steel framework without causing damage to any of the existing buildings
and installations. The paper summarizes the results of the numerical study performed by the authors on
the frame structure. The frames were designed to withstand horizontal and vertical loads and also to
satisfy the ULS and SLS criteria. The frames have fixed base connections, tapered columns, hunched and
kingpost truss rafters and a pitch roof angle of 30.
1 INTRODUCTION
649
Nagy Zsolt et al.
1.2 Constrains
Due to the destination and the particular position of the building  limited access due to the fact that
the ice rink is situated between two buildings and the river Mureú flows alongside the third side, there
were the following constrains:
Access for erection only from one side, without the access on ice pad structure due to the
existing refrigeration system;
To keep the existing buildings;
To control the designed assemblies self weight, in order to facilitate the erection.
Also the pressure of a short deadline acted as constraint. The site conditions and the proposed
architecture should be seen in Figure 1.
Figure 1. The site conditions and the architecture of the building (view from access side)
650
Nagy Zsolt et al.
651
Nagy Zsolt et al.
Load combination for ultimate limit state (ULS) and serviceability limit state (SLS) according
to CR02005 (EN 1990).
3.2 Design of the main structurelinear elastic analysis (LEA)
The design of the steel structure was performed following the Romanian code STAS 10108/078 [6].
For strength, stability and stiffness requirements of the structural elements the prescription of SR
EN199311[4], SREN199318[5] and P100/2006 [3] were also used.
In the case of large spanned structures, the vertical deflection under gravitational loads represents one
of the major constraints in the design process. In order to keep under control the deformations of the
frames, fixed base connections, tapered columns and hunched kingpost truss rafter solution were chosen
[6]. The rafters were extended on both sides over the annexes, increasing both the vertical and horizontal
stiffness of the frame. A suitable horizontal and vertical bracing system were provided in order to control
structural flexibility, eigen values and deflections of the main structure. Fly braces were provided at the
inner flange of the rafter in order to improve the flexuraltorsional buckling resistance of these elements.
Having class 3 section of the structural elements, linear elastic structural analysis was performed,
using a seismic behavior factor of q=1 according to P1002006 [3]. Even with q=1, the combinations of
actions for seismic design situations were not the dominant load combinations. The design checks of the
structural elements for ULS include persistent or transient design situations (fundamental combinations)
where snow loads play the key role.
For SLS design checks of the structural elements fundamental and exceptional load combinations
were used. Performing a dynamic 3D analysis of the structure, with the structural masses concentrated on
joints, first longitudinal eigen period of Tlong=0.588 sec and first transversal eigen period of Ttransv=0.448
sec were obtained (see Figure 4).
First longitudinal vibration mode T=0.588 sec First transversal vibration mode T=0.448 sec
Figure 4. Eigen vibration modes and periods
The maximum transversal and longitudinal sway displacement for SLS check under seismic loads
according to P1002006 are:
d rSLS 0,005 h
d rSLS
,x 0.014 d ,a
112.5 mm (1)
Q q 0.4 1.0
d rSLS 0,005 h
d rSLS
,y 0.047 d ,a
137.5 mm (2)
Q q 0.4 1.0
The maximum vertical deflection of the rafter for SLS check under snow load is:
L
f 161.4mm d f a 166.7 mm (3)
300
652
Nagy Zsolt et al.
In order to have an overview about the real behavior of the structure, a finite element linear elastic
analysis (FEM) of the transverse frame has been performed with Ansys computer program. The elements
of the frame were modeled using shell finite elements (Shell 43 see Figure 5). The forces on the rafter
were applied as point loads (points where purlins are fixed on the frame). The connections between
structural elements raftertocolumn, beamtocolumn, raftertorafter, column base connections were
considered fully rigid. The results of the detailed linearelastic analysis (LEA) confirmed the previously
evaluated ULS and SLS results. The recorded vertical displacement in case of FEM linear elastic analysis
was 152 mm (instead of 203 mm –linear elastic analysis). It might be emphasized that the resulted
structure is more rigid in case of FEM analysis, explained by the shift of the neutral axis along the
elements with variable cross sections (i.e. tapered column, and hunched rafter). Figure 6 shows the stress
distribution along the transverse frame, where we can observe the maximum stress concentration around
the joint of the king post rafter and the hunched frame rafter.
Figure 6. Stress distribution along the transverse frame under gravity load combinationslinear elastic
analysis
653
Nagy Zsolt et al.
The resulted maximum stress does not exceed 252 N/mm2. There were many concerns about the
stress distribution in the connection of the king post truss with the rafter. As it can be observed from
Figure 6 the distribution of the stresses does not exceed the maximum allowable yield limit.
3.3 Design of the main structure non linear elasticplastic analysis
In Figure 8, it is illustrated the way in which the initial bow (out of plane) imperfection is considered
in the nonlinearelastic analysis (GMNIA). Three types of lateral restraints of the rafter were considered
separately in the analyses (see Figure 7 [7],[8]). Types 2 simulate the purlin/sheeting effect, when the
purlin can be connected with one or two bolts, respectively. Type 3 is the same with type 2 with an
additional fly brace. Type 1, the reference case, actually means no lateral restrains introduced by purlins
and side rails.
Figure 9. Stress distribution along the transverse frame under gravity load combinationsnon linear
elasticplastic analysis
To simplify the computational model, in the analysis the lateral restraints were considered axially
rigid. The values of the applied imperfection is 167 mm (50 m span frame) for initial bow imperfection
(er), l/150 corresponding to curve c, for plastic analysis, according to clause 5.3.2 (3)a) of EN19931.1
[1]. The material behavior was introduced by a bilinear elasticperfectly plastic model, with a yielding
limit of 355 N/mm2. In Figure 10 are illustrated the capacity curves for different type of analysis. As it
654
Nagy Zsolt et al.
was expected the lateral restrains of the rafter played an important role in the total capacity of the main
frame. Also it must be emphasized that there is more than 25% structural capacity reserve.
4 CONCLUSION
The paper illustrates the successful application of the steel structure for a large span using a simple
portal frame shape, combined with a king post truss rafter. A wide range of design parameters are briefly
summarized. The paper emphasizes the whole design process, assisted by FE analysis  in order to
perform supplementary stability checks of the framed structure. Due to the unusual shape of the
transverse frame, there were many concerns about its real behavior under gravitational load, the most
important ones in this particular case. For this purpose a linear elastic analysis (LEA) followed by a
nonlinear elasticplastic analysis (GMNIA) were performed in order to determine the real behavior of the
frames. From structural point of view a good agreement between 3D structural analysis and LEAFEM
has been found. GMNIA analysis confirm at least 25% overstrenght of the structure by applying the
chosen structural solution and lateral restraints of the main rafter.
Even with behavior factor q=1, the combinations of actions for earthquake design situations were not
the dominant load combinations. In the design checks of the structural elements, gravity loads played the
key role.
Figure 11. The whole structural model and the actual stage of the building
655
Nagy Zsolt et al.
REFERENCES
[1] EN 199311 Eurocode 3: Design of steel structures Part 1.1: General rules and rules for buildings;
[2] CR02005: Cod de proiectare pentru bazele proiectarii structurilor in constructii (Design Code.
Basis of design. Romanian design code).
[3] P1002006: Cod de proiectare seismica P100. Partea IPrevederi de proiectare pentru cladiri
(EN19981).
[4] SREN 199311: Eurocod 3: Proiectarea structurilor de oĠel Partea 11: Reguli generale úi reguli
pentru clădiri
[5] SREN199318:Eurocod 3: Proiectarea structurilor de oĠel. Partea 18: Proiectarea îmbinărilor
[6] H. C. Schulitz, W. Sobek, K. J. Habermann  Steel Construction Manual, Birkhauser Verlag 2000,
ISBN no. 3764361816
[7] D. Dubina, I. M. Cristutiu, V. Ungureanu, Zs. Nagy: Stability and ductility performances of
light steel industrial building portal frames, 3rd European Conference of Steel Structures,
Eurosteel 2002, CoimbraPortugal, sept. 2002, pp 635643 (2002);
[8] D. Dubina, I. M. Cristutiu: Buckling strength of pitchedroof portal frames of Class 3 and
Class 4 tapered sections, International Conference on Steel and Composite Structures  Eurosteel
2005, MaastricthHolland, 711 june 2005, pp 635643 (2005);
656
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Buckling, Elliptical Hollow Sections, Oval Hollow Sections, Postbuckling, Steel Structures.
Abstract. The local postbuckling behaviour of elliptical hollow section (EHS) tubes under compression
is analysed in this paper. It is found that EHS tubes with low to moderate aspect ratios can support loads
up to their limit loads but are imperfection sensitive (shelltype behaviour), while EHS tubes with
moderate to high aspect ratios can carry loads higher than their limits loads (platetype behaviour) and
are imperfection insensitive. For increasing EHS aspect ratio, it is found that the compressive stresses
accumulate near the zones of minimum radius of curvature while the zones of maximum radius of
curvature experience a relatively low compressive stress level. Thus, it is likely to apply the effective
width concept to EHS tubes with moderate to high aspect ratio.
1 INTRODUCTION
EHS steel tubes are now available as hotrolled structural products [1,2] and represent an interesting
solution for many visible applications in steel construction, particularly for glass facades. These shapes
are included in the new edition of EN 10210 [3] and are available in a standard range of dimensions. In
response to the emergence and commercial availability of EHS tubes, several recent investigations on
their buckling behaviour and strength have been published. Gardner and Chan [4] and Chan and Gardner
[5,6] assessed the nonlinear behaviour of hotrolled EHS tubes by means of experimental and numerical
analyses and proposed structural design rules. They found that the slenderness limits for pure
compression set out in EC3 for circular hollow section (CHS) classification can be safely adopted for
EHS, based on the equivalent diameter of the point of the EHS with maximum radius. Zhu and Wilkinson
[7] also performed shell finite element analyses to evaluate the buckling and postbuckling behaviour of
EHS in compression. Silvestre [8] developed a formulation of Generalised Beam Theory (GBT) to
analyse the elastic buckling behaviour of members with noncircular hollow sections (NCHS) and applied
it to study the behaviour of EHS shells and tubes under compression, particularly the variation of the
critical buckling stress with the member length and crosssection geometry. RuizTeran and Gardner [9]
have also examined the buckling response of EHS tubes in compression and proposed analytical formulae
to accurately predict the critical stress. Thus, the main objective of this paper is to unveil the mechanics
of the elastic local postbuckling behaviour of EHS tubes and to explain in a detailed fashion the
transition between the shelltype (imperfection sensitive) behaviour of EHS tubes with low eccentricity
and the platetype (imperfection insensitive) behaviour of EHS tubes with high eccentricity.
2 PARAMETRIC STUDY
An indepth study on the influence of the EHS aspect ratio on the variation and nature of the post
buckling equilibrium path, ensuing stress distributions and imperfection sensitivity is presented. A
reference EHS stub column with length L = 300 mm, thickness t = 4 mm and fully fixed supports is
657
Nuno Silvestre and Leroy Gardner
considered. The EHS geometry is characterised by the major axis width 2a and minor axis width 2b,
which are considered here as the dimensions of the EHS midline (i.e., 2a + t and 2b + t are the outer
dimensions). Based on a commercially available [1,2] reference geometry (2a = 150 mm, 2b = 75 mm, a/b
= 2.0), five further EHS configurations were generated and studied. These were obtained by (i) keeping
the crosssection perimeter unaltered (P = 363 mm) and (ii) varying the aspect ratio a/b from 1.10 to 5.0.
The six EHS geometries considered in this paper are represented in figure 1 and are characterised by an
equal crosssection area A=1450 mm2 and thus an equal amount of steel in each column.
a/b=1.10 a/b=1.25 a/b=1.50 a/b=2.00 a/b=3.00 a/b=5.00
Figure 2: Critical buckling modes and corresponding critical stresses σcr (in N/mm2)
658
Nuno Silvestre and Leroy Gardner
The material behaviour of the steel tube was assumed to be homogeneous, isotropic and linear elastic, which
is fully characterised by the values of Young’s modulus (E = 210000 N/mm2) and Poisson’s ratio (ν = 0.3).
Initial geometrical imperfections were included in the models in the form of the most relevant (critical) local
buckling mode shapes, incorporated into the tube initial geometry by means the initial conditions ABAQUS
command. Therefore, preliminary buckling (eigenvalue) analyses were carried out for all the six columns in order
to obtain the local buckling stress values and the corresponding buckling mode configurations, which are
depicted in figure 2. For the initial studies, a small imperfection amplitude equal to 0.1 mm (2.5% of EHS
thickness) was adopted, while imperfection sensitivity is investigated further in Section 6 of this paper. No
residual stresses were incorporated into the numerical analyses, since they were deemed to be of very low
magnitude in hotfinished elliptical tubes [5].
5000
4600 N/mm
2 0.6
4000
2
3555 N/mm
0.4
3000 Imperfection amplitude:
1% thickness
2130 2.5% thickness
2000 10% thickness
0.2 25% thickness
1202
50% thickness
1000 100% thickness
E/L
1
0 0.0
0 5 10 15 20 25 30 0.0 2.5 5.0 7.5 10.0 12.5 15.0
Axial shortening u (mm) Axial shortening u (mm)
Figure 3: (a) Applied stress σ vs. axial shortening u for several a/b ratios (imperfection = 2.5% thickness)
and (b) critical stress ratio σ/σcr vs. axial shortening u for a range of imperfections (a/b=2).
659
Nuno Silvestre and Leroy Gardner
σ /σcr
2.0
a/b=5.00
1.8 a/b=3.00
a/b=2.00
1.6 a/b=1.50
a/b=1.25
a/b=1.10
1.4
Limit point of the Bifurcation point of
1.2 imperfect tube the perfect tube
1.0
0.8
0.6
0.4
0.2 1
1
0 ε /εcr
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Figure 4: Relationship between the critical stress (σ/σcr) and strain (ε/εcr) ratio for varying a/b.
From careful observation of figures 3(a) and 4 the following comments can be made:
(i) The six tubes possess an equal crosssection area and, therefore, the slope of the primary path is given
by E/L and is equal for all six tubes; the primary paths are also perfectly linear until the limit stress
σlim is reached. As expected, the limit stress σlim is lower than the corresponding critical stress σcr due
to the influence of the imperfections. The variation of limit stress σlim with aspect ratio is shown in
table 1. In fact, the ratio between the limit stress and the critical stress (σlim/σcr) shows little variation
with a/b and remains around 0.85 for the adopted imperfection amplitude of 2.5% of the EHS
thickness. This suggests that the imperfection sensitivity of EHS tubes is not strongly dependent on
the aspect ratio a/b of the crosssection, within the examined range of 1.10 < a/b < 5.00.
(ii) After reaching the limit stress σlim, the six nonlinear paths show very distinct responses. The
descending branch (decrease of both applied stress and axial shortening) is much more pronounced
for tubes with lower aspect ratio a/b, as shown in the graph of σ/σcr versus ε/εcr given in figure 4.
(iii) Figure 3(a) shows that all the tubes are associated with postcritical curves that possess local minima, which
move horizontally towards the linear primary path as the aspect ratio increases. Beyond the local minima the
postbuckling paths (for a/b ≥ 1.10) exhibit a positive slope (increase in applied stress and axial shortening)
and are stable. It may be observed that the maximum slope of the ascending postbuckling branch increases
with aspect ratio a/b, this increase being more substantial for low to moderate aspect ratios (a/b ≤ 2) and less
pronounced for moderate to high aspect ratio (a/b ≥ 2). Furthermore, the slope of the ascending post
buckling path can reach values up to 40% of the initial slope of the linear primary path; a value similar to that
for flat simplysupported plates with unrestrained edges.
(iv) From points (i) to (iii), it may be concluded that the maximum applied stress σmax that an elastic EHS tube
with low to moderate aspect ratio (a/b ≤ 1.5) can support is its limit stress σlim whereas, the maximum applied
stress σmax that an elastic EHS tube with moderate to high aspect ratio (a/b ≥ 2.0) can carry is higher than its
limit stress σlim (see figure 3(a)). The initially unstable postbuckling response exhibited by all six tubes
investigated (with an imperfection of 2.5% of the section thickness), means that snapthrough behaviour is
experienced at the limit stress. However, figure 3(a) shows that the snapthrough reduces with increasing
aspect ratio a/b. For instance, the very eccentric tube with a/b = 5.0 experiences, at the limit stress level
σlim=1202 N/mm2, a very small snap between u = 1.86 mm and u = 2.22 mm. Conversely, the moderately
eccentric tube with a/b = 2.0 experiences, at the limit stress level σlim=3555 N/mm2, a larger snap between u
= 4.98 mm and u = 12.85 mm.
The deformed configurations of the EHS tubes in the postbuckling regimes are shown in figure 5,
where deformation may be seen to be concentrated towards the midheight of the specimens. Initial
geometrical imperfections were imposed with an inward deformation of the flatter region (i.e. maximum
local radius of curvature) of the EHS at midheight (see figure 2). From figure 5, it may be seen that the
660
Nuno Silvestre and Leroy Gardner
661
Nuno Silvestre and Leroy Gardner
σ (N/mm2) σ (N/mm2)
25000 25000
σ/σcr σ/σcr
0.408 0.511
a/b=1.10 0.858 a/b=1.25 0.875
20000 0.807 20000 0.746
0.659 0.659
0.557 0.567
0.487 0.503
15000 0.436 15000 0.457
0.393 0.415
0.350 0.412
0.371 0.462
10000 10000
5000 5000
0 0
0 90 180 270 360 0 90 180 270 360
θ (°) θ (°)
5000 5000
σ (N/mm2) σ (N/mm2)
25000 25000
σ/σcr σ/σcr
a/b=1.50 0.508 a/b=2.00
0.403
20000 0.831 20000 0.888
0.681 0.670
θ
0.584 0.589
0.508 0.532
15000 0.466 15000 0.514
0.485 0.568
0.542 0.693
0.723 0.859
10000 10000 1.220
5000 5000
0 0
0 90 180 270 360 0 90 180 270 360
θ (°) θ (°)
5000 5000
σ (N/mm ) 2
σ (N/mm2)
25000 25000
a/b=5.00
20000 a/b=3.00 σ/σcr 20000 σ/σcr
0.401 0.400
0.819 0.797
0.612 0.732
15000 0.658 0.862
0.956
15000
1.200
1.300 1.760
1.630 2.510
2.020 3.370
10000 2.250 10000 3.840
5000 5000
0 0
0 90 180 270 360 0 90 180 270 360
θ (°) θ (°)
5000 5000
Figure 6: Evolution of normal stresses σ(θ) with σ/σcr for the six different aspect ratios a/b.
662
Nuno Silvestre and Leroy Gardner
For a/b = 1.10, the stress decreases and changes sign, with the points of minimum EHS radius of
curvature (θ = 0° and 180°) being in tension at the maximum axial shortening (black line). For 1.25 ≤
a/b ≤ 2.0, the stress decreases and changes sign, but then tends to increase again with increasing
displacement. For 3.0 ≤ a/b ≤ 5.0, the stress slightly decreases after the peak but then increases
significantly, and the points θ = 0° and 180° are never in tension. The points of maximum
compressive stress are not at the minimum EHS radius of curvature (θ = 0°, 180°) but are
immediately adjacent to these points at θ = 1030°, 150170°, 190210° and 330350°.
(iii) As noted above, nonlinear distributions of σ(θ) vary markedly with aspect ratio a/b. For low to
moderate aspect ratios (a/b ≤ 1.5), the stress distribution in the zones of minimum EHS radius of
curvature (θ < 60°, 120° < θ < 240°, θ > 300°) is distinctly nonlinear with θ, with compressive and
tensile stresses occurring over different ranges of θ. For increasing aspect ratio (a/b ≥ 1.5), the stress
distribution in the zones of minimum EHS radius of curvature (θ < 60°, 120° < θ < 240°, θ > 300°)
becomes more uniform. This may be seen most clearly for the EHS tube with a/b = 5, where the
compressive stresses are almost uniform inside the ranges θ < 60°, 120° < θ < 240° and θ > 300°, for
the maximum axial shortening (black line).
