cover

cover next page >

Cover

title : Connections in Steel Structures : Behaviour, Strength, and

Design
author : Bjorhovde, Reider.; Brozzetti, Jacques.; Colson, Andr•e.
publisher : Taylor & Francis Routledge
isbn10 | asin :
print isbn13 : 9780203497890
ebook isbn13 : 9780585461199
language : English
subject Welded joints--Congresses, Bolted joints--Congresses, Building,
Iron and steel--Congresses.
publication date : 1988
lcc : TA492.W4S66 1988eb
ddc : 672.5/2042
subject : Welded joints--Congresses, Bolted joints--Congresses, Building,
Iron and steel--Congresses.
cover next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/cover.html07/11/2009 12:02:08 PM
page_i

< previous page page_i next page >

Page i
CONNECTIONS IN STEEL STRUCTURES
Behaviour, Strength and Design
< previous page page_i next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_i.html07/11/2009 12:02:36 PM
 page_ii

< previous page page_ii next page >

Page ii
Proceedings of a state-of-the-art workshop on Connections and the Behaviour, Strength and Design of Steel Structures
held at the Laboratoire de Mécanique et Technologie, Ecole Normale Supérieure, Cachan, France, 25–27 May 1987.
< previous page page_ii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_ii.html07/11/2009 12:02:38 PM
 page_iii

< previous page page_iii next page >

Page iii
CONNECTIONS IN STEEL STRUCTURES
Behaviour, Strength and Design
Edited by
REIDAR BJORHOVDE
University of Pittsburgh, Pittsburgh, Pennsylvania, USA
JACQUES BROZZETTI
Centre Technique Industriel de la Construction Métallique, St. Rémy lès
Chevreuse, France
and
ANDRÉ COLSON
Laboratoire de Mécanique et Technologie, Ecole Normale Supérieure de
Cachan, Université Paris VI, Cachan, France

ELSEVIER APPLIED SCIENCE

LONDON and NEW YORK
< previous page page_iii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_iii.html07/11/2009 12:02:39 PM
 page_iv

< previous page page_iv next page >

Page iv
ELSEVIER APPLIED SCIENCE PUBLISHERS LTD
Crown House, Linton Road, Barking, Essex IG11 8JU, England
This edition published in the Taylor & Francis e-Library, 2003.
Sole Distributor in the USA and Canada
ELSEVIER SCIENCE PUBLISHING CO., INC.
52 Vanderbilt Avenue, New York, NY 10017, USA
WITH 21 TABLES AND 274 ILLUSTRATIONS
© ELSEVIER APPLIED SCIENCE PUBLISHERS LTD 1988
British Library Cataloguing in Publication Data
Connections in steel structures: behaviour, strength and design: proceedings
of a state-of-the-art workshop on connections and the behaviour, strength
and design of steel structures held at Laboratoire de Mécanique et Technologie Ecole Normale Supérieure, Cachan,
France, May 25–27, 1987
edited by Reidar Bjorhovde, Jacques Brozzetti, André Colson.
1. Building, Iron and steel—Joints
I. Bjorhovde, Reidar II. Brozzetti, Jacques III. Colson, André
624.1′821 TA684
Library of Congress Cataloging in Publication Data
State-of-the-Art Workshop on Connections and the Behaviour, Strength,
and Design of Steel Structures (1987: Cachan, France)
Connections in steel structures.
“Proceedings of a State-of-the-Art Workshop on Connections and the
Behaviour, Strength, and Design of Steel Structures held at Laboratoire de
Mécanique et Technologie, Ecole Normale Supérieure, Cachan, France,
May 25–27, 1987”—Half t.p.
1. Welded joints—Congresses. 2. Bolted joints—Congresses. 3. Building, Iron and steel—Congresses.
I. Bjorhovde, Reidar. II. Brozzetti, Jacques. III. Colson, André. IV.
Title.
TA492.W4S66 1987 672.5′2042 87–36401
ISBN 0-203-49789-9 Master e-book ISBN
ISBN 0-203-23099-X (OEB Format)
ISBN 1-85166-177-8 (Print Edition)
No responsibility is assumed by the publisher for any injury and/or damage to persons or property
as a matter of products liability, negligence or otherwise, or from any use or operation of any
methods, products, instructions or ideas contained in the material herein.
Special regulations for readers in the USA
This publication has been registered with the Copyright Clearance Center Inc. (CCC), Salem,
Massachusetts. Information can be obtained from the CCC about conditions under which
photocopies of parts of this publication may be made in the USA. All other copyright questions,
including photocopying outside the USA, should be referred to the publisher.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording,
or otherwise, without the prior written permission of the publisher.
< previous page page_iv next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_iv.html07/11/2009 12:02:41 PM
 page_v

< previous page page_v next page >

Page v
PREFACE
This book is the proceedings of an international workshop that was organized by the editors for the period 25–27 May
1987 at the Laboratoire de Mécanique et Technologie, Ecole Normale Supérieure de Cachan, France, in collaboration
with the Centre Technique Industriel de la Construction Métallique, St Rémy-lès-Chevreuse, France.
Financial support was provided by a grant from the Ministère de la Recherche et de l’Enseignement Supérieur in the
framework of the civil engineering research programme.
The editors are sincerely appreciative of the help of the many individuals who contributed to the success of the
conference. In particular, thanks are extended to:
— M.le Recteur Marcel Bonvalet, Directeur de l’Ecole Normale Supérieure de Cachan;
— M.Lucien Wahl, Directeur Général du Centre Technique Industriel de la Construction Métallique; and
— M.Mircea Predeleanu, Directeur du Laboratoire de Mécanique et Technologie.
The support and contributions of the workshop participants are further acknowledged, as is the input and assistance of
Elsevier Applied Science Publishers Ltd.
REIDAR BJORHOVDE
JACQUES BROZZETTI
ANDRÉ COLSON
Cachan, France
< previous page page_v next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_v.html07/11/2009 12:02:43 PM
 page_vi

< previous page page_vi next page >

Page vi
This page intentionally left blank.
< previous page page_vi next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_vi.html07/11/2009 12:02:44 PM
 page_vii

< previous page page_vii next page >

Page vii
CONTENTS
Preface v
List of Contributors xiii
Introductory Notes 1
Session 1:
Local Analysis of Joints
Bolted and welded connections in steel structures 8
J.STARK, F.BIJLAARD and G.SEDLACEK
Strength of connections in bending compared to ultimate behaviour rather than elasto-plastic theory 18
A.GRANSTRÖM
Experimental study of the non-linear behaviour of beam-to-column bolted joints 26
J.JANSS, J.P.JASPART and R.MAQUOI
Prediction of moment-rotation behaviour of semi-rigid beam-to-column connections 33
J.B.RADZIMINSKI and A.AZIZINAMINI
Experimental analysis of end plate connections 41
R.ZANDONINI and P.ZANON
Bolted end-plate connections in steel structures 52
H.AGERSKOV
< previous page page_vii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_vii.html07/11/2009 12:02:46 PM
 page_viii

< previous page page_viii next page >

Page viii
Strength and behaviour of in-plane weak axis joints and of 3-D joints 60
J.JANSS, J.P.JASPART and R.MAQUOI
Analysis of joints between structural hollow sections 69
J.WARDENIER
Strength of welded connections with hollow section bracings and I or H sections as chord 77
J.M.ARIBERT
Numerical simulation of the behaviour up to collapse of two welded unstiffened one-side flange connections 85
W.A.SIBAI and F.FREY
Discussion of Local Analysis Papers 93
Session 2A:
Mathematical Models
Finite element modelling of connections 96
D.BEAULIEU and A.PICARD
Three-dimensional physical and mathematical modeling of connections 104
A.COLSON
Mathematical model for semi-rigid joints under cyclic loads 112
F.M.MAZZOLANI
Moment-rotation relation of top- and seat-angle with double web-angle connections 121
N.KISHI, W.F.CHEN, K.G.MATSUOKA and S.G.NOMACHI
Moment-rotation relation of single/double web-angle connections 135
N.KISHI, W.F.CHEN, K.G.MATSUOKA and S.G.NOMACHI
Modelling of column-base behaviour 150
R.E.MELCHERS
On the non-linear behavior of joints in steel frames 158
F.TSCHEMMERNEGG
Semi-rigid connections in frames, trusses and grids 166
Ch.STUTZKI, J.LOPETEGUI and G.SEDLACEK
< previous page page_viii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_viii.html07/11/2009 12:02:48 PM
 page_ix

< previous page page_ix next page >

Page ix
Flexibility formulae and modelling of joint behaviour in girders made of rectangular hollow sections 175
A.CZECHOWSKI, J.KORDJAK and J.BRÓDKA
Discussion of Mathematical Modelling Papers 183
Session 2B:
Classification
Structural properties of connections in steel frames 186
J.W.B.STARK and F.S.K.BIJLAARD
Semi-rigid connections in isolation and in frames 195
J.B.DAVISON, P.A.KIRBY and D.A.NETHERCOT
Standardised method for measuring three-dimensional response of semi-rigid joints 203
M.CELIKAG and P.A.KIRBY
Discussion of Classification Papers 211
Session 3:
Frame Analysis
Frame analysis including semi-rigid connections and P-Delta effects 214
S.A.IOANNIDES
Non-linear analysis of plane frame structures with semi-rigid connections 222
Y.GALEA
Nonlinear inelastic analysis of flexibly-connected steel frames 231
M.H.ACKROYD
A finite element for the analysis of semi-rigid frames 238
C.POGGI and R.ZANDONINI
Discussion of Frame Analysis Papers 248
Session 4:
Frame Stability and Simplified Methods
Column base plate connections 252
A.PICARD and D.BEAULIEU
Simplified design approach for end-restrained columns in frames 263
R.BJORHOVDE
< previous page page_ix next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_ix.html07/11/2009 12:02:50 PM
 page_x

< previous page page_x next page >

Page x
Influence of semi-rigid connections on the overall stability of steel frames 272
F.M.MAZZOLANI
Influence of semi-rigid and partial strength joints on the behaviour of frames 280
F.S.K.BIJLAARD and H.H.SNIJDER
Large scale tests on column subassemblages and frames 291
P.A.KIRBY, J.B.DAVISON and D.A.NETHERCOT
Simplicity in flexibly-connected frame analysis 300
K.H.GERSTLE and N.E.COOK JR
Simplified design of flexibly-connected building frames 309
M.H.ACKROYD
Discussion of Papers on Frame Stability and Simplified Methods 318
Session 5:
Design Requirements
LRFD connection design in the USA 322
R.O.DISQUE
Structural steel connections in Canada: Current requirements and future trends 330
M.I.GILMOR
Practical considerations in the design of frames with PR connections 339
S.D.LINDSEY
Observations on the seismic behaviour of steel connections after the Mexico earthquakes of 1985 343
E.MARTINEZ-ROMERO
Modified Eurocode 3 design recommendations for hollow section lattice girder joints 353
J.WARDENIER
The response of steel structures with semi-rigid connections to seismic actions 362
G.SEDLACEK, M.-S.KOO and G.BALLIO
Discussion of Design Requirements Papers 375
< previous page page_x next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_x.html07/11/2009 12:02:51 PM
page_xi

< previous page page_xi next page >

Page xi
Session 6:
Data Base Organisation
Needs, concepts and organization: a discussion 378
Data sheet format presentation 380
R.BJORHOVDE and J.BROZZETTI
Classification of connections 388
R.BJORHOVDE, J.BROZZETTI and A.COLSON
Research and Development Needs
Introductory comments 394
Research and development needs 394
< previous page page_xi next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xi.html07/11/2009 12:02:52 PM
 page_xii

< previous page page_xii next page >

Page xii
This page intentionally left blank.
< previous page page_xii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xii.html07/11/2009 12:02:53 PM
page_xiii

< previous page page_xiii next page >

Page xiii
LIST OF CONTRIBUTORS
M.H.ACKROYD
Techtrol Ltd, 3870 Cote Vertu, St Laurent, Québec H4R 1VR, Canada
H.AGERSKOV
Department of Structural Engineering, Building 118, Technical University of Denmark, DK-2800 Lyngby, Denmark
J.M.ARIBERT
INSA, 20 Avenue des Buttes de Coesmes, 35043 Rennes Cedex, France
A.AZIZINAMINI
Construction Technology Laboratories, Inc., Skokie, Illinois 60077, USA
G.BALLIO
Department of Structural Engineering, Politecnico di Milano, Piazza Leonardo da Vinci 32, I-20133 Milan, Italy
D.BEAULIEU
Département de Génie Civil, Université Laval, Ste Foy, Québec G1K 7P4, Canada
F.S.K.BIJLAARD
TNO Institute for Building Materials and Structures, PO Box 49, 2600 AA Delft, The Netherlands
R.BJORHOVDE
Department of Civil Engineering, University of Pittsburgh, 949 Benedum Hall, Pittsburgh, Pennsylvania 15261, USA
J.BRÓDKA
Metal Structures Research and Design Centre ‘Mostostal’, Ul. Krucza 16/22, 00–926 Warsaw, Poland
< previous page page_xiii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xiii.html07/11/2009 12:02:56 PM
page_xiv

< previous page page_xiv next page >

Page xiv
J.BROZZETTI
CTICM, Domaine St Paul, 78470 St Rémy-lès-Chevreuse, France
M.CELIKAG
Department of Civil and Structural Engineering, The University, Mappin Street, Sheffield S1 3JD, UK
W.F.CHEN
School of Civil Engineering, Purdue University, West Lafayette, Indiana 47907, USA
A.COLSON
Laboratoire de Mécanique et Technologie, ENS de Cachan, 61 Avenue du Président Wilson, 94230 Cachan, France
N.E.COOK JR
Department of Civil, Environmental and Architectural Engineering, University of Colorado, Boulder, Colorado 80309,
USA
A.CZECHOWSKI
Metal Structures Research and Design Centre ‘Mostostal’, Ul. Krucza 16/22, 00–926 Warsaw, Poland
J.B.DAVISON
Department of Civil and Structural Engineering, The University, Mappin Street, Sheffield S1 3JD, UK
R.O.DISQUE
American Institute of Steel Construction, 400 North Michigan Avenue, Chicago, Illinois 60611, USA
F.FREY
IREM/DGC, Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland
Y.GALEA
CTICM, Domaine St Paul, 78470 St Rémy-lès-Chevreuse, France
K.H.GERSTLE
Department of Civil, Environmental and Architectural Engineering, University of Colorado, Boulder, Colorado 80309,
USA
< previous page page_xiv next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xiv.html07/11/2009 12:03:05 PM
page_xv

< previous page page_xv next page >

Page xv
M.I.GILMOR
Canadian Institute of Steel Construction, 201 Consumers Road, Suite 300, Willowdale, Ontario M2J 4G8, Canada
A.GRANSTRÖM
AG Konstruktioner, Björns Backe 3, S-182 35 Danderyd, Sweden
S.A.IOANNIDES
Structural Affiliates International, Inc., 2416 Hillsboro Road, Nashville, Tennessee 37212, USA
J.JANSS
Centre de Recherches de l’Industrie des Fabrications Métalliques, Liège, Belgium
J.P.JASPART
Institut du Génie Civil, Université de Liège, 6 Quai Banning, 4000 Liège, Belgium
P.A.KIRBY
Department of Civil and Structural Engineering, The University, Mappin Street, Sheffield S1 3JD, UK
N.KISHI
Muroran Institute of Technology, Muroran 050, Japan
M.-S.KOO
Lehrstuhl für Stahlbau, RWTH Aachen, Mies-von-der-Rohe-Strasse 1, D-5100 Aachen, Federal Republic of Germany
J.KORDJAK
Metal Structures Research and Design Centre ‘Mostostal’, Ul. Krucza 16/22, 00–926 Warsaw, Poland
S.D.LINDSEY
Stanley D.Lindsey & Associates Ltd, 1906 West End Road, Nashville, Tennessee 37205, USA
J.LOPETEGUI
Lehrstuhl für Stahlbau, RWTH Aachen, Mies-von-der-Rohe-Strasse 1, D-5100 Aachen, Federal Republic of Germany
< previous page page_xv next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xv.html07/11/2009 12:03:06 PM
 page_xvi

< previous page page_xvi next page >

Page xvi
R.MAQUOI
Institut du Génie Civil, Université de Liège, 6 Quai Banning, 4000 Liège, Belgium
E.MARTINEZ ROMERO
E.Martinez Romero SA, Av. Nuevo Leon 54–2nd floor, 06140 Mexico DF, Mexico
K.G.MATSUOKA
Muroran Institute of Technology, Muroran 050, Japan
F.M.MAZZOLANI
Engineering Faculty, Istituto di Tecnica delle Costruzioni, University of Naples, Piazzale V. Tecchio, I-80125 Naples,
Italy
R.MELCHERS
Department of Civil Engineering and Surveying, University of Newcastle, Newcastle, New South Wales 2308, Australia
D.A.NETHERCOT
Department of Civil and Structural Engineering, The University, Mappin Street, Sheffield S1 3JD, UK
S.G.NOMACHI
College of Industrial Technology, Nihon University, Izumi, Narashino 275, Japan
A.PICARD
Département de Génie Civil, Université Laval, Ste Foy, Québec G1K 7P4, Canada
C.POGGI
Department of Structural Engineering, Politecnico di Milano, Piazza Leonardo da Vinci 31, I-20133, Milan, Italy
J.B.RADZIMINSKI
Department of Civil Engineering, University of South Carolina, Columbia, South Carolina 29208, USA
G.SEDLACEK
Lehrstuhl für Stahlbau, RWTH Aachen, Mies-von-der-Rohe-Strasse 1, D-5100 Aachen, Federal Republic of Germany
< previous page page_xvi next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xvi.html07/11/2009 12:03:07 PM
page_xvii

< previous page page_xvii next page >

Page xvii
W.A.SIBAI
IREM/DGC, Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland
H.H.SNIJDER
TNO Institute for Building Materials and Structures, PO Box 49, 2600 AA Delft, The Netherlands
J.W.B.STARK
TNO Institute for Building Materials and Structures, PO Box 49, 2600 AA Delft, The Netherlands
Ch.STUTZKI
Lehrstuhl für Stahlbau, RWTH Aachen, Mies-von-der-Rohe-Strasse 1, D-5100 Aachen, Federal Republic of Germany
F.TSCHEMMERNEGG
Universität Innsbruck, Technikerstrasse 13, A-6020 Innsbruck, Austria
J.WARDENIER
Delft University of Technology, Stevinweg 1, PO Box 5048, 2600 GA Delft, The Netherlands
R.ZANDONINI
Department of Engineering, University of Trento, Mesiani di Povo, I-38050 Trento, Italy
P.ZANON
Department of Engineering, University of Trento, Mesiani di Povo, I-38050 Trento, Italy
< previous page page_xvii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xvii.html07/11/2009 12:03:09 PM
 page_xviii

< previous page page_xviii next page >

Page xviii
This page intentionally left blank.
< previous page page_xviii next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_xviii.html07/11/2009 12:03:23 PM
 page_1

< previous page page_1 next page >

Page 1
INTRODUCTORY NOTES
Reidar Bjorhovde
Department of Civil Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania, U.S.A.
Jacques Brozzetti
C.T.I.C.M.
Saint Rémy lès Chevreuse, France
and
André Colson
Laboratoire de Mécanique et Technologie
Ecole Normale Supérieure
Université Paris VI
Cachan, France
When the idea of an International Workshop on Connections and the Behaviour, Strength and Design of Steel Structures
was first conceived, it reflected the recognition that whereas much research and development work has been conducted
on steel structures in general, and on the miscellaneous components and connections in particular, relatively little
attention has been paid to the increasingly important aspects of the unified, complete structure. Part of this deficiency is
clearly related to the complexity of the structures, and the fact that most researchers have devoted their attention to
determining the characteristics of the individual elements. It is also related to the analytical and experimental problems
that have been encountered by investigators of topics that are highly non-linear and three-dimensional in nature, for
which neither computational nor testing facilities of adequate capacity have been available until but in recent years.
The advent of very powerful computers, coupled with the development of sophisticated testing and data acquisition
equipment, have made for the completion of research studies of increasingly more complicated member, connection and,
to a certain extent, overall frame response problems. The former difficulties of being capable of solving the problems,
given the proper tools, are therefore now being removed at an
< previous page page_1 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_1.html07/11/2009 12:03:25 PM
 page_2