(iv) For EHS tubes with low aspect ratio a/b, the development of tension stresses in the zones of minimum
EHS radius of curvature lead to a “softening effect” in behaviour of the tube, since the average stress
along the EHS midline reduces as the tensile stresses grow. This is the reason for the almost horizontal
postbuckling branches following the initial drop in load (see figure 3(a)) exhibited by the EHS tubes
with a/b = 1.1 and 1.25; their postbuckling behaviour is similar to that exhibited by circular shells and
they do not possess any postcritical stiffness. Conversely, for EHS tubes with moderate to high
eccentricity (or aspect ratio a/b), the development of high compressive stresses in the zones of minimum
EHS radius of curvature lead to a “hardening effect” in behaviour of the tube, since the average stress
along the EHS midline now tends to increase as the compressive stresses grow. This is the reason for the
ascending and stable postbuckling branches (see figure 3(a)) exhibited by the of EHS tube with a/b =
1.5, 2.0, 3.0 and 5.0. Consequently, their postbuckling behaviour is closer to that exhibited by flat plates
and they do possess notable postcritical stiffness.
(v) On the basis of the above findings, it may be concluded that an approach based on the “effective
width concept”, widely used for the strength analysis of flat plates, may be adapted to the design of
EHS tubes with moderate to high aspect ratios. This procedure is outside of the scope of the present
paper, but is the subject of ongoing research.
6 IMPERFECTION SENSITIVITY
In previous sections, a constant imperfection amplitude of 2.5% of the section thickness has been
adopted. In this section, the imperfection sensitivity of EHS tubes under compression is examined. Each
of the six tubes was analysed for six imperfection amplitudes: ξ = 0.04 mm (1% of the thickness), ξ =
0.10 mm (2.5% of the thickness), ξ = 0.4 mm (10% of the thickness), ξ = 1.0 mm (25% of the thickness),
ξ = 2.0 mm (50% of the thickness) and ξ = 4.0 mm (100% of the thickness). For the tube with a/b = 2,
figure 3(b) shows the equilibrium paths (critical stress ratio σ/σcr versus axial shortening u) obtained for
the several imperfection amplitudes. It is clear that, for imperfection amplitudes lower than 25% of the
thickness, the equilibrium paths possess a limit point, and the limit stress σlim decreases significantly with
increasing imperfection amplitude. Conversely, for imperfection amplitudes higher than 50% of the tube
thickness, (i) the equilibrium paths always ascend and (ii) there is no limit stress. This behavioural aspect
has far reaching implications for the imperfection sensitivity of EHS tubes: due to its stable and
ascending postbuckling branch, a moderately (or highly) eccentric EHS tube may or may not be
imperfection sensitive depending on the range of imperfections being considered. For a given aspect ratio a/b,
there is always a “bound imperfection amplitude” (ξb) that separates the ranges of imperfection amplitude where
the tube is imperfection sensitive (ξ < ξb) and insensitive (ξ > ξb). From figure 3(b), for the EHS tube with a/b =
2, this ξb value should lie between 25% and 50% of the thickness. Using a trialanderror procedure, the exact
value of the “bound imperfection amplitude” was found to be ξb = 1.2mm = 30% of the thickness.
663
Nuno Silvestre and Leroy Gardner
7 CONCLUSIONS
The elastic local postbuckling behaviour of tubes with elliptical hollow sections (EHS) under
compression was analysed in this paper. The obtained numerical results were then presented and
analysed; the following conclusions are drawn:
(i) Nonlinear equilibrium paths  The maximum applied stress that an elastic EHS tube with low to
moderate aspect ratio (a/b ≤ 1.5) can support is its limit stress σlim, while the maximum applied stress
that an elastic EHS tube with moderate to high aspect ratio (a/b ≥ 2.0) can carry is higher than its
limit stress σlim. It was observed that the slope of the ascending branch increases with aspect ratio a/b
and can reach values up to 40% of the initial slope of the linear primary path.
(ii) Normal stress distributions  For increasing aspect ratio a/b, the compressive stresses grow and
accumulate near the zones of minimum radius of curvature while the zones of maximum radius of
curvature possess an approximately uniform and relatively low compressive stress level. Therefore, it
is expected that an approach based on the “effective width concept” widely used for the strength
analysis of flat plates may be adapted to the design of EHS tubes with moderate to high aspect ratios.
(iii) Imperfection sensitivity  For a given aspect ratio a/b, there is a “bound imperfection amplitude” ξb
that separates the ranges of imperfection amplitude where the EHS tube is imperfection sensitive (ξ <
ξb) and insensitive (ξ > ξb). Moreover, it was shown that the imperfection sensitivity of EHS tubes
significantly drops for increasing aspect ratio a/b, ranging between shelltype behaviour (strongly
imperfection sensitive) and platetype behaviour (imperfection insensitive).
REFERENCES
[1] Corus. Celsius 355s Ovals. Internet: http://www.corusgroup.com, 2006.
[2] Interpipe – The Hollow Section Company. Elliptical Hollow Sections to S355 J2H, Internet:
http://www.interpipe.co.uk/, 2007.
[3] CEN – Comité Européen de Normalisation. EN 102102: Hot finished structural hollow sections of
nonalloy and fine grain steels – Part 2: Tolerances, dimensions and sectional properties, 2006.
[4] Gardner, L., Chan, T.M., “Crosssection classification of elliptical hollow sections”, Steel and Composite
Structures, 7(3), 185200, 2007.
[5] Chan, T.M., Gardner, L., “Compressive resistance of hotrolled elliptical hollow sections”, Engineering
Structures, 30(2), 522532, 2008.
[6] Chan, T.M., Gardner, L., “Flexural buckling of elliptical hollow section columns”, Journal of
Structural EngineeringASCE, 135(5), 546557, 2009.
[7] Zhu, Y., Wilkinson, T., “Finite element analysis of structural steel elliptical hollow sections in
compression”, Research Report No R874, Centre for Advanced Structural Engineering, The University of
Sydney, 2007.
[8] Silvestre, N., “Buckling behaviour of elliptical cylindrical shells and tubes under compression”,
International Journal of Solids and Structures, 45(16), 44274447, 2008.
[9] RuizTeran, A.M., Gardner, L., “Elastic buckling of elliptical tubes”, ThinWalled Structures, 46(11),
13041318, 2008.
[10] DS Simulia Inc. ABAQUS Standard (version 6.7), 2007.
664
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
* Department of Civil and Environmental Engineering, Imperial College London, London SW7 2AZ
emails: marios.theofanous@imperial.ac.uk, leroy.gardner@imperial.ac.uk
Keywords: Stainless Steel Structures, Continuous Beams, Plastic Design, Continuous Strength Method.
Abstract. Despite the high material ductility of structural stainless steels and the existence of a Class 1
limit in the European structural stainless steel design code EN 199314 [1], plastic design is not
permitted for stainless steel structures, which leads to uneconomic design. The present paper investigates
the applicability of inelastic design procedures to indeterminate stainless steel structures. Five three
point bending tests and ten twospan continuous beam tests on stainless steel square and rectangular
hollow sections are reported herein. Analysis of the results reveals that current design provisions are
overly conservative and significant moment redistribution and hence material savings can be achieved if
inelastic design procedures are followed at both crosssectional level and system level.
1 INTRODUCTION
The need for metallic structures to resist high loads that have a small probability of occurrence in an
economic way necessitates the exploitation of the inelastic range of the material’s stressstrain curve,
provided that they possess sufficient ductility. Modern structural design guidance specifies the extent to
which the exploitation of the material’s inelastic range is allowed, following the crosssection
classification procedure. The European structural design codes for stainless steel EN 199314 [1] and
carbon steel EN 199311 [2] specify four behavioural classes of crosssections according to their
susceptibility to local buckling. Indeterminate structures employing carbon steel crosssections classified
as Class 1 may be plastically designed. Despite the high material ductility of structural stainless steels [3]
and the existence of a Class 1 limit in [1], plastic design is not permitted for stainless steel structures,
which leads to uneconomic design.
In this paper the applicability of inelastic design procedures to stainless steel indeterminate structures
is investigated. Five threepoint bending tests and ten twospan continuous beam tests on stainless steel
SHS and RHS are reported. The experimental response of both the simply supported beams and the
continuous beams is then compared with the predictions of EN 199314 [1]. Analysis of the results
reveals that current design provisions are overly conservative, since they do not account for material
strainhardening and the significant moment redistribution (in the case of the continuous beams) taking
place before collapse occurs. Hence material savings can be achieved if inelastic design procedures are
followed at both crosssection level and system level. To this end, the continuous strength method
(CSM), outlined in [4][6], which allows for the actual material response at crosssectional level, is
adapted to stainless steel indeterminate structures, resulting in more favourable strength predictions.
2 EXPERIMENTAL INVESTIGATION
An experimental investigation into the structural response of stainless steel simple and continuous
beams has been carried out in the Structures Laboratory at Imperial College London. The employed
crosssections were SHS and RHS in grade EN 1.4301/1.4307 stainless steel with nominal sizes of
50×50×3, 60×60×3, 100×100×3 and 60×40×3. The specimens were extracted from the same lengths as
665
Marios Theofanous and Leroy Gardner
the ones utilised in the experimental study reported in [7]. The tensile coupon test data reported in [7] are
utilised herein, as no further material coupon tests were conducted. The obtained tensile flat material
properties are shown in Table 1, where E0 is the Young’s modulus, V0.2 and V1.0 are the proof stresses at
0.2% and 1% offset strains, respectively, and n and n'0.2,1.0 are strain hardening exponents, utilised in the
two stage RambergOsgood model [8][10]. The 0.2% proof stress ı0.2, obtained from tensile flat
coupons, is utilized to obtain the elastic and plastic moment resistances (Mel and Mpl respectively).
Five threepoint bending tests were initially performed, to provide fundamental flexural performance
data, which were utilised to assess the suitability of current design provisions codified in EN 199314
(2006). Subsequently ten twospan continuous beam tests (fivepoint bending) were conducted, which
enabled the study of stainless steel indeterminate structures and an assessment of the current codified
provisions. Performing both simply supported and continuous beam tests on the same crosssections
enables the assessment of the effect of moment redistribution on ultimate capacity.
2.1 Simply supported beam tests
Five simply supported beam tests have been conducted in the threepoint bending configuration. One
test was conducted for each of the three SHS employed, whilst two tests were conducted for the RHS
60×40×3 specimen, one about the major axis and one about the minor axis. All beams had a total length
of 1200 mm and were simply supported between rollers, which allowed axial displacement of the beams’
ends. The rollers were placed 50 mm inward from each beam end. Wooden blocks were placed within the
tubes at the loading point to prevent web crippling. The applied crosshead movement rate was 3 mm/min.
Prior to testing, measurements of the geometry of the specimens were taken, which are summarised
in Table 2, where the experimentally obtained ultimate moment Mu and the Mu/Mel and the Mu/Mpl ratios
are also included. In Table 2, B and D are the outside width and depth of the crosssection respectively, t
is the mean section thickness and ri is the internal corner radius. A typical failure mode, exhibiting local
buckling of the compression flange and the upper part of the web, is shown in Figure 1.
Table 2: Measured dimensions and test results from 3point bending tests.
Axis of B D t ri Mu
Specimen Mu/Mel Mu/Mpl
bending (mm) (mm) (mm) (mm) (kNm)
SHS 50×50×3 Major 50.18 50.24 2.76 1.53 7.00 1.68 1.41
SHS 60×60×3 Major 60.37 60.63 2.79 3.50 8.74 1.62 1.36
SHS 100×100×3 Major 99.85 99.93 2.78 2.13 18.77 1.35 1.16
RHS 60×40×3MA Major 40.00 60.11 2.75 1.88 7.99 1.84 1.49
RHS 60×40×3MI Minor 60.10 39.95 2.75 1.88 5.69 1.66 1.41
666
Marios Theofanous and Leroy Gardner
All tests were displacementcontrolled with a loading rate of 3mm/min in terms of vertical crosshead
movement. Two symmetrical loading configurations were employed to vary the required rotation
capacity and moment redistribution before collapse. In the first configuration, denoted ‘1/2 span’ in Table
3, the loads were applied at midspan, whilst in the second configuration, ‘denoted 1/3 span’, the loads
were applied at a distance equal to 366.7 mm (1/3 of the clear span length) from the central support. The
1/3 span configuration is shown in Figure 2, where the employed instrumentation is also depicted.
Wooden blocks were inserted at the supports and at the loading points of each specimen and the loads
and reactions were applied through a steel block of thickness 15 mm and width 30 mm, to prevent local
bearing failure.
Loading jack
Spreader beam
Beam specimen
Strain gauge LVDT5 200 LVDT6
LVDT1 LVDT3
LVDT2 LVDT4
LVDT7 LVDT8
Load Cell
100 733.3 366.7 366.7 733.3 100
Figure 2: Test configuration ‘1/3 span’  loads applied at 366.7 mm from central support.
The employed instrumentation consisted of a load cell at the central support, eight LVDTs and six
strain gauges, as shown in Figure 3. The load cell was utilised to measure the reaction force at the central
667
Marios Theofanous and Leroy Gardner
support, which is necessary to determine the stress condition of each specimen, due to their static
indeterminacy. The strain gauges were affixed at the midwidth of the top and bottom flanges at a
distance of 60 mm from each loading point and from the central support point. Their readings verified
that no net axial load occurred in the specimens and hence the end rollers did not provide any axial
restraint. Six LVDTs were employed in pairs at the ends of the specimens and the central support, as
shown in Figure 2, to measure the end rotations and the rotation of the plastic hinge at the central support,
whilst two additional LVDTs were employed at the loading points to measure the vertical displacement.
The applied load and crosshead movement were also recorded. All readings were taken at 2 second
intervals.
The key experimental results are summarised in Table 4, including the ultimate load Fu and the
plastic rotation at ultimate load normalised by the corresponding elastic rotation at ultimate load,
șpl,max/șel,max. The load corresponding to the formation of the first plastic hinge at the central support,
denoted Fh1, and the theoretical collapse load Fcoll are also included. The load Fh1 was determined based
on elastic calculations, whereas Fcoll was determined by classical plastic analysis procedures, assuming
rigidplastic material (and momentrotation) response. All specimens failed by developing three distinct
plastic hinges, one at the central support and one at each loading point. A typical failure mode for the 1/2
span arrangement is displayed in Figure 3.
668
Marios Theofanous and Leroy Gardner
Table 5: Codified and proposed classification and effective width formulae for simply supported beams.
EN 199314 [1] Revised slenderness limits [6]
Specimen
Class Mpred/Mu Class Mpred/Mu
SHS 50×50×3 1 0.71 1 0.71
SHS 60×60×3 1 0.73 1 0.73
SHS 100×100×3 4 0.65 4 0.68
RHS 60×40×3MA 1 0.67 1 0.67
RHS 60×40×3MI 3 0.60 1 0.71
MEAN 0.67 0.70
COV 0.08 0.04
Table 6: Codified and proposed classification and effective width formulae for continuous beams.
669
Marios Theofanous and Leroy Gardner
supported beams is very wellpredicted and a low COV is observed. For the continuous beams, the CSM
gives more favourable strength predictions compared to the classification procedure, but failure to
account for moment redistribution results in excessive conservatism. Moreover, a relatively large COV is
observed, due to the dependency of the effect of moment redistribution on the crosssection slenderness.
Table 7: Assessment of the CSM for simply supported and continuous beams.
Fpred/Fu
Specimen
Simply supported beams Continuous beams
SHS 50×50×31 0.90 0.68
SHS 50×50×32  0.56
SHS 60×60×31 0.95 0.73
SHS 60×60×32  0.77
SHS 100×100×31 0.91 0.89
SHS 100×100×32  0.90
RHS 60×40×3MA1 0.87 0.64
RHS 60×40×3MA2  0.64
RHS 60×40×3MI1 0.87 0.67
RHS 60×40×3MI2  0.56
MEAN 0.90 0.71
COV 0.04 0.17
670
Marios Theofanous and Leroy Gardner
Table 8: Assessment of codified and proposed classification for continuous beams allowing for plastic design.
EN 199314 [1] Revised slenderness limits [6]
Specimen
Class Fpred/Fu Class Fpred/Fu
SHS 50×50×31 1 0.68 1 0.68
SHS 50×50×32 1 0.68 1 0.68
SHS 60×60×31 1 0.72 1 0.72
SHS 60×60×32 1 0.76 1 0.76
SHS 100×100×31 4 0.68 4 0.71
SHS 100×100×32 4 0.68 4 0.72
RHS 60×40×3MA1 1 0.63 1 0.63
RHS 60×40×3MA2 1 0.63 1 0.63
RHS 60×40×3MI1 3 0.52 1 0.69
RHS 60×40×3MI2 3 0.43 1 0.71
MEAN 0.64 0.69
COV 0.15 0.06
The novelty of the method lies in departing from the traditional rigidplastic material response and
assuming that the ultimate moment capacity of the first plastic hinge can be fully exploited. In essence,
the method utilises the upper bound theorem of limit analysis and relies on the determination of a suitable
collapse mechanism. The moment capacity at the location of the plastic hinges is calculated by means of
the CSM; for the first plastic hinge the full deformation capacity is exploited, whilst for subsequent
plastic hinges, the deformation capacity is a fraction of the deformation capacity at the first hinge,
proportional to the plastic rotation ratio as determined from kinematics.
The accuracy of the CSM for indeterminate structures is assessed in Table 9, where all crosssections
classified as Class 1 according to the revised slenderness limits proposed in [6] have been treated with
this method. The SHS 100×100×3 specimens, which have a slender (Class 4) crosssection, have been
treated with the conventional CSM; hence the effect of moment redistribution has not been considered for
these sections. Overall, significant enhancement in design efficiency and good agreement with the test
671
Marios Theofanous and Leroy Gardner
results is observed as evidenced by the low COV of 0.06. Further research into the topic is underway to
determine the slenderness range within which the proposed method can be safely applied.
4 CONCLUSIONS
An experimental study comprising five threepoint bending tests and ten twospan continuous beam
tests (fivepoint bending) has been conducted and the conservatism embedded in the provisions for
stainless steel indeterminate structures codified in EN 199314 [1] has been highlighted. The application
of conventional plastic analysis to stainless steel indeterminate structures and the accuracy of the CSM
have been investigated. It was concluded that both material strainhardening at crosssectional level (at
the location of the plastic hinges) and moment redistribution occurring in indeterminate structures,
comprising sections with sufficient deformation capacity, are significant and should therefore be
accounted for in design. A recently proposed adaptation of the CSM for carbon steel indeterminate
structures [11] has been further investigated and applied to stainless steel indeterminate structures,
yielding excellent results for stocky crosssections. Hence CSM for indeterminate structures emerges as a
promising design approach for stainless steel continuous beams.
REFERENCES
[1] EN 199314. Eurocode 3: Design of steel structures  Part 1.4: General rules – Supplementary
rules for stainless steel. CEN, 2006.
[2] EN 199311. Eurocode 3: Design of steel structures  Part 1.1: General rules – General rules and
rules for buildings. CEN, 2005.
[3] Gardner, L. “The use of stainless steel in structures”. Progress in Structural Engineering and
Materials, 7 (2), 4555, 2005.
[4] Gardner, L. and Ashraf, M. “Structural design for nonlinear metallic materials”. Engineering
Structures, 28 (6), 926934, 2006.
[5] Ashraf, M., Gardner, L. and Nethercot, D. A. “Structural stainless steel design: Resistance based on
deformation capacity”. Journal of Structural Engineering, ASCE, 134 (3), 402411, 2008.