< previous page page_2 next page >

Page 2
accelerating pace. However, this poses an array of problems of a different nature for the researchers, but especially for
design standard writers and their users, the design engineers. Specifically, the volume of information is becoming large
to the point where it is no longer realistically possible to keep abreast of all new developments. Secondly, methods and
procedures vary substantially from one locale to the next, with the result that interpretation of the findings and how they
may apply to a particular case will be a dubious undertaking. Finally, with studies being carried out in universities,
research laboratories and companies around the world, some amount of duplication of research is essentially
unavoidable.
On the background of the problems that have been outlined in the preceding, the idea of an intensive workshop was
conceived. It was felt that with the gathering of a group of researchers with extensive experience in the study of steel
structures and their connections, to present and discuss their own ongoing research project as well as those of others, it
would be feasible to arrive at an assessment of the state-of-the-art. In addition, having established the status of current
knowledge, an important function of the workshop would be to address the criteria and topics of future research, as well
as to discuss suitable vehicles for gathering and maintaining a coherent data base for connections. The latter activity
would be particularly important for designers and code writers, to serve as a source of reliable information on
connections.
To achieve the most in-depth, intensive discussions of the problems, it was decided to limit the workshop attendance to a
relatively small group of internationaly recognized researchers, code writers and design engineers with extensive
background in the study of connections and steel structures. It was felt that this was essential to the success of the
workshop, on the premise that the attendees would not need to be educated by each other in the traditional technical
conference format, but rather would be able to have truly in-depth exchanges of opinions, methods, procedures, results,
and future directions. For the same reasons the wokshop program was organized such that the discussion time could be
maximized, with primary emphasis on the informal exchange of ideas and understanding. The proceedings reflect the
importance of the discussions and the evaluations of research needs and data collection efforts, to the effect that
individual chapters address these items following the written contributions of the invitees for each technical session.
It was decided to organize the workshop into technical sessions with the following primary subject coverage:
Session No 1 : Local Analysis of Joints
Session No 2 : Modeling of Load-Deflection Behaviour
Session No 3 : Methods of Frame Analysis
Session No 4 : Frame Stability and Simplified Methods
Session No 5 : Design Requirements and Codes
Session No 6 : Data Base Organization and Summary
The following comments highlight the topics that were to be discussed within each session, along with the anticipated
outcome.
< previous page page_2 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_2.html07/11/2009 12:03:28 PM
 page_3

< previous page page_3 next page >

Page 3
The session on local analysis of joints was intended for a review of current theoretical and experimental research dealing
with the local behaviour of actual connections under load. Local behaviour was defined as the detailed and full-range
response of the connection, including all connecting parts, mechanical fasteners, welds, and so on. Such studies are
essential for the understanding of the failure mechanisms of the connection, as well as for identifying the main load and
deformation controlling parameters. The three primary characteristics of non-linear behaviour can also be identified,
namely, stiffness, ultimate strength, and ductility.
The expected outcome of the local analysis session was the acquisition of information on the most important numerical
and experimental research work on local connection behaviour, including load transfer mechanisms and failure modes. It
would also establish familiarity with the types of connections that are used in actual structures around the world, and
facilitate correlation efforts.
Due to the complexity of connection strength and behaviour, it was decided to divide the session on modeling of load-
deflection behaviour into two sub-sessions. Their topics of coverage and anticipated outcome are discussed in the
following.
The first sub-session on modeling was aimed at mathematical modeling with parameter identification, as well as
comparison with test results. Specifically, rather than dealing with the local conditions, as were the topics of the first
session, the discussions and papers of this modeling session were intended to reflect the overall connection response.
Special emphasis was therefore placed on developing an inventory of connection moment-rotation mathematical models.
The latter may have been based on curve-fitting techniques or statistical analyses of test results; it may also have been
possible to derive M-Φ-equations that are associated with the physical characteristics of the connections. Further, in
recognition of the non-linear response of all realistic connections, contributors were asked to consider the global
characterization of semi-rigid joint behaviour, including for column footings. Finally, the correlation between connection
model and test results is important, and it was also intended to obtain an evaluation of the effects of the sense of loading,
i.e. monotonic and cyclic load effects would be significant considerations.
The expected outcome of the first modeling sub-session was the development of an inventory of mathematical models of
semi-rigid connections, including a collection of moment-rotation (or similar) curves and the attendant primary variables
for strength, behaviour and ductility.
The second sub-session on connection modeling was aimed at the classification problem, with an introduction to the data
base concept. Thus, it was felt that sufficient experimental and other information is currently available for several
connections in various countries. The expected outcome of the session might be the development of a classification
system for beam-to-column connections, including considerations such as initial stiffness, ultimate strength, and rotation
capacity at service and ultimate load levels.
< previous page page_3 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_3.html07/11/2009 12:03:29 PM
 page_4

< previous page page_4 next page >

Page 4
The session dealing with the subject of frame analysis was intended for a presentation of the methods of structural
analysis which specifically include the constitutive law of the connections. Procedures are currently available which
incorporate first and second order geometric and material effects for the connections, the structural members, and the
entire frame, but it was felt that it would be important to establish exactly what the individual solutions can accomplish,
and how well they correlate with tests (if any) and other analytical solutions.
The anticipated outcome of the frame analysis session was the establishment of a repertory of computer programs,
including definitions of the primary assumptions and variables. It was also expected that some frame test results would
be presented, to lend further support to the theoretical research results.
Recognizing the unique problems of frame and member stability, the session addressing the topics of frame stability and
simplified methods was intended for dealing with: (1) the influence of semi-rigid connection characteristics on the
stability of members and frames, including serviceability and ultimate limit state considerations, and (2) the development
of simplified methods of calculation for column and frame stability, as well as for member forces. It was felt that the
simplified solutions might be more adaptable to practical usage, such as would be considered by design engineers.
The expected outcome of the frame stability session was the presentation of stability solution procedures for members
and frames, taking into account known stiffness characteristics of the connections. The computer solutions that were
thought of would also be expected to facilitate the computations of member stress resultants and displacements
throughout the frame. It was also thought to be possible that these results might indicate approaches that could take into
account second order (P-Delta) effects, without necessarly having to go through complete second order frame solutions.
The technical session addressing design requirements and codes was aimed at dealing with subjects such as the status of
incorporating semi-rigid connection and frame response concepts into design codes as well as practice. This would
necessarily have to include some considerations of cyclic loading, such as may be encountered under seismic conditions,
as well as material requirements. A particular group of concerns would be those associated with serviceability and
ultimate limit states, the questions of story and overall lateral deflections (drift) of frames, and the format of limit states
design criteria for structures with semi-rigid connections. This should also consider the questions of necessary and
available rotation capacities.
The outcome of the discussions of code requirements was envisioned to be an assessment of the state-of-the-art of code
development in various countries, along with indications of what is perceived to be needed for practical design criteria.
In particular, it was known that actual structural designs had been performed by at least one company in the United
States and possibly others elsewhere; it would be of significant interest to have the assessments of such designers of the
current knowledge and future needs.
< previous page page_4 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_4.html07/11/2009 12:03:30 PM
 page_5

< previous page page_5 next page >

Page 5
In addition to providing a final review of all of the papers and the respective discussions of the various technical
sessions, the final session was intended as the outlet for two primary groups of presentations, namely, (1) a discussion
and evaluation of a potential data base organization and (2) the reports of the session reporters and research reporters.
The former subject was of significant interest, especially to the effect that it was hoped that a proper organizational
structure would facilitate future exchanges of information, as well as avoid any duplication of work. The research
reporters and their function were considered crucial to the workshop, as these reports would focus only on the work that
needs to be undertaken, to remove current deficiencies of available data and to add to data bases as well as methods of
analysis and design. It is noted that the use of research reporters constitutes the first time such assignments have been
utilized in a technical conference setting; it might be considered by other groups for similar activities in the future.
< previous page page_5 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_5.html07/11/2009 12:03:31 PM
 page_6

< previous page page_6 next page >

Page 6
This page intentionally left blank.
< previous page page_6 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_6.html07/11/2009 12:03:32 PM
page_7

< previous page page_7 next page >

Page 7
SESSION No 1
LOCAL ANALYSIS OF JOINTS
< previous page page_7 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_7.html07/11/2009 12:03:34 PM
 page_8

< previous page page_8 next page >

Page 8
BOLTED AND WELDED CONNECTIONS IN STEEL STRUCTURES
Jan STARK
Eindhoven University of Technology, Eindhoven, The Netherlands
Frans BIJLAARD
TNO Institute for Building Materials and Structures, Delft, The Netherlands
Gerhard SEDLACEK
University of Aachen, Aachen, West Germany
ABSTRACT
On European (Eurocode 3) as well as on world level (ISO/TC 167) work is going on to come to international
harmonisation of design procedures for steel structures. Both activities focus on a presentation of design codes in a limit
state design (LRFD) format. A brief review will be given of the design rules for bolts, as now proposed in the drafts of
the international documents. To validate the strength functions and to determine suitable values for the model factors, a
statistical re-evaluation of available test information is carried out. Preliminary results of this study are presented and
compared with the proposed new design rules.
INTRODUCTION
The 1984 draft of Eurocode 3 is now under revision on the basis of the comments from the EEC member states. The final
draft is planned for the end of 1988. The European Commission has recognized the need for background information that
confirms and explains the rules presented in EC3. The authors of this paper are charged with the task to provide such
information for the rules on fasteners and joints. The results of this study are also relevant for the work on an
international standard in ISO/TC 167. The conversion of traditional working stress design methods for connections into a
limit state design format was not easy. It was found that many of the rules in existing national standards were based on
engineering judgement, more than on a consistent evaluation of experimental evidence. The following procedure is used
for the determination of the design rules on fasteners in EC3. Based on observation of actual behaviour in tests and on
theoretical considerations a ‘‘design model” is selected. Then by statistical interpretation of all available test data i.e.
regression analysis the efficiency of the model is checked. Eventually the design model has to be adapted until the
correlation of the theoretical values and the test data is sufficient. Then through manipulation with the statistical data of
the test population, the model factor γRd can be determined. Dividing the strength function by the model factor γRd
gives the design expression.
A first attempt for this evaluation, based on a procedure proposed by the EC Coordination Group [1, 2], is reported in
[3]. The results showed clearly that the procedure was not adequate for the purpose of checking strength functions and
had to be adapted. In this paper a revised evaluation procedure for the same strength functions as in [3] is presented.
DESIGN MODELS FOR BOLTS
Bolts in a joint can be loaded in tension, shear or a combination of tension and shear.
< previous page page_8 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_8.html07/11/2009 12:03:36 PM
 page_9

< previous page page_9 next page >

Page 9
Design model for a bolt loaded in tension
If a bolt is subjected to increasing tensile load fracture will occur in the threaded portion. By definition the stress area of
the bolt As is the cross sectional area of a smooth cylinder which, subjected to tension, has the same failure load as the
threaded part of a bolt from the same

Fig. 1 Bolt loaded in tension

material. The formula for As, given in Fig. 2, is internationally standardised.

Fig. 2 Definition of stress area

It is evident that, based on the definition of As, the strength function for a bolt in tension will be taken as:

where fub is the ultimate tensile strength of the bolt material.

In a joint so called prying forces may occur due to flexibility of the connected parts. This is illustrated in Fig. 3.

Fig. 3 Prying forces Q in a T-stub joint

In conventional design methods the occurence of prying forces is sometimes taken into account in the design expression
for Ft. Principally this is not correct and therefore in EC3 it is suggested to account for prying forces on the loading side
of the safety verification.
< previous page page_9 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_9.html07/11/2009 12:03:38 PM
 page_10

< previous page page_10 next page >

Page 10
Design models for bolts loaded in shear
The maximum resistance of a joint loaded in shear may correspond to four different failure modes as illustrated in Fig. 4.

Fig. 4 Possible failure modes of bolted joint loaded in shear

For each of these failure modes, a “design model” is selected. The strength function for shear failure of the bolt (Fig. 4,
mode a) is based on the pure shear strength of the bolt material. Based on experimental evidence, the pure shear strength
τub, is taken as 0.7 fub. The shear plane may pass through the threaded portion of the bolt or through the shank (see Fig.
5). Dependent on the case considered the strength function per shear plane is:

Fig. 5 Bolt loaded in shear

If the end distance of the bolt hole is sufficiently large to avoid occurence of the shear mode of failure shown in Fig. 4c,
the bearing pressure between the bolt and the plate will cause ovalization of the hole under increasing load. Although
even a significant ovalization of the hole is not an ultimate limit state in the strict sense, it is considered necessary to give
a limit for the mean bearing stress to control deformation. According to [4] the mean ultimate bearing stress may be
taken as 3 fy if the plates are in compression, where fy is the yield stress level of the plate material. The strength function
for bearing (hole elongation) will be:

where d is the shank diameter and t is the plate thickness.

< previous page page_10 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_10.html07/11/2009 12:03:40 PM
 page_11

< previous page page_11 next page >

Page 11

Fig. 6 Simplified model for shear failure of the plate

In EC3 it is assumed that this value is also applicable for plates in tension if the end and edge distances are sufficiently
large. In case of smaller end distances the shear mode (Fig. 4, mode c) will occur. To determine the main parameters of
the strength function for this case, the simplified model shown in Fig. 6 is used.
Assume that the length of the longitudinal shear planes is approximately 0.9 times the end distance e1. According to the
Von Mises yield criterion, the design shear stress is . The design load will then be:

where α is a function of
So the design function takes the same form as for hole elongation with the “bearing stress” 3 fy replaced by α fy. This

design function is used in draft EC3 where α is taken as (see Fig. 7). It is important to realise that the
design function for bearing in fact covers two different failure modes.

Fig. 7 Factor α for the “bearing strength” as a function of the end distance
In the design function for shear failure of the plate, the yield shear stress τy is introduced, following existing practice.
However, failure will normally not occur before the shear stress is equal to the ultimate shear strength τu. For the
revision of EC3 therefore comparison studies will be carried out to investigate whether the strength function for Fb can
be expressed in terms of fu rather than fy. Shear failure of the plate can also occur when the pitch P between two bolts in
a row is relatively small. To account for this failure mode, the factor α in the strength function for bearing is also made

dependent of (see Table 1).

< previous page page_11 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_11.html07/11/2009 12:03:42 PM
 page_12

< previous page page_12 next page >

Page 12

Fig. 8 Illustration of influence of edge distance on shear failure mode of the plate
A small edge distance e2 will cause a splitting type of failure as illustrated in Fig. 8. This reduces the capacity. At the
present not much test data are available from test specimen with smaller edge distances then 1.5 d. In draft EC3 the edge
distance is therefore limited to a minimum of 1.5 d.
The strength function for tensile failure of the connected parts (Fig. 4, mode d) is based on the assumption that failure
occurs if the net section An is fully stressed upto the tensile material strength fu. So the strength function is:

To prevent excessive deformations of the member, also the mean stress in the gross cross-section at ultimate load should
be less than the yield strength, with . In the case of unsymmetrical or unsymmetrically connected members such
as angles (see Fig. 9), tests show that rupture at an edge may start well before the mean stress in the gross section is fu.
In that case a reduced effective net section is taken into account in the strength function, as illustrated in Fig. 9.

Fig. 9 Illustration of reduced effective net section for an angle connected on one leg only
Design model for bolts under the combined action of tension and shear
If the shear plane passes through the threaded portion of the bolt (Fig. 5), failure will most certainly occur over the shear
plane. The strength function is based on a simple interaction formula:
< previous page page_12 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_12.html07/11/2009 12:03:43 PM
 page_13

< previous page page_13 next page >

Page 13

where and are the design capacities with respect to tension and shear and should be determined for the stress area
As (see Table 1), T is the normal force and V is the shear force acting on the bolt.
If the shear plane passes through the shank, dependent on the ratio of tension and shear, failure may either occur in the

shank (shear plane) or the threaded part. In the interaction formula and may be determined for the cross section A
of the shank, but additionally the tensile capacity of the threaded part should be checked with:

In Table 1 the strength functions according to some recently published international documents are summarised in
comparison with the strength functions assumed for the evaluation of test results.
Table 1 Summary of design rules for bolted connections
Capacities ECCS TC 10 Draft ISO 167 Draft EC3 Based on evaluation of test results
[5] [6] [7]
Tensile capacity smallest of:

Shear capacity

Bearing capacity
value of α to be determined from
test

Tensile of capacity
of connected
members
A fy A fy

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_13.html (1 of 2)07/11/2009 12:03:47 PM

 page_13

Plate or or or or

Angle, 1 bolt

Angle, 2 or more β An fy not covered β An fy

bolts

< previous page page_13 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_13.html (2 of 2)07/11/2009 12:03:47 PM

 page_14

< previous page page_14 next page >

Page 14
EVALUATION OF TEST RESULTS
To validate the strength functions and to determine suitable values for the model factors a re-evaluation of available test
information is carried out. Test results from different sources are systematically stored in a data base at the University of
Aachen. The total collection of test results on bolts amounts at the present more than 1900. By a joint CEB-ECCS-
coordination group, a uniform evaluation procedure for test results is developed, which is independent of the material
and the structure considered [1, 2]. This procedure is adopted by the Eurocode organisation. But as mentioned in the
introduction this procedure had to be adapted and the revised procedure is described in [8].
For the case that for each bolt and eventually the connected plate the actual material strength and the geometric
properties are known, the procedure as used in [8] is described. For every specimen i the experimental value rvi is
compared with the value rti, calculated with the relevant strength function (without γRd) and taking into account the
actual properties of the specimen. For a chosen test population, the values of rvi and rti can be plotted in a diagram (Fig.
10).

Fig. 10 Sample mean and correction for deviation

If the strength function is exact and complete, all the points (rvi, rti) lie on the bisector of the angle between the axes of
the diagram and the correlation coefficient . However, in general the points will scatter. So the mean value
and the standard deviation of the test values rvi respectively the calculated values rti are determined:

Tests:

Theory:

The correlation coefficient then follows from:

If the value of is greater than 0.9 the correlation is considered sufficient. Then a mean value correction is carried out.
The straight line is determined, which intersects the origin and represents the mean values of the tests.

< previous page page_14 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_14.html07/11/2009 12:03:50 PM
 page_15

< previous page page_15 next page >

Page 15
The coefficient of this line follows from:

The formula for this line of mean values is then multiplied by a coefficient δk to find the characteristic line (5% fractile).
The value of δk can be calculated with the statistical data of the population [8].
Finally with good approximation the model factor γRd can be calculated:

Not all test reports provide the data on actual material and geometric properties of test specimen as required for the
straight forward procedure described before. Two additional cases can be distinguished:
– Simple values of the material strengths in the sample are given, but these can not be related exclusively to individual
test specimen;
– Only the characteristic values of the material strength are given.
The procedure has been modified for these cases. Instead of the strength of each specimen, the mean value of the sample
is used and the standard deviation of the coefficient δk is increased properly. Comparison calculations show that the
modified procedure gives satisfactory results.
PRELIMINARY RESULTS OF THE EVALUATION FOR BOLTS
Only as an indication some preliminary results are given. At present for the evaluation all relevant test data in the data
base are used without any critical review of the test reports. In Fig. 11 the rvi-rti diagram is given for shear tests on 8.8
bolts, where the shear plane passes throuth the threaded protion. Fig. 12 shows the ratio between rvi and rti for the same
type of test, but for different bolt grades.

Fig. 11
Shear failure of 8.8 bolts
Shear plane through threaded portion

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_15.html (1 of 2)07/11/2009 12:03:52 PM

 page_15

Fig. 12
Shear failure of bolts of different grades with shear plane through threaded portion
< previous page page_15 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_15.html (2 of 2)07/11/2009 12:03:52 PM

 page_16

< previous page page_16 next page >

Page 16
This figure clearly shows the tendency that higher grades have lower shear capacities. Based on this information it
should be considered to reduce for 10.9 bolts the ultimate shear strength of 0.7 fu, as assumed in the strength function
(Table 1).
CONCLUDING REMARKS
For welded connections the same procedure will be followed in evaluation of test results. At the moment about 510 test
results are available mainly from the international test serie. An example is shown in Fig. 13. The design strength was
based on:

where F is the tensile force on the connection, Σ a l is the sum of the throat thickness times the length of the welds and α
depends on the steel grade ( for Fe 360; for Fe 420; for Fe 510).