[6] Gardner, L. and Theofanous, M. “Discrete and continuous treatment of local buckling in stainless
steel elements”. Journal of Constructional Steel Research, 64 (11), 12071216, 2008.
[7] Nip, K.H., Gardner, L. and Elghazouli, A.Y. “Cyclic testing and numerical modelling of carbon
steel and stainless steel tubular bracing members”. Engineering Structures, 32 (2), 424441, 2010.
[8] Mirambell, E. and Real, E. “On the calculation of deflections in structural stainless steel beams: an
experimental and numerical investigation”. Journal of Constructional Steel Research, 54 (1), 109
133, 2000.
[9] Rasmussen, K.J.R. “Fullrange stressstrain curves for stainless steel alloys”. Journal of
Constructional Steel Research, 59 (1), 4761, 2003.
[10] Gardner, L. and Nethercot, D.A. “Experiments on stainless steel hollow sections  Part 1: Material
and crosssectional behaviour”. Journal of Constructional Steel Research, 60 (9), 12911318, 2004.
[11] Gardner, L. and Wang, F. “Influence of strain hardening on the behaviour and design of steel
structures”. International Journal of Structural Stability and Dynamics, (submitted).
672
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
1 INTRODUCTION
Conventional analyses of steel frame structures are usually carried out under the assumption that the
beamcolumn connections are either fully rigid or ideally pinned. However, most of the connections used
in current practice are semirigid type whose behaviour lies between these two extreme cases. The
predicted response of the idealized structure may be quite unrealistic compared to that of the actual
structure if connection stiffness is ignored in the analysis and design procedures.
The semirigid connections have important function in structural steel design, because influence
substantially the moment distribution in beams and columns and negatively affect the stability of the
frame, since they increase the drift of the frame and cause a decrease in effective stiffness of the member.
So, the disregard of the actual behaviour of the connections can lead to unrealistic predictions of response
and resistance of structures.
The important attributes that affect the behaviour of semirigid steel frames structures are connection,
geometric and material nonlinearities. The connection nonlinearity is given by the nonlinear moment
rotation relationship of semirigid connections. The geometric nonlinearity includes secondorder effects
associated with the PG and P' effects and geometric imperfections. And finally, material nonlinearity
includes spread of yielding or plasticity associated with the influence of residual stresses. The realistic
modelling of a steel frame requires the use of these attributes if an accurate response is to be obtained.
One way to account for all these effects in semirigid frame design is through the use of an advanced
analysis. Advanced Analysis is a method that can sufficiently capture the limit state of strength and
stability of a structural system and its individual members, so that separate checks of the capacity of
members are not required. With technological advances in computational area has been possible to
employ advanced analysis techniques directly in the offices of engineering design.
During the past 20 years, researches efforts have been devoted to the development and validation of
several nonlinear inelastic analysis methods for steel frames with semirigid connections, as the studies
presented by [1][9]. The behaviour of semirigid connections has been progressively incorporated in
673
R. da Silva, A. Lavall
structural analyses, resulting in more realistic analysis of the response global of structures, allowing a
design accurate and certainly more economical.
This paper presents the development of a finite element for use in the secondorder inelastic analysis
of partially (PR) and fully (FR) restrained planar steel frames. The finite element considers the spread of
plasticity within the cross section and along the member length, several residual stresses distributions,
shear deformation of members, PG and P' effects. Nonlinear spring elements are used to include
partially restrained connections. The behaviour of the connections is modelled using multilinearized
momentrotation curves. The formulation considering Timoshenko theory and self equilibrated residual
stresses is based on updated Lagrangian formulation. The Corotacional technique is used to obtain the
element’s tangent stiffness matrix. A computer program associated with the finite element model is
developed. Numerical examples are presented and the results are compared with those previously
published by others researchers with the objective to validate the finite element model for the Advanced
Inelastic Analysis.
674
R. da Silva, A. Lavall
The natural and Cartesian degrees of freedom of the element are defined, respectively, by:
The relations between natural and Cartesian degrees of freedom are important and listed below:
q1 lc lr
°
® q2 D a T a T c p3 M c M r (2)
°q D T T p6 Mc M r
¯ 3 b b c
The virtual power theorem is used in the development of the finiteelement stiffness:
where dVr is the volume element in the reference configuration, ı the normal stress, W the shear stress, įİ
virtual longitudinal deformation and įJ virtual distortion of a fiber.
The virtual longitudinal deformation and virtual distortion are respectively:
GH H ,D qD ,iGpi ; GJ J ,D qD ,iGpi (6)
675
R. da Silva, A. Lavall
§ ·
Pi ¨ ³ V H ,D dVr ³ W J ,D dVr ¸qD ,i QD qD ,i
¨V ¸ (7)
© r Vr ¹
dP wP dp dp
kt (8)
dt wp dt dt
where, kt is the tangential stiffness matrix of element in Cartesian coordinates. The components kij are
obtained through differentiation of Pi with respect to Cartesian coordinate’s pj:
wPi
kij qD ,i QD ,E qE , j QD qD ,ij (9)
wpi
§ dV dW ·
QD ,E ³Vr ¨¨ H ,D H ,E VH ,DE J ,D J ,E W J ,DE ¸¸ dVr (10)
© dH dJ ¹
§ § dV dW · · § ·
k ij qD ,i ¨¨ ³ ¨¨ H ,D
d H
H ,E J ,D
d J
J ,E ¸¸dVr ¸¸ q E , j qD ,i ¨¨ ³ V H ,DE W J ,DE dVr ¸¸q E , j QD qD ,ij (11)
© Vr © ¹ ¹ © Vr ¹
The first term of the equation represents the constitutive part, the second and third parts represent the
PG and P' effects, respectively.
5 NUMERICAL EXAMPLES
676
R. da Silva, A. Lavall
shown in figure 3. The uniformly distributed load is modelled as a set of equivalent nodal loads. The
frame is analyzed in secondorder elastic theory. Beam and columns are made of profiles I. The area and
the moment of inertia of the beam are, respectively, equals to 43 cm2 and 2770 cm4. The area and the
moment of inertia of columns are equals to 33,4 cm2 and 1510 cm4, respectively. The lengths and
numbering of bars are also shown in figure 3. The Young’s modulus is 21000 kN/cm². Beam and
columns were modelled, respectively, with ten and four finite elements and the cross sections were
divided into twenty layers.
1 kN/cm
30 kN
2 3 4 5
400 cm
1 6
600 cm
Figure 3: Portal frame with semirigid connections
The top and seat angle with double web angle connections, C23 and C45, are identical at the ends of
the beam. The behaviour and data of the connections are presented by [11], considering the three
parameter model. At this paper, the behaviour of the semirigid connections is represented by multilinear
curves, according figure 1, with data shown in table 1.
Figure 4 shows the graphic of moment versus relative rotation for the windward connection (C23)
and leeward connection (C45), for all loading increments. It is observed that, when the frame is subjected
firstly to uniformly distributed load, the connections presented same behaviour. The bending moment is
equal to 7245 kNcm and the relative rotation is equal to 0,08880 rad, for total vertical loading. When
lateral load is applied, the windward connection (C23) unloads, showing a linear behaviour with slope
equal to the initial stiffness and the leeward connection (C45) continues load, ie, continues to rotate in the
same direction, with slope based on tangent stiffness, as shown in Figure 4. For the structure completely
loaded, the connection C45 presents moment equal to 7437 kNcm and relative rotation equal to
0,12982 rad. The connection C23, due to the unloading caused by the lateral force, presents moment
equal to 2395 kNcm and relative rotation of 0,08676 rad. It can be concluded that, the loading
characteristics are very different from the unloading characteristics of the connections. The connection
behaviour is very much affected by the history and direction of the loads applied sequentially.
8000
7000
6000
Moment (kNcm)
5000
4000 Windward
3000 connection
2000 Leeward
1000 connection
0
0,00 0,02 0,04 0,06 0,08 0,10 0,12 0,14
Relative Rotation (rad)
Figure 4: Behaviour of partially restrained connection under loading and unloading
677
R. da Silva, A. Lavall
41,19 kN
39,90 kN
38,61 kN
10 @ 457,2 cm = 4572 cm
37,28 kN
35,76 kN
33,27 kN
32,34 kN
29,40 kN
914,4 cm 914,4 cm
The figure 6 shows the loaddeflection behaviour of the frame with endplate connections, with
different rigidity and moment capacity, namely C1, C2 and C3 connections, until the failure. In graphics
of this figure, the abscissa axis represents the maximum lateral sway of the top of the frame and the
678
R. da Silva, A. Lavall
ordinate axis denotes the load level. It is observed that, the load originally proposed, according figure 5,
was gradually expanded until the strain of steel reached the limit of 21Hy. Results of the loaddeflection
behaviour, obtained by the developed program, are compared with results of [13], obtained by a computer
program able of performing a secondorder inelastic analysis of planar steel structures based on the
refined plastic hinge method. It can be noted that the results of the program, developed based on the
plasticzone method, showed a good correlation with the results obtained by [13]. A difference less than
5% in the ultimate load between the analyses was obtained for all structural models.
Frame with C1 connection Frame with C2 connection Frame with C3 connection
2,5 (2,35) 2,5 (2,44) 2,60 (2,51)
2,0 (2,30) (2,35) (2,40)
2,0 2,08
Load Factor
Load Factor
Load Factor
1,5 1,5 1,56
1,0 1,0 1,04
Zhou Zhou Zhou
0,5 Program 0,5 Program 0,52 Program
0,0 0,0 0,00
0 15 30 45 60 75 90 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70
Lateral Deflection (cm) Lateral Deflection (cm) Lateral Deflection (cm)
Figure 6: Loaddisplacement behaviour at the top of the frame
Figure 7 shows, comparatively, the loaddeflection behaviour at the top of frame, obtained by the
program developed, considering conventional rigid connections and the C1, C2 and C3 connections. The
results show that the frame with the C1 connection has larger deflection, resulting in the more flexible
structure between models analyzed. The frame with the C3 connection presents deflection and load factor
values very close to the conventional model with perfectly rigid connections. It can be concluded that the
properties of connections have significant influence on the strength, stiffness, and ductility of the frame.
2,7
2,4
2,1
1,8
Load Factor
1,5
1,2 Frame with C1 connection
0,9 Frame with C2 connection
0,6 Frame with C3 connection
0,3 Rigid Frame
0,0
0 10 20 30 40 50 60 70 80 90
Lateral Deflection (cm)
Figure 7: Loaddisplacement behaviour at the top of frame obtained by the program developed
It can be noted that, when connections in a frame become stiffer, the response of the semirigid frame
get close to the rigid frame. So, endplate connections can be regarded semirigid or rigid depending on
their rigidity. A frame with endplate connections can be regarded as a rigid frame if its connections are
rigid enough.
6 CONCLUSION
A computer program for Advanced Inelastic Analysis of partially (PR) and fully (FR) restrained
planar steel frames, considering the geometric, material and connections nonlinearities, is developed. The
finite element considers PG and P' effects, shear deformations of members through the Timoshenko
679
R. da Silva, A. Lavall
theory and spread of plasticity. The frame element, made up of layers, enables to identify the plastic
region through the cross section and along member length and to consider any kind of residual stresses
distribution.
The method of advanced analysis showed to be very efficient in the analysis of the behaviour of steel
structures involving semirigid connections from the initial stage of loading until the final stage of
collapse. The results indicate that semirigid connections in steel structures have fundamental importance,
since greatly affect the behaviour of the structure. The developed method considering the nonlinear
behaviour of the connections through MTr multilinear curve showed to be suitable for these analyses.
ACKNOWLEDGMENTS
The authors are grateful to FAPEMIG – Fundação de Amparo à Pesquisa do Estado de Minas Gerais,
for its support to carry out this research.
REFERENCES
[1] Liew, J.Y.R., White, D.W., Chen, W. F. “SecondOrder Refined Plastic Hinge Analysis of Frame
Design: Part I”. Journal of Structural Engineering, 119(11), 31963216, 1993.
[2] Chen, W.F. and Toma, S. Advanced Analysis of Steel Frames: Theory, Software, and Applications.
CRCPress, Boca Raton, 1994.
[3] Kruger, T. S., van Rensburg, B. W. J., du Plessis, G. M. “Nonlinear Analysis of Structural Steel
Frames” Journal of Constructional Steel Research, 34, 1995.
[4] Kim, S.E. and Chen, W.F. “Practical advanced analysis for semirigid frame design”. Journal
Engineering. Fourth Quarter, 129141, 1996.
[5] Chen, W.F., Goto, Y., Liew, J.Y.R. Stability Design of SemiRigid Frames. John Wiley e Sons,
Inc., New York, 1996.
[6] Foley, C.M. and Vinnakota, S. “Inelastic Behaviour of Multistory Partially Restrained Steel
Frames. Part I”. Journal of Structural Engineering, 125(8), 854861, 1999.
[7] Foley, C.M. & Vinnakota, S. “Inelastic Behaviour of Multistory Partially Restrained Steel Frames.
Part II”. Journal of Structural Engineering, 125(8), 862869, 1999.
[8] Gizejowski, M.A., Barszcz, A.M., Branicki, C.J., Uzoegbo, H.C. “Review of analysis methods for
inelastic design of steel semicontinuous frames”. Journal of Constructional Steel Research, 62, 81
92, 2006.
[9] Liu, Y., Xu, L., Grierson, D.E. “Compoundelement modeling accounting for semirigid
connections and member plasticity”. Engineering Structures, 30, 1292–1307, 2008.
[10] Chen, W.F and Kishi, N. “SemiRigid Steel BeamtoColumn Connections: Data Base and
Modeling”. Journal of Structural Engineering, 115(7), 105119, 1989.
[11] Kishi N., Ahmed, A., Yabuki, N., Chen, W.F. “Nonlinear Finite element Analysis of Topand Seat
Angle with Double WebAngle Connections”. Journal of Structural Engineering and Mechanics,
12(2),201214, 2001.
[12] Lavall, A.C.C. Uma Formulação Teórica Consistente para a Análise Nãolinear de Pórticos Planos
pelo Método dos Elementos Finitos Considerando Barras com Imperfeições Iniciais e Tensões
Residuais nas Seções Transversais. Ph.D.Thesis. School of Civil Engineering, São Carlos, 1996.
[13] Zhou, F. ModelBased Simulation of Steel Frames with Endplate Connections. Ph.D. Thesis,
Department of Civil and Environmental Engineering, University of Cincinnati, Ohio, p. 133, 2005.
680
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Robustness, robust design, alternate load paths, ductility, composite joints, catenary action
Abstract. The paper highlights the ductility demand of beamtocolumn connections in the frame of
progressive collapse assessment of steel or composite structures considering sudden column loss. For
progressive collapse mitigation different design strategies are available to increase the collapse
resistance of a building. A very effective design strategy in buildings is the provision of alternate load
paths. Besides strengthening the structural system alternate load paths could be also realized by allowing
change of bearing mechanism within the structural elements. Therefore the structural system has to
undergo large deformations resulting in high demands on ductility of members and joints. The ductile
joint configurations, presented within this paper, allow for redistribution of internal forces within the
structural system by enabling large deformations. So they are contributing to the redundancy of steel or
composite frame structures due to their beneficial properties concerning ductility supply, the possibility
to activate plastic reserves as well as energy absorption capacity.
1 INTRODUCTION
Depending on the public or commercial relevance of a building today it is no longer sufficient for
engineers to consider only basic design criteria for planning of structural framework. Engineers are
increasingly required to consider progressive collapse mitigation as additional design criteria. For
building structures the design strategy of alternate load path is therefore quite effective. The alternate load
path method is realized for that matter by activating plastic system reserves and by transition from
flexural loading to membrane tensile action in the members and joints initiating of catenary action.
Therefore the joints have to be designed in detail and all single joint components have to be adjusted in
such a manner that under bending and tensile loading at each time of loading the weakest component has
to be always ductile. This is feasible with only small additional effort by using the inherent plastic
reserves of the material steel.
681
Ulrike Kuhlmann and Lars Rölle
Particularly in cases in which collapsing parts cause impact loading on key elements. Designing for such
high impact loads is, in most cases, not possible. Such conditions are particularly found in structures of
primarily vertical alignment, such as buildings structures.
2.2 Redundancy as robustness measure
In order to be effective as robustness measure alternate load paths have to be designed sufficiently
strong to transfer the actual occurring forces, including e.g. overstrength effects. An alternate load path
may on the one hand be formed within the structural system (global level) by e.g. strengthening of
transfer girders or by bracing a full floor level to suspend the loads above the damaged part like an
outrigger, see Figure 1a. On the other hand redundancy can be achieved by allowing force redistribution
within a structural member (local level). Structural steel and composite buildings with inherent
sufficiently ductile material behavior allow large deformations when local failure occurs. Large
deformations result in large plastic strain rates of material which enables the activation of additional
plastic material reserves. So on local level the material steel has the capability to form plastic hinges
which all activating also plastic system reserves by redistributing. On global level the redundancy of steel
structures for progressive collapse mitigation may form alternate load paths. e.g. by activation of catenary
action in the horizontal members that means by transition from flexural loading to membrane tensile
acting in the members and joints, see Figure 1b. Therefore a highly ductile behavior of all structural
members combined with sufficient strength is necessary. In framed structures the joints are in general the
weakest link and therefore special focus is on the joint design to avoid a premature failure of the
connections during the procedure of force redistribution.
a) b)
Figure 1: Alternate load paths by a) strong transfer girders or b) catenary action
Steel and composite members benefit from the ductile material behavior of structural steel. So steel
has the capability to combine strength, ductility and energy absorption capacity which are basic properties
for designing robust and redundant buildings.
Plastic material reserves of steel depend on the distance between the level of the nominal values and the
actual values as well as on the ratio of fu/fy. In a structural robustness analysis the actual material
properties are of main interest. Information about actual material resistance of steel is e.g. available in the
probabilistic model code of the Joint Committee of Structural Safety [6].
2.4 Ductility demands for members and joints
For common steel profiles in structural engineering depending on the rotation capacity of the cross
section diverse categories of ductility classes exist. So the capability of the crosssection to undergo
locally a total plastification i.e. to develop a plastic hinge and to assure additionally sufficient rotation
capacity without premature stability failure (class 1 crosssection) is ensured by slenderness limits of
crosssectional parts. Therefore for plastic analysis of a steel structure including redistribution the
requirements according the various codes are to use only those crosssections with sufficient moment
bearing capacity as well as rotation capacity. If rigid and fullstrength joints are used the plastic hinges
are located in the beams. So the total required deformation and rotation capacity to activate the membrane
effect in the direct affected part of the structure has to be offered by the beam members. But fullstrength
joints cause much additional effort and they are costly.
Ductility demands for joints are decisive for partialstrength joints which have less resistance than the
beams but also reduced fabrication costs compared to fullstrength joints. Using partialstrength joint
682
Ulrike Kuhlmann and Lars Rölle
configurations the plastic hinges are developing initially in the joint which requires high rotation
capacities of the joints. Therefore a detailed joint design is necessary considering the interaction of all
joint components including overstrength effects to ensure that for the whole loading sequence of the joint
the decisive weakest component is always ductile, see also chapter 4.
For bolted connections there is interplay of hardening or overstrength effects and the various
deformation capacities of the single components. By ensuring that especially the components “endplate in
bending” and “column flange in bending” have a certain ductility additional membrane effects on local
level (in the TStub) may be activated leading to a further increase of the resistance provided there is
sufficient bearing capacity of the bolts.
Figure 2: Composite frame under the event column loss with a) 2Deffect and b) 3Deffect
For framed composite structures without a continuous slab (floor system as single spans with only
minimum reinforcement) the single slabs are not transversely tied together. For this reason in case of
column loss the membrane action may only be activated in plane of the directly affected frame. So only
the composite main beam system is able to redistribute forces and offer alternate load paths.