Fig. 13 Weld failure

REFERENCES
1. MAIER, W., KERSKEN-BRADLEY, M., “Experimentelle Bestimmung der Tragfähigkeit von Bauteilen—
Grundideen der Auswertung” Festschrift Prof.Dr.-Ing. Bodo Heimeshoff zum 60. Geburtstag, Lehrstuhl für
Baukonstruktion und Holzbau, TU München, 1986.
2. ‘‘Auswertung von Versuchsergebnissen bei Berücksichtigung von Rechenmodellen—Zur Beschreibung der
Einflussgrössen”, Forschungsbericht Institut für Bautechnik, Berlin IV 1–5–396/84.
3. STARK, J.W.B., BIJLAARD, F.S.K., SEDLACEK, G., “Bolted and welded connections in steel structures”. To be
presented at the International Conference on Steel Structures, Cardiff, 8–10, 1987.
4. “European Recommendations for Steel Construction”, European Convention for Constructional Steelwork (ECCS),
Brussels, March 1978, ECCS-EG 77–2E.
5. “European Recommendations for bolted connections: in structural steelwork”, European Convention for
Constructional Steelwork, Technical
< previous page page_16 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_16.html07/11/2009 12:03:55 PM
 page_17

< previous page page_17 next page >

Page 17

Committee 10: Bolted and Welded Connections, Publication No 38, Fourth Edition, March 1985.
6. “Steel Structures, Materials and Design”, Draft Document ISO/TC/167/SC1 Doc. N114E, Norges
Byggstandardiseringsrad, Oslo, January 1986.
7. “Eurocode No 3: Common Unified Rules for Steel Structures”, Commission of the European Communities, Brussels
1984, Report EUR 8849 DE, EN, FR.
8. “Procedure to determine the characteristic strength functions from tests”. Proposal for Part 1 of the background
documents for the design rules in Eurocode 3.
< previous page page_17 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_17.html07/11/2009 12:03:56 PM
 page_18

< previous page page_18 next page >

Page 18
STRENGTH OF CONNECTIONS IN BENDING COMPARED TO ULTIMATE BEHAVIOUR RATHER THAN
ELASTO-PLASTIC THEORY
Anders Granström
AG Konstruktioner,
Björns backe 3,
S-182 35 Danderyd,
SWEDEN
ABSTRACT
In the design of structures, large structural elements such as beams and columns are generally designed according to
elastic or elasto-plastic theory based on the yield strength measured on small tensile specimens. When it comes to
connections these theories are often extremely conservative, particularly where bending is the governing failure mode.
In this paper the reason for the great strength of connections is discussed, and a few test results are presented that
illustrate the magnitude of the strength increase in relation to elasto-plastic theory.
INTRODUCTION
The yield limit or the 0.2% limit is generally used as a design criterion for beams and columns, the strength of which are
not governed by elastic instability. That procedure is justified, since only moderate straining is necessary to eliminate
even great residual stresses created by the rolling and welding processes.
It is well known that steel, due to strain hardening, exhibits a greater strength than corresponds to complete yield of a
cross section. However, strain hardening generally starts at rather great strain values, see table 1 [1]. In a beam there is
little use of this extra strength since it is paired with excessive yielding in large areas and large deflection, which may be
unacceptable even in an ultimate state condition.
In connections, however, excessive straining is local and affects the over-all deformations in the large structure to a
much smaller extent. Therefore strain hardening should be considered, and may offer a great contribution to the design
strength of a connection. Actually, it was found in [2] that the ultimate strength in bending of a detail may be
< previous page page_18 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_18.html07/11/2009 12:03:58 PM
 page_19

< previous page page_19 next page >

Page 19
TABLE 1
Results from tensile & compressive tests on various structural steels
SS Plate/ beam Tensile tests Compressive tests
Steel mm
No ReL N/mm2ReH N/mm2Rm N/ Strain hard, ReL N/ ReH N/mm2Strain hard,
mm2 starts at % mm2 starts at %
1412 HE200A 297 306 469 1.5 306 317 1.2
1412 P1 15 259 270 426 1.3 312 329 1.2
2624 P1 12 755 805 800 2.8 767 792
considerably greater than UTS. The reason for this behaviour is that the true fracture stress in a tensile test, due to
contraction, fig 1, is about twice as great as UTS. This fact is also shown in fig 2 [3] for a common structural steel,
designated I, and a cold finished steel, designated II.
With plates in bending the contraction of the tensile zone is restrained by the compressive zone. Therefore the potential
factor of strength increase should be about 2 related to UTS, and about 2 to 3 related to yield, depending on the type of
steel.

Figure 1. Tensile test at failure, where the fracture stress calculated on the initial area Ao is slightly less than UTS
< previous page page_19 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_19.html07/11/2009 12:04:00 PM
 page_20

< previous page page_20 next page >

Page 20

Figure 2. Principal stress-strain relationships according to [3]

RESULTS
In this chapter a brief view on actual test results is given, including details in tension, bending, tension & bending, and
compression. Also welds in compression are included.
Details in tension
In [2] details similar to that shown in fig 3, cut out from larger details, and with various types of welds, were tested to
failure in tension. The average ultimate stress in the connection, with little variation, was about 1.7 times the yield
strength (YS) as measured in a tensile test. This is fairly close to UTS from the tensile tests, which varied between 1.6
and 1.8 times YS.
Details in bending
Also in [2], details similar to that shown in fig 3 with various types of welds. The loading span was in most cases 300 or
500mm. Most
< previous page page_20 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_20.html07/11/2009 12:04:02 PM
 page_21

< previous page page_21 next page >

Page 21
details were loaded to large deformations without fracture. Only in one case, with giant fillet welds (a=17mm) one of the
horizontal members failed, in brittle fracture at 2.1 times YS. Many of the other details were loaded up to 2.6 times YP
without fracture.
Another investigation [4] with thick plate in bending showed an ultimate stress of 2.1 times YS, see fig4.

Figure 3. Mild steel details tested in bending

Figure 4. Bending test according to [4]

< previous page page_21 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_21.html07/11/2009 12:04:03 PM
 page_22

< previous page page_22 next page >

Page 22
Even in bolted end-plate connections [5], where the strains close to the tension member are considerably greater than at
the yield line near the bolts, average ultimate bending stresses of 3.5 times YS for mild steel, and 1.8 times YS (and
UTS) for ductile EHS, were evaluated neglecting the different magnitude of strain at the two yield lines.
Details in tension & bending
The details loaded in tension or bending according to fig 3 were later straightened, machined flush, and bolted together
as end-plate connections [6]. The maximum stresses at failure of the connections were calculated, conservatively
assuming equal compressive and tensile stress and equal stress in the tension member and at the bolt line, with minor

Figure 5. Uneven strain distribution in bolted end-plate connection

< previous page page_22 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_22.html07/11/2009 12:04:05 PM
 page_23

< previous page page_23 next page >

Page 23
corrections for geometric distortion during the test and different YS of the end plate, fig 6. In spite of the great mean
tensile stress, also shown in fig 6, the maximum stresses are of the magnitude of 3—quite compatible with the ultimate
stresses in pure bending.
Considering the mode of fracture and the magnitude of the ultimate stress, a correction for shear stresses is obviously not
useful. Factors that should be considered should rather be large strain concentration, heavy welding, and thick plate.

Figure 6. Mean stresses and maximum stresses related to yield strength

< previous page page_23 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_23.html07/11/2009 12:04:06 PM
 page_24

< previous page page_24 next page >

Page 24

Figure 7. Compression test configuration

Details in compression
Equally high strength was found by the author for details in pure compression, and even greater strength was found for
small welds, in an investigation aimed to decide acceptable lack of fit (gap) in connections with transmission of forces
by assumed contact pressure. The test results were never published, but some findings were used in ECCS
Recommendations [7] and a few more were referred to in IIW Recommendations [8].
The tests were executed as indicated in fig7. The guide angles were used since initial plastic buckling started soon after
yield in the thin compressive member. With guide angles, as indicated in fig 7, the load was possible to increase to levels
corresponding to 750 Mpa on the 10mm & 800 Mpa on the 15mm compression member. Without any visible sign of
fracture, the compressive member was then more or less liquidized, “floating out’’ around the guide bolts a couple of
mm. These stresses correspond to about 3 times YS of the mild steel compression members.
< previous page page_24 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_24.html07/11/2009 12:04:08 PM
 page_25

< previous page page_25 next page >

Page 25
In the case of the 3.5mm fillet welds, the maximum stress reached occasionally 2000 Mpa calculated as mean
compressive stresses over the throat sections, which corresponds to 4 times the specified UTS for tensile specimens. Still
the gap was not completely closed, and the welds sustained at subsequent tensile loading the specified UTS.
ACKNOWLEDGEMENTS
The author wish to thank the Swedish Institute of Steel Construction, which financially supported most of the
investigations referred to in this paper, as well as the work with the paper itself. The author also wish to thank the
Swedish Council for Building Research, which financially made it possible for the author to go to the “workshop” in
Cachan to present and discuss the results.
REFERENCES
1. Unpublished comercial investigation made by the author.
2. Granström, A., Methods of assessing the Risk for Lamellar Tearing. The relevance of laboratory tests in assessing the
strength of large structures , publ. 58, Swedish Institute of Steel Construction, Stockholm 1977.
3. Baehre, R., Historik—Material—Produkter. Kompendium i Stalbyggnad del I, Institutionen för Brobyggnad/
Stalbyggnad, Royal Institute of Technology, Stockholm 1971.
4. Höglund, T., Om BSK och Detal jutformning av Stalkonstruktioner. FoU bygg/Väg—och vattenbyggaren. 1983, no
10, pp. 11–15, ISSN 0042–2177.
5. Granström, A., Bolted End-plate Connections II. EHS Steel & Asymmetric Bolt Locations, report 86:5, Swedish
Institute of Steel Construction, Stockholm, September 1984, ISBN 91–85644–37–4.
6. Granström, A., The Strength of Bolted End-plate Connections. Ultimate Design Method & Lamellar Tearing
Consideration, report 15:13, Swedish Institute of Steel Construction, Stockholm, February 1979, ISBN 91–85644–03-X.
7. ECCS Commission IX, Recommendations for Steel Constructions. ECCS-EG-76–1E, Vol. I, June 1976, p 153.
8. IIW Commission XV, Design Rules for Welded Joints with Gaps Loaded in Compression,, Welding in the World.
1977, 15, No 3/4, pp 73–78, ISSN 0043–2288.
< previous page page_25 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_25.html07/11/2009 12:04:10 PM
 page_26

< previous page page_26 next page >

Page 26
EXPERIMENTAL STUDY OF THE NON-LINEAR BEHAVIOUR OF BEAM-TO-COLUMN BOLTED JOINTS
J.Janss
Centre de Recherchesde l’Industrie des Fabrications Métalliques (CRIF)
Liège
BELGIUM
J.P.Jaspart
and
R.Maquoi
M.S.M., Institut du Génie Civil, Université de Liège
BELGIUM
ABSTRACT
The paper is aimed to compare the behaviour of several types strong axis of beam-to-column bolted joints, which are
commonly used in practice.
INTRODUCTION
An experimental investigation was recently performed at the University of Liège, with the financial support of IRSIA
and ARBED [1]. Three types of connections between a beam and a column flange (strong axis connection) were
considered: i) end plates, ii) double web cleat, and iii) flange cleats (fig. 1). All the connections are bolted ones with
bolts of quality 10.9 (H.S.bolts), preloaded to 0.8 times the material yield stress or, in some cases, with a lower torque
moment (controlled hand tightening). The basic data are listed in table 1; all the mechanical and geometrical properties
of specimens were measured but are not listed here. The testing arrangement is shown in figure 2. The load P is
increased step by step either up to collapse of the connection, or up to a maximum deflection (20 cm) of the cantilever
end. The deflections are measured at the ends of the column and cantilever beam, as well as the slopes of beam and
column axis in the joint. From these measurements, the relative rotations of the connection, of the joint and that due to
the shear flexibility of the column web are deduced (fig. 3).
< previous page page_26 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_26.html07/11/2009 12:04:13 PM
 page_27

< previous page page_27 next page >

Page 27
Table 1—Basic data for the test specimens
Test Column Beam Relative Stiffness Connection Loading Axial force Bolt tight. Clearance d
no C B B/C (fig. 2) N (mm)
01 HEB160 IPE200 0, 78 a S 300kN C –
02 HEB160 IPE200 0, 78 b S – C 0
03 HEB160 IPE200 0, 78 c S – C 15
04 HEB160 IPE200 0, 78 a S 700kN C –
05 HEB160 IPE200 0, 78 b S – C 15
06 HEB160 IPE200 0, 78 c S – M 15
07 HEB160 IPE200 0, 78 a S – C –
08 HEB160 IPE200 0, 78 b Cyc – C 0
09 HEB160 IPE200 0, 78 c S – C 15
010 HEB160 IPE300 3, 35 a S – C –
011 HEB160 IPE300 3, 35 b S – C 15
012 HEB 160 IPE300 3, 35 c S – C 15
013 IPE240 IPE200 0, 50 a S – C –
014 IPE300 IPE200 0, 23 a S – C –
015 HEB160 IPE200 0, 78 a Cyc – C –
016 HEB160 IPE200 0, 78 b S – M 15
017 HEB160 IPE200 0, 78 c Cyc – C 15
018 HEB160 IPE300 3, 35 a Cyc – C –
019 HEB160 IPE300 3, 35 b Cyc – C 0
020 HEB160 IPE300 3, 35 c Cyc – C 15
021 IPE240 IPE200 0, 50 a Cyc – C –
022 IPE300 IPE200 0, 23 a Cyc – C –
-: The axial force is negligible or d has no meaning
S: static loading; Cyc: cyclic loading
C: preloading to 0.8 fy; M: controlled hand tightening

Figure 1—Types of joints

< previous page page_27 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_27.html07/11/2009 12:04:15 PM
 page_28

< previous page page_28 next page >

Page 28

Figure 2—Testing arrangement

Figure 3— Curve: a) of the connection; b) of the joint; c) of the web column—(end plates).
RESULTS OF THE STATIC TESTS
Connections with end plates.
The relative rotations of the connection and of the joint are nearly linear at the beginning of the loading (fig. 3.a,b). Then
occurs yielding, which is followed by a strain hardening of quasi constant stiffness. The column web behaviour is
different at all; the web is first nearly infinitely rigid, then yields rather suddenly (fig. 3c). From the test results
< previous page page_28 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_28.html07/11/2009 12:04:17 PM
 page_29

< previous page page_29 next page >

Page 29
the influence of the main parameters can be summarized as follows:
a) The axial force has no sensible effect on the yield shear force in the range i.e.
while the yielding moment of the column flange is affected so that the linear portion of the curve is reduced; this
observation is in accordance with Packer’s and Morris’ conclusions [2] but in contradiction with Zoetemeijer’ and
Munter’ ones [3];
b) The initial stiffness of the joint seems just be little influenced by the axial force;
c) It is necessary to distinguish between both behaviours of the column web and of the connection and to modelize them
separately, as it appears clearly from figures 4. a-c. In test 07, the column web and the connection contribute the
rotation in similar proportions, while, in test 010, the deformation is mainly due to the shear flexibility of the column
web, which is weaker than the connection. It can also be observed that the use of the full elastic strength of the beam
would imply a marked non linear behaviour of the joint.

Figure 4—Curve: a) of the connection; b) of the joint; c) of the web column—(end plates).

d) The initial stiffness in tests 07 and 014—specimens fitted with same end plates—is shown dependent on the
dimensions of the column flange (fig. 5). The more marked non linear behaviour of specimen 014 is due to a weaker
column flange and to the yielding of the compressed zone of the web column. With respect to similar stiffness,
specimen 013 benefits from the stronger rigidity of IPE240 column flange. Unlike test 07, the collapse of the
connections 013 and 014 is reached, due to yielding in the compressed part of the column web, and the associated
rigidity becomes nearly zero.
< previous page page_29 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_29.html07/11/2009 12:04:19 PM
 page_30

< previous page page_30 next page >

Page 30

Fig. 5-Connection curves for tests 07–013–014.

Fig. 6-Effect of bolt tightening and of beam depth on connection curves (double web cleat)

Fig. 7-Connection curve for test 011 (double web cleat).

Connections with double web cleat.
The response of the tests carried out on such connections is quite different from the previous one. The linear behaviour
with a nearly constant stiffness is followed by successive slips of the bolts due to hole clearance. When all the bolts are
bearing, the holes are ovalized and the cleats distort and yield. Following conclusions can be drawn:
a) The curves corresponding to similar specimens 05 and 016 (d=15 mm) are plotted in figure 6. Only the bolt
tightening is different: preload to 0.8 fy for specimen 05 and controlled hand tightening for specimen 016. The level
of tightening does not alterate neither the initial stiffness, nor the behaviour after slip; only the bending moment
associated to the first slip varies and does it proportionnally to the preload in the bolts (fig. 6).
b) By reducing the clearance d (fig. 1) from 15 mm to zero (specimen 02), the transmission of the applied bending
moment is modified with the result of a much higher initial stiffness (fig. 6). Indeed compression is transmitted by
direct bearing between beam-and column flanges, while tension is supported by friction, then by shear after slip
occurs, and last by bearing pressure. The non linearity before slip is due to distorsion and yielding of the angles in the
tensile zone, and last by the slip initiation in the upper bolts in the beam.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_30.html (1 of 2)07/11/2009 12:04:21 PM

 page_30

c) A deeper beam (IPE300) allows the use of deeper angles with four bolts (specimen 011), with the result of an initial
stiffness nearly equal to that of specimen 010 (connection with end plates). The clearance d=15mm allows the beam-
and column flanges bearing and explains the increase of rigidity at the end of the test (fig. 7).
< previous page page_30 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_30.html (2 of 2)07/11/2009 12:04:21 PM

 page_31

< previous page page_31 next page >

Page 31
Connections with flange cleats.
The curves dealing with tests no 03–06–012 are shown in figure 8. It can observed that:
a) The initial stiffness of tests 03 and 06 is not affected by the level of tightening and that the ultimate strength is quasi
independent of this level. However, the slip is initiated earlier for test 06.
b) With a deeper beam, (test 012), the initial stiffness does not change significantly; so for the overall shape of the
curve.

Fig. 8— curves for tests 03–06–012 (flange cleats).

Fig. 9—Comparison of curves for 3 types of connections.

Comparison between different types of connections.
Figure 9 shows for instance the curves associated respectively to the three types of connections for a joint between
a HE160B column and an IPE300 beam.
CYCLIC TESTS
The cyclic tests were performed in accordance with a standard procedure [5]. At the present time, the results have not yet
been fully investigated. Therefore only a cyclic test is compared to a static one, both being carried out on identical
specimens (fig. 10).
< previous page page_31 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_31.html07/11/2009 12:04:23 PM
 page_32

< previous page page_32 next page >

Page 32

Figure 10—Comparison between static and cyclic tests.

MODELLING OF CURVES
An ambitious goal would be to define the curve of a joint from the whole geometry and the mechanical properties
of the connection and of the connected elements. Such an attempt was made for connections with end plates [4] but has
to be somewhat modified [6] and improved in order to be in a quite satisfactory agreement with experimental results.
The authors are intended to contribute such improvements and extend the procedure to other kinds of connections and to
shear column webs.
REFERENCES
1. Jaspart, J.P., Essais sur assemblages poutre-colonne d’axe fort. CRIF, Bruxelles (in preparation).
2. Packer, J.A. and Morris, L.J., A Limit State Design for the Tension Region of Bolted Beam-to-Column Connections.
The Structural Engineer, Vol. 55, 10, October 1977, pp. 446–458.
3. Zoetemeijer, P. and Munter, H., Influence of an Axial Load in the Column on the Behaviour of an Unstiffened Beam
to Column End Plate Connection. Stevin Laboratory Report No 6–84–1, Delft University of Technology, January 1984.
4. Yee, Y.L. and Melchers, R.E., Moment-Rotation Curves for Bolted Connections. Jl. Struc. Engrg., A.S.C.E. Vol. 112,
3, March 1986, pp. 615–635.
5. E.C.C.S., Recommended Testing Procedure for Assessing the Behaviour of Structural Steel Elements under Cyclic
Loads. E.C.C.S. Publication no 45, 1986.
6. Maquoi, R. and Jaspart, J.P., Moment rotation Curves for Bolted Connections—Discussion of [4]. (to appear in Jl.
Struct. Engrg.).
< previous page page_32 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_32.html07/11/2009 12:04:24 PM
 page_33

< previous page page_33 next page >

Page 33
PREDICTION OF MOMENT-ROTATION BEHAVIOR OF SEMI-RIGID BEAM-TO-COLUMN CONNECTIONS
James B.Radziminski
Department of Civil Engineering
University of South Carolina
Columbia, SC 29208
USA
and
Atorod Azizinamini
Construction Technology Laboratories, Inc.
Skokie, IL 60077
USA
ABSTRACT
This paper presents the results of a study of the static behavior of semi-rigid steel beam-to-column connections. From the
experimental investigation, the geometric parameters which most significantly affect the moment-rotation performance
of the connections have been evaluated. An analytical model was developed to predict the complete non-linear static
moment-rotation behavior of the connections. The model was found to compare reasonably well with the test data for the
range of connection stiffnesses examined.
INTRODUCTION
The analysis of steel frame building structures with fully rigid beam-to-column connections is a straightforward one, and
buildings utilizing such framing systems have been used extensively in design practice. More recently, increasing
interest has been expressed in the use of semi-rigid (partially restrained) connections for building frames, spurred by
potential economies that may be realized from simpler connection details, and by “tuning” the connections (i.e.,
adjusting connection stiffnesses) to optimize the distribution of moments in the connected members.
Semi-rigid connections are characterized by a decidedly non-linear moment-rotation ( ) behavior which, in turn, can
contribute significantly to the performance of the complete structure, including strength limits and drift considerations
[1–4]. In certain instances, definition of the complete relationship may not be necessary, and an estimate of the
initial stiffness of the connection (initial slope of the curve) will suffice. For example, under service gravity live
load and wind load fluctuations, corresponding to a nominally elastic response of the structure, frame analyses using
linearized connection stiffnesses may be adequate [5,6]. Under extreme loading conditions, however, such as those
imposed in a seismic event, use of the full non-linear curve (complete hysteresis loop) is required to assess
structural performance.
In the study reported herein, the static moment-rotation behavior of connections consisting of bolted top and seat flange
angles, and double web angles, was investigated. The objectives of the study have been to experimentally determine the
geometric parameters which most significantly affect connection performance, and, from these data, to develop
analytical models to predict the static behavior. Models used to predict the
< previous page page_33 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_33.html07/11/2009 12:04:25 PM
 page_34

< previous page page_34 next page >

Page 34
initial stiffness of the connections are reported elsewhere [7]. In this paper, a semi-empirical model for predicting the
complete non-linear curve of the connections is presented.
EXPERIMENTAL INVESTIGATION
Description of Test Program
The test members consisted of a pair of beam sections attached to a centrally positioned stub column. The connection
elements were comprised of top and seat angles bolted to the flanges of the beams and supporting column, together with
double web angles bolted to the beam web and to the column flanges. The top and seat angles were the same for a given
test, and the web angles were centered on the beam. ASTM A36 steel was used for the members and connection
elements; all fasteners were 3/4-inch or 7/8-inch diameter ASTM A325 heavy hex, high strength bolts. A325 hardened
washers were used under the turned elements.
Two beam sizes, W14X38 and W8X21, were used in the test program. For the 14-inch sections, the overall beam length
was 20 feet, and for the 8-inch sections, 12 feet, so that the span-to-depth ratio was slightly less than 20 in each case. The
stub column for the W14X38 beams was a W12X96 section; a W12X58 column section was used with the W8X21
beams. Heavy column sections were selected to minimize column panel zone deformation, thereby confining the
deformations to the connection elements. Details of the W14X38 and W8X21 connections are shown in Figs. 1 and 2,
respectively.
A pair of duplicate specimens was tested simultaneously using the framing arrangement shown in Fig. 3. A 55 kip,
hydraulically actuated ram was used to apply load to the test members through the central stub column. In all of the tests,
the controlled input variables were the rate and magnitude of actuator displacement. Complete details of the test
procedures, including specimen preparation and types of instrumentation used, are reported in Ref. [8].

Figure 1. Details of Connection for W14X38 Beam

< previous page page_34 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_34.html07/11/2009 12:04:26 PM
 page_35

< previous page page_35 next page >

Page 35

Figure 2. Details of Connection for W8X21 Beam

Figure 3. Schematic of Test Member and Loading System

Test Results
Eighteen specimens were tested in the experimental investigation. The geometric properties that were varied in the
parametric study included: the depth of the beam sections, the thickness and length of the beam flange angles, the gage
and spacing of the bolts in the leg of the flange angles connected to the column flange, the thickness and length of the
web angles, and the bolt diameter. Details of the geometries of the individual test specimens are reported in Table 1.
A summary of the test results is presented in Table 2. Table 2 includes the initial stiffness of the connection, the slope
and intercept moment of a secant line from the origin intersecting the curve at a rotation of 4X10−3 radians, and the
moment and tangent slope at 24X10−3 radians. The latter slope, although not intended to represent a final, constant slope
for a specific connection, does allow comparisons to be made among the various connections, as well as quantifying the
degree of softening of a connection relative to its initial stiffness. Similarly,
< previous page page_35 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_35.html07/11/2009 12:04:27 PM
 page_36

< previous page page_36 next page >

Page 36
the reported secant slope and moment are indicative of the relative behavior of the connections during the early part of
the loading cycle.
Table 1
Schedule of Test Specimens
Top and Bottom Flange Angles Web Angles
Specimen Beam Bolt Diameter Angle Length Gage in Leg Angle Length
Number Section (inches) (inches) on Column (inches)
Flange
(inches)
14S1 W14x38 3/4 L6x4x3/8 8 2 1/2 2L4x3 8 1/2
1/2x1/4
14S2 W14x38 3/4 L6x4x1/2 8 2 1/2 2L4x3 8 1/2
1/2x1/4
1453 W14x38 3/4 L6x4x3/8 8 2 1/2 2L4x3 5 1/2*
1/2x1/4
1454 W14x38 3/4 L6x4x3/8 8 2 1/2 2L4x3 8 1/2
1/2x3/8
851 W8x21 3/4 L6x3 6 2 2L4x3 5 1/2
1/2x5/16 1/2x1/4
852 W8x21 3/4 L6x3 1/2x3/8 6 2 2L4x3 5 1/2
1/2x1/4
853 W8x21 3/4 L6x3 8 2 2L4x3 51/2
1/2x5/16 1/2x1/4
854 W8x21 3/4 L6x6x3/8 6 4 1/2 2L4x3 5 1/2
1/2x1/4
855 W8x21 3/4 L6x4x3/8 8 2 1/2 2L4x3 5 1/2
1/2x1/4
856 W8x21 3/4 L6x4x5/16 6 2 1/2 2L4x3 5 1/2
1/2x1/4
857 W8x21 3/4 L6x4x3/8 6 2 1/2 2L4x3 5 1/2
1/2x1/4
1455 W14x38 7/8 L6x4x3/8 8 2 1/2 2L4x3 8 1/2
1/2x1/4
14S6 W14x38 7/8 L6x4x1/2 8 2 1/2 2L4x3 8 1/2
1/2x1/4
1458 W14x38 7/8 L6x4x5/8 8 2 1/2 2L4x3 8 1/2
1/2x1/4
14S9 W14x38 7/8 L6x4x1/2 8 2 1/2 2L4x3 8 1/2
1/2x1/4
858 W8x21 7/8 L6x3 6 2 2L4x3 5 1/2
1/2x5/16 1/2x1/4
8S9 W8x21 7/8 L6x3 1/2x3/8 6 2 2L4x3 5 1/2
1/2x1/4
8510 W8x21 7/8 L6x3 1/2x1/2 6 2 2L4x3 51/2
1/2x1/4
*Two Bolts at 3-inch spacing, Mounted in top two holes on stub column. Fig. 1.
Table 2
Summary of Test Results
Specimen Initial Slope of Slope of Secant Line to Moment at Slope of Curve at Moment at
Number Curve Curve at 4.0x10−3 4.0x10−3 24x10–3 radians 24x10−3
radians radians (k-in.) radians (k-in.)

14S1 195.0 108.7 435 5.8 688

14S2 295.0 151.8 607 12.6 (947)*
14S3 115.9 88.8 355 7.2 652
14S4 221.9 124.0 496 8.3 822

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_36.html (1 of 2)07/11/2009 12:04:29 PM

page_36

14S5 247 117.4 467 10.8 763

14S6 286 146.9 589 10.3 1053
14S8 579 256.9 1027 15.0 1561
1459 258 154.1 624 9.3 1024
8S1 66.7 44.3 177 4.1 329
8S2 123.4 69.0 276 1.5 (384)**
8S3 104.7 64.3 257 4.0 422
8S4 15.3 14.4 57.5 2.2 165
855 76.7 47.9 191.5 2.7 337
856 39.5 30.0 120 3.2 244
8S7 48.0 40.8 163 3.2 381
858 70 44.4 183 3.8 380
8S9 104 54.5 226 4.4 423
8S10 427 104.1 434 6.9 (634)***
* Major slip at 12x10−3 and 20x10−3 radians
** Major slip at 16x10−3 radians
*** Major Slip at 5.3x10–3 radians
< previous page page_36 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_36.html (2 of 2)07/11/2009 12:04:29 PM

 page_37

< previous page page_37 next page >

Page 37
The complete moment-rotation curves for several of the test connections are shown in Figs. 4 and 5. In all of the tests,
the connections exhibited an response that became non-linear early in the loading sequence. This is attributed,
primarily, to local yielding and eventual plastic hinge formation at each toe of the fillet in the angle attached to the
tension flange of the beam. Another hinge developed in the vicinity of the bolt line on the leg of the flange angle
attached to the column, together with progressive plastic hinging in the outstanding legs of the web angles. With the
exception of three specimens that exhibited slip in the connection angles, all of the test connections were able to develop
continually increasing moments through the full range of rotations imposed during the tests. During the latter period of
loading, a nearly constant or only very gradually decreasing positive slope of the moment-rotation curve was exhibited
by all but one of the specimens.

Figure 4. Typical Moment-Rotation Curves for W14X38 Specimens

Figure 5. Typical Moment-Rotation Curves for W8X21 Specimens

< previous page page_37 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_37.html07/11/2009 12:04:30 PM
 page_38

< previous page page_38 next page >

Page 38
From the experimental investigation [7, 8], it was found that, of the geometric parameters studied, those that most
significantly affect the initial stiffness and moment-rotation performance of the connections are: The depth of the beam
section to which the connections were framed, the thickness of the flange angles, and the gage (including the effect of
bolt diameter) in the leg of the flange angle attached to the column. Variations in the length of the flange angles, and the
length and thickness of the web angles, had a less pronounced effect on connection response.
PREDICTION OF MOMENT-ROTATION BEHAVIOR
In 1969, Sommer [9] developed a technique to predict the moment-rotation curve for flexible type connections. The use
of Sommer’s method can lead to predictions of a negative slope. To overcome this problem, a modified procedure
was developed [7]; the analytical model, which uses the test data from the parametric study described above, is of the
form:

[1]
where: =rotation of end of beam with respect to column face
M=moment developed by beam-to-column connection
Ci=empirically determined coefficients

and
where: Pi=geometric parameters affecting relationship

The parameters (P) selected for the model were the depth of the supported beam (d), the thickness (t) and length (1) of
the flange angles, the ‘‘effective” gage in the leg of the flange angles attached to the column flange (where g
is the gage from heel of angle and db is the bolt diameter), and the thickness (tc) of the web angles.
Using the curve-fitting process described in Refs. [7,8], the following numerical values of the exponents αi and
coefficients Ci were obtained:
P1=t α1=−1.1280877
P2=d α2=−1.2870455
P3=tc α3=−0.41454097
P4=1 α4=−0.69412158
P5=g−db/2 α5=1.34994572
and C1=0.2232429X10−4
C2=0.1850728X10−7
C3=0.3188976X10−11
Comparisons of the moment-rotation curves predicted by Eq. [1] with the test results for several test specimens are
shown in Figs. 6,7. In general, with the exception of the stiffest connections in both the W14X38 and W8X21 test series,
the model offered reasonable approximations to the moment-rotation curves of the connections up to the limits of
rotation examined in the test program. For the stiffest connections of each beam size, the
< previous page page_38 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_38.html07/11/2009 12:04:34 PM
 page_39

< previous page page_39 next page >

Page 39
model underestimates the moments developed at the larger connection rotations. It is recommended, therefore, that Eq.
[1] be restricted to connections in which the flange angle thickness is of the same order as the thickness of the beam
flange; use of this expression should not be extrapolated to predicting the response of significantly stiffer
connections nor to connections consisting of top and seat angles only.

Figure 6. Predicted Moment-Rotation Curves for W14X38 Specimens

Figure 7. Predicted Moment-Rotation Curves for W8X21 Specimens

SUMMARY, CONCLUSIONS
A test program was conducted to evaluate the static moment-rotation performance of semi-rigid beam-to-column
connections consisting of top and seat flange angles, and double web angles. Initial stiffnesses and complete curves
were quantified for connections of varying geometry.
< previous page page_39 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_39.html07/11/2009 12:04:36 PM
 page_40

< previous page page_40 next page >

Page 40
From the test program it was found that, for the type of connection studied, the geometric parameters that most
significantly affect moment-rotation behavior are: the depth of the beam section to which the connection elements are
framed, the thickness of the flange angles, and the effective gage in the leg of the flange angles attached to the column.
Using the test results, a parametric model was developed to predict the complete non-linear static moment-rotation
behavior of the connections. The model offers reasonable predictions of the response of the connections; however,
for the stiffest connections tested with each beam size, it underestimates the moments developed at the larger rotations.
Until additional data are available and further refinements to the model are possible, it is recommended that the
expression reported herein be confined to the estimation of curves for connections in which the thickness of the
flange angles does not exceed appreciably the thickness of the flange of the supported beam.
ACKNOWLEDGEMENT
The support of the National Science Foundation, under Grant Nos. PFR-7923520 and CEE-8115014, is gratefully
acknowledged.
REFERENCES
1. Frye, M.J., and Morris, G.A., “Analysis of Flexibly-Connected Steel Frames,” Canadian Journal of Civil Engineers,
Vol. 2, 1975.
2. Akbari, F., “Drift of Flexibly Connected Frames,” Computers and Structures, Vol. 15, No. 2, 1982.
3. Ackroyd, M.H. and Gerstle, K.H.., “Behavior of Type 2 Steel Frames,” Journal of the Structural Division, ASCE, Vol.
108, No. ST7, Proc. Paper 17207, July 1982.
4. Lindsey, S.D., Ioannides, S.A., and Goverdhan, A.V., “The Effect of Connection Flexibility on Steel Members and
Frame Stability,’’ in Connection Flexibility and Steel Frames, W.-F.Chen, ed., ASCE, 1985.
5. “Type 2 Construction with Wind Moment Connections: A Return to Simplicity,” American Iron and Steel Institute,
Washington DC, 1976.
6. Stelmack, T.W., Marley, M.J., and Gerstle, K.H., “Analysis and Tests of Flexibly Connected Steel Frames,” Journal of
Structural Engineering, ASCE, Vol. 112, No. 7, July 1986.
7. Azizinamini, A., “Monotonic Response of Semi-Rigid Steel Beam to Column Connections,” M.S. Thesis, University
of South Carolina, Columbia , SC, July, 1982.
8. Azizinamini, A., Bradburn, J.H., and Radziminski, J.B., “Static and Cyclic Behavior of Semi-Rigid Steel Beam-
Column Connections,” Structural Research Studies, Department of Civil Engineering, University of South Carolina,
Columbia, SC, March 1985.
9. Sommer, W.H., “Behavior of Welded Header Plate Connections,” M.S. Thesis, University of Toronto, Ontario, 1969.
< previous page page_40 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_40.html07/11/2009 12:04:37 PM
 page_41

< previous page page_41 next page >

Page 41
EXPERIMENTAL ANALYSIS OF END PLATE CONNECTIONS
R.Zandonini and P.Zanon
Department of Engineering, University of Trento
Mesiano di Povo—I-38050 Trento, Italy
ABSTRACT
The prediction of joint rotational behaviour is a preliminary step in the analysis of semi-rigid frames. Mechanical models
may be used and the behaviour of the whole joint obtained by “summing up” the contributions of the individual
components. In order to develop simple models suitable to predict such component contributions results from ad hoc
experimental research are needed. The paper presents an experimental technique adopted for that purpose. Results
related to extended end plate connections are reported and their evaluation in order to characterize the moment rotation
relationship discussed.
INTRODUCTION
The great majority of the investigations related to steel joints were, till recently, aimed at defining their ultimate strength
capacity. The knowledge of the behaviour of the joints over its complete range were not of particular interest. Frame
design was in fact based on the simple ideal models of rigid and pinned joints. Late research work, both experimental
and numerical [1–5], contributed to a better understanding of the influence of joint flexibility on the performance of the
whole frame. The possibility of considering the joint behaviour as a design parameter, allowing advantages in terms of
cost effectiveness and/or structural reliability to be achieved, is indicated. The practicality of such an approach requires
the capability of predicting the joint response.
This problem seems not yet solved. In effect, the number of governing mechanical and geometrical parameters (even for
connections with simple detailing), suggests the suitability of an approach, which separate the contribution to the whole
joint behaviour of its individual components (fig. 1): i.e. column panel zone (fig. 1b), column flanges (fig. 1c),
connection elements (fig. 1d). The joint response is then obtained by “summing up” these contributions (fig. 1e). The
feasibility of this procedure has already been established [6,7]. Moreover, such an approach has the advantage to favour
a straight definition of the joint “rotation”, a quantity which was given in the past a nonidentical meaning by different
researchers.
< previous page page_41 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_41.html07/11/2009 12:04:38 PM
 page_42

< previous page page_42 next page >

Page 42

Fig. 1
The authors are conducting a research aimed at developing simple mechanical models, allowing for the prediction of the
joint moment rotation curves. With reference to this philosophy by components, the first phase of the study is devoted to
the behaviour of the sole connections [8]. This paper intends to highlight the experimental technique adopted and present
a first series of results related to extended end-plate connections. Their evaluation is finally discussed in order to point
out the parameters useful for the prediction and to single out the contribution of the different components (end plate and
bolts). This data gives the preliminary information needed to select suitable mechanical models and represent the
necessary benchmark to their validation.
ROTATION OF A CONNECTION AND COMPONENTS CONTRIBUTION
The rotation of the connection (fig. 1d) can be defined with reference to the end of the beam, as indicated in figure 2,
under the assumption that this cross section remains plane. In reality, because of the yielding induced by the high contact
pressures and/or by the interaction with the end plate in the tension zone, this assumption is generally not fulfilled.

Fig. 2
Nevertheless, it seems the most appropriate for its practicality, due to the simplicity and the consistency with the
classical beam theory. In the same figure a criterion is graphically defined, which enables a singling out of the
contribution to the connection rotation ( ) of the deformability of the end plate ( ) and of the bolts ( ).
TESTS ON EXTENDED END PLATE CONNECTIONS
The geometrical characteristics of the connections tested are reported in figure 3. End plates with extension on one side
or on two sides of the beam were considered. Within the same type of connection only the plate thickness was varied,
allowing different ratios between the plate and bolts stiffness to be obtained.
< previous page page_42 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_42.html07/11/2009 12:04:40 PM
 page_43

< previous page page_43 next page >

Page 43
The specimen was connected to a counterbeam with negligible deformability. Displacement transducers located as in
figure 4 made it possible to determine: a) the connection rotation (four transducers A in fig. 4a); b) the rotation of beam
cross section at 300 mm from the plane of the connection (two transducers B in fig. 4a); c) the rotation due to bolt
deformation (four transducers C in fig. 4b). Finally, displacement transducers with the sensitive pin entering a concentric
hole in the tension bolts (fig. 5) enabled a control on the initial tightening conditions and a following of the bolt axial
deformation. Bolts were pretensioned up to a force equal to the 40% of the nominal tensile strength; this state of stress
roughly approximates the snug tight condition. The specimens were subject to a horizontal force applied to the upper end
of the beam stub so to give rise in the connection to bending moment and shear. In order to obtain a realistic appraisal of
the connection response, the load is increased by a series of subsequent loading and unloading cycles.