The slab in this case is unable to activate additional membrane effects transversally to the frame
plane. So only 2Dbehavior may be assumed within a large displacement analysis, see Figure 2a.
Having a continuous RC slab in the composite structure including a uniform amount of reinforcement
the slab is connecting the single frames transversally. Consequently for the event of column loss the slab
is not only contributing to the resistance of the composite beam in the plane of the directly affected frame
but also acting as a tie in transversal direction. The RC slab provides ties in two horizontal directions and
enables therefore 3Dbehavior for the case of column loss (see Figure 2b). However, for effective opera
tion the continuity and anchorage of the ties is obligatory. Activating membrane action in longitudinal
direction within the composite beams and joints and additionally in transversal direction within the RC
slab the redundancy of the composite frame is clearly increased in comparison to 2D systems.
683
Ulrike Kuhlmann and Lars Rölle
deformability define the global redundancy of the structure. In comparison to nominally pinned joints
there are only small extra costs (material + labor) but much more redundancy of the structure, so by only
small additional efforts the effectiveness concerning progressive collapse mitigation is improved [10].
4.2 Design of the steel elements of the joint
For the design of pure steel joints or for the steel elements of composite joints the adjustment of the
single joint components is of high importance in order to design highly ductile joint configurations.
Therefore the parameters mainly influencing the joint behavior have to be treated with special care. Table
1 describes the qualitative degree of influence concerning the rotation capacity as well as the bearing
capacity of the various parameters investigated for the joint components. So small changes of some of
these parameters might positively influence the ductility of the joint in a significant way whereas at the
same time the bearing capacity is only decreasing marginally. The listed parameters are mainly
influencing the components “endplate in bending” and “column flange in bending” which are able to
activate additional local membrane effects under large deformations. So local additional bearing effects
may compensate the decrease of the joint bending capacity when reducing e.g. the endplate thickness or
the steel grade.
Table 1: Qualitative influences of main steel joint parameters
Parameter influence on rotation capacity Influence on bearing capacity
ratio dbolt/tendplate + 
bolt arrangement + 
steel grade endplate + 
ratio tendplate/tcolumn flange o n.n
+ disproportionate high o proportionate  little
a) b)
Figure 3: a) Influence of ratio fu/fy for the available ultimate strain of a reinforced concrete bar in tension,
b) Influence of stud and reinforcement arrangement in the hogging moment zone of the slab [11].
684
Ulrike Kuhlmann and Lars Rölle
The reinforcement ratio is influencing the moment resistance and the deformation/rotation capacity of the
joint. By increasing the amount of reinforcement the deformation capacity is also increased. The reason is
that the reinforcement ratio is significantly influencing the steel stress Vsr1 of the rebar when a first crack
has formed. The ratio of Vsr1/fyk is important for the available plastic strain Hsmu of the reinforced concrete
slab under tension, where fyk is the yield strength of the rebars. For high deformation capacities of the slab
a higher reinforcement ratio is advantageous because a rising ratio causes steel stresses Vsr1 when a first
crack has formed which are well below the yield strength.
Beside the reinforcement ratio and class the arrangement of the shear studs in the hogging moment
region are influencing the deformation capacity of the slab. More precisely the distance of the first shear
stud to the column profile is decisive for the available expansion length of the rebars. By increasing the
distance of the first stud the length for activating plastic strain in the reinforcement is clearly increased
resulting in increase of deformation capacity. It is pointed out that also a discontinuous amount of
reinforcement within the “tension bar” in the joint region should be provided to profit from the modified
stud arrangement, see Figure 3b.
4.4 Overstrength effects
According to the basic design criteria (ULS + SLS) members and joints are designed assuming
nominal material values. This is justified by the present safety concept. However for large displacement
analysis considering only nominal values may lead to results which are nonconservative.
So aside of the plastic behavior of the material and the stability sensitivity of the sections which
dominate the ductility of the members the joint behavior is decisive. Composed of various components
the aim should be that only ductile components control the overall joint behavior. For this not only the
component behavior itself is of importance but the interplay of the various components considering also
possible overstrength effects play an important role.
Figure 4 gives the example of a joint composed of a ductile and a brittle component, e.g. the endplate
in bending acting together with bolts which usually fail in a brittle manner. The design according to the
nominal values of strength leads to a moment rotation curve of the joint also acting ductile, see case a).
However the actual values of strength may exceed the nominal values (overstrength effects) so that no
longer the ductile component dominates the failure load, but the brittle one, see case b). As a consequence
the overall behavior of the joint shows a very limited rotation capacity. Disregarding overstrength effects
the joint may lead to only limited ductility as shown and as consequence no redistribution of forces can
take place that means the structure has only reduced redundancy.
685
Ulrike Kuhlmann and Lars Rölle
by designing partialstrength joints with sufficient ductility. Within a national research project [9]
parameters influencing the ductility of bolted beamtocolumn connections were investigated.
Furthermore the influence of overstrength effects on the resistance and rotation capacity of the joint was
analyzed. In a current diploma thesis [7] a composite structure is analyzed for the event column loss.
5.2 Experimental investigations
Within the two mentioned research projects the joint deformability and ductility as well as the
combined bending and tensile resistance have been investigated.
The performed steel joint tests mainly aimed at the investigations of increasing the joint ductility by
varying different parameters. The main parameters influencing the deformability in the tension zone of
the joint are the ratio of the endplate thickness and the bolt diameter (under consideration of the individu
al material strength) and the arrangement of the bolts depending on the distance to the web, see Table 1.
By decreasing the ratio of the bolt diameter and the endplate thickness the rotation capacity is in
creased. By modifying the bolt arrangement particular by increasing the distance of the bolts to the beam
web and beam flange the rotation capacity is also increased. A test series in [9] examined the influence of
the steel strength and the simultaneous activation of the components endplate and column flange in
bending. The resulting momentrotation curves are given in Figure 5a. By reducing the steel grade of the
endplate and the column flange the rotation capacity is also increased accompanied by only small
decrease of the resistance as, see e.g. test curves Z6 and Z3 in Figure 5a.
a) b)
Figure 5: a) Influence of the steel grade to the rotation capacity and b) measured MNinteraction of the
composite joint tests
The objective of the composite joint tests was the determination of the simultaneous momenttensile
resistance within the joint. The tests simulated the loading procedure from pure bending state to a mixed
bending and tensile state up to a pure tensile state at the end. The tests were successfully following the
whole theoretical MNcurve (as shown in Figure 5b). The failure of the joints always occurred under
mainly pure tensile exposure. From the results of the composite joint tests under combined bending and
tension exposure it can be concluded that having a highly ductile joint behavior due to welladvised
adjustment of the single components the transition from pure bending state up to a membrane state in the
joint is feasible. The design of the joint specimens considered already overstrength effects and the bolts
were intentionally oversized to exclude premature brittle failure of the connection. The results of the
joints tests have been also confirmed by a substructure tests executed by the project partner ULg (Liége,
Belgium). Within this substructure test the activation of catenary action, after the event column loss
happened, was possible due to the highly ductile performance of the joints [8]. Failure was mainly
induced by the concrete slab: for the hogging moment joints by increased cracks and final rupture of the
reinforcement, for the sagging moment joints by crushing of the concrete. In addition a remarkable
residual resistance and ductility remained when the concrete slab had already failed.
5.3 Numerical investigations
The numerical simulations were executed by the Finite Element software ANSYS [1]. First
recalculations were made to verify the FEModel at the tests results and afterwards parametrical studies
686
Ulrike Kuhlmann and Lars Rölle
followed to extend the range of parameters as well as investigate the actual influence of material over
strength effects on the joint behavior. The influence of the material properties on the joint behavior
(ductility and bearing capacity) depending on stochastic distribution of the material strength was
investigated in a first step by considering various combinations of characteristic, see Figure 6ac [9].
Numerical simulations were used due to the fact that the local membrane effect in the Tstub of the
components “endplate in bending” or “column flange in bending” is not yet implemented in the analytical
approach of the component method acc. to EN 199318 [1][4].
a) b) c)
Figure 6: Deformed shape of steel joint considering a) nominal material values and b) overstrength
effects and c) Moment–rotation curves for varied material strengths
In terms of resistance the overstrength effects usually cause an additional material reserve which can
be activated in the case of progressive collapse analyses. But considering connections where different
types of steel grade are assembled the overstrength effects may result in unintentional negative effects.
Particular limited ductility is the main phenomena as the distribution of the available joint rotation
depending on the statistical spread of the material properties in the diagram of Figure 6 shows.
Within a current diploma thesis numerical simulations on global level at steel and composite
structures have been performed to analyze the collapse resistance of the structure as well as requirements
for the implemented partialstrength joint configurations. Another aspect is also to determine the
additional positive contribution of the continuous RC slab in two horizontal directions, see Figure 7. [7]
First results showed that such a composite beamcolumn structure is able to resist the event of a
column loss under the accidental load combination for about 7080% utilization of ULS loading. The
identified requirements for the partialstrength joints concerning ductility and MNresistance are also
feasible and within the range of the available rotation capacity and strength determined by the
experimental investigation.
687
Ulrike Kuhlmann and Lars Rölle
REFERENCES
[1] ANSYS User’s manual. ANSYS Mechanical Solutions Release 11.0. 2007.
[2] EN 199111: Eurocode 1 – Actions on structures – Part 17: General Actions  Accidental actions
due to impact and explosions, CEN, 2002
[3] EN 199211 Eurocode 2: Design of concrete structures  Part 11: General rules and rules for
buildings. CEN, 2005
[4] EN 199318: Eurocode 3. Design of Steel Structures – Part 18: Design of Joints, CEN, 2005
[5] GSA 2003: US General Services Administration: Progressive Collapse Analysis and Design
Guidelines for New Federal Office Buildings and Major Modernization Projects, GSA. 2003
[6] JCSS: Probabilistic Model Code: Part III Resistance Models – Steel. Joint Committee on Structural
Safety, 2001.
[7] Kleiner, A. Untersuchung der Robustheit und Kollapsresistenz von Stahl und
Verbundrahmentragwerken – Numerische Simulation eines Stützenausfalls. Universität Stuttgart,
Mitteilung des Instituts für Konstruktion und Entwurf Nr. 200943X, 2009
[8] Kuhlmann, U. et al. (2008): Robust structures by joint ductility. Publishable Report. ContractNo.
RFSCR04046, 2008.
[9] Kuhlmann, U., Rölle, L. (2008): Duktilitätskriterien für typisierte Stirnplattenverbindungen,
Schlussbericht. DAStForschungsvorhaben im Auftrag der AiF. AiFVorhaben Nr. 14627,
Dezember 2008.
[10] Rölle, L.; Kuhlmann, U. (2009): Alternate load path method for robust design by ductile steel and
composite joints. Proceedings of IABSE Symposium Bangkok. September 2009
[11] Schäfer, M. (2005): Zum Rotationsnachweis teiltragfähiger Verbundknoten in verschieblichen
Verbundrahmen. Dissertation. Universität Stuttgart, Mitteilung des Instituts für Konstruktion und
Entwurf Nr. 20051, 2005.
[12] Starossek, U., Wolff, M.: Progressive Collapse  Design Strategies. Proceedings of the IABSE
Symposium Lisbon, 2005.
688
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
József Szalai*
Abstract. Stability analysis and design has always played a key role in the process of verification of steel
structures. The possible analysis methods and design procedures have long history with a plentiful
literature providing various proposals for the engineers. This paper concentrates on the use of different
types of eigenvalue analysis as a simple and powerful tool for stability design. Nowadays almost all the
engineering software products have some kind of eigenvalue analysis options so these tools are easily
available for the practicing engineers providing them a deeper look on the structural behavior. Various
types of application possibilities are reviewed and new methods are proposed supporting the most upto
date standard procedures of different levels from the isolated member design to the partial or global
structural stability design. The suitable theoretical (both mathematical and mechanical) background is
developed and the numerical procedure is implemented. The technique is applicable for a wide range of
structural types and stability problems making the automatic effective length calculation possible in
general without the use of any iterative process or tabulated values for certain cases. An application
example is presented showing the comprehensiveness of the methods, and special efficiency indicators
are presented in order to supply information about the adequacy of the applied design method.
1 INTRODUCTION
There are two available methods for stability design of steel structures provided by the EN 199311
[1]:
isolated member approach: Sections 6.3.1, 6.3.2, 6.3.3
general method: Section 6.3.4
The first one is valid for uniform members only and based on structural member isolation and
buckling mode separation. The main difficulties of this method arise from these two simplifications, the
member isolation is usually handled by applying suitable effective length factors – considering the
appropriate support and restraint conditions – while the mode separation is solved by special interaction
factors creating the connection between the pure loading and buckling cases. There are a great number of
papers on both topics including some theoretical investigations about the mechanical basics [2, 3] and
several proposals for the practical application [4, 5]. The EN 199311 regularizes only the calculation of
interaction factors the problems coming up from the member isolation are not dealt with in the standard.
In the general method these two simplifications are eliminated by examining a complete structural part
and calculating only one slenderness belonging to the real, compound loading and buckling situation.
Although in the recent version of EN 199311 there are several restrictions on the application field of
this method however on the other hand there are heavy research and development efforts on extending its
applicability [6, 7] and this method is expected to cover much larger area of practical problems then the
isolated member based conventional procedures. It is also important to note that in case of the general
method the calculation of the generalized structural slenderness requires more complicated analysis which
can usually provided only by application of some software package. It seems evident is that in both
689
József Szalai
methods the key question is the determination of the slenderness values and especially the calculation of
the appropriate elastic critical values (critical forces or critical load levels). This aspect is very poorly
treated in the structural standards and accordingly the practicing engineer is fully responsible for it.
Moreover the calculated elastic critical values are usually very important and have a significant influence
on the final result of stability design.
In this paper the possibilities of eigenvalue analysis for the calculation of elastic critical values are
examined from the point of view of standard stability design according to the recent version of EN 1993
11. Different application methods are introduced adapted to the different design approaches, and special
indicator factors are developed highlighting the relevance of the used type of eigenvalue analysis. An
application example is presented to show the practical working of the different methods.
where KE is the elastic stiffness matrix, KG is the second order geometric stiffness matrix, D is the
eigenvalue and U is the corresponding eigenvector. In the mechanical interpretation the eigenvalue
denotes the elastic critical load level and the eigenvector shows the eigenshape (eigenmode) or buckling
shape (buckling mode). It is important to note that the mechanical meaning and accuracy of the calculated
eigenmodes and eigenvalues highly depend on the definition of stiffness matrices. In this paper special
decomposition techniques are applied for the compiled geometric stiffness matrix – while the elastic
stiffness matrix is always formed on the complete structural model – to suitably calculate the elastic
critical load levels necessary for the certain design approaches. Upon these techniques the following
eigenvalue analysis are proposed:
Complete Eigenvalue Analysis (CEA) – the geometric stiffness matrix is compiled on the whole
structural model
Partial Eigenvalue Analysis (PEA) – the geometric stiffness matrix is compiled only on a
separated part of the structural model
Selected Eigenmode Analysis (SEA) – the geometric stiffness matrix is compiled only on
selected displacement degreeoffreedoms
The CEA is the mostly known and commonly used analysis technique, the resulted critical load levels
and corresponding buckling modes apply to the whole global structural model – even it is apparently
restricted to a part of it – and consider the compound loading case. In the further text we write the CEA in
the following form:
K E D cr
C
K GC U cr
C
0 (2)
The CEA method is applicable for the critical load level calculation for the general stability design
method.
If a part of the complete model is examined and intended to design for stability the PEA can be used.
In this method the structural model is divided into two parts: (P1) a relevant part and (P2) a remaining
part. Accordingly the complete geometric stiffness matrix can be decomposed:
K GC K GP1 K GP 2 (3)
One can obtain an eigenvalue solution for the relevant part by solving the following equation:
690
József Szalai
K E
D crP1 K GP1 U crP1 0 (4)
In a mechanical interpretation this calculation yields special buckling shapes which are induced by the
internal forces acting only in the members which are part of the relevant substructure while the initial
stiffness of the whole structure is considered (as a restraint condition for the examined substructure). In
that sense this solution is similar to the ones applying fictitious springs at the joining parts of the relevant
substructure modeling the restraints from the remainder of the whole structure. The PEA method also
provides global type solution for the real, combined buckling situation so it is also applicable for the
general stability design.
The third proposed method (SEA) is directly developed for the isolated member approach since it is
able to calculate the separated buckling modes. In this case the geometric stiffness matrix is compiled in
such a way that the rows and columns associated with the displacement degreeoffreedom necessary for
the relevant buckling mode are considered and the rest of the matrix is neglected. For example if the
subject of the analysis is the lateral buckling about the minor axis (axis ‘z’ according to EN 199311)
then those terms are kept only in the geometric stiffness which contain the second order compression
effect for the selected displacements (uy and Iz). Further if this buckling mode is required only on an
isolated member then this reduced compilation is done only on this element. Consequently the second
order effects of the system are concentrated so as to be able to experience only the relevant buckling
mode. Considering however the complete first order elastic stiffness matrix of the whole structure the
appropriate restraints (and accordingly the necessary effective lengths) of the isolated member can be
calculated quite accurately. The SEA method for the selected ith buckling mode (for instance Ncr,z) is
written as follows:
K E
D crNcr,z K GNcr,z U crNcr,z 0 (5)
(1) for a certain eigenmode what are the relevant members (relevant model portion)?
(2) for certain members (certain model portion) which is the most relevant eigenmode?
691
József Szalai
For the problem (1) let the energy be calculated for members or model portions (k denotes the relevant
member or model portion):
1 C k C
Eik U cr ,i K E U cr ,i (7)
2
In this case the elastic stiffness of the relevant model part is considered only. Obviously the sum of all the
energy values of model parts gives the total energy of the complete model, i.e. (having m number of
model parts):
m
Ei ¦ Eik (8)
k 1
Accordingly an ERI can be constructed so as to show the relevance of the separated model portions
considering the ith eigenmode as a percentage:
C k C
Eik U cr ,i K E U cr ,i
ERI1ik 100 100 C C
(9)
Ei U cr ,i K E U cr ,i
Problem (2) is more complicated but also more significant, since in the case of a complex structural
model it is usual, that different eigenmodes describe the buckling behavior of distinct parts of the model.
For that reason a scaling procedure is necessary in order to select the appropriate eigenmode for the
stability design. To develop a proper scaling factor let us examine the basis of the stability design
approach of the EN 199311 which is the buckling curve based reduction factor. The mechanical model
for the buckling reduction factors is the AyrtonPerry formula. In this model the failure is associated with
the load level at which the second order maximum elastic stress of the geometrically imperfect member
reaches the yield stress. Consequently the reduction factor depends mainly on the amplified imperfection
which has usually a shape equal to an appropriate eigenmode or a combination of them. The scaling
factor should therefore consider this effect to show the importance of the eigenmodes in accordance with
the mechanics of the buckling reduction factors. Firstly the basic amplitudes for the eigenmodes are
determined by normalizing using the elastic stiffness matrix:
1 N U crN , j K GC U crN , j D cr
C
1
,i
Ei U cr ,i K E U crN ,i (10)
2 U crN ,i K GC U crN ,i D cr
C
,j
As a result the greater the critical factor the less the amplitude of the eigenmode is, this is realistic when
considering the eigenmodes as geometric imperfection. It is known [6] that the geometrical imperfections
having the shape of an eigenmode cause the following additional amplified second order displacements:
1
U add U imp (11)
D cr 1
If the normalized eigenmodes of Eq. (10) are considered as imperfections in Eq. (11) then a further
scaling factor can be created by calculating the energy of this amplified imperfection in the kth model
portion:
T
1 §¨ 1 · § 1 ·
E ik U crN ,i ¸ K Ek ¨ U crN ,i ¸ (12)
¨
2 © D cr ,i 1 ¸ ¨ ¸
¹ © D cr ,i 1 ¹
692
József Szalai
Finally if it is considered that enough number (m) of eigenmodes is calculated (i.e. the last eigenvalue –
what is the highest elastic critical load level – is sufficiently high) then the following ERI can be
constructed showing the relative significance of the ith eigenmode for the kth model portion:
2
§ 1 · N k N
¨ ¸ U K U
Eik ¨ D 1 ¸ cr ,i E cr ,i
ERI2ik 100 100 © cr ,i ¹ (13)
m 2
m § · N k N
¦ Eik ¦ ¨ D 1 ¸ U cr,i K E U cr,i
¨ 1 ¸
i 1 © cr ,i ¹
i 1
K E D cr
P1 C
P1
,i K G U cr ,i K E
D crP1,i K GP1 U cr
P1
,i D cr ,i K G U cr ,i
P1 P 2 P1
D cr
P1 P 2 P1
,i K G U cr ,i (14)
Eq. (14) expresses an error force vector generated by the partial buckling mode on the remaining model
portion (P2). Obviously if this term is significant then the partial buckling mode is possibly irrelevant
which shows that the current model portion has no distinct buckling problem. On the other hand if this
force approaches zero then the buckling mode is dominant and accurate for the selected model portion.
To develop a straightforward indicator we use again an energy format for Eq. (14):
4 APPLICATION EXAMPLE
In the following example the CEA method is used and the ERI1 and ERI2 factors are examined.
B1
C2
C1
693
József Szalai
In Fig. 1 the structural model is illustrated, which consists of two columns (C1 and C2) of HEA200
and a beam (B1) of HEA220 loaded by a 30 kN/m line load. According to the support condition four
cases are evaluated (the beam is always simply supported) summarized by Table 1.
All the eigenvalue analysis are calculated by ConSteel software [9], [10] using a 7 degreeoffreedom
finite element model. Figs. 25 show the calculated eigenmodes and elastic critical factors for Case1
Case4 – all buckling modes form some kind of outof plane buckling –, and Tables 25 contain the ERI1
and ERI2 factors evaluated for the beam and the two columns (for ERI1 the columns and for ERI2 the
rows give the 100% value).
694
József Szalai
It can be seen that the first and dominant buckling mode is always the lateraltorsional buckling of the
B1 beam coupled with some form of flexural buckling of the columns. The corresponding critical load
factors increases from Case1 to Case4 denoting the significance of the applied additional restraints for the
supports of the columns however from the ERI factors it becomes clear that the contribution of the
columns to the certain buckling modes is significantly different. From the ERI1 values it can be
concluded that in the first buckling mode the B1 beam is always dominant and for the other modes the
dominant member is also highlighted in the tables. Looking at however the ERI2 values the followings
can be determined for a certain member:
which is the most relevant buckling mode;
which is the appropriate critical load factor for the calculation of member slenderness in the
stability design.
In Case1 the ERI2 values show the strong dominance of the member B1 in the first mode, C1 in the
second mode and C2 in the fourth mode. This is a consequence of the pure shape of the buckling modes,
since the columns have pinned supports. In Case2 the fix support of the column C1 has a considerable
restraining effect on the beam increasing the first critical load factor accordingly in this case the first
mode is more dominant for the column C1 than the third one which show an isolated buckling mode for
this member. The column C2 is not really effected by the additional restraint, so the dominant mode
695
József Szalai
remains the fourth one with almost the same critical load factor value. In Case3 the situation is quite the
same, the column C1 has the dominant isolated buckling mode with same critical load factor as in Case1,
and for column C2 the first mode is the most relevant. In Case4 from the ERI1 factors it is clear that the
complete model contributes to the first buckling mode and the ERI2 values are explicitly show that this
mode is the most relevant for all the members. Naturally the meaning of the ERI2 values can be refined
by increasing the number of calculated eigenmodes which is usually necessary in case of larger structural
models. This simple and straightforward example is intended to show the mechanical meaning of the
different indicator factors.
5 CONCLUSIONS
One of the most important issues in stability design which is out of the field of standard regulations is
the calculation of elastic critical forces or load levels for the determination of slenderness values. In this
paper several methods are presented for this problem using the eigenvalue analysis based approaches. For
different structural arrangements different types of buckling modes can be dominant and moreover the
modern structural standards provide several different possibilities for the stability design. To yield
appropriate slenderness for the different problems and design methods three approaches are proposed: the
complete eigenvalue analysis, the partial eigenvalue analysis and the selected eigenmode analysis. Further
supporting the selection of the relevant approach and eigenmode special indicator factors are developed.
In an application example the practical working with the proposed eigenvalue analysis approaches and
indicator factors are presented.
REFERENCES
[1] European Standard, EuroCode 3. Design of Steel Structures – Part11: General rules and rules for
buildings, EN 199311, 2005.
[2] Boissonade, N., Jaspart, J.P., Muzeau, J.P., Villette, M., “Improvement of the interaction formulae
for beam columns in Eurocode 3”. Computers and Structures, 80, 23752385, 2002.
[3] Greiner, R., Lindner, J. “Interaction formulae for members subjected to bending and axial
compression in EUROCODE 3—the Method 2 approach”, Journal of Constructional Steel
Research, 62, 757770, 2006.
[4] Serna, M.A., López, A., Puente, I., Yong, D.J. “Equvivalent uniform moment factors for lateral
torsional buckling of steel members”, Journal of Constructional Steel Research, 62, 566580, 2006.
[5] DongHo C., Hoon Y., “Iterative system buckling analysis, considering a fictitious axial force to
determine effective length factors for multistory frames”. Engineering Structures, 30, 560570,
2009.
[6] Szalai, J., Papp, F. “On the theoretical background of the generalization of AyrtonPerry type
resistance formulas”. Journal of Constructional Steel Research, 66(5), 670679, 2010.
[7] Sedlacek, G., Müller, C. “Zur Vereinheitlichung der Stabilitatsregeln im Eurocode 3“. Stahlbau,
73, 733744, 2004.
[8] Perelmuter, A.V., Slivker, V.I. “The problem of interpretations of the stability analysis results”,
European Conference of Computational Mechanics, Cracow, Poland, 2629 June, 2001.
[9] Papp, F., Iványi, M. “ConSteel as the prototype of a CAD/CAM oriented program for concurrent
design of beamcolumn structrures”. Proc. 5th International Conference on Computational Steel
Structures Technology, Leuven, Belgium, 2000.
[10] Website: www.consteel.hu
696
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Stainless steel structures, Residual stress, Roll forming, Neutron diffraction.
Abstract. During the manufacture of roll formed structural members the production and storage of sheet
materials, as well as their subsequent forming causes plastic deformation in varying degrees around the
resulting cross section. Plastic deformation causes both an increase in material strength in the section
material through cold working and it also affects the residual or internal stress distribution present
throughout the resulting structural section. Both the material strength and the residual stress distribution
influence the structural behavior of the cross section, therefore it is important to map both these
properties in order to achieve efficient structural design. Destructive techniques have commonly been
used to map residual stresses in structural sections. To achieve a high resolution of measurements these
techniques are extremely labor intensive and sensitive to the measurement technique and it is almost
impossible to measure the strain relaxation that occurs in three orthogonal components by this process.
Nondestructive residual stress measurements are relatively infrequently used for structural engineering
applications. The presented experimental program demonstrates the applicability of the nondestructive
technique of neutron diffraction for mapping residual stresses in structural members at four locations
through the thickness of a roll formed stainless steel section. The measurements were made using the
ENGINX instrument at the UK’s pulsed neutron source: ISIS in Oxford.
1 INTRODUCTION
Cold formed structural sections are a comparatively novel type of structural section that started to be
more widely used for construction in the 1940s. They now comprise 15% and 13% of all new structures
in the housing market in the USA and Australia respectively [1] and expansion in the UK markets is
being actively encouraged for both environmental and economic reasons [2]. Since cold formed sections
are produced by plastically deforming metal sheets at room temperature the thickness of the sections can
be less than the minimum thickness required to retain the high temperatures essential to manufacture the
more conventional hot rolled sections. Hence cold formed sections which are used to carry light loads and
span short distances can be lighter and structurally more efficient than the hot rolled alternatives.
Roll formed sections are the most prevalent type of cold formed section. During the manufacture of
roll formed sections plastic deformation can occur at several stages and plastic deformation will have
been experienced to varying degrees around the resulting cross section. Through a process termed cold
working the regions of the section that have experienced plastic deformation exhibit an increase in
material strength and a decrease in ductility [3][4]. The resulting distribution of material strength around
roll formed sections has been mapped and used in structural design codes of both carbon steel and
stainless steel roll formed sections to increase the material efficiency [5][6]. However the plastic
deformation which causes the increase in material strength also influences the distribution of residual or
697
Rachel B. Cruise and Anna M. Paradowska
internal stresses. Residual stresses are stresses that exist within a structural member in its unloaded state
and their magnitude and distribution can affect the structural behaviour of structural cross sections.
Since hot rolled sections have been used in the construction industry for a much longer period of time
than roll formed sections there is a deeper understanding of their structural behaviour. Furthermore the
techniques for measuring factors that can influence structural behaviour, such as residual stresses, have
been developed to capture the significant aspects of the magnitudes and distributions observed in hot
rolled sections. However, these techniques have been shown to give an incomplete picture of the
magnitude and distribution of residual stresses in roll formed sections [7][9]. This paper will therefore
present data from a pilot study that used a nondestructive technique, namely neutron diffraction, to
measure residual stress distributions in a roll formed stainless steel box section to demonstrate the
potential of adopting this technique.
3 ROLL FORMING
Roll forming is a highly automated and therefore efficient production process. There are two types of
sheet material that are commonly used as the starting material for roll forming: hot and cold rolled sheet
material. Stainless steel can be rolled whilst hot to produce hot rolled sheet of a minimum thickness of
approximately 3mm (see stage 1 in figure 1). If thinner sheet material is required, since the sheet will be
too thin to retain the heat needed to allow for hot rolling to occur, the stainless steel is passed through
rollers whilst it is at room temperature, therefore plastically deforming the sheet to reduce its thickness.
This process produces cold rolled sheet material.
For reasons of efficient storage and to enable the sheet material to be used as the starting material of
this completely automated section forming route both hot and cold rolled sheet material are wound into
coils as shown in stage 2 of figure 1. To manufacture roll formed sections the coil material is unwound
(see stage 3 in figure 1) and then fed into shaping rollers which plastically deform the sheet material into
the required cross sectional shape. To roll form a box section the sheet is rolled into a circular cross
section, welded closed and then this tube is crushed into a rectangular cross section as shown in stages 4
6 in figure 1.
Roll formed structural sections can therefore experience plastic deformation at three stages in their
manufacture. Firstly plastic deformation can be experienced in cold rolled sheet production, secondly
during the coiling and uncoiling of the sheet material and finally during the forming of the cross section.
698
Rachel B. Cruise and Anna M. Paradowska
4 COLD WORKING
In general cold working, or the increase of material strength through plastic deformation, can be
explained by considering the effect of plastic deformation on the ordered arrangement of atoms in a
metallic lattice. Plastic deformation can be described on the atomic scale as the movement of planes of
atoms with respect to one another in the metallic lattice. In carbon steel and stainless steel this causes an
increase in dislocations in the metallic lattice. The creation of more dislocations in the metallic lattice
increases the number of obstacles to planes of atoms moving. Therefore the cold worked material is
observed on the macro scale to increase its resistance to further plastic deformation and so exhibit an
increase in material strength [11].
For stainless steel with an austenitic microstructure an increase in dislocations is not the only
mechanism that can increase the material strength of the cold worked material. The arrangement of atoms
in an austenitic microstructure, prior to experiencing cold working, can be described by the unit cell
shown in figure 2a. When this unit cell is duplicated and stacked together the arrangement of atoms in the
metallic lattice of an austenitic microstructure is described. This particular unit cell is termed a Face
Centred Cubic (FCC) unit cell.
a) Face Centred Cubic (FCC) unit cell of b) Body Centred Cubic (BCC) unit cell
the austenitic microstructure. of the martensitic microstructure.
Figure 2: Microstructures in cold worked austenitic stainless steel.
The austenitic microstructure of stainless steel grade 1.4301 is a metastable microstructure which
means that work done to the material through plastic deformation will cause the austenitic microstructure
to, in part, transform into a martensitic microstructure. The unit cell of the martensitic microstructure is
shown in figure 2b and it is termed a Body Centred Cubic (BCC) unit cell. The BCC unit cell is smaller
and has a higher ratio of volume of atom to volume of unit cell compared to the FCC unit cell. This ratio
is commonly termed the Atomic Packing Factor (APF) and it is an indication of the density of the unit
cell. In addition, unlike the FCC unit cell, the BCC unit cell has within its geometry no planes where the
atoms are as tightly packed together as possible. This is of significance because owing to the geometry of
these close packed planes they can easily slide past one another and the lack of these in the BCC unit cell
causes the martensitic microstructure to give the cold worked stainless steel its increase in strength and
reduction in ductility [12]. The relationship between the two microstructures, and therefore the
transformation that occurs during plastic deformation, is shown in figure 3 by identifying the atoms that
699
Rachel B. Cruise and Anna M. Paradowska
will create the unit cell of the martensitic microstructure within the metallic lattice of the austenitic
microstructure.
Atoms that will form the x
martensitic unit cell
y
z
A B C D
50 mm Weld
6 mm
Austenitic
microstructure 100 mm
Figure 3: Transformation from an austenitic Figure 4: Test sample.
microstructure to a martensitic microstructure.
5 RESIDUAL STRESSES
Uneven plastic deformation also creates residual or internal stress distributions that equilibrate over
the whole cross section. Residual stresses are defined at three different scales by the distance over which
they equilibrate. Type I residual stresses relate to the macro scale, where equilibrium is achieved over
distances that relate to the scale of the structural cross section. It is this type of residual stress that is
considered to have the greatest importance for structural behavior. Type II and type III residual stresses
relate to the micro scale. Type II residual stresses are defined as equilibrating over several metallic grains
(regions where the metallic lattice is continuous) and type III residual stresses are defined as equilibrating
within metallic grains [13].
Residual stresses ıx, ıy and ız act in three orthogonal directions; normal to the surface of the section,
transverse to the section length and along the section length, respectively. This coordinate system is
defined in figure 4. Because of the influence that residual stresses can have on structural behavior it is
important to measure the magnitude and distribution of residual stresses in cold formed sections and there
are two distinct types of techniques which have been employed to date: destructive and nondestructive.
700
Rachel B. Cruise and Anna M. Paradowska
x
x
z
ız,m
İz,m
Membrane residual stress, ız,m +
x
x İz,b
z
ız,b
 İz,b Ȝ
Bending residual stress, ız,b =
x
d ș
Combined residual stress, ız ız
Using the sectioning technique, combined with an electrolytic technique to remove layers of section
material, longitudinal residual stresses, ız were measured at different depths through the thickness of a
cold formed carbon steel box section in [7]. From these measurements it was observed that the
longitudinal residual stress distribution through the thickness of the section was not a linear variation as
commonly assumed [14] and therefore that the membrane and bending residual stresses could not be the
only residual stresses to exist in the section. Through this study [7] a third residual stress component
termed the layering residual stress was identified, which is not released and therefore not measured during
sectioning, since it has no resultant axial force or moment. This unmeasured layering residual stress is
important to quantify to determine peak residual stresses in the section material. Analytical models that
map the coiling, uncoiling and cold forming of stainless steel and carbon steel sheet material into
structural sections [8][9] have also determined that the variation of longitudinal residual stresses, ız
through the thickness of a cold formed section does not conform to the assumed linear model.
Longitudinal residual stresses, ız in roll formed stainless steel sections were determined through the
sectioning technique in [15] where both a linear and a rectangular block through thickness distribution
were assumed to calculate the longitudinal bending residual stresses, ız,b. It was observed that for
sectioned material with a rectangular cross section, there was a difference of two thirds in the magnitude
of the bending stresses between the two assumed distributions. This study showed that assuming a linear
through thickness residual stress distribution can cause large errors in determining the longitudinal
residual stresses.
Furthermore residual strains released normal to the surface of the section, İx and transverse to the
length of the section, İy are not easily quantified in the same location as the longitudinal strain, İz through
the use of the sectioning technique. However all strain components contribute to the normal, transverse
and longitudinal residual stresses (ıx, ıy and ız respectively) through the three dimensional definition of
Hookes’ Law, given in equations 13. Where E is the Young’s modulus and Ȟ is Possion’s ratio.
σx =
E
(1 +ν )(1 − 2ν )
[ε x (1 −ν ) +ν (ε y + ε z )] (1)
σy =
E
(1 +ν )(1 − 2ν )
[ε (1 −ν ) +ν (ε
y x ]
+εz) (2)
σz =
E
[ (
ε (1 − ν ) + ν ε x + ε y
(1 + ν )(1 − 2ν ) z
)] (3)
701
Rachel B. Cruise and Anna M. Paradowska
Using sectioning to measure residual stresses has the disadvantage that the complete residual stress
distribution is not fully released and therefore not measured and the strains in the normal, transverse and
longitudinal directions are hard to measure simultaneously to correctly determine the corresponding
residual stresses. Also the method of removing material can affect the residual stress pattern through
plastic deformation and heating that might occur during mechanical interventions.
8 NEUTRON DIFFRACTION
Axonometric Plan view:
view:
North The south x The north
collimator collimator y collimator
South collects collects
collimator neutrons that neutrons that
determine determine
atomic atomic
Neutron spacings in the spacings in
source Apertures normal the transverse
Neutron
Sample direction, x. direction, y.
source
a) Sample positioned to measure atomic spacing in the normal and transverse direction.
Axonometric Plan view:
view: North
The south The north
collimator
collimator x collimator
South z collects
collimator collects
neutrons that neutrons that
determine determine
Sample atomic atomic
Neutron spacings in the spacings in
source Apertures normal the
Neutron
direction, x. longitudinal
source direction, z.
b) Sample positioned to measure atomic spacing in the normal and longitudinal direction.
Figure 7: Test setup.
702
Rachel B. Cruise and Anna M. Paradowska
The nondestructive technique of neutron diffraction uses the interaction of a neutron beam and the
specimen’s atomic structure, as governed by Bragg’s law, to measure the spacing between atomic planes,
d. Bragg’s law is given in equation 4 and the variables Ȝ, d and ș, are defined in figure 6.
nλ = 2d sin θ (4)
When n in equation 4 is an integer the diffracted neutrons interfere constructively. This causes the
collimators either side of the test sample to detect a peak of neutrons at atomic spacings characteristic of
the arrangement of atomic planes in the microstructure under observation. Just such neutron diffraction
measurements were performed during a three day pilot study using the ENGINX instrument at ISIS.
Through thickness residual stress distributions were measured in four locations AD around a roll formed
austenitic stainless steel grade 1.4301 box section of dimensions 100×50×6 mm, as shown in figure 4. At
each location, AD, seven diffraction measurements were made over a 2×2×2mm3 gauge volume at
intervals of 0.5 mm through the thickness of the section. Measurements were made with the test sample
held in two orientations in order to measure atomic spacings in three orthogonal directions, as illustrated
in figure 7. The atomic spacings measured at different locations in the test sample were compared with a
stress free atomic spacing measured in 2×2×2 mm3 cubes, cut using an Electric Discharge Machine from
locations AD in the same cross section. The atomic strains held in the roll formed cross section were thus
determined and converted to residual stresses using equations 13 and material data obtained from tensile
coupon tests performed on material cut from locations AD in the test sample.
9 RESULTS
Figure 8 shows a longitudinal through thickness residual stress distribution taken from location D
with vertical error bars and horizontal lines indicating the overlapping of each measurement.