Fig. 3

Fig. 4a Fig. 4b

Fig. 5
EXPERIMENTAL RESULTS AND THEIR EVALUATION
The rotational behaviour of the overall connection is described by the moment—rotation characteristics. The figure 6

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_43.html (1 of 2)07/11/2009 12:04:42 PM

 page_43

gather all the relationships obtained for the specimens tested, grouped by type of connections. The average

value of the plastic moments of the beam is also reported, determined on the base of the experimental yield
strength and the actual dimensions of the cross section. It may be remarked that all the connections showed a moment
capacity greater than the plastic
< previous page page_43 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_43.html (2 of 2)07/11/2009 12:04:42 PM

 page_44

< previous page page_44 next page >

Page 44
moment of the beam, though their ultimate strength increased with the end plate thickness from 181.5 kNm (EP1–1) to
225.7 kNm (EP1–5) for the EP tests and from 202.8 kNm (EPB1–2) to 227.5 kNm (EPB1–5) for the EPB tests. A
comparison between the curves related to the corresponding specimens of the two series shows that the presence of the
extension of the end plate on the compression side has a limited influence on the connection behaviour and strength.

Fig. 6
The significant strains in the beam stub, within the hardening range, caused the inelastic buckling of the compressive
flange (and of the adjacent zone of the web for the specimens with thicker end plate) before the moment capacity of the
connection was achieved. The rotation capacity is at the contrary adversely affected by the increase of the end
plate thickness: decreases respectively from 0.08 to 0.02 rad (EP connections) and from 0.05 to 0.01 (EPB
connections).
Although these differences in strength and ductility, the moment rotation relationships do have a very similar shape,
characterized by a linear elastic response up to a moment value ranging from the 40% (thin plates) to the 60% (thick
plates) of the ultimate capacity followed by a behaviour remarkably nonlinear. The shape and the extension of the curve
in the latter range depends upon the connection component, which principally is affected by yielding. When the end plate
is relatively thin, the connection rotation in the nonlinear range is mainly due to the elastic-plastic bending of this
component (fig. 7): for the connection EP1–1 the end plate contributes more than the 90% of the overall rotation. The
rotation capacity of the whole connection is then substantially affected by the local ductility of the tension zone of the
end plate. This

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_44.html (1 of 2)07/11/2009 12:04:43 PM

 page_44

Fig. 7
< previous page page_44 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_44.html (2 of 2)07/11/2009 12:04:43 PM

 page_45

< previous page page_45 next page >

Page 45
ductility is generally very high, but may be considerably reduced as a consequence of the process of welding the beam
flange and the plate, which modifies the material characteristics in a zone were important plastic strains should develop.
The contribution of the bolts becomes more and more significant (up to about 35% for the test EP1–5) with the increase
of the end plate thickness. The limited ductility, typical of high strength bolts, tends then to affect negatively the rotation
capacity of the overall connection.
The measures related to the bolts elongation pointed out some interesting features of their behaviour; in particular, as it
appears from the curves in figures 8 and 9, related to the average deformation histories of the bolts external and internal
to the beam stub respectively for the connections and :

Fig. 8a

Fig. 8b

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_45.html (1 of 2)07/11/2009 12:04:45 PM

 page_45

Fig. 9a

Fig. 9b
a) The initial preload significantly decreases, and often tends to vanish, since the very first loading cycles, though the
bolt stress remains in the
< previous page page_45 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_45.html (2 of 2)07/11/2009 12:04:45 PM

 page_46

< previous page page_46 next page >

Page 46
elastic range well below the elastic limit. This loss of preloading seems then to depend principally upon the permanent
deformation of the washers.
b) The bolts internal to the beam are substantially more stressed than the external bolts. However, the increase of the end
plate thickness reduces this difference in the stress state.
c) The end plate thickness affects the type of the bolt stress state, as a comparison between the curves related to the
external bolts (figg. 8b for the test EP1–1 and 9b for the test EP1–5) already suggest. The external bolts in the specimen
EP1–1 experienced remarkable plastic deformations, while their behaviour in the specimen EP1–5 shows a limited
nonlinearity, although the ultimate strength of the connection is attained under a higher applied moment. If an attempt is
made to determine the axial force in the bolt by relating the deformation measured during the connection tests to the
constitutive law determined via tensile tests on individual bolts, the results indicate that yielding in the bolts of
connections with thin end plates is beginning when the axial load is noticeably lower than the yield strength value
deduced from the tensile tests. This discrepancy is on the contrary modest for the connections with the thickest end
plates These results point out that the significant deformations of thin end plates tend to
subject the bolts to a nonegligible bending, as in effect experimentally observed.
The presence of this bending moment greatly influences the yield and the ultimate strength of the bolt and therefore
explains the above mentioned differences in the behaviour of the external bolts of tests EP1–1 and EP1–5. As to the
internal bolts, the plate, due to its greater stiffness, imposes at the same time higher axial forces but lower bending
moments; the type of behaviour is then less affected by the plate thickness (compare also figures 8a and 9a).
The knowledge of the rotation of the beam cross section at 300 mm from the connection plane (fig. 4a) allows also the
flexural response of the end part of the beam to be determined. The plastic deformation of the beam was always
important, as it appears from figure 10. These results emphasize the need that, in experimental research, rotation
measuring techniques are adopted, which allow the flexibility inherent to the sole connection response to be singled out
from that of the end part of the beam.

Fig. 10
The results as a whole represent a consistent set of data, which makes it possible to draw some first general conclusions
on the mechanical models more appropriate for prediction purposes. The additional restraint, given to the end plate in the
zone internal to the beam by the presence of the beam web, influences to a large extent the connection behaviour. Such a
restraint causes a nonnegligible dissymmetry, with respect to the beam tension flange, both of the plate deformation and
of the stress state of the same plate and of the bolts. Consequently, the models which symmetrize the plate and bolts
behaviour.
< previous page page_46 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_46.html07/11/2009 12:04:46 PM
 page_47

< previous page page_47 next page >

Page 47
such as the T stub model, may not always be able to predict the connection response with satisfactory accuracy. In
particular, they cannot (at least they cannot directly) recognize the different stress state of the internal and external bolts.
The complexity of the observed behaviour suggests that a greater accuracy can be achieved by models, which account in
a more realistic way for the end plate behaviour, by dealing separately with the external and the internal part. The former
part can be simply modelled as a cantilever element clamped at the beam flange and subject to the bolt reaction and to
the possible prying forces. The part internal to the beam has to be considered as a plate with appropriate edge restraints.
A first validation of this modelling procedure is given by the results related to the prediction of the connection plastic
moment (see the folowing section). The plastic mechanism presented in figure 11 (and obtained by modifying the
proposal of reference 9) allows this moment to be determined with an error lower than the 9%. For thick end plates the
plastic mechanism of the internal part of the plate may not be activated, because it would correspond to a force in the
bolt greater than its tensile strength. A further improvement may be obtained, by refining, on the base of the
experimental data, a method proposed by one of the authors [10].

Fig. 11
MODELLING OF THE M—Φ CHARACTERISTICS
The use of simple mechanical models in order to predict the moment rotation relationships requires the definition of the
parameters which can be regarded as “characteristics’’, i.e. suitable to describe the rotational behaviour of the
connection. The model accuracy is then checked on the base of its capability of evaluating these parameters with
satisfactory precision. The experimental results were therefore thoroughly analysed with the aim at defining the
characteristic parameters for end plate connections.
Figure 12 shows the rotational response of a specimen to the first loading cycles. The effect of the loss of the preloading
of the bolts on the overall connection behaviour is apparent. In the initial phase, two different stiffness values
characterize the response related respectively to the conditions of plate in contact (ki) and of plate separated from the
flange of the counterbeam (ki,red). The loading (and unloading) branch defined by the stiffness ki shortens up to vanish
after a few cycles. The stiffness ki,red (which remains almost constant for all the loading and unloading branches)
characterizes completely the elastic branches of the subsequent cycles. The inelastic behaviour is highly nonlinear and
shows a remarkable strain hardening; however, it seems possible to identify a characteristic value of the stiffness, kst,
defined as the mean value of the stiffnesses of all the inelastic parts of the loading branches (fig. 12).
< previous page page_47 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_47.html07/11/2009 12:04:47 PM
 page_48

< previous page page_48 next page >

Page 48
The values of the stiffnesses ki, ki,red, and kst determined for the connections tested are reported in table 1, where also
the contributions of the end plate and of the bolts are singled out.
In order to describe completely the connection characteristic it is also necessary to define the values of the elastic limit
moment Me and of the plastic moment Mp. The latter definition is not clear-cut, when based on the experimental results.
It seems appropriate to refer to a conventional value and define the plastic moment as the moment at which a significant
increase of the plastic deformation of the connection becomes apparent. This moment level may be practically obtained
by means of the stiffness parameters previously defined. Two boundary values of Mp (Mp,inf, and Mp,sup) may then be
determined on the base respectively of the initial stiffness ki and its reduced value ki,red, following the procedure
illustrated in figure 13. The two evaluations of the plastic moment differ only slightly for the connections tested; the
mean value was then assumed and is reported in table 2. The table 2 gathers the values of Me and Mp; the ultimate
moments Mu of the connection, the plastic moment of the beam stub Mp,b, and the rotation corresponding to Mu are
also reported.

Fig. 12

Fig. 13

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_48.html (1 of 2)07/11/2009 12:04:48 PM

 page_48

Fig. 14
The knowledge of the above mentioned parameters enables a modelling of the connection behaviour as illustrated in
figure 14. However, representation with higher level of accuracy can also be used.
An analysis of tables 1 and 2 provides useful indications for the development of mechanical models suitable for
predicting the response of the connection.
In table 1 the contributions to the connection stiffness supplied by the
< previous page page_48 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_48.html (2 of 2)07/11/2009 12:04:48 PM

 page_49

< previous page page_49 next page >

Page 49
end plate and by the bolts are singled out. These contributions were computed on the basis of the relationships
determined for the single components (fig. 10). The overall stiffness, defined with reference to the moment-rotation
curve of the whole connection, was verified to be in very good agreement with the value deduceable by considering the
two component as arranged “in series”. Values of ki are mean values related to the cycles with maximum applied
moment lower than the elastic limit moment. They are influenced by the bolt preloading and by initial contact
conditions. As to the reduced stiffness ki,red, it may be noticed that:
a) the bolt contribution depends on plate stiffness and increases with this parameter; b) the increment of the plate
contribution is less significant than expected on the basis of the variation in thickness, indicating an important influence
of the plate restraint conditions, which change with the plate flexibility (also in relation to the bolts flexibility); c) in all
cases the plate deformation contributes more to the joint flexibility; however the bolt contribution is never negligible.
Finally, as to stiffness kst it may be noticed that: a) no correlation exists between the bolt contribution and the plate
thickness; b) plate deformation is fundamental for the assessment of the overall stiffness. Elastic limit moments Me,cn
reported in table 2 range between the 55% and the 80% of the plastic moment Mp,cn and from the 40% and the 62% of
the ultimate moment of the connection Mu,cn. Plastic moments are an almost constant percentage (between the 70% and
the 80%) of the ultimate moment capacity and are close to the plastic moment of the beam.
Furthermore, if results are evaluated by component, as done for the stiffness parameters, the analysis of the curves
enables a determination of the levels of moment, for which the onset of yield is achieved for the plate and for the bolts.
Moment levels at which a sharp increase of the plastic deformation shows up for the two components can also be
defined.
CONCLUDING REMARKS
The capability of assessing the contribution given by each joint component, is important for the development of methods
suitable for the prediction of the joint response. An experimental procedure allowing the collection of the needed data for
the setting up of such an approach with reference to the sole connection is proposed.
The procedure was applied in a series of tests on end plate connections; the response of each component (i.e. the end
plate and the bolts) was singled out together with its influence on the connection behaviour. The results evaluation made
it possible to determine the main stiffness and strength parameters for the modelling of the connection behaviour. The
parameters “characteristic” of the overall relationship can be obtained by means of simple mechanical models for
the single components. The present phase of the research is devoted to the definition of such models
ACKNOWLEDGEMENTS
The research is made possible by a grant of the Italian Ministry of Education (M.P.I.).
< previous page page_49 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_49.html07/11/2009 12:04:49 PM
 page_50

< previous page page_50 next page >

Page 50
REFERENCES
1. Davison J.B., Kirby P.A. and Nethercot D.A., Column Behaviour in PR Construction—Experimental Behaviour,
ASCE Convention, New Orleans, September 1986.
2. Zandonini R., Steel Frames with Semi-Rigid Joints, Proc. 2nd Regional Colloquium on Stability of Steel Structures,
Vol. I/2, Budapest, September 1986, pp. 373–380.
3. Anderson D., Bijlaard F.S.K., Nethercot D.A. and Zandonini R., Analysis and Design of Steel Frames with Semi-
Rigid Connections, IABSE survey, to appear.
4. Cosenza E., De Luca A. and Faella C., Inelastic Buckling and Postcritical Behaviour of Steel Structures with Semi-
Rigid Joints, Proc. 2nd Regional Colloquium on Stability of Steel Structures, Vol. I/2, Budapest, September 1986, pp.
279–285.
5. Bijlaard F.S.K. and Zoetemeijer P., Influence of Joint Characteristics on the Structural Response of Frames, In Steel
Structures; Recent Research Advances and Their Applications to Design, ed. M.N.Pavlovic, Elsevier Applied Science
Publishers, London, 1986, pp. 109–134.
6. Yee Y.L. and Melchers R.E., Moment-Rotation Curves for Bolted Connections, J. of Structural Engineering, ASCE,
vol. 112, ST3, March 1986.
7. Klein H., Das elastisch—plastische Last-Verformungsverhalten M-Φ steifenloser, geschweisster Knoten fur die
Berechnung von Stahlrahmen mit HEB-Stutzen, Doctoral Thesis, University of Innsbruck, 1985.
8. Zandonini R. and Zanon P., Semi-Rigid and Flexible Connections: an Experimental Investigation, Proc. of the Int.
Conf. Steel Structures; Recent Research Advances and Their Applications to Design, Part I, Budva, Yugoslavia,
September 1986, pp. 273–282.
9. Surtees J.O. and Mann A.P., End Plate Connections in Plastically Designed Structures, Proc. Conf. on Joints in
Structures, Vol.1, University of Sheffield, July 1970.
10. Zanon P., Bolted Connections: the Ultimate Limit State Behaviour of End Plates, Costruzioni Metalliche, No. 1,
1985.
TABLE 1
ki(1),(3) ki,red(1) kst(1)
TEST Plate Bolts Conn. Plate Bolts Conn. Plate Bolts Conn.
EP 1–1 330 893 244 51 169 39 1.22 19.80 1.15
–2 353 850 249 64 146 45 1.60 10.00 1.38
–3 383 950 273 86 156 56 3.28 14.50 2.67
–4 408 872 277 99 186 65 5.20 13.50 3.76
–5 476 973 320 134 189 78 5.70 11.36 3.79
EPB 1–2(2) 415 788 272 75 144 49 1.77 8.00 1.45
–3 380 748 252 128 166 72 3.67 10.60 2.72
–4 441 894 295 148 202 85 8.00 9.20 4.29
–5 452 745 281 172 216 96 7.40 11.50 4.52
(1)All stiffnesses k are expressed in (KNm/rad)×10+3
(2)EPB 1–1 Is not reported btciust of a premature failure due to bolt stripping
(3)Values for tests EPB are obtained neglecting the effect of the initial bow of the end plate below the compression
flange.
< previous page page_50 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_50.html07/11/2009 12:04:50 PM
page_51

< previous page page_51 next page >

Page 51
TABLE 2
beam(1) Connection(1)
TEST Mp,b Me,cn Me,cn/Mu,cn Me,cn /Mp,cn Mp,cn Me,cn /Mu,cn Mp/Mp,b Mu,cn Me,cn /Mp,b u, cn
EP 1–1 176.1 73.6 0.406 0.622 118.4 0.652 0.672 181.5 1.03 0.08
1–2 175.8 81.8 0.410 0.555 147.3 0.738 0.838 199.5 1.13 0.05
1–3 170.6 98.1 0.480 0.628 156.3 0.765 0.916 204.4 1.19 0.03
1–4 173.3 130.8 0.588 0.765 170.9 0.768 0.986 222.4 1.28 0.02
1–5 176.7 139.0 0.616 0.781 177.9 0.788 1.007 225.7 1.28 0.02
EPB 1–2 175.8 89.9 0.443 0.558 161.2 0.798 0.917 202.8 1.15 0.05
1–3 170.6 98.5 0.482 0.591 166.7 0.816 0.977 204.4 1.19 0.02
1–4 173.3 106.3 0.485 0.608 174.7 0.797 1.008 219.1 1.26 0.01
1–5 176.7 122.6 0.539 0.705 173.9 0.764 0.984 227.5 1.29 0.02
(1)Moments M are expressed in KNm, rotations are expressed in radiants.
< previous page page_51 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_51.html07/11/2009 12:04:53 PM
 page_52

< previous page page_52 next page >

Page 52
BOLTED END-PLATE CONNECTIONS IN STEEL STRUCTURES
Henning Agerskov
Department of Structural Engineering
Technical University of Denmark
DK-2800 Lyngby
Denmark
ABSTRACT
High-strength bolted end-plate connections in steel structures have been studied in several investigations at the Technical
university of Denmark. The present paper gives a short summary of the main results obtained in these investigations. In
an analytical investigation of end-plate connections in beam sections, a design method was developed, and the basic
equations of this method are given in the paper. Special attention was paid to studying the effect of prying action in the
connection. Also T-stub connections and end-plate connections in circular sections were studied to clarify the strength
and stiffness characteristics. For all three types of connections, experimental investigations were carried out to verify the
analytical results obtained.
INTRODUCTION
In recent years, extensive and increasing use has been made of high-strength bolted end-plate connections in steel
structures. The situation today is that this type of connection is presumably the most widely used beam-to-column
connection in building structures.
However, full utilization of end-plate connections requires an understanding of the behaviour of the connection when
subjected to external loading. Due to a number of complications, e.g. the effect of bolt pretension, prying action, and the
many parameters governing the total behaviour of the connection, end-plate connections have been studied in a large
number of investigations during the last 10–15 years.
The present paper gives a short summary of the analytical and experimental investigations of bolted end-plate
connections, carried out at the Department of Structural Engineering of the Technical University of Denmark. End-plate
connections in beam sections, as those shown in Fig. 1, as well as T-stub connections and end-plate connections in
circular sections, were studied.
One of the problems investigated was the development of prying forces in end-plate or T-stub connections when
subjected to tension.
< previous page page_52 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_52.html07/11/2009 12:04:55 PM
 page_53

< previous page page_53 next page >

Page 53

Figure 1. Beam-to-column and beam-to-beam end-plate connections.

Very divergent answers to the question of the size of the prying forces have been given over the years, ranging from a
total negligence of prying forces to more than a 100 % increase in bolt forces due to prying action.
END-PLATE CONNECTIONS IN BEAM SECTIONS
In the investigation of end-plate connections in beam sections, a theory was developed that may be used to determine the
yield load of the connection [1,2]. On the basis of this theory, the influence of the different variables affecting the
behaviour of the connection was studied in an investigation in which a large number of connections were analysed on
computer. This analysis showed that diagrams for direct use in the design could be established.
The purpose of the tests carried out in the experimental investigation was to study the strength and deformation
characteristics of bolted end-plate connections. The experimental results show good agreement with the results obtained
from the proposed design method.
Basic Equations of Proposed Design Method
In the following, a short survey is given of the basic equations resulting from the theory developed. With respect to the
derivation of the equations and the basis of the theory, see Refs. [1,2].
The analytical model of the connection under consideration is shown in Fig. 2. The external tension in the beam flange is
denoted by 2F and the bolt force by B. A prying force of magnitude Q is assumed to act at the end of the plate. Forces F
and Q are tributary to each bolt. Fig. 3 shows the geometry of the bolts used in the connection.
The yield load of the connection is considered as the load at which the reduced yield moment is reached in the end-plate
at the toe of the weld. Three basic cases are possible in the theory. In the first, separation takes place in the bolt lines
before the reduced yield moment is reached. The situation shown in Fig. 2 corresponds to this case. In the second case,
the reduced yield moment is reached before separation occurs at the bolt lines. The difference between the first and
second case is that in the latter, a compressive force will exist between the plates in the bolt lines at the instant at which
the reduced yield moment is reached. In the third case, which may arise when a relatively heavy end-plate is used, the
bending deformation of the plate is so small that separation also occurs at the edges of the plate at the yield load,
meaning that no prying forces exist.
< previous page page_53 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_53.html07/11/2009 12:04:56 PM
 page_54

< previous page page_54 next page >

Page 54

Figure 2. Analytical model of connection.