The atomic spacings measured at locations AD in the test sample were characteristic of an austenitic
stainless steel microstructure with no detection of the presence of a martensitic microstructure. This could
be because the martensitic microstructure created during cold working is small or it could be very
localized and its presence was not detected due to the use of a large gauge volume.
10 CONCLUSIONS
Despite the significant overlap of the through thickness measurements this pilot study successfully
demonstrates that neutron diffraction can be used to obtain through thickness residual stress data to a
good accuracy and that the variation of longitudinal residual stresses through the thickness of roll formed
stainless steel sections is clearly not linear as conventionally assumed. Further measurements using a
smaller gauge volume are planned to allow the measurement of through thickness residual stress
703
Rachel B. Cruise and Anna M. Paradowska
distributions to a higher resolution and thereby reduce any smoothing effect of the overlapping
measurements and also increase the chance of detecting any martensite present.
ACKNOWLEDGEMENTS
The author would like to acknowledge the hard work and technical expertise of Dr. Shu Yan Zhang
and to thank the Science and Technology Facilities Council for their financial and technical support.
REFERENCES
[1] Way A.G.J., PopoOla S.O., Biddle A.R. and Lawson R.M., Durability of Light Steel Framing in
Residential Building, Second Edition, SCI, Silwood Park, Report no: P262, 2009.
[2] National Audit Office, Using modern methods of construction to build homes more quickly and
efficiently, 2005.
[3] Karren K.W., “Corner properties of coldformed steel shapes”, Journal of the Structural Division,
ASCE, 93(ST1), 401432, 1967.
[4] Cruise R.B. and Gardner L., “Strength enhancements induced during cold forming of stainless steel
sections”, Journal of Constructional Steel Research, 64 (11), 13101316, 2008.
[5] EN 199314, Eurocode 3: Design of steel structures  Part 1.4: General rules. Supplementary
Rules for Stainless steels, European standard, CEN, 2006.
[6] EN 199313, Eurocode 3: Design of steel structures  Part 1.3: General rules. Supplementary
rules for coldformed members and sheeting, European standard, CEN, 2006.
[7] Key P.W. and Hancock G.J., “A theoretical investigation of the column behaviour of coldformed
square hollow sections”, ThinWalled structures, 16(14), 3164, 1993.
[8] Moen C.D., Igusa T. and Schafer B.W., “Prediction of residual stresses and strain in coldformed
steel members”, ThinWalled structures, 46(11), 12741289, 2008.
[9] Quach W.M., Residual stresses in cold formed steel sections and their effect on column behavior,
PhD Thesis, Dept. of Civil and Structural Engineering, Hong Kong Polytechnic University, 2005.
[10] Gardner L., Cruise R.B., Sok C.P., Krishnan K. and Ministro J., “Life cycle costing of metallic
structures”, Proceedings of the Institution of Civil Engineers, Engineering Sustainability,
160(ES4), 167177, 2007.
[11] Edwards L. and Endean M. (eds.), Manufacturing with Materials, ButterworthHeinemann Ltd.,
1999.
[12] Spencer K., Embury J.D., Conlon K.T., Véron M. and Bréchet Y., “Strengthening via the formation
of straininduced martensite in stainless steels”, Materials Science and Engineering, A 387389,
873881, 2004.
[13] Withers P.J. and Bhadeshia H.K.D.H., “Residual stress: Part 2  Nature and Origins”, Materials
Science and Technology, 17:4, 366375, 2001.
[14] Schafer B.W. and Peköz T., “Computational modelling of coldformed steel”, Journal of
Constructional Steel Research, 47(3), 193210, 1998.
[15] Cruise R.B. and Gardner L., “Residual stress analysis of structural stainless steel sections”, Journal
of Constructional Steel Research, 64 (3), 352366, 2008.
[16] Jandera M., Gardner L. and Machacek J., “Residual stresses in cold rolled stainless steel hollow
sections”, Journal of Constructional Steel Research. 64(11), 12551263, 2008.
[17] Webster G.A. and Wimpory R.C., “Nondestructive measurement of residual stress by neutron
diffraction”, Journal of Material Processing Technology, 117(3), 395399, 2001.
704
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
László G. Vigh*
Keywords: patch load, steel and aluminium Ibeams, curved flangetoweb connection, nonlinear
numerical simulation, interaction of bending and transverse load resistance, Eurocode.
Abstract. Current design methods for calculating transverse load resistance of web plates neglect the
effect of the flangetoweb connection. This simplification is on the conservative side and in deep plate
girders its influence is practically negligible. However, in case of hotrolled or extruded profiles, the
fillet corner shaping through its geometry, rigidity and strength may highly increase the patch load
resistance. Using nonlinear numerical analysis, the author completed a parametric study on simply
supported girders subjected to transverse load in order to study a) the effect of the connection and b) the
capacity in interaction of transverse load and bending. The study clearly confirms the beneficial
influence of the curved corners. To take this effect into account the author proposes a modification in the
Eurocode formulation. The results of the simulation prove the validity of the proposed method.
1 INTRODUCTION
It is well known that transverse (or patch) load resistance of plate girders is influenced by the load
case (patch load, opposite patch load, end patch load), the web and flange material properties (i.e. yield
limits), the dimensions (a, hw) and thickness (tw) of the unstiffened web panel, and the flange dimensions
(bf, tf). Additionally, the loaded length (ss) has major importance in the resistance, too. These details are
typically considered in the design method as well, just like in current formulations of corresponding
Eurocodes, such as Eurocode 3 Part 15 for steel plated structures (EC3 [1]) and Eurocode 9 Part 11 for
aluminium structures (EC9 [2]).
However, effect of the connection between the flange and web, thus the effect of the fillet curved
corners is fully neglected. Even though this simplification is on the conservative side and reasonably
accurate for deep slender girders, this type of connection shaping may highly increase the resistance, as it
is confirmed in [3]. Firstly, – similarly to the calculation of local plate buckling under compression – one
may claim to consider the fillet reducing the web height. Secondly, it widens the effective loaded length
of the web. Thirdly, it may highly influence the plastic hinge capacity of the flanges; that is the base of
the mechanism solution model of patch loading originally recommended by Roberts et al [4].
Reviewing the development of the actual code formula, the reason why this effect is out of
consideration can be found. (Note that EC9 applies the same method as given in the steel standard EC3.)
According to Lagerqvist et al [5], the calibration of the semiempirical design method was completed on
the basis of 388 test specimens made of steel, including 358 welded girders, 11 European and 19
American rolled beams. Most of these tested girders come with high, slender web and only few cases
represent rolled/extruded profiles with stocky webs and relatively large curved corners. Consequently,
the calibration is directly valid for slender webs only. Needless to say, welded connection that has
smaller extent than practically applied curved corners of rolled profiles results in much smaller influence
on the resistance. It can be also stated that the higher and more slender the web is, the less the effect of
705
László G. Vigh
the edge boundary condition is. Note that this effect can be much higher in case of aluminium, because
extruded profile usually comes with larger radius due to fabrication and material reasons.
The author completed a parametric numerical study on various steel and aluminium Ibeams
configurations and the corner effect is quantified. Based on the results, the author proposes a simple
modification in the Eurocode method to account for this detail and its beneficial effect.
where Wpl and Wel are the elastic and plastic section modulus, respectively; αM,1 stands for the
correction factor to account for the plastic overstrength including strain hardening; fy and fo are the
characteristic yield strength and the proof strength, respectively; while γM is the partial safety factor.
Both EC3 and EC9 prescribe exactly the same mechanismsolution based procedure for the
transverse load resistance calculation. As per EC9, for simple patch load case the method follows:
ª 2
t w3 § h · º t3
critical load: Fcr = 0.9k F E = 0.9 «6 + 2¨ w ¸ » E w (2)
hw «¬ © a ¹ »¼ hw
( )
dimensionless parameters: m1 = f of b f / ( f ow t w ); m2 = 0.02 hw / t f ( )
2
(3/a,b)
(
effective loaded length: l y = s s + 2t f 1 + m1 + m2 ≤ a ) (4)
l y t w f ow
slenderness: λF = (5)
Fcr
reduction factor: χ F = 0.5 / λ F ≤ 1.0 (6)
f ow
transverse load resistance: FRd = χ F l y (7)
γ M1
where E is Young’s modulus; kF is the buckling coefficient; a is the length of the unstiffened web
panel (Figure 1); hw and tw are the web height and thickness; bf and tf are the flange width and thickness;
ss is the stiff bearing length; fof and fow are the proof strength of the flange and the web, respectively.
The interaction of design bending moment MEd and transverse loading FEd shall be checked
through the following interaction formula:
FEd M
+ 0.8 Ed ≤ 1.4 (8)
FRd M c , Rd
For simplicity, this study does not deal with the complex interaction of shear, bending and transverse
load. (Note that influence of shear load on the patch load resistance is currently not covered by the basic
method of Eurocode. A useful method is discussed in [6].) Additionally note that the following results are
corresponding to the simple patch load case; the author did not deal with opposite and end patch loading.
706
László G. Vigh
3.1 Programme
The author completed a parametric study in order to quantify the effect of the curved corners on
patch load resistance of rolled steel and extruded aluminium girders. Simply supported girders (with I
shape crosssection shown in Figure 1) subjected to transverse concentrated load at midspan is
considered. Varying parameters are the section geometry, the span a (= length of unstiffened web panel),
the loaded length ss and the radius r of the curved corner. The parametric study programme is
summarized in Table 1. One series of analysis was carried out assuming HEA sections made of steel
grade S235 and one with modified HEA sections made of a specific AlMgSi alloy. In the aluminium
case, profile modification – namely, change of flange width – aimed to obtain ductile sections.
Altogether, the analysis series include more than 150 cases.
The crosssections are ductile (Class 1) in each case. Varying the span permits of analysing cases of
dominant bending failure, dominant web crippling or their interaction. Three different loaded lengths are
investigated: 0 mm, 50 mm and 200 mm, respectively. The basic radius for the different sections are
12 mm, 18 mm and 27 mm, respectively; these values are multiplied by 0, 1, 1.5 and 2.
For simplification, interaction with shear is not discussed here: cases where the influence of shear on
design bending resistance is larger than 5% are excluded.
bf
tf
h
tw
a/2
707
Table 1: Programme and results
"HEA 100" / "HEA 10075" "HEA 200" / "HEA 200100" "HEA 300" / "HEA 300140"
FRd,FEM FRd,FEM FRd,FEM
h bf tf tw r ss a h bf tf tw r ss a h bf tf tw r ss a
Steel Alu Steel Alu Steel Alu
[mm] [k N] [mm] [k N] [mm] [k N]
150 112.9 101.5 500 204.0 158.8 1000 371.8 277.3
300 106.1 93.9 0 1200 180.0 131.6 0 2000 334.5 233.4
0
500 95.2 83.5 2000 149.0 103.6 3000 293.0 195.3
750 81.5 71.9 500 256.1 204.8 1000 441.0 333.2
300 139.3 129.9 0 50 1200 231.0 173.3 0 50 2000 402.0 286.8
0
50 500 123.8 118.6 2000 188.0 122.8 3000 350.6 230.2
750 104.1 91.7 500 391.0 346.3 1000 645.0 
300 211.0 209.4 200 1200 325.0 230.7 200 2000 598.8 386.9
200 500 147.0 154.8 2000 212.0 135.5 3000 435.0 262.6
750 124.0 112.4 200 500 245.0 200.7 300 1000 460.1 355.5
150 135.8 130.3 (St) 0 1200 212.0 163.8 (St) 0 2000 404.2 300.0
100
300 125.0 116.7 2000 173.8 124.2 3000 348.1 244.3
(St) 0 190 10 6.5 290 14 8.5
500 109.7 101.9 500 303.4 251.6 1000 537.9 419.0
96 8 5 750 93.0 84.4 100 18 50 1200 269.6 207.9 140 27 50 2000 482.0 355.7
150 189.9 184.2 (Al) 2000 205.2 139.7 (Al) 3000 409.8 270.9
75
(Al) 12 300 155.0 154.7 500 417.0 401.3 1000 691.0 622.3
50
500 134.0 135.3 200 1200 329.0 256.3 200 2000 641.0 441.4
708
750 110.9 99.5 2000 222.0 150.7 3000 463.0 293.9
300 223.0 246.8 500 282.2 240.1 1000 535.7 434.0
László G. Vigh
200 500 148.0 168.2 27 0 1200 243.0 194.5 40.5 0 2000 471.5 362.4
750 127.0 121.9 2000 197.0 142.4 3000 404.1 286.5
500 123.0 116.8 500 331.1 288.4 1000 626.5 529.7
18 0
750 103.2 94.6 36 0 1200 283.7 231.7 54 0 2000 558.0 437.4
150 187.2 190.7 2000 221.0 164.8 3000 465.0 338.1
300 159.0 163.2
24 0
500 137.7 134.7 h – total section height; bf – flange width; tf– flange thickness; tw – web thickness; r – radius of corner;
750 114.2 106.9 ss – loaded length; a – span (= length of unstiffened web panel); FRd,FEM – FEM load capacity
Table 2: Comparison to r = 0.
Steel Alu
r=
% avg % % avg %
r 5  24 16 8  29 20
1.5 r 27  44 36 32  57 46
2r 40  67 54 4991 74
László G. Vigh
250
Stress (MPa)
shell elements 200 E = 210 GPa; Et = 21 MPa
t2 fy = 235 MPa
with varying
150
tw thickness
100 AlMgSi
S235
50
axis of shell
elements 0
0 0.05 0.1 0.15
Note: these are not body elements Strain ()
a) approximation of curved corners b) material models
Figure 2: Modeling details.
3.3 Discussion of results
3.3.1 Results
The load capacities obtained by the analyses are tabulated in Table 1. Utilization factor for bending
moment is calculated as the ratio of the ultimate load FRD (or MRd) computed from the numerical analysis
to the design bending moment resistance Mc,Rd calculated in accordance with Eurocodes. Utilization
factor for transverse load resistance (FRD to Fpatch,RD) is similarly determined. The interaction results are
illustrated by the help of these parameters in Figure 3 where the EC interaction curve is also plotted.
3.3.2 Reliability of the results
The applied nonlinear numerical analysis is an alternative design procedure allowed for by both
Eurocodes. In Figure 3/a, the interaction points corresponding to the reference case r = 0 lie in the
vicinity of the standard interaction curve, which in general confirms the validity of the results. On
average the deviation is within 10%. However, in many cases the numerical simulation gives
conservative result compared to the standard procedure. Different interpretations may be given for this
observation:
1) On the one hand, this may indicate that the numerical model is conservative, i.e. especially the
applied geometrical imperfection, or the way of joint discretization, etc. is conservative.
2) Reference comparison perhaps should be done to cases of normal radius, as the calibrated design
method may indirectly include some connection effect.
3) The fact that zero loaded length cannot be kept in experimental environment queries the
reliability of the standard method: it may overestimate the capacity for cases ss = 0. When
excluding these cases, only 5 of the points fall below the standard interaction curve.
As a consequence, further study is required to accurately evaluate the reliability of the alternative
design methods.
Despite the discussed uncertainties, it can be stated that the numerical results are in accordance with
the standardized method and the analysis with different radius and loaded length gives a solid base for the
following comparative study.
3.3.3 Effect of curved corners
The results confirm that the curved corner may highly influence the transverse load resistance as well
as the resistance in interaction with bending. Compared to the reference cases r = 0, even the
consideration of normal radius leads to notable increase in the capacity, as Figure 3/a,b and Table 2
prove. When using double radius, up to 67% and 91% increase can be achieved in case of steel and
aluminium, respectively.
709
László G. Vigh
Load vs. deflection curves and deformed shapes of Figure 4 well demonstrate the quantitative and
qualitative change in capacity and in nature of behaviour. For example, compare postultimate behaviour
of cases ss = 200 mm in Figure 4/a: in case of no radius web crippling dominates as indicated by the
sudden drop in the postpeak range, while the existence of curved corner leads to governing bending
failure. In the latter case, due to the ductile (Class 1) section, long yield plateau can develop, followed by
the capacity drop due to instability at the very end of the curve.
The larger influence in case of aluminium alloy can be explained by two reasons:
1) The selected aluminium profiles are more sensitive to web crippling than the steel ones, which is
also reflected by the reduction factor χF calculated in accordance with EC: it ranges within
0.9~1.0 for the steel and within 0.53~1.0 for the aluminium cases, respectively.
2) Strain hardening is considered in the aluminium calculations. Note that the manual calculation
also accounts for the strain hardening in the bending moment resistance formulation, but not in
the transverse load resistance.
The results thus promise that improved capacity values in the steel cases can possibly be achieved by
implementing advanced (more accurate and realistic) nonlinear simulation.
2 EC 2.5
r= 1 r
r= 0
1.8
r= 1 r r = 1.5 r
1.6 2
r = 1.5 r
r= 2 r
1.4 r= 2 r
FRd / Fpatch,Rd
FRd / FRd,r=0
1.2 1.5
0.8 1
0.6
0.4 0.5
0.2
0 0
0 0.5 1 1.5 0 0.5 1 1.5
MRd / Mc,Rd MRd / Mc,Rd
700
600
500
Load (kN)
400 ss = 0;
300 ss = 0; r = 0
r = 1.5r
ss = 0; r = 1
200 ss = 0; r = 1.5
ss = 0; r = 2
100 ss = 200; r = 0
ss = 200; r = 1
0
0 5 10 15
ss = 200;
r=0
Deflection (mm)
a) load vs. deflection curves b) deformed web at ultimate load
Figure 4: Typical simulation results – steel, HEA300, a = 2 m.
710
László G. Vigh
bf
tf
hw R
tw
a) effective web height hw
n.a
tf n.a tf’
Mpl,2
Mpl,1
Mpl,1 = Mpl,2
b1) equal section modulus
tf tf’’
A1 A2
A1 = A2
b2) equal area method
b) effective flange thickness tf or tf’’ ’
2 EC
1.8 radius neglected
eq. area
1.6
eq. section mod.
1.4
FRd / Fpatch,Rd
1.2
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5
MRd / Mc,Rd
711
László G. Vigh
On the other hand, the plastic hinge of the flanges occurring in the ultimate stage (plastic hinge
mechanism) can extend to the curved corner area. Thus, the curved corner can be considered as part of
the flange. Since the resistance against transverse loading is dominated by the flange plastic hinge, one
has to calculate the plastic bending capacity of a fictive section including the flange and the
accompanying corner area. Without rebuilding the existing design formula, this can be easily achieved by
introducing an effective flange thickness providing the same local plastic capacity (Figure 5/b1). This is
referred as equal section modulus method hereafter. As a simplification, the effective thickness can be
conservatively calculated by simply smearing the curved corner area to the flange (Figure 5/b2, referred
as equal area method). This latter method gives smaller effective thickness than the previous, more
accurate one; consequently, it is always on the safe side.
Introducing these two modifications into the design method, the interaction relation shown in
Figure 6 is obtained for the studied configurations. Regardless to the mentioned uncertainties (knifeedge
load, etc.), it is concluded that the proposed modification gives more accurate evaluation of the patch
load resistance.
5 CONCLUSIONS
Parametric study is completed on simply supported girders made of rolled steel or extruded
aluminium profiles, subjected to transverse load. Based on the results, the following conclusions are
found:
Influence of the curvedcorner webtoflange joint on the transverse load resistance can be significant
in case of stocky webs; the increase in capacity may reach 6090%.
To take this beneficial effect into account, the author proposed a simple modification in the current
Eurocode design method. The modified procedure utilizes the clear web height and effective flange
thickness; thus, more accurately representing the actual connection rigidity and plastic flange strength.
The proposed procedure well estimates the transverse load – bending interaction capacity computed
by nonlinear numerical simulation.