Figure 3. Geometry of bolts used in connection.

Fig. 4 shows the forces acting on the end-plate at the yield load of the connection in the three basic cases.
a. Separation before Yield Moment
The two basic equations are

(1)
and

(2)
Eq. 1 expresses the fact that the moment at the toe of the weld at the yield load must be equal to the moment
corresponding to full plastification of the section. Eq. 2 states that the deflection in the bolt line of the middle surface of
the end-plate must be the same in deformation of the plate as in deformation of the bolt.
According to Fig. 2, the distance from the edge of the end-plate to the bolt line is a and the distance from the bolt line to

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_54.html (1 of 2)07/11/2009 12:04:58 PM

 page_54

the toe of the weld is b. The thickness of the end-plate is t, the length of the end-plate tributary to each bolt is w, σy is
the yield stress of the plate material, B0 is the initial bolt tension, and is the bolt force at which separation occurs in
the bolt lines. With and
is found from
.
The lengths ls, lt, ln, and lw, referring to the bolts used in the connection, are shown in Fig. 3. In Eq. 2, As is the area of
the bolt shaft, and α1, α2 are determined from and in which and .

According to Fig. 4(a), the prying force Q is . The unknowns, F, B, and Q can be
determined from the above equations. The magnitude of the prying force may be given in dimensionless form by the
ratio .
< previous page page_54 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_54.html (2 of 2)07/11/2009 12:04:58 PM

 page_55

< previous page page_55 next page >

Page 55

Figure 4. Forces acting on end-plate in the three basic cases: a) separation occurs at bolt lines before yield moment is
reached; b) yield moment is reached before separation at bolt lines; and c) no prying action in connection.
b. Yield Moment before Separation
The two basic equations in this case are

(3)
and

(4)
in which C is the compressive force between the end-plates at the bolt line, corresponding to the yield load of the
connection. The relationship between the bolt force, B, and the plate compression, C, at the yield load of the connection
is, [1,2]

(5)
The prying force, Q, is . Insertion of C and Q in Eqs. 3 and 4 gives two equations, from which B
and F may be determined. a, b, w, t, σy, B0, As, 1, α1, and α2 are as defined in the preceding section.
c. No Prying Action
In this case, the prying force is and thus The compressive force between the end-
plates at the bolt line is . With the bolt force, B, equal to the external force, F, the yield load of the connection is
determined from

(6)
in which b, w, t, and σy are as defined in the preceding section.
The theory developed includes consideration of the deformations of the end-plates and the bolts. This means that the
results ob-
< previous page page_55 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_55.html07/11/2009 12:05:00 PM
 page_56

< previous page page_56 next page >

Page 56
tained may be applied in establishing analytical load-deflection curves for end-plate connections, which was also
demonstrated in a recent investigation by Yee and Melchers, [5].
Prying Action
The effect of prying action has been the subject of many investigations over the years, and considerable differences in
results have been obtained. This also appears from recent investigations. Thornton [3] gives examples where prying
action results in a 100% increase in bolt forces, whereas Krishnamurthy [4] found in his investigations that prying action
may be totally neglected. Yee and Melchers [5] recommend a fixed value of 33% increase, while Grundy, Thomas and
Bennetts [6] use a fixed value of 20% increase in bolt forces due to prying action.
In the present investigation, the influence of the various factors affecting the magnitude of the prying forces was studied.
A computer analysis of a large number of connections (about 2800) was carried out, varying the different parameters
within areas of practical interest. From this analysis it appears that, generally, an increase in the prying action in the
connection, determined through the value of the ratio β, will result from an increase in: (1) the distance from the bolt line
to the toe of the weld, b; (2) the bolt diameter, dB; and (3) the yield stress of the bolt material, σyB, i.e., initial bolt
tension. Generally, a reduction of the value of β will be the result of an increase in: (1) the distance from the edge of the
end-plate to the bolt line, a; (2) the length of the end-plate tributary to each bolt, w; (3) the plate thickness, t; and (4) the
yield stress of the plate material, σy.
On the basis of the results obtained in the computer analysis, combinations of the various parameters were investigated
in order to establish diagrams giving the prying ratio as a direct function of certain combinations of the independent
variables. It was ascertained that the prying ratio may be determined, with sufficient accuracy for practical purposes,
from the values of the ratio b/a and a dimensionless quantity, γ, defined as

(7)

Figure 5. Prying ratio at yield load of connection.

< previous page page_56 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_56.html07/11/2009 12:05:01 PM
 page_57

< previous page page_57 next page >

Page 57
The curves of the diagrams resulting from this investigation have been combined in a single diagram, shown in Fig. 5.
The curves give the prying ratio at the yield load of a connection with specified values of b/a and γ. The curves have
been drawn for values of . Smaller values of γ are of no practical interest. For no prying
appears in the connection, and
Experimental Investigation
In the experimental part of the investigation of end-plate connections in beam sections, four series of tests were carried
out. In these, high-strength bolted end-plate connections in rolled IPE-and HEB-sections were investigated. In all the
tests, the bolt forces were determined by means of strain gages. The tests showed that the tensile force in the connection
may, with sufficient accuracy for practical use, be considered as uniformly distributed over all the bolts in the tensile
zone. Furthermore, a comparison of the test results with the results from the suggested design method showed good
agreement between the values obtained, the deviation in bolt forces being, in most cases, only a few per cent.
The load-deflection behaviour of the connections was also studied in the investigation, and load-deflection curves are
available from all the tests carried out. With respect to details of the experimental investigation and the results obtained,
reference is made to Refs. [1,2].
T-STUB CONNECTIONS
The investigations of T-stub connections carried out at the Technical University of Denmark were part of the study of the
tensile zone of end-plate connections. Two series of tests on high-strength bolted T-stub connections with either 4 or 8
bolts were performed. In the tests, the bolt forces, the load-deflection curves, the yield load, and the ultimate load of the
connections, were determined and compared with corresponding analytical results.
One of the main aims of the experimental part of the study was to clarify whether end-plate connections may, with
sufficient accuracy for practical purposes, be designed as T-connections, with the end-plate behaving as the T-stub
flange and the beam flange as the T-stub web. The tests of the present investigation showed this to be the case. For
further details of the results obtained in the investigations on T-stub connections, see Refs. [1,2,7].
END-PLATE CONNECTIONS IN CIRCULAR SECTIONS
Bolted connections in circular sections—solid round bars or tubes—may also advantageously be carried out as end-plate
connections. In recent years, extensive use has been made of circular sections in the construction of steel lattice
structures, e.g. in the form of towers (television, radio, and high voltage transmission towers), roof structures (space
structures and lightweight plane trusses), and cranes (tower cranes and overhead travelling cranes).
The aim of this investigation has been to clarify the strength and stiffness characteristics of high-strength bolted end-
plate connections in circular sections. The analytical and experimental inve-
< previous page page_57 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_57.html07/11/2009 12:05:02 PM
 page_58

< previous page page_58 next page >

Page 58

Figure 6. Bolted end-plate connection in circular section.

stigations carried out concentrated on connections of a type as shown in Fig. 6.
In the analytical investigation, a theory was developed that permits the analysis of bolted end-plate connections in
circular sections, including determination of the yield load of the connection, bolt forces, the effect of prying action, etc.
In the experimental part of the study, a test series was carried out using bolted end-plate connections with varying round
bar diameter and end-plate thickness. The stresses in the end-plate and the bolt forces were determined by means of
strain gages.
A comparison of the results obtained from this test series with those obtained by means of the suggested design method
shows little difference between the values obtained, the maximum deviation in bolt forces being only about 8%.
The yield load for the connection, as defined in the suggested design method, is also in this case considered to be the
load at which the reduced yield moment is reached in the end-plate at the toe of the weld. The use of this definition of the
yield load will result in some strength reserve due to strain hardening, but it also means that heavy plastic deformations
of the end-plate will be avoided. In the connections with thin end-plates, the ultimate load was found to be more than
100% greater than the theoretical yield load. For the connections with heavier end-plates, the difference was found to be
25–75%.
It is the intention to continue the present investigation of end-plate connections in circular hollow sections.
More details about this investigation and the results obtained can be found in Refs. [8,9,10].
CONCLUSIONS
In the investigations carried out at the Technical university of Denmark various types of end-plate connections have been
studied. For the three different connection types (end-plate connections in beam sections, T-stub connections, and end-
plate connections in circular sections) analytical investigations have been performed, resulting in design methods that
can be used to determine the strength, i.e. the yield load of the connection. In a number of investigations, yieldline
methods have been applied, and these methods are useful in determining the ultimate load of end-plate connections.
< previous page page_58 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_58.html07/11/2009 12:05:03 PM
 page_59

< previous page page_59 next page >

Page 59
However, they do not provide any information about the deformations in the connection. The analytical treatment of the
present investigations includes consideration of the deformations of the individual connection elements. Thus, these
methods can be applied in the analytical determination of load-deflection curves, as also demonstrated recently by Yee
and Melchers [5]. Furthermore, the analytical methods of the present investigations may be useful in the study of
dynamically loaded structures and of structures subjected to fatigue.
For all connection types studied, test series were carried out to verify the analytical results. Yield loads and ultimate
loads were determined, as well as bolt forces and stresses in the end-plates. The development of prying action in the
connections was studied, and the deformations of the connections were determined. The test results showed that the
proposed design methods provide sufficient accuracy for practical purposes. The experimental results obtained
concerning connection deformations may contribute to a data base of load-deflection curves, if established.
REFERENCES
1. Agerskov, H., High-Strength Bolted Connections Subject to Prying, Journ. of the Struct. Div., ASCE, Vol. 102, No.
ST 1, Jan. 1976.
2. Agerskov, H., Analysis of Bolted Connections Subject to Prying, Journ. of the Struct. Div., ASCE, Vol. 103, No. ST
11, Nov. 1977.
3. Thornton, W.A., Prying Action—a General Treatment, Engineering Journal, AISC, 2nd Quarter, 1985.
4. Krishnamurthy, N., Loading and Evaluation of Splice-Plate Connections, ASCE Convention, Atlanta, Oct. 1979,
Preprint No. 3696.
5. Yee, Y.L. and Melchers, R.E., Moment-Rotation Curves for Bolted Connections, Journ. of Struct. Engrg., ASCE, Vol.
112, No. 3, March 1986.
6. Grundy, P., Thomas, I.R. and Bennetts, I.D., Beam-to-Column Moment Connections, Journ. of the Struct. Div.,
ASCE, Vol. 106, No. ST 1, Jan. 1980.
7. Agerskov, H., Tests on High-Strength Bolted T-Stub Connections, Report No. R 106, Dept. of Struct. Engrg., Techn.
Univ. of Denmark, 1979.
8. Agerskov, H. and Bjørnbak-Hansen, J., Bolted End Plate Connections in Round Bar Steel Structures, Report No. R
147, Dept. of Struct. Engrg., Techn. Univ. of Denmark, 1982.
9. Agerskov, H. and Bjørnbak-Hansen, J., Bolted Connections in Round Bar Steel Structures, Journ. of Struct. Engrg.,
ASCE, Vol. 111, No. 4, April 1985.
10. Bauingenieur, Kopfplattenverbindungen in zugbeanspruchten Rundstäben, Kurze Technische Berichte nach [9], Vol.
61, No. 6, June 1986.
< previous page page_59 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_59.html07/11/2009 12:05:04 PM
 page_60

< previous page page_60 next page >

Page 60
STRENGTH AND BEHAVIOUR OF IN-PLANE WEAK AXIS JOINTS AND OF 3-D JOINTS
J.Janss
Centre de Recherches de l’Industrie des Fabrications Métalliques (CRIF)
Liege
BELGIUM
J.P.Jaspart
and
R.Maquoi
M.S.M., Institut du Génie Civil, Université de Liège
BELGIUM
ABSTRACT
A limited number of tests have been carried out on weak axis—and on 3-D beam-to-column connections. On base of the
first results obtained, some conclusions are drawn.
INTRODUCTION
The behaviour of plane- or space frames may be analysed by means of computer programs. Some of the latter can
account for the semi-rigidity of the beam-to-column joints. For beams (fig. 1.a) that are connected to the column flanges
(strong axis connection), the information regarding the behaviour of the joints is available. It is largely missing when the
beam (fig. 1.b) is connected to the column web (weak axis connection) [1] [2]. The question is still more intricate when
3-D joints (fig. 1.c) because of the interaction in the column web between both strong- and weak axis connections [3].
With a view to collect some data on weak axis—and 3-D connections, some tests were carried out at the University of
Liège. It is only reported on those, the results of which have already been fully investigated.
The authors are indebted to IRSIA and ARBED for their financial assistance in this research work.
< previous page page_60 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_60.html07/11/2009 12:05:05 PM
 page_61

< previous page page_61 next page >

Page 61

Figure 1—Types of beam-to-column joints: (a) strong axis connection; (b) weak axis connection; (c) 3-D connection.
WEAK AXIS CONNECTIONS.
When defining the test specimens, three main parameters were accounted for: the type of connection, the relative beam-
to-column stiffness and the slenderness ha/a of the column web.

Figure 2—Types of connections: (a) end plates; (b) double web cleat (c) flange cleats

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_61.html (1 of 2)07/11/2009 12:05:06 PM

page_61

Figure 3—Test arrangement.

< previous page page_61 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_61.html (2 of 2)07/11/2009 12:05:06 PM

 page_62

< previous page page_62 next page >

Page 62
Three kinds of bolted connections were considered because of their current use in practice: end plates, double web cleat
and flange cleats (figure 2). In order to exhaust the carrying capacity of the column web and to assess the behaviour of
the latter, rather thick angles or end plates and preloaded high strength bolts (quality 10.9) were used. All the mechanical
and geometrical properties of columns, beams and connections were measured.
The basic data describing the set of test specimens are summarized in table 1. For each type of joint, four beam-to-
column stiffness ratios are examined; that allows also to change the slenderness ratio ha/a of the column web, that is
obviously a governing parameter. It will be noticed that HE180A and HE160B sections have a same weak axis inertia,
while a very different web slenderness.
Table 1—Basic data for test specimens with weak axis connections
Test Column C Beam B Column Web-slenderness ha/a Relative stiffness B/C Type of connection (fig. 2)
A1 IPE240 IPE160 35, 5 0, 33 a)
A2 IPE240 IPE160 35, 5 0, 33 b)
A3 IPE240 IPE160 35, 5 0, 33 c)
A4 IPE300 IPE160 39, 2 0, 70 a)
A5 IPE300 IPE160 39, 2 0, 70 b)
A6 IPE300 IPE160 39, 2 0, 70 c)
A7 HEA180 IPE160 25, 3 1, 06 a)
A8 HEA180 IPE160 25, 3 1, 06 b)
A9 HEA180 IPE160 25, 3 1, 06 c)
A10 HEB160 IPE160 16, 8 1, 02 a)
A11 HEB160 IPE160 16, 8 1, 02 b)
A12 HEB 160 IPE160 16, 8 1, 02 c)
The test arrangement is illustrated in figure 3. The load is applied at the cantilever beam and is increased progressively
either up to collapse of the connection, or up to a limiting vertical deflection (40 cm) of the cantilever due to
requirements of the testing facilities. The depth of the column is similar to an usual storey height in a multi-storey frame
while the beam length allows to emphasize the bending moment with regard to the shear force.
The vertical displacements at both ends of the cantilever beam and the transverse displacements of the column are
measured during the tests. With the geometrical and mechanical properties of the sections used, these deflections allow
for the computation of the relative joint rotation. This rotation is composed of the rotation of the connection proper (slip,
distorsion of connecting elements, elongation of bolt shanks,…) and of the rotation due to the shear flexibility of the
column web. Dedicated measurement set-ups allow to distinguish between both rotation components. For instance,
figures 4 to 7 are dealing with the double web cleat connection A2; the rotation of the connection itself is divided into
one part due to slip and another one resulting from the distorsion of angles.
< previous page page_62 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_62.html07/11/2009 12:05:07 PM
page_63

< previous page page_63 next page >

Page 63

Figure 4—Test A2: Joint Curve

Figure 5—Test A2: Curve due to slip

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_63.html (1 of 2)07/11/2009 12:05:09 PM

 page_63

Figure 6—Test A2: Curve due to angles distorsion

Figure 7—Test A2: Curve due to the flexibility of the column web
The assessment of the results will be limited to the tests on joints with end plates, that are the sole to have been
investigated in deep at the present time. A complete report will be prepared by the end of 1987 [5]. Because of their
thickness (30 mm), the end plates are so stiff that they may be considered as not deformable. In addition, at the loading
levels reached at collapse, the effect of the elongation of bolt shanks can be disregarded with respect to that of column
web yielding. Thus the joint rota-
< previous page page_63 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_63.html (2 of 2)07/11/2009 12:05:09 PM

 page_64

< previous page page_64 next page >

Page 64
tion is nearly due exclusively to the flexibility of the column web, especially as soon as the onset of yielding. For this
type of connection an obvious similarity of the curve of the joint and of that dealing with the flexibility of the
column web is observable (fig. 8). At the first beginning of the loading, the behaviour remains linear elastic and exhibits
an initial stiffness Ki, that is reproduced and recovered when the test specimen will be unloaded and then loaded again,
whichever the loading level reached. Then occurs a rather fast yielding of the column web, that is last followed by a
strain hardening range, the stiffness Kp of which is quasi constant. According to Yee and Melchers [4], the yield bending
moment Mye will be defined as the one given by the intercept of the strain hardening slope with the Ordinate axis (fig.
8).

Fig. 8-Definition of the initial stiffness Ki, strain-hardening Kp stiffness and yield bending moment Mye.

Fig. 9-Comparison of web behaviour of different connections with end plates.

The curves related to four tests on bolted joints with end plates are shown in figure 9. Both tests A1 and A4 (using
IPE sections as column) are represented by two nearly identical curves; they provide a same value of the initial stiffness
Ki, of the strain hardening stiffness Kp and of the yield bending moment Mye. The strain hardening stiffness Kp looks
like approximately constant for the four tests, i.e. whichever the slenderness ha/a of the column web in the range 16–36.
The initial stiffness Ki and the yield moment Mye are shown to increase rapidly when the slenderness ha/a decreases.
The experimental values of Mye are plotted in figure 10 against the web slenderness.
< previous page page_64 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_64.html07/11/2009 12:05:10 PM
 page_65

< previous page page_65 next page >

Page 65

Fig. 10—Yield bending moment vs. slenderness of the column web.

Unlike for strong axis connections, no design model exists—at the authors’ knowledge—for weak axis connections,
which would provide the ultimate carrying capacity of the joint. By observing the yield lines in the column web, an
attempt of establishing such a design model was made (fig. 11). Though the effect of the web slenderness is qualitatively
well reflected (fig. 10), some improvements of the model are shown necessary with a view to a better quantitative
agreement. The consideration of more sophisticated yield lines patterns in the column web as well as the account of
membrane force—due to the out-of-plane deflection of this web—in the yield criterion are in progress, so that a realistic
design model can be hopefully expected soon.