Further study is needed on the relation of the basic standardized procedure and the numerical
simulation with respect to reliability. The research shall include study on the role of imperfection, knife
edge loading case, material modelling.
The method should be validated to other load application cases, as well.
Interaction of transverse load, bending and shear is additionally subject to further research.
REFERENCES
[1] EN 199315:2005 Eurocode 3 – Design of steel structures – Part 15: Plated structural elements,
Final Draft, CEN, Brussels, 2005.
[2] EN 199911:2007 Eurocode 9 – Design of aluminium structures – Part 11: General structural
rules, CEN, Brussels, 2007.
[3] Vigh, L.G., Virtual and real test based analysis and design of nonconventional thinwalled metal
structures, PhD dissertation, Budapest University of Technology and Economics, 2006.
[4] Roberts, T.M., “Slender plate girders subjected to edge loading”, Proc. Inst. Civil Engineers, Part
2, Vol. 71, 805819, Sept. 1981.
[5] Lagerqvist, O. and Johansson, B., “Resistance of Igirders to Concentrated Loads”, Journal of
Construction Steel Research, 39 (2), 87119, 1996.
[6] Kövesdi, B., Kuhlmann, U., Dunai, L., “Combined shear and patch loading of girders with
corrugated webs”, Periodica Polytechnica Civil Engineering, 2010. (submitted for publication)
[7] ANSYS Structural Analysis Guide, Online Documentation ANSYS Inc., 2005.
712
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. The purpose of this paper is to present solutions for the elastic and inelastic lateraltorsional
buckling of steel webtapered beamcolumns using two different computational procedures and compare
these solutions with the AISC Specification for tapered members.
1 INTRODUCTION
Webtapered members are structural members commonly used in the typical onestory preengineered
building. Appreciable savings in materials and in the cost of structural framing can be assumed by the
use of elements having a tapering depth or flanges.
In the United States of America, the last specification that addressed tapered members is the 1999
American Institute of Steel Construction Specification [1] for web tapered members which was based on
a study performed in 1966. The contributors to the study were the Column Research Council, presently
known as the Structural Stability Research Council, and the Welding Research Council, under the
technical guidance of Lee et al. [3] at the University of New York at Buffalo. The general design
approach used in the 1999 Specification is to apply modification factors to convert the tapered members
into appropriately proportioned prismatic members so that the prismatic AISC equations may be applied.
From the practitioner’s point of view, the 1999 AISC design equations for tapered members represent the
use of existing basic formulas for prismatic members altered with the use of an additional factor.
Furthermore, the additional factor will give the designer an inherent feeling for the increase in strength
over a prismatic section.
At the same time, the “easy to use” 1999 AISC Specification is restricted to doubly symmetric I
shaped sections. The reason for this limitation was the inability to uncouple the torsional and flexural
deformations due to varying locations of the shear center for singlysymmetric sections during Lee’s
study. The development was also limited to members with small tapering angles. According to Lee et al.
[5], Boley showed that the methods used by Lee and his colleagues to compute normal stresses are
reasonably accurate as long as the tapering angle is less than 15 degrees. For practical considerations, the
limiting tapering ratio has been further restricted to 6. Moreover, the development is limited to members
with flanges of an equal and constant area with webs that are not slender. However, what is of interest is
that the current practice in the lowrise metal building industry is the use of flanges of unequal area and
slender webs. Therefore, the 1999 AISC Specification does not appear to provide equations for web
tapered Ishaped beam geometries of proportions that are consistent with what has been the industry
standard for metal buildings.
Jimenez et al.[5] and other researchers have performed new studies on the topic of inelastic stability
of tapered members and have shown that the 1999 AISC equations predict unconservative results when
determining the lateraltorsional buckling strength of tapered beams and beamcolumns for certain
713
Gabriel A. Jiménez
slenderness values of typical tapered members. The current AISC Specification [2] does not explicitly
define the use of the AISC provisions for tapered members. In 2006 White et al. [6] performed a
prototype study on how to use the current AISC Specification to tapered members. The findings of their
prototype study appeared to generate reasonable solutions; however additional verifications with other
versions of the code as well as experimental results are needed.
The general behavior of a typical beamcolumn is illustrated in Figure 1, where the relationship
between the applied endmoment Mo and the resulting endslope θ is shown for a wideflange member
bent about its strongaxis, in which the length as well as the axial force P is assumed to remain constant
as the moment Mo is increased from zero to its maximum value and past the maximum moment into the
unloading zone.
The optimum performance of the beamcolumn is reached if failure is due to excessive bending in the
plane of the applied moment, and this behavior is represented by the upper branch of the curve in Figure
1. The corresponding maximum moment is Mo1max. If no lateral bracing is provided, failure will be due to
lateraltorsional buckling and the resulting moment is Mo2max represented by the lower branch of the
curve in Figure 1. The additional incremental moment represented by Mo2max beyond Mocr is small, and,
therefore, the bifurcation point is considered to reasonably determine the buckling limit to the beam
column. The work described in this paper deals with the determination of the value of Mocr for web
tapered beamcolumns.
d 2v ª zº
B x (z ) + Pv − Mo« ρ + (1 − ρ ) » = 0 (1a)
dz 2
¬ L ¼
d 2u ª z½ º
By + Pu − β «Mo ® ρ + (1 − ρ ) ¾ − P y o (z )» = 0 (1b)
¬ ¯ L¿
2
dz ¼
714
Gabriel A. Jiménez
(1c)
The beamcolumn prescribed by the above differential equations is shown in Figure 2. It is subjected to
end bending moments Mo at z = L and ρMo at z = 0, where “z” is the coordinate axis along the
undeformed centroidal axis and “L” is the length of the member. The coefficient “ρ” is the ratio of the
end moments. The deformations of the shear center are: “u” in the xdirection, “v” in the ydirection and
the crosssection twists about the shear center an angle “β”. In Figure the smaller end will be denoted as
end A and the larger end as end B.
plane of symmetry; K( z) = ³ σs d A
2
where: σ = is the stress on any crosssectional element dA
A
(positive in compression) and “s” is the distance of element dA from the shear center. These coefficients
vary with respect to the coordinate “z” to account for the nonuniform variation of the crosssection
properties along the length of the column. Also, when the beamcolumn is in the inelastic range the
coefficients will vary with the different patterns of the yielding.
715
Gabriel A. Jiménez
[ ] [
ui −1[By(i )] + ui Ph 2 − 2By(i ) + ui+1[By(i )] + βi λ (i )h 2 = 0 ] (2b)
[ ] ª 2Mo(1 − ρ )h 2 º
ui−1 − λ (i )h 2 + ui « [ ]
» + ui +1 λ (i )h + βi−2 [− C w (i )] +
2
¬« n ¼»
ª § K (i ) ·¸º ª § K (i ) ·¸½°º
°
βi −1 «2C w (i ) + CT (i )h 2 ¨1 − » + βi +1 «− ®2C w (i ) + CT (i )h 2 ¨1 − »+
¸¾
(2c)
«¬ ¨
© CT (i ) ¹»
¸ « °̄ © CT (i ) ¹°¿»¼
¨
¼ ¬
βi+ 2 [C w (i )] = 0
ª iº
where : λ (i ) = Py o (i ) − M (i ), M (i ) = Mo « ρ + (1 − ρ ) »
¬ n¼
The ends of the beamcolumn are allowed to rotate, the end sections are free to warp, and the ends of the
member are not permitted to twist or to translate. These boundary conditions can be written as follows:
uo = 0, un = 0, β1 = β1 , βn+1 =  βn1 ,βo = 0,βn = 0.
This leads to a set of simultaneous algebraic equations in the lateral displacement u and the rotation β at a
number of discrete points spaced at h = L/n, in which n is an odd number to which the beamcolumn is
u ½
divided. This set of simultaneous equations may be written in matrix form: [ A] ® ¾ = 0 . In this equation
¯β ¿
the matrix [A] is a set of the coefficients Aij representing combinations of the crosssection properties (By,
CT(z), Cw(z), yo(z), and K( z) ), the load parameters ( P and Mo) and the length of the member (L). In
order to compute the stiffness of a crosssection it is necessary to know how much of the section is plastic
and how much of the section is elastic, and where the corresponding regions are located on the cross
section. The nondimensionalized M/My, φ/φy, P/Py, relationships about the strongaxis for an Ishape
section have been determined by Jimenez [4]. Figure 4 shows these relationships for the following cases
of yielding:
In outline form, the steps that are used in computing the critical moment Mocr for steel webtapered beam
columns are as follows:
716
Gabriel A. Jiménez
This process is repeated for different load levels until a zero value for the determinant is found.
To create the finite element model using ANSYS, a commercially available finite element program,
several steps had to be performed including element selection, laying out the mesh and determining
boundary conditions. The finite element mesh is comprised of BEAM188 elements. BEAM188 elements
are suitable for analyzing slender to moderately stubby/thick structures. This element is based on
Timoshenko beam theory. Shear deformation effects are included. The BEAM188 is a quadratic beam
element in 3D. This element is wellsuited for linear, large rotation, and/or large strain nonlinear
applications. Furthermore, the provided stress stiffness terms enable the elements to analyze flexural,
lateral, and torsional stability. The crosssection associated with the element may be linearly tapered.
Elasticity and plasticity models are supported.
0.8 M
0.7
0.6
(M/Mp) A
0.5 Finite
Difference M
0.4 Ansys
0.3 p
0.2
0.1
0
0 20 40 60 80 100 120
(L/rx)A
717
Gabriel A. Jiménez
It appears that the ANSYS solution produces more conservative results for the slenderness ratios between
40 to 90. This behavior is due to the gradual yielding in the ANSYS model versus the four defined
yielded patterns from Figure 4 utilized in the FD method. Figure 6 corresponds to the overall buckled
shape of the tapered beamcolumn as depicted by ANSYS. Yielding of both flanges at the smaller end is
evident. A closeup look of the smaller end is depicted in Figure 7.
Figure 7: Close up view of yielded flanges/web at the smaller end of a tapered beamcolumn.
Comparisons were made between the 1999AISCLRFD code and this study. Figure 8 illustrates a
typical case of a tapered beamcolumn subjected to compressive axial load and end moments for lateral
torsional buckling (LTB). The beamcolumn problem is treated in the 1999 Specification in the form of
an interaction equation. It can be seen that for values of M/Mp greater than about 0.4 the predictions of
the interaction equation are unsafe. The unconservative results predicted by the use of the beam equation
in the AISC Specification are typical for different tapering ratios with unsafe discrepancies up to 25 %
between the advanced analysis and the Specification.
718
Gabriel A. Jiménez
6 CONCLUSION
This paper presents studies for the outofplane behavior of tapered beamcolumns using ANSYS and
the Finite Difference method. It was shown that the ANSYS solution produces slightly more conservative
results due to the progressive yielding of the flanges and web. It was found that for medium to short
beams and beamcolumns, the 1999 Specification [1] predicts strengths on the nonconservative side,
with maximum discrepancies of about 25% between advanced analysis approaches and those given the
specifications. Jimenez and Galambos [5] suggested an alternate set of equations to better predict the
strength capacity of tapered beams and beamcolumns.
Furthermore, additional studies are needed to evaluate the use of the 2005 AISC provisions [2] with
previous specifications for tapered members.
REFERENCES
[1] American Institute of Steel Construction, Load and resistance factor design specification for
structural steel buildings, AISC, 1999
[2] American Institute of Steel Construction, Load and resistance factor design specification for
structural steel buildings, AISC, 2005
[3] Lee G. Morrell M. and Ketter R., “Design of tapered members”, Welding Research Council
Bulletin, 173, 132, 1972.
[4] Jimenez G., “Inelastic stability of tapered structural members”, Doctoral Dissertation, Minneapolis,
Minnesota, 1998.
[5] Jimenez G., and Galambos T.V., “Inelastic lateraltorsional buckling of tapered beamcolumns”,
Proceedings, Annual Technical Session, Structural Stability Research Council, 1998.
[6] White D., and Kim, Y. D., “A prototype application of the AISC (2005) Stability and Design
Provisions to metal building structural systems,” Metal Building Manufacturers Association, 2006.
719
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Keywords: Lateral torsional buckling, local hinges, firstyield design, warping torsion of open thin
walled members.
Abstract. Continuous steel beams used for girders, purlins, crane girders, etc. are subjected to lateral
torsion buckling. In many cases construction joints of those beams are carried out as moment hinges
rather than as rigid connections. The study is concerned with the effect of these hinges on the lateral
stability behaviour and their effect on the design. Based on a numerical approach critical lateral
buckling loads are calculated for different hinge locations and moment distributions. Finally load
capacities obtained from equivalent slenderness approach are compared to those from firstyield
criterion with internal forces from theory 2nd order analysis.
1 INTRODUCTION
Figure 1 shows common type of construction joints for continuous Ibeams which will be assembled
on the construction site. These joints are able to transfer shear forces but only small fractions of the
bending, torsion and warping moments associated with 3dimensional loading and nonuniform torsion.
Therefore in the structural analysis it is assumed for the hinge that lateral displacements v and w are
compatible and
My Mz Mx MZ 0 (1)
721
P. Osterrieder et al.
While moment hinges may be easily considered in any commercial computer program for inplane
stress analysis the allowance of local static boundary conditions in the eigenvalue calculation required for
lateral buckling design is rather unusual. To account for local boundary conditions within a displacement
based finite element formulation basically two alternatives are available. Applying as usual static
condensation to local element stiffness matrix leads to coupled elastic and geometric element matrices
which are not suitable for numerical eigenvalue solution applying vector iteration. Introducing instead
double nodes and incorporating coupling conditions during assembly of the system stiffness results in
uncoupled elastic and geometric system stiffness matrices. The latter approach has been applied to a FE
program for nonlinear analysis of 3D beams with open thinwalled cross sections [1]. Bifurcation loads
from this numerical approach are applied to check for lateral buckling according to Eurocode 3 [2,
paragraph 6.3.2.2]. Design loads from this approach are compared to ultimate loads obtained by an elastic
theory 2nd order analysis considering geometric outofplane imperfections according to [2, paragraph
5.3.2].
L = 10 m, IPE360 section
q applied at centroid
Critical loads for a cantilever beams which is equivalent to the structural system in figure 2 with a
span of 5m are calculated from analytical equations in [3, eqns.(9.14), (9.17)] as qcr = 57,2 kN ( ș'x 0 at
both ends) and qcr = 27.9 kN ( ș'x z 0 at both ends) with ș'x as the warping displacement. Numerical
results for the cantilever beam obtained from [1] are identical to those given for the entire beam with the
hinge in figure 2.
Specific attention is required, when the joint design in addition cannot transfer shear forces in y
direction at all or only with considerable web deflections. In this case the left and the right side of the
beam is essentially uncoupled and so lateral buckling behaviour of the two substructures.
Figure 3: Lateral torsional buckling mode of mono symmetric beam with moment hinge
722
P. Osterrieder et al.
If the cross section is not doubly symmetric as shown in figure 3, displacements v und w in the
direction of the major y and zaxes are related to the shear centre of the cross section.
Most common qcr is calculated alternatively in engineering practice for a virtually cut out single span
beam only with appropriate geometric and static boundary conditions at the respective supports. The
lateral torsional stability is governed by weakest – the most slender – beam. The interaction between the
single spans depends not only on the major axis bending moments but especially on the boundary
conditions for outofplane bending (v, șz) and torsion (șx, ș'x ). While in most real structures lateral
deformations v and torsional rotations șx at the supports will be restrained, boundary conditions for șz and
ș'x depend strongly on the outofplane bending and torsional stiffness of the adjacent spans. To illustrate
the sensitivity of the buckling load qcr with respect to the b.c. critical loads are calculated for the three
single spans in fig. 4 with and without warping restraint at the ends and related to the critical value qcr =
24,94 kN/m for the entire system. Results in table 1 are for L1 = L2 = 8m, L = 10m, Į = 0.3 and IPE330
section. To avoid kinematics in the eigenvalue analyses for span II the rotation at the right end about the
major yaxis is completely restrained.
For critical loads in row 2 it is assumed that the rotation șz about the weak zaxis is completely
restrained for all beams at the intermediate supports. Results in row 3 are calculated for unrestrained
rotations șz in a for the left beam, in b for the right beam and the rotation spring stiffness
EI z ª kNm º
kT z 3 620, 6 « » (2)
L1 ¬ rad ¼
at both supports of the beam with hinge. From the results in table 1 it is obvious, that buckling of the
middle span governs the stability problem and further that the middle span is elastically restrained against
outofplane bending and warping by both outer spans. More general it can be concluded that for most
practical problems of continuous beams the span containing the hinge will be relevant for stability
723
P. Osterrieder et al.
design. Thus the critical load may be approximated considering only the span with the moment hinge.
Fully warping restraint at both supports leads to the upper bound and free warping to the lower bound for
the critical eigenvalue.
From the smallest critical load the critical moment and the dimensionless lateral torsional buckling
slenderness
Wy f y
OLT (3)
M cr
Figure 5: Bending moment distribution for single span beam with hinge
For a single span beam with a moment hinge as specified in eqn. (1) at distance ĮL from the left end
(fig. 5) the moment Mb at the right end depends further on the moment Ma and the uniformly distributed
load q
§ qL2 Ma ·
Mb E¨ ¸ (4)
© 2 D ¹
As long as the hinge is close to the centre of the span the left and the right segments will interact in
lateral torsional buckling depending on the bending moment distribution. When the hinge gets closer to
the right or left bearing, the shorter beam segment will support the longer segment in the outofplane
behaviour. For a hinge very close to one of the bearings the outofplane boundary conditions at the hinge
for the remaining longer segment may be approximated as
v 0 T z z 0 T x z 0 T x' z 0 (5)
Figure 6 shows a comparison of critical loads and associated eigenmodes for a beam with uniformly
distributed load q, a hinge at Į = 0.3 and equivalent bending moment distributions about major yaxis in
724
P. Osterrieder et al.
both systems. The total buckling mode is governed by the weaker beam segment and differs from the
partial mode only slightly with an increase in the critical load for the partial system of 11%.
Outofplane B.C.
qcr = 7.61 kN/m = 1.11·6.87
IPE300
Figure 6: Buckling modes and buckling loads for total and partial structure
Results in table 2 for a beam with a total span of 10 m, IPE300 section, varying hinge location Į, out
ofplane b.c. as shown in fig. 6 and bending moment Ma = 0 confirm this observation more generally.
qcr L3
qcr (6)
EI z GI t
for beams with hinge b.c. given in eqn. (1) and șz = 0 at both supports are plotted over the stiffness
coefficient
S EIZ
k (7)
L GI t
In [3] and similar in [4] it has been shown, that critical loads presented in this dimensionless form are
applicable to almost any beam with hotrolled doubly symmetric Isection. From figs. 7b to 9b it follows
that for beams restrained against warping at the ends the relation between the dimensionless stiffness and
the dimensionless critical load is almost linear. For beams with free warping b.c. at both ends the relation
is found to be highly nonlinear with almost asymptotic behaviour for increasing stiffness.
725
P. Osterrieder et al.
726
P. Osterrieder et al.
§ ·M d 1,0
M pl ,y 1 °
EC 3
M R,d F LT
JM1
¨
¨ ) ) 2 EO 2
¸ pl ,y
¸ JM1
®d 1 ) LT 2 º (8)
0,5 ¬ª1 D LT OLT OLT ,0 EOLT ¼
© LT LT LT ¹ ° O2
¯ LT
The following investigation was carried out for beams with variable span L, moment hinges and
IPE300 section (section class 1 [2, table 5.2], buckling curve b [2, table 6.4] and imperfection coefficient
ĮLT = 0,34 [2, table 6.4]). Further it is conservatively assumed that ȕ = 0,75, Ȗ M1 = 1,1, ȜLT,0 = 0, 4 and
f = 1 . For lateral buckling capacities with Į = 0.5 (fig. 10a) Ma = Ma = qL2/8 and for Į = 0.4 (fig. 10b)
and Ma = qL2/12 the moment Mb = 0.175qL2. Again boundary conditions at both ends are such that
rotations about zaxis are restrained and warping unconstrained.
Figures 10a and 10b show dimensionless load capacities depending on the dimensionless buckling
slenderness O LT . Curves qu,EC3 in fig. 10 are obtained by dividing the ultimate load derived from eqn.
(8) by the load qpl associated with fully plastic moment. They agree with the Ȥ LT distribution. It is to
notice that eqn. (8) does not take into account the effect of shear forces. Therefore the cross section
resistance has to be checked additionally and will restrict the capacity  specifically for small buckling
slendernesses  with respect to provisions in [2, 6.2.8].
4.2 Lateral torsional buckling resistance based on theory 2nd order analysis
Figure 11: Lateral torsional buckling mode of tapered beam with moment hinge
727
P. Osterrieder et al.
For stability design of more general structures like continuous beams with discrete or continuous
elastic support, arbitrary boundary conditions, intentionally outofplane loading and variable cross
section (see fig. 11) a theory 2nd order threedimensional stress analysis with geometric outofplane
imperfections leads to a generally applicable approach.
For class 1 cross sections two criterions [5], [6] are available for definition of the ultimate load
capacity
 theory 2nd order first yield criterion
 theory 2nd order first hinge criterion
For first yield design the vonMises stress ıv anywhere along the beam axis within the cross section,
obtained from a theory second order analysis of the geometrically imperfect structure with linear elastic
material behaviour must satisfy
fy
max ıVII ( ı xII,d )2 3( Ĳ dII )2 d (9)
ȖM 0
In (9) VIIx,d is the theory 2nd order axial stress due to the combined action an axial force N, bending
moments M y and Mz about principal axis and the warping moment MȦ from nonuniform torsion. The
theory 2nd order shear stress WdII is calculated from shear forces Vy and Vz and the St. Venant’s torsion
moment MTP. Graphs qu,el,v in figs. 10 are obtained by dividing the ultimate load derived from eqn. (9)
by the qpl. From comparison with the qu,el,x curve, which neglects shear stresses in eqn. (9) it is obvious,
that in the stocky slenderness area the capacity is essentially limited by the shear stresses. In the moderate
slender area the first yield criterion leads, depending on the moment distribution, to somewhat higher
capacities than the equivalent slenderness approach of EC3. For very slender structures shear stresses do
not count and capacities are very similar.
5 CONCLUSION
A procedure has been developed for lateral torsional buckling design of continuous beams with
moment hinges. It has been alternatively applied for equivalent slenderness procedure in EC3 and for first
yield criterion on the basis of a geometrically nonlinear theory 2nd order stress analysis including
geometric imperfections.
REFERENCES
[1] BTIIProgramm, Biegetorsionstheorie II.Ordnung, Friedrich + Lochner GmbH  Software für Statik
und Tragwerksplanung, Version 03/09, Stuttgart/Dresden.
[2] DIN EN 199311, Eurocode 3: Bemessung und Konstruktion von Stahlbauten, Teil 11:
Allgemeine Bemessungsregeln und Regeln für den Hochbau, 2005.
[3] Trahair, N.S., FlexuralTorsional Buckling of Structures, CRC Press Boca Ration, 1993, ISBN
0849377633.
[4] Lindner, J., Stabilisierung von Trägern durch Trapezbleche, Stahlbau 1/1987, Ernst & Sohn
[5] Osterrieder, P., Voigt, M., Saal, H., Vergleichende Betrachtungen zum Biegedrillknicknachweis
nach DIN 18800 Teil 2 (Ausgabe März 1988), Stahlbau 58 (1989), Heft 11, 341 – 347
[6] Osterrieder, P., Kretzschmar, J., Firsthinge analysis for lateral buckling design of open thinwalled
steel members, Journal of Constructional Steel Research 62 (2006), pp. 3543
728
8. PLATED STRUCTURES AND BOX GIRDERS
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. This paper focuses on the design of steel plated girders under combinations of transverse
loading, bending moment and shear force. In the member states of the European Union the design of
slender steel plates is covered by EN 199315:2006. Although conclusions from literature show a rather
significant interaction between transverse loading and shear force (FV), no consideration of this type of
interaction is made in Section 7.2 of EN 199315:2006. In order to close that gap experimental and
numerical studies were undertaken to analyse the stability behaviour and to identify the influence of key
parameters. Based on that an FV proposal is developed which is completed by the consideration of a
bending moment so that finally a fully usable FMV interaction equation is proposed.
1 INTRODUCTION
Steel plated structures occur as part of slender structural systems due to their advantageous strength
toweight ratio which allows especially aesthetical solutions. Transverse stiffeners are usually provided at
locations where forces are applied locally. However, this is not possible if the position of the load
introduction is transient e.g. in case of bridge girders being incrementally launched or for deep crane
runway beams. In both cases high transverse forces have to be introduced into the slender steel webs of
the girder, often with high bending moment and shear force at the same time, see figure 1.
Although conclusions from literature show a rather significant interaction between transverse loading
and shear force (FV), no consideration of this type of interaction is made in Section 7.2 of
EN 199315 [1]. In order to close that gap experimental and numerical studies were undertaken to
analyse the stability behaviour and to identify the influence of key parameters. Based on that an FV
proposal is developed which is completed by the consideration of a bending moment so that finally a fully
usable FMV interaction equation is proposed.
Before current proposals are evaluated and improvements are proposed, considerations on the general
formulation of an interaction equation and on the choice of the verification point are presented in the
following sections.
731
Ulrike Kuhlmann and Benjamin Braun
Limited number of available data. In general, there is only a limited number of available data
points particularly from experiments because the ratio of loading is added as an important
parameter so that in order to be able to identify the parameter variation within a given load ratio a
similarly large number of specimens should be tested than for a basic load scenario.
Reference value for basic loading. The reference strength for basic loading coming from
resistance models usually has a variation itself. The interaction equation can therefore only be as
good as the resistance model for the reference strength. It is desirable to know the experimental
ultimate load from basic loading for each interaction test series, otherwise an assumption based on
a resistance model has to be made.
In the following, the interaction between transverse loading and shear force is exemplarily used to
illustrate the aforementioned difficulties and to explain the decisions which were taken in the formulation
of the proposals later on. Figure 2a) shows the FV interaction with reference strengths based on basic
loadings from experimental and numerical studies in which nothing else than the load parameter was
varied in comparison to the interaction case. In contrast to this figure 2b) shows the same interaction data
but with reference strengths according to EN resistances. It can be shown that the data increasingly
scatters for the EN reference strengths due to the variation of the design model itself. In order to draw a
concise conclusion on the interaction behaviour, it would be necessary to eliminate the effect of the
reference strength's design model. In the FV study, this was done while studying the effect of parameters
where experimental and numerical reference strengths are also referenced. Thus we can state that a
procedure similar to figure 2a) is better suited to analyse the interaction behaviour.
Besides that the quality of the reference strength influences the interaction data, so it is not the best
way to evaluate or define an interaction equation based on design models for reference strengths. Imagine
that only the experimental data would be available in figure 2b). In that case only two data points lie in
the quadrant which is relevant for interaction and the interaction equation would be less strict than in the
case when the numerical data are additionally considered. The parameters of the experiments are covered
732
Ulrike Kuhlmann and Benjamin Braun
rather by the means and upper fractiles of the resistance models which would lead to an underestimation
of interaction. It can be shown that if parameters are chosen such that they cover the lower tail of the
resistance models variation, as done in the numerical simulation, interaction becomes more severe. Of
course, if the definition of interaction equations is coupled to application ranges which restrict the use e.g.
to the parameters of the experiments, interaction may be defined more loose. However, in view of general
applicability and safety, the reference strengths should be based on corresponding experimental and
numerical basic loadings whenever possible.
733
Ulrike Kuhlmann and Benjamin Braun
Regardless of girders without and with longitudinal stiffeners the bending moment resistance is
determined according to Section 4 and the shear resistance according to Section 5, both EN 199315 [1].
In order to evaluate the quality of the proposed interaction equations with regard to the different patch
load resistance models statistical analyses are performed. In the statistical evaluation the test result Re
which can be of experimental or numerical origin is consistently compared to the calculated resistance Rt
of the chosen engineering model under the same load ratio. From a constant load ratio two scalar load
amplification factors can be determined. The quotient of the two scalar load amplification factors
represents a key figure and on that basis a vectorial comparison is carried out for each pair of tested and
calculated resistances.
The evaluation of the different resistance models showed that the proposed interaction equation is
safe sided not only for girders without but also with longitudinal stiffeners. The statistical evaluation is
given in table 1. Detailed results of the study can be found in [9].
734
Ulrike Kuhlmann and Benjamin Braun
comprehensive summary of interaction equations can be found e.g. in [10]. The statistical evaluation of
the different interaction equations based on EN reference strengths is given in table 2. The Roberts
proposal is the most conservative one, whereas the Bergfelt proposal is the most favourable one. However
it can be shown that all proposals perform similar and that the trilinear EN approach is simple though
appropriate. And although the EN interaction equation was determined on the basis of EN reference
strengths, it can be shown that it could be further used for welded sections even if the advanced resistance
model of Gozzi is used. However, the objective to propose a single FMV interaction equation led to the
development of a FM interaction equation which can be consistently merged with the FV proposal, see
equation (1).
Following the principles which were set up in sections 2 and 3, an interaction equation based on the
general format according to equation (2) has been developed.
c
F § M ·
¨ ¸ d 1.0 (2)
FR ¨© M R ¸¹
Equation (2) is fitted as lower bound curve to a small FE study of Davaine and the FM database
which has been evaluated with the advanced resistance models of Gozzi and Clarin. The comparison is
shown in figures 4 and 5. The difference between both figures is the reference strength which has been
chosen for the bending moment resistance. Data points inside the interaction curve can be disregarded
since they are close to basic loading cases so that their deviation is considered as inherent to the resistance
models for the reference strengths. In figure 4 the reference strengths for the bending moment resistance
MR is based on the relevant crosssection class, i.e. Mpl or Mel. The parameter c is determined and
rounded off to a single decimal place so that c = 5.0. It can be shown that for high levels of bending
moment the interaction curve hardly catches the distribution of data points. For that reason, in a second
step the plastic moment resistance irrespective of the crosssection class was chosen as reference strength,
as it is similarly used in the MV interaction of Section 7.1, EN 199315. The parameter c is determined
and rounded off to a single decimal place so that c = 3.6. The results are shown in figure 5. It can be
shown that the data is slightly more homogenouos though hardly perceptible. The statistical evaluation of
both proposals which is given in tables 4 and 5 supports this. In table 3 the results of current EN rules are
given and a comparison shows that in both cases an improvement exists which can be identified by
comparing especially the standard deviation. However, in terms of statistical quality both proposals are
almost identical.
This consistency and the data scatter which is perceived to be slightly more homogenuous leads to the
adoption of the plastic bending moment resistance as reference value. Thus, the consistent definition of
the FV and FM interaction equations as continuous function enables the merging of both criteria. The
full FMV interaction equation becomes equation (3). In addition the resistance criteria of the basic
loadings according to section 4.6, 5.5 and 6.6, EN 199315, should be met. The resulting interaction
surface is illustrated in figure 6.
735
Ulrike Kuhlmann and Benjamin Braun
Table 4: Statistical evaluation of the FM interaction equation with MR = MR,EN and c = 5.0.
Model Mean value Standard deviation Lower 5%fractile
Gozzi (2007) 1.458 0.229 1.081
Davaine (2005) 1.404 0.204 1.067
Clarin (2007) 1.462 0.243 1.063
736
Ulrike Kuhlmann and Benjamin Braun
Table 5: Statistical evaluation of the FM interaction equation with MR = Mpl,R and c = 3.6.
Model Mean value Standard deviation Lower 5%fractile
Gozzi (2007) 1.466 0.229 1.089
Davaine (2005) 1.414 0.204 1.078
Clarin (2007) 1.466 0.242 1.068
3.6 1.6
F §¨ M ·
¸
§ V 0.5 F ·
¨ ¸ d 1.0 (3)
FR ¨© M pl ,R ¸ ¨ V ¸
¹ © b ,R ¹
where
FR is the transverse loading resistance of the cross section according to Section 6, EN 19931
5, or according to the advanced resistance models by Gozzi [4], Davaine [5] and Clarin[6];
Mpl,R is the plastic resistance of the cross section consisting of the effective area of the flanges
and the fully effective web irrespective of its section class;
Vb,R is the shear resistance of the cross section according to Section 5, EN 199315.
7 CONCLUSIONS
Based on the research work of the COMBRI research project [3] and beyond [11], evident and
necessary improvements regarding the interaction criteria of steel plates are reported in this paper. At the
beginning thorough considerations on the formulation of interaction criteria and verification points were
made in general. In detail a new interaction equation for the effective width method in case of transverse
loading, bending moment and shear force has been proposed which is summed up below.
First a comparison of the experimental and numerical FV results with known tests from literature
showed that the interaction between transverse loading and shear force is not negligible. A comparison
with proposals from literature showed that only few approaches exist which do not appropriately describe
the interaction behaviour e.g. with regard to the influence of the long loading lengths. Especially for
longer loading lengths the formulation of a new interaction equation was required, see equation (1). By
choosing the verification point at the centerline of the transverse loading the part of the shear force which
737
Ulrike Kuhlmann and Benjamin Braun
is induced by the transverse loading and which is already included in the resistance model can be
accounted for. As a result not only the smallest data scatter is found but also a conclusive subdivision of
the interaction case into the basic loadings "transverse loading" and "shear force" is possible which makes
an interaction verification for a transverse loading without additional shear force obsolete.
Plates under transverse loading are unavoidably subjected to bending moment so that this interaction
has already been addressed in a number of research works which have been thoroughly evaluated in this
work. The performance of all proposals is similar and it could be shown that the trilinear EN approach is
simple though appropriate. However, the objective to propose a single FMV interaction equation led to
the development of a FM interaction equation which can be consistently merged with the FV proposal,
see equation (3). The verification point is naturally chosen at the centerline of the transverse loading
which is the location where also the maximum bending moment occurs.
The new formulation is based on the experimental and numerical data set from own work and from
literature and a statistical evaluation proves the applicability of the equation not only to current EN
resistance models but also to the improved resistance models developed by Gozzi, Davaine and Clarin,
for unstiffened and for longitudinally stiffened girders.
REFERENCES
[1] EN 199315, Eurocode 3: Design of steel structures  Part 15: Plated structural elements, 2006.
[2] Kuhlmann, U., Johansson, B., Raoul, J., Braun, B., Gozzi, J., Clarin, M., Davaine, L. and Martin,
P.O., “A Survey on Patch Loading Models for Bridge Launching”, Proc. of the IABSE Symposium,
Weimar, Germany, 2007.
[3] COMBRI, Competitive Steel and Composite Bridges by Improved Steel Plated Structures, Final
Report, RFCS contract no. RFSCR03018, 2007.
[4] Gozzi, J., Patch Loading Resistance of Plated Girders  Ultimate and serviceability limit state,
Doctoral thesis 2007:30, Division of Steel Structures, Luleå University of Technology, Sweden,
2007.
[5] Davaine, L., Formulation de la résistance au lancement d’une âme métallique de pont raidie
longitudinalement  Résistance dite de "Patch Loading", Doctoral Thesis, L’Institut National des
Sciences Appliquées de Rennes, France, 2005.
[6] Clarin, M., Plate Buckling Resistance  Patch Loading of Longitudinally Stiffened Webs and Local
Buckling, Doctoral thesis 2007:31, Division of Steel Structures, Luleå University of Technology,
Sweden, 2007.
[7] ANSYS® v10.0, ANSYS Inc., Canonsburg, Pennsylvania, USA.
[8] Roberts, T.M. and Shahabian, F., “Design procedures for combined shear and patch loading”, Proc.
of the Institution of Civil Engineers, Structures & Buildings, 140(August), 219225, 2005.
[9] Kuhlmann, U., Braun, B. and Johansson, B., “The interaction behaviour of slender plates subjected
to shear force and patch loading”, Proc. of the 5th International Conference on Thinwalled
Structures, Brisbane, Australien, 2008.
[10] Lagerqvist, O., Patch Loading  Resistance of steel girders subjected to concentrated forces,
Doctoral thesis 1994:159 D, Division of Steel Structures, Luleå University of Technology, Sweden,
1994.
[11] Braun, B., Design of steel plates under combined loading, Doctoral thesis, Institute for Structural
Design, University of Stuttgart (in preparation).
738
SDSS’Rio 2010 STABILITY AND DUCTILITY OF STEEL STRUCTURES
E. Batista, P. Vellasco, L. de Lima (Eds.)
Rio de Janeiro, Brazil, September 8  10, 2010
Abstract. The structural case of hybrid girders subjected to patch loading is treated identically than
the one for homogenous specimens in EN199315. The EN199315 formulation is based upon a
theoretical model which consists of a plastic resistance Fy partly reduced by a resistance function χF,
this latter accounting for instability. The EN199315 formulation predicts that the ultimate load
capacity of steel plate girders subjected to patch loading increases, among others, with the hybrid
grade fyf/fyw. In this work, an appraisal of the theoretical and numerical responses of hybrid and
homogenous specimens subjected to patch loading is presented. Some peculiarities concerning the
treatment of hybrid girders are pinpointed though. It is numerically demonstrated that the moment
capacity of the flanges (and thus, fyf/fyw) does not play any role in the resistance of girders predicted in
EN199315. Accordingly, a design proposal which enhances the current formulation is provided.
1 INTRODUCTION
A girder is deemed as being hybrid when it is fabricated with different steel strengths for the flange
and web panels. In hybrid design, the nominal yield strength of one or both flanges is larger than the
nominal yield strength of the web. This type of girder is popular as the girder yields a greater flexural
capacity at lower cost and weight compared to a homogeneous girder [1]. On the other hand, patch
loading phenomena has been widely analyzed since the early sixties. Experimental and theoretical
analyses have pinpointed the typical failure mechanisms of girders subjected to patch loading and
consequently ultimate load predictions are nowadays available [2][4]. Broadly speaking, two
magnitudes have been given to describe the resistance of members subjected to this sort of loads. The
former defines a plastic resistance Fy of the member whereas the latter, an elastic critical load Fcr. The
former has been generally obtained by limit analysis whereas the latter, by theoretical formulae properly
calibrated with numerical simulations. The factual situations to which these members are subjected lie
inside a blurred transition between yielding and instability. It is well known that the root square of the
ratio between the plastic resistance Fy and the elastic critical load Fcr is commonly referred to as the
slenderness parameter. Admittedly, there exists a direct relation between this slenderness and the actual
failure mode. This relation has been labeled in the European guidelines for the design of plated
structural elements EN199315 [5] as the resistance function. The patch loading phenomena has been
harmonized to this procedure.
Despite the vast amount of research devoted separately to both hybrid girders and the patch loading
field, the research work that matches both subjects is rather scant [6][8]. In this paper, the resistance of
hybrid steel plate girders subjected to patch loading is dealt with simultaneously. Theoretically, it is
demonstrated that the EN199315 formulation predicts the resistance of plate girders subjected to patch
loading as a monotonic increasing function with, among other parameters, the flange yield strength fyf
(and consequently, with the hybrid grade fyf/fyw). Numerically, a vast study aimed at comparing
numerical and theoretical results is presented. In this study, it is found that the results obtained with the
739
Rolando Chacón et al.
EN199315 provisions do not reproduce satisfactorily the trends obtained numerically. Alternatively, a
design proposal in accordance with the procedure implemented in EN199315 and aimed at correcting
the aforementioned anomaly is proposed.
Bien plus que des documents.
Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.
Annulez à tout moment.