Figure 11—Test A10: Yield lines pattern

< previous page page_65 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_65.html07/11/2009 12:05:11 PM
 page_66

< previous page page_66 next page >

Page 66
3-D CONNECTIONS
The results of tests on semi-rigid joints with strong axis connections (serie 0) are presented in another paper [6]. It is
briefly reported hereabove on those dealing with weak axis connections (serie A). For a specified type of test specimen,
both sets of results give two limiting points of an interaction curve (fig. 12), the aim of which is to show the carrying
capacity of a column web involved in a 3-D joint. Any other point of the interaction curve would require tests on the
associated 3-D specimen. A limited number of such tests were performed at the University of Liège; the web slenderness
and the type of connection were selected as the governing parameters. The basic data for the test specimens are listed in
table 2.

Figure 12—(a) Test I1: Limiting points of the interaction curve, (b) Curve due to the shear flexibility (Test 04) of
the column web; (c) Curve of the connection (Test A10).
< previous page page_66 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_66.html07/11/2009 12:05:12 PM
 page_67

< previous page page_67 next page >

Page 67
Table 2—Basic data for test specimens with 3-D connections
Test Column Weak axis beam P1 Weak axis connection (fig. 2) Strong axis beam P2 Strong axis Connection (fig. 2)
I1 HEB160 IPE160 a IPE200 a
I2 HEB160 IPE160 b IPE200 a
I3 HEB160 IPE160 c IPE200 a
I4 IPE300 IPE160 a IPE200 a
I5 IPE300 IPE160 b IPE200 a
I6 IPE300 IPE160 c IPE200 a
The test arrangements were quite similar to those used for the tests on joints with weak axis connections. The tests were
performed by keeping constant the load P2 at the end of the weak axis beam while the load P1 at the end of the strong
axis beam is increased progressively. The value of P2 is associated to the onset of the bending moment Mye, derived
from the test on in-plane weak axis connection. Measurements are made similarly to those carried out for the testing of
semi-rigid joints with strong axis connections (serie 0) [6].
The behaviour of the column web, when a 3-D connection, can then be compared to that when strong axis connection
only. Thus the curves due to the shear flexibility of the column web are plotted in figure I3 for test 07 (strong axis
connection) and I1 (3-D connection). It is observable that the rotation remains very small till the onset of yielding in the
web of the column fitted with a strong axis connection. The effect of shear deformation is much more marked when an
additional loaded weak axis beam exists, because of a more progressive yielding of the web. However the influence of
the weak axis beam tends to be masked when approaching the ultimate limit state.

Figure I3—Comparison of the shear flexibility of the column web between test I1 and test 07.

Figure 14—Interaction curve for a 3-D connection

< previous page page_67 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_67.html07/11/2009 12:05:13 PM
 page_68

< previous page page_68 next page >

Page 68
Such a test on a 3-D joint, carried out as indicated hereabove, enable to determine one additional point of the associated
interaction curve and to sketch the shape of this latter (fig. 14). Obviously more experimental points are needed to assess
more precisely this interaction curve and to suggest a design model for 3-D joints; therefore additional tests will be
performed soon in order to implement the information.
REFERENCES.
1. Rentschler, G.P. and Chen, W.F., Tests of Beam-to-Column Web Moment Connections. J1. Struct. Div., A.S.C.E.,
ST5, Mai, 1980, 1005–1022.
2. Vircik, J., Marek, P. and Hudak, J., Limit States of Carrying Capacity Demonstration of the Reliability Conditions on
a Selected Structural Detail. Design Limit States of Steel Structures—1st International Correspondence Conference—
Preliminary Report. Edited by J.Melcher, Brno, 1983.
3. Massonnet, Ch. and Save, M., Calcul Plastique des Constructions. Volume 1 Structures dépendant d’un paramètre, 3d
edit., Nélissen Edit., Liège, 439–441.
4. Yee, Y.L. and Melchers, R.E., Moment-Rotation Curves for Bolted Connections. J1.Struct. Engrg., A.S.C.E., Vol.
112, 3 March 1986, 615–635.
5. Jaspart, J.P., Essais sur assemblages poutre-colonne d’axe faible et tridimensionnels. C.R.I.F., Bruxelles (in
preparation).
6. Janss, J., Jaspart, J.P. and Maquoi, R., Experimental Study of the Non Linear Behaviour of Beam-to-Column Bolted
Joints, Proceedings, Workshop on Connections and the Behaviour, Strength and Design of Steel Structures, Cachan,
May 25–28, 1987.
< previous page page_68 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_68.html07/11/2009 12:05:14 PM
 page_69

< previous page page_69 next page >

Page 69
ANALYSIS OF JOINTS BETWEEN STRUCTURAL HOLLOW SECTIONS
J.Wardenier
Department of Civil Engineering
Delft University of Technology
Delft, The Netherlands
ABSTRACT
Structural hollow sections with various shapes are used in various combinations also with open sections. Furthermore,
various joint configurations are possible. The large variety in sections, combinations and joint configurations make that
many failure modes are observed depending on the geometry variables and loading.
A unified analysis is essential for a uniform presentation of the design rules. In this paper the criteria of failure, failure
modes and the analysis are briefly discussed as a background for those not familiar with hollow section joints.
INTRODUCTION
Recent research in Europe, Japan, USA and Canada has added new evidence for the analysis of joints between hollow
sections and combinations with open sections. In these programmes various types of axially loaded welded joints have
been investigated. The main objective was to enlarge the range of validity of the IIW recommendations [1] and to
provide test evidence to check the effective width criteria.
The IIW recommendations cover T, Y, X, N, K and KT joints for various member combinations (figures 1 and 2).
Since research and analyses are carried out in various countries, a major objective is to achieve consistent
recommendations. It is important that the designer understands the behaviour, failure modes and the resulting design
recommendations. This means also that similar failure modes for different joints should be treated in a similar way.
Although there re-
< previous page page_69 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_69.html07/11/2009 12:05:15 PM
 page_70

< previous page page_70 next page >

Page 70

Figure 1. Types of joints

Figure 2. Member combinations

main areas which need further investigation, the current modified recommendations show a large degree of consistency.
CRITERIA OF FAILURE
The static strength can be characterized by the criteria shown in fig. 3, i.e.:
– ultimate load capacity (5)
– deformation criterion (2) or (3)
– visually observed cracks (4)
< previous page page_70 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_70.html07/11/2009 12:05:17 PM
 page_71

< previous page page_71 next page >

Page 71
Because of the non-linear load deflection behaviour there is no international agreement about the determination of the
yield load or regarding the deformation limit. For most types of joints the ultimate load is clearly defined.

Figure 3. Failure criteria

For the design recommendations of the International Institute of Welding and also for those included in Eurocode 3, in
principle the ultimate strength is taken as basis for the determination of the design strength (K, N, KT joints and other
joints loaded in compression). The range of validity of the resulting formulae has been limited if the joint deflection at
working load would exceed 1% of the chord diameter or chord width.
For joints loaded in tension in principle the same strength is given as for joints loaded in compression (buckling not
governing). This means a strength which nearly agrees with that belonging to the kink in the load-deformation diagram
(see figure 3). For joints which show excessive deformations at ultimate load, e.g. T and X joints between rectangular
hollow sections with the strength has been based on the analytical yield strength determined with a yield
line model neglecting membrane and strain hardening effects. In this way the associated deformations are indirectly
limited, however, in a conservative way.
Furthermore, in general the range of validity of the formulae has been limited to avoid crack initiation below the design
load.
Various proposals exist for the determination of a deformation limit or for the determination of a yield load, but none has
been accepted internationally. For example, Kurobane uses the kink in the load-deflection
< previous page page_71 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_71.html07/11/2009 12:05:18 PM
 page_72

< previous page page_72 next page >

Page 72
diagram if plotted on a log-log scale; Mang is using 80% of the load at which the maximum curvature in the load
deflection diagram appears; Yura relates the deformation limit to the deformation of a member with a particular length
based on offshore applications [7,8,9].
There is a need to describe the deformation for the various types of joints as a function of the joint parameters and the
loading. This is especially important for joints loaded by bending moments or combined loadings. In particular for
multiplanar joints with multiplanar loading more evidence is needed.
MODES OF FAILURE
Depending on the type of joint, joint parameters and loading conditions several basic modes of failure can occur, as
illustrated in fig. 4 for joints between rectangular hollow sections:
a. plastic failure of the chord face or chord cross section
b. rupture of the braces from the chord (punching shear)
c. cracking of the braces near the joint
d. shear failure of the chord
e. local buckling in compressive areas of the joint members
f. weld failure
g. lamellar tearing
Many times failure occurs by combinations of the basic types mentioned above.
Due to the locally high peak stresses it is essential that weld failures are avoided. Thus the welds should be stronger than
the connected members.
Lamellar tearing can be avoided by choosing approapriate steel qualities and welding processes, especially for heavy
wall thicknesses.
Local buckling in the joint can be avoided by limitation of the validity range of the recommendations. This is important
since the secondary bending moments in welded trusses of hollow sections are rather difficult to estimate. This is caused
by the non-uniform stiffness along the perimeter at the intersection between chord and connecting brace.
< previous page page_72 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_72.html07/11/2009 12:05:19 PM
 page_73

< previous page page_73 next page >

Page 73

Figure 4. Modes of failure for rectangular hollow section joints

ANALYTICAL MODELS
For the analysis of the various failure modes analytical models are used to determine the governing joint parameters for
the ultimate strength, e.g.:
a. Ring model (CC joints)
b. Punching shear model (CC-CR-RR joints)
c. Effective width model (CR-RR-RI-CI joints)
d. Yield line model (CR-RR joints)
e. Chord web bearing models (RR-RI-CI joints)
f. Model for shear failure of the chord (all K and N gap joints)
For indication some of these models are shown in figs. 5, 6 and 7. A detailed description of these models is given in [3,
6]. Due to the complex joint behaviour these models cannot fully describe the influence of all influencing parameters.
TEST EVIDENCE
All types of joints discussed in this paper have been extensively tested in many countries all over the world [3,5,6].
Based on the analytical models and the test evidence, the ultimate strength formulae have been determined.
< previous page page_73 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_73.html07/11/2009 12:05:21 PM
 page_74

< previous page page_74 next page >

Page 74

Figure 5. Ring model for CC joints

Figure 6. Punching shear model for CC and RR joints

Figure 7. Chord side wall bearing model for RR joints

< previous page page_74 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_74.html07/11/2009 12:05:22 PM
 page_75

< previous page page_75 next page >

Page 75
EVALUATION TO DESIGN RULES
Purely experimental design rules should, in principle, be avoided unless no other information is available and the
limitations are clearly defined. However, only for particular failure modes pure analytical design rules can be used. For
the more complex joints the design rules are of a semi-empirical nature.
The strength functions which are based on an analytical analysis and which describe a ‘‘yield” strength are not treated in
a statistical manner. As far as these joints show a ductile behaviour the analytical strength function is adopted as design
strength.
From the semi-empirical ultimate strength formulae characteristic strengths have been obtained using a semi-
probabilistic approach [3,4,5,6,7 and 8]. All variables have been taken into account, e.g. scatter in test results, but also
the variation in mechanical properties and dimensions and the fabrication tolerances. The resulting formulae are
sometimes simplified for practical use.
The characteristic joint strength Nk, representing the strength that 95% of the joints are expected to reach, is divided by a
partial joint factor γm to obtain the design strength . The partial joint safety factor γm adopted in the IIW and Cidect
recommendations varies between 1 and 1.25 depending on the joint deformation capacity and the reserve in strength.
RECENT ADOPTED MODIFICATIONS TO THE IIW RECOMMENDATIONS
Recently, modifications have been adopted to the IIW recommendations. For joints between circular hollow sections the
formula for the design strength of K joints has been modified, resulting in a somewhat lower strength for overlap joints
with a high chord diameter to thickness ratio.
For joints with a rectangular section or an I section as chord, the range of validity has been extended, particularly the
width to wall thickness ratio. Depending on the type of joint, plastic design sections or compact sections can now be
used for the compression braces. For these joints also the effective width criteria have been modified. The updated
version of the IIW rules will be published in the course of 1987/1988.
< previous page page_75 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_75.html07/11/2009 12:05:26 PM
 page_76

< previous page page_76 next page >

Page 76
FUTURE RESEARCH
Subjects of further investigation are:
– multiplanar joints under multiplanar loading
– interaction between axial loading and bending moments
– beam to column connections
– stiffness and deformation aspects (behaviour and limits)
– local buckling behaviour
– influence weld configuration and weld size
– stress concentration factors (RR-CI-RI joints).
REFERENCES
1. IIW-XV-E, Design Recommendations for Hollow Section Joints—Predominantly statically loaded—IIW Doc XV-
491–81/XIII-1003–81 (Revised version).
2. ECSC, Eurocode Nr. 3: Common Unified Rules for Steel Structures, Commission of the European Communities,
Report EUR 8849.
3. Wardenier, J., Hollow Section Joints, Delft University Press, 1982.
4. Packer, J.A. and Kremer, J.S.M., Reliability of Tubular Joints in Offshore Structures. 4th International Conference on
Struc. Safety and Reliability, Kobe, Japan, May 1985.
5. Packer, J.A., Birkemoe, P.C. and Tucker, W.J., Canadian Implementation of Cidect Monograph Nr. 6, University of
Toronto, Dept. of Civil Engineering, Publ. Nr. 84–06.
6. Giddings, T.W. and Wardenier, J., The Strength and Behaviour of Statically Loaded Welded Connections in Structural
Hollow Sections, Monograph Nr. 6, Cidect, 1986.
7. T. Koning, C.H.M.de and Wardenier, J., The Static Strength of Welded Joints between Structural Hollow Sections or
between Structural Hollow Sections and H-sections. (Evaluation study). Parts 1, 2 and 3. TNO-IBBC and Stevin Reports
6–84–18/19/20.
8. Makino, Y. and Kurobane, Y. Strength and Deformation Capacity of Circular Tubular Joints. Proceedings of the
International Meeting “Safety Criteria in Design of Tubular Structures”, Tokyo, July 1986. Maruzen Co., Ltd., Tokyo.
9. Yura, J.A., Zettlemoyer, N. and Edwards, I.F. Ultimate Capacity Equations for Tubular Joints, Offshore Technology
Conference Proceedings, Vol. 1, Nr. 3890, 1980.
< previous page page_76 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_76.html07/11/2009 12:05:27 PM
 page_77

< previous page page_77 next page >

Page 77
STRENGTH OF WELDED CONNECTIONS WITH HOLLOW SECTION BRACINGS AND I OR H SECTION
AS CHORD
J.M.ARIBERT
Structures Laboratory
Institut National des Sciences Appliquées (INSA)
35043—Rennes Cedex—France
ABSTRACT
The topic relates to the failure by cracking of the bracing elements, considering three different types of connection. It is
shown how to determine accurate design formulae towards this failure, using both experimental tests and theoretical
modelling by plastic mechanisms.
INTRODUCTION
This paper deals with the ultimate strength of a particular type of semi-rigid welded connection which is used in Europe
for the construction of lattice plane girders. This type of connection results from the association of a chord with an open
section (I or H) and bracing members with hollow sections (rectangular or circular). Several failure modes have been
observed for these welded connections, but one of these modes must be specially controled for design; generally, it is
called “failure by transverse cracking of the bracings”, but in reality this failure mode includes two phenomena: the first
one is a sudden fracture of the weld when the bracing member is in tension; the second one is a local buckling of the end
of the bracing member when this member is in compression. Concerning this transverse cracking failure, it is possible to
obtain accurate design formulae by combining an empirical approach and a theoretical analysis based on plastic
mechanisms. The aim of this paper is to give some ideas about a such procedure, essentially for the X type connection in
tension; however, a more general application of the procedure is briefly shown for two other types: the K type
connection and the X type connection in compression.
As shown in figure 1, the stress distribution at the bottom of the bracing member is absolutely not uniform because the
flexural stiffness of the chord’s flange is stronger near the web than near its edges. In order to express this varying
stiffness, most of the norms use the practical concept of “effective width be”; consequently, the strength towards the
cracking failure is calculated by:

(1)
(for compression and for tension).
Generally, bei is given by:
< previous page page_77 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_77.html07/11/2009 12:05:28 PM
 page_78

< previous page page_78 next page >

Page 78

Figure 1. Geometry of the X type connection

(2)
where tw and tf are the thicknesses respectively of the web and the flange of the chord. The constant ci depends on the
used norm; for example, the French Standard NF—P—22–255 [1] recommends the following values:

(3)
with (compression) and (tension). It must be emphasized that the associated formulae (1), (2) and
(3) take into account a safety factor of 1.4; it is important to know this point for comparisons which might be made
between the application of the previous regulation formula, or others, and experimental results.
In order to possess a better knowledge of the cracking failure, a specific experimental programme of about fifty
connections of different types (X, K and N) with different sections has been carried out at the INSA of
Rennes during the period 1981–1984. The studied parameters were the geometrical dimensions of the sections (t1,2, b1,2
and h1,2 for the bracing members, tw, tf, and bf for the chord), the yield stresses of the members (σyo, σy1,2) and the
angles θ1,2 between bracings and chord. A complete information about the data and results of all these tests may be
found at the references [2, 3].
PROPOSED FORMULATION FOR X TYPE CONNECTION LOADED IN TENSION
Only from the analysis of the experimental results, we have been able to determine the main influence of the previous
parameters on the ultimate load, and we have proposed a new design formula for the X type connections in tension:

(4)
In figure 2, a comparison between experimental ultimate loads Nu2 and the corresponding loads given by the
formula (4) shows a satisfactory agreement
< previous page page_78 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_78.html07/11/2009 12:05:30 PM
 page_79

< previous page page_79 next page >

Page 79

Figure 2. Comparison between experimental results and the formula (4)

if the safety factor 1.4 is taken into account; more exactly, the mean relative deviation of the points towards the line
is 9 % whereas it was 20 % with the formula (1), (2) and (3).
We have supplied the empirical formula (4) with a theoretical modelling which allows to calculate Nu2 by means of a
specific mechanism by yield lines. This mechanism is located in the chord’s flange; it is schematized at the figure 3 in
the case of a rectangular hollow section with θ2=90°. The dimension k can be calculated by minimizing the internal
work dissipated in the yield lines AC, AD and AB; furthermore, the uniformly distributed load qu is equal to the yield
stress σy2 multiplied by the thickness t2 because the mechanism of the flange must be kinematicly consistent with the
yield deformation of the bracing member. From this value of qu, we can deduce the unknown length x of the loaded part
which is parallel to the chord’s web:

(5)
then, if the ultimate strength of the connection is calculated by:

(6)
If we have of course:

(7)
A remark may be made about the distributed load q; in reality, this load cannot be uniform along the length x because the

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_79.html (1 of 2)07/11/2009 12:05:31 PM

 page_79

stress distribution has an elasto-plastic

< previous page page_79 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_79.html (2 of 2)07/11/2009 12:05:31 PM

 page_80

< previous page page_80 next page >

Page 80

Figure 3. Yield pattern for the X type connection in tension

nature. But the hypothesis with q uniform is valid for a Limit Analysis because the vertical displacement along the length
x is constant for the studied mechanism; in a way, x may be considered as an equivalent length for a rigid-plastic
behaviour.
The previous modelling has been easy generalized [4] when the bracing member is not perpendicular to the chord
( ); the distribution of the load q is only different.
Many comparisons have been made between the calculation model and the experimental approach; they confirm fully
the good accuracy of the empirical formula (4) from a point of view both local and global. For example, the figure 4

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_80.html (1 of 2)07/11/2009 12:05:33 PM

 page_80

Figure 4. Example of local comparison between the empirical formulae and the calculation model
< previous page page_80 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_80.html (2 of 2)07/11/2009 12:05:33 PM

 page_81

< previous page page_81 next page >

Page 81
shows such a comparison for a given connection, but with a variable thickness tf of the chord’s flange; we can see that
the empirical formula (4) is clearly closer to the model results than the norm’s formula (1), (2) and (3). In another way,
the figure 5 deals with a global comparison for various manufactured profiles, up to a

Figure 5. Global comparison between the empirical formulae and the calculation model

depth equal to 600 millimeters. The horizontal axis of the diagram relates to the ultimate load given by the
theoretical model; the vertical axis is meant to represent either the norm’s load Nc2 or the load given by the
empirical formula (4). Here again, the formula (4) is in perfect agreement with the model, for the whole field of
application, whereas the norm’s formula shows a real deviation.
GENERALIZATION TO OTHER TYPES OF CONNECTION
Hereafter we present only two generalizations, but a more complete study may be consulted at the reference [4].
For the K type connections with overlap, experimental results have allowed to propose the following formula for the
strength of the bracing member in tension:

(8)
This sophisticated formula includes three terms: the first one is equal to half the ultimate load of the bracing in tension if
it is considered alone; the second term relates to the strengthening due to the bracing in compression. And the third term
results from another strengthening due to the effect of the overlap area 2 hr tr.
< previous page page_81 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_81.html07/11/2009 12:05:34 PM
 page_82

< previous page page_82 next page >

Page 82
A model by yield lines has been found in good agreement with the formula (8). As shown in figure 6, the main
characteristics of this model are the following ones:

Figure 6. Yield pattern for the K type connection with overlap

1/ the exact position of the yield line BE (defined by the dimension n) is calculated by minimizing the ultimate load of
the bracing member.
2/ The distributed load along the perimeter at the bottom of the bracing member is assumed to have three different
constant values according to the part of this perimeter: the values qu sin θ2 and qu sin2 θ2 are consistent with the yield
deformation of the bracing member near the lines AD and BE; as for the middle value q, its calculation results from the
equality between the internal work and the external work of the mechanism.
3/ The relative displacement between the two bracing members is only possible if the overlap area is locally yielded by
shearing.
The second generalization that we are going to present now, deals with the failure by buckling of the X type connection
loaded in compression. Experimental results have led to the formula:

(9)
in which the influence of the angle parameter θ1 is knowingly neglected on account of the nature of the failure.
The formula (9) has been checked by means of a mechanism including two parts, as shown in figure 7: one part is in the
chord’s flange with a pattern like that of the
< previous page page_82 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_82.html07/11/2009 12:05:35 PM
 page_83

< previous page page_83 next page >

Page 83

Figure 7. Yield pattern for the X type connection in compression

connection in tension (fig. 3); the second part of the mechanism takes place near the bottom of the bracing member
where second order geometrical effects must be considered for the calculation (so, at the figure 7, ∆H and ∆V are finite
displacements before buckling). The kinematic consistence between these two parts requires the equality between the
vertical displacement of the flange’s edge and the height variation δ∆V of the yield pattern in the bracing when buckling
occurs. Finally, the calculated ultimate load Nu1 of the connection becomes a function of two indeterminate parameters:
the initial value λ for the height of the yield pattern in the bracing, and the angle expressing the wall deformation of
the bracing before buckling. Founding upon some experimental observations, we have adopted equal to b1/4; by
another way, we have calculated for each type of connection, assuming that the displacement of the flange’s edge is
equal to bf/100.
CONCLUSION
The theoretical modelling turns out to be in satisfactory agreement with all the available experimental results; besides, it
provides a mechanical interpretation to the design formulae which were empirical up to that date. It also allows to
develop a certain critical analysis of both the accuracy and the field of application of these formulae; if the failure
mechanisms of the different parts of a connection were well-known, this field of application could be better defined by a
rational way (for example, see the figure 4).
< previous page page_83 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_83.html07/11/2009 12:05:36 PM
 page_84

< previous page page_84 next page >

Page 84
REFERENCES
1. Afnor—Norme française NF P 22–255—Assemblages soudés de profils creux ronds ou rectangulaires sur profils de
type I et H. Dec. 1979.
2. Aribert, J.M., Lachal A., Mouty J. and Petit L., Failure of the welded connection between RHS bracing and H section
—Welding of tubular structures—Pergamon Press 1984, pp 459–469.
3. Aribert, J.M., Lachal A. et Sofwan A.—Assemblages de profils creux sur profil ouvert—Etude de la ruine au niveau
de la liaison soudée—Construction Métallique no 2, 1984, pp. 5–41.
4. Aribert, J.M., Lachal A. et Sofwan A.—Modélisation par mécanismes de la ruine d’assemblages de profils creux sur
profil ouvert—Construction Métallique no 4, 1984, pp. 3–32.
< previous page page_84 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_84.html07/11/2009 12:05:37 PM
 page_85

< previous page page_85 next page >

Page 85
NUMERICAL SIMULATION OF THE BEHAVIOUR UP TO COLLAPSE OF TWO WELDED UNSTIFFENED
ONE-SIDE FLANGE CONNECTIONS
Miss Wajd ATAMAZ SIBAI and François FREY
IREM, Department of Civil Engineering, Swiss Federal Institute
of Technology, CH-1015 Lausanne, Switzerland
ABSTRACT
Unstiffened one-side welded flange connections of hot-rolled H sections have been tested by Klein under simple bending
loading; moment-rotation curves and plastic capacity of the nodes are also proposed. Here, two of Klein’s nodes are
computed up to collapse with a nonlinear finite element code. The numerical simulation emphasizes the importance of
strain hardening; it also helps one to understand the detailed mechanical behaviour of the node, from which nonlinear
spring constants can be deduced with a view to performing a beam analysis. The two nodes are also computed under
pure bending action, which leads to a more severe loading condition than simultaneous bending and shear; hence it is
concluded that Klein’s proposals should be slightly improved.
INTRODUCTION
Within the framework of a Swiss-Austrian joint research project, Klein [1] studied unstiffened welded flange
connections of hot-rolled I and H sections in an experimental way. In his thesis, he proposed simple expressions for
‘‘beam moment—beam rotation” curves (Mb-α, see Fig. 7a and 10) and plastic limit loads of such connections.
We present here the detailed numerical study of two of these connections. Material and geometrical nonlinear effects are
taken into account, although the latter is far less important than the former. We use the computer code FINELG [2] of
IREM and MSM (University of Liège).
The aim of the numerical simulation is twofold:
1. - to understand deeply the mechanical behaviour of the connection up to collapse;
2. - to derive a model including nonlinear springs, which expresses the
< previous page page_85 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_85.html07/11/2009 12:05:38 PM
 page_86

< previous page page_86 next page >

Page 86
semi-rigid “moment—relative rotation” M-Φ behaviour with a view to performing a beam analysis of the connections
and, further, of frames.
The selected specimens, named NR4 and NR16 in [1], are unsymmetrical one-side connections consisting of an IPE 330
beam welded to a HEB 160 column for NR4, and of a HEB 500 beam welded to a HEB 300 column for NR16 (Fig. 1).
The test loading is a concentrated force at the beam tip; thus the connection is submitted to both bending and shear; in
practice, however, the ratio of shear is generally less important; for comparison purposes, a second load case of pure
bending has been numerically performed.

Figure 1. Test set-up (after [1]) and free-body diagram of the node. Simple bending load case:
pure bending load case:
NUMERICAL MODEL
The geometry is discretized into a 3D model of shell elements (webs and flanges) and beam elements (end plates and
stiffeners). Fig. 6 gives an idea of the mesh. A geometrical imperfection is introduced in the column web at the beam
compressed flange level, allowing for possible web buckling (Fig. 1). The out-of-plane magnitude is taken from the
rolling tolerances.
Steel is supposed to follow a piecewise linear law shown in Fig. 2; yield stress is taken from [1], strain hardening
parameters and ultimate stress from [3]. The 2D elastoplastic state of stress is dealt with using the incremental flow
theory and von Mises yield condition. Parabolic
< previous page page_86 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_86.html07/11/2009 12:05:53 PM
 page_87

< previous page page_87 next page >

Page 87
patterns of rolling residual normal stresses in flanges and webs are taken according to the ECCS recommendations [4].
Welding imperfections are not considered. Complete data may be found in [5].

Figure 2. Stress-strain curve (mild steel).

NUMERICAL RESULTS
Moment-deflection curves Mb-v are shown in Figs. 3 and 4 (for Mb and v, see Fig. 1); simple bending curves are to be
compared with the experimental ones. We conclude that:
a. - the agreement with the experimental results is satisfactory;
b. - pure bending is more severe than simple bending;
c. - the initial stiffness is larger than the experimental one;
d. - strain-hardening occurs very quickly (point B);
e. - unloading is linear elastic (Fig.4).

Figure 3. curve for NR4.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_87.html (1 of 2)07/11/2009 12:05:59 PM

 page_87

Figure 4. curve for NR16.

< previous page page_87 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_87.html (2 of 2)07/11/2009 12:05:59 PM

 page_88

< previous page page_88 next page >

Page 88
Figure 5 shows the progressive spread of plasticity in the webs of connection NR4 loaded by simple bending. The
flanges (not shown) are comparatively not very plastified. Hence,
f. - collapse of the connections happens mainly by total plastification of that part of the column web which is adjacent to
the beam.

Figure 5. Web plastification of NR4 connection at various load levels (simple bending).

Global and detailed web deformation of connection NR4 at collapse is shown in Fig. 6. Although it is rather complex,
the two main causes of the node flexibility [1] are clearly visible:
g. - local deformation near points B and C, i.e. at the beam flanges junction;
h. - shear deformation.
More detailed results may be found in Ref. [5].
CONNECTION FLEXIBILITY AND SPRING MODEL
If we simplify the global deformation of the ABCD web panel of Fig. 6c, assuming straight edges, we obtain the
quadrangle of Fig. 7a; this defor-
< previous page page_88 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_88.html07/11/2009 12:06:00 PM
 page_89

< previous page page_89 next page >

Page 89

Figure 6. Web deformation of connection NR4 at maximum load (simple bending case; displacements magnified 3x); (a)
full structure; (b) connection region; (c) enlarged ABCD “column-beam” web panel.
mation may be divided into two parts:
1.- the transversal effect of the beam flange forces Fb (one elongates AB, the other shortens CD) results in a relative
rotation φ between the beam and the column axes; this rotation concentrates mainly along edge BC (Fig. 7b) and
provides a first law ;
2.- the shear effect, due to the shear force Vn in the web panel

results in a relative rotation γ between the beam and the column axes (Fig. 7c); this rotation occurs mainly along edges
AB and CD and makes it

Figure 7. Global deformation of the connection (a) decomposed into the beam flange effect (b) and the shear effect (c).
< previous page page_89 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_89.html07/11/2009 12:06:01 PM
 page_90

< previous page page_90 next page >

Page 90
possible to establish a second law.
This latter rotation is larger than the former; moreover, in pure bending, and thus the shear force Vn increases.
This explains why pure bending deforms the connections more than simple bending.
The two laws and have been derived from the numerical model and are plotted in Fig. 8. They can be
used in a T-sping model of

Figure 8. Relative rotation laws for connection NR16.

Fig. 9. T-spring model for beam analysis.

the connection; this model consists of three short beam elements and springs (Fig. 9). It simulates the whole ‘‘M-Φ”
connection behaviour and can be used in a standard beam finite element numerical model.
THE T-SPRING/BEAM MODEL APPLIED TO THE TESTS
The effective rotation a of edge BC (Fig. 8a) is plotted in Fig. 10 as a function of the corresponding beam moment Mb.
Experimental [1], proposed [1], finite element and T-spring/beam curves may be compared:
(a) the T-sping/beam model is accurate;
(b) the tabulated values proposed in [1] make no distinction between:
(b1) yield limit (fy) and strain hardening,
(b2) simple and pure bending.
DISCUSSION AND CONCLUSION
A 3D finite element computation is a fruitful complement to laboratory tests. A simple and accurate three node T-spring
model, which applies to
< previous page page_90 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_90.html07/11/2009 12:06:02 PM
 page_91

< previous page page_91 next page >

Page 91

Figure 10. Mb-α curves for NR4 and NR16 connections.

beam analysis of fixed and sway frames, is derived. Any additional flexibility (due to bolted connection for instance) can
easily be added to the spring laws. A two node L-spring or four node X-spring model can also be developed for other
types of connections.
Reducing the one-side T connection flexibility to a single beam-to-column spring (Fig. 11) appears tempting but
dubious. The law can be transformed into an equivalent law such that beam deflections are equal in
both models [5]. Hence, the whole

Figure 11. The 3 spring model (a) reduced to an equivalent 1 spring model (b).
M-Φ flexibility is concentrated into a single spring, but the second order effects in the columns, for instance, are
erroneously and unsafely reduced ( see Fig. 11b).
Klein’s proposal [1] should be refined to distinguish:
1. - simple plastic behaviour from strain hardening one,
2. - simple bending from pure bending.
Indeed, the column shear force is smaller in practice than in the laboratory tests; the ratio Vc/Fb is about 0.3 in the
< previous page page_91 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_91.html07/11/2009 12:06:04 PM
 page_92

< previous page page_92 next page >

Page 92
tests, but probably less than or equal to 0.1 in current building practice. Note, moreover, that this ratio varies during the
loading history of a structure up to collapse.
In conclusion, the authors are of the opinion that the flexibility of the connection should be based on pure bending
action.
REFERENCES
1. Klein H., Das elastisch-plastische Last-Verformungsverhalten steifenloser geschweisster Knoten; Dr. Dissertation,
Univ. Innsbruck, 1985.
2. FINELG, Nonlinear finite element analysis program, User’s manual; IREM Int. Rep. 86/6, July 1986.
3. Alpsten G.A., Variations in mechanical and cross-sectional properties of steel. ASCE-IABSE Int. Conf. on Planning
and Design of Tall Buildings, Vol. Ib-9, 1972.
4. Manual on Stability of Steel Structures, ECCS Publ. No. 22, Bruxelles, 1976.
5. Atamaz Sibaï W., Simulation numérique du comportement de deux assemblages de rive soudés non raidis; IREM Int.
Rep. 87/5, June 1987.
ACKNOWLEDGMENT
At the time this paper was written, the authors were unaware of the simultaneous publication of Braun’s thesis [6]. This
thesis is a sequel, extension and partly reformulation of Klein’s work. It appears that Braun’s results are in good
agreement with the work presented here, especially in that range of small relative node rotations which is of practical
interest. More about this subject should appear in a further paper.
6. Braun Ch., Das Momenten-Rotationsverhalten von geschweissten steifenlosen Rahmenknoten bei
Querkraftbeanspruchung; Dr. Dissertation, Univ. Innsbruck, Juni 1987.
< previous page page_92 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_92.html07/11/2009 12:06:05 PM
 page_93

< previous page page_93 next page >

Page 93
DISCUSSION OF LOCAL ANALYSIS PAPERS
A significant amount of attention was given to the prying force phenomenon, and distinctly different methods of
computation and design code incorporation appeared to be in use in various countries. There seemed to be general
agreement that prying can constitute an important additional load in tension-type bolted connections, although comments
were made to the effect that prying has traditionally been overestimated. It is clear, however, that with prying as a
deformation—induced force in many types of connections, some recognition must be given to the attendant reduction in
the available external load-carrying capacity of the bolt. In other words, prying reduces the bolt capacity, but only
because some of it is taken to resist the deformation—induced effect. The ultimate strength of the bolt itself is not
influenced.
It appears that design practices in Europe, as reflected by the recently issued Eurocode 3—“Steel Structures” and North
America (AISC Specifications for U.S.A.; C.S.A. Standard for Canada) are similar in the way prying is treated. Thus,
although there are researchers and designers that feel that prying need not be considered, the general consensus now is to
incorporate the prying effect on the load side of the design equation. However, changes may be forthcoming, especially
considering the importance of fatigue behaviour of tension connections, and the fact that prying in such cases will work
to reduce the overall tensile strength of the bolt. Further, the Danish results cover a large range of connection types, with
good correlation.
Several speakers and discussers dealt with the subjects of material strength and behaviour in general, and their
application to connection performance and design in particular. Suggestions were advanced that the ultimate stress of the
steel should form the basis for connection design, rather than the yield stress. However, several comments were made to
the effect that this way lead to unacceptably low margins of safety against local steel cracking, due to the high degree of
restraint that is likely to prevail. This is particularly accentuated for higher strength steels, where the ratio of ultimate
stress to yield stress may drop well below 1, 30.
Exceptions to the above obervations certainly apply in the design of bolted connections in bearing, as pointed out by
numerous speakers and discussers. In these cases it is critical to bear in mind the deformation
< previous page page_93 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_93.html07/11/2009 12:06:06 PM
 page_94

< previous page page_94 next page >

Page 94
capacity of the joint, particularly in the vicinity of the last bolt in a row. Previous American allowable stress design
criteria used a bearing stress of 1,5 times the ultimate value; this has recently been reduced to 1,2, due to large hole
deformations in the plate or member material at the higher stress. European criteria currently utilize a comparable
bearing stress value of 3,0 times the yield stress; this will likely be changed in the future, to reflect the use of the ultimate
stress. A value of 2,0 to 2,5 may be adopted; the former will be comparable current North American practice.
Modeling methods and solutions received significant attention, and it was clear that numerous and highly advanced finite
element programs and procedures are now available for the solution of one-, two- and three-dimensional local analyses.
Material and geometric properties are considered, as are the gradual spread of yielding, stability effects, and so on. It was
noted that parametric investigations of this type will permit the determination of the most important factors for the
behaviour and strength of the connections, as well as the separation of the individual effects. For example, several
speakers pointed out that it is important and useful to determine how much of the total rotation in an end-plate
connection that can be attributed to end-plate bending, how much to bolt elongation, how much to column flange
bending, and so on. Finally, although some three-dimensional experimental work has been performed for beam-to-
column assemblies, with reasonable correlation between tests and theory, much work needs to be done to improve the
total data base.
Several contributors adressed the needs for consideration of serviceability and ultimate limit states for connections,
although no general consensus nor conclusions were arrived at. It was felt that it may be premature to consider such
criteria, since the responses of connections vary so widely, and the controlling factors can be widely different. For
example, the question of local cracking was mentioned in several cases, as caused by weld contraction strains through
lamellar tearing and other weld problems. Clearly, this focuses attention on the needs for understanding the response of
the material itself under a variety of conditions, but most of all under the type of restraint that it will only experience in
the connection regions of a structure. It would appear to be better to define overall modes of failure, and to identify the
ones that are preferable in terms of ductility, deformation levels, and overall behaviour. This will remove some of the
problems that can be encountered when purely theoretical solutions appear to predict unreasonably high local stresses,
for example, especially if the latter is based on a computational model of limited sophistication.
Summarizing, numerous advanced models of behaviour are currently available for the analysis of connections, and the
correlation with test results is generally good. The tools for solving most, if not all types of connection analysis problems
therefore seem to be at hand.
< previous page page_94 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_94.html07/11/2009 12:06:07 PM
page_95

< previous page page_95 next page >

Page 95
SESSION No 2A
MATHEMATICAL MODELS
< previous page page_95 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_95.html07/11/2009 12:06:08 PM
 page_96

< previous page page_96 next page >

Page 96
FINITE ELEMENT MODELING OF CONNECTIONS
D.Beaulieu and A.Picard
Département de Génie civil
Université Laval
Sainte-Foy, Québec, Canada
G1K 7P4
ABSTRACT
Researchers traditionally depended almost exclusively on expensive laboratory tests to understand the behavior of
connections in steel structures and to derive empirical rules for design. Finite element modeling appears today as a
valuable and less costly alternative (or more precisely a complement) to laboratory testing. The finite element method is
no longer an esoteric and inaccessible tool in view of recent developments in the fields of numerical analysis, computer
graphics and computer power. In this short paper, we intend to describe the research and development efforts done at
Laval University which leaded to modeling of steel connections by the finite element method.
INTRODUCTION
The experimental approach to study the behavior of connections in steel structures will certainly remain the most popular
for still some years but because of the high costs involved, researchers are increasingly looking for less costly but
acceptable alternatives. The most obvious alternative is modeling by the finite element method.
Due to the highly complex nature of connections and the large number of parameters involved, numerous tests are
required before an adequate set of empirical formulae is developed for the design of a specific type of connection. It
appears to be more rational and more economical to develop numerical models to play with the various parameters and
to check the accuracy of the numerical models against the results of an appropriate number of experimental tests. Not
only are experimental tests needed to validate the models but they are also required for calibration purposes. Using finite
element models therefore does not mean discarding or discrediting laboratory testing and never will.
Such a vision is shared by many researchers but it is not until lately, with the development of more powerful elements
and numerical
< previous page page_96 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_96.html07/11/2009 12:06:09 PM
 page_97

< previous page page_97 next page >

Page 97
techniques, and the rapid evolution of computers and computer graphics that it is believed possible. To come up with
acceptable models, however, is not a simple task and requires several years of research and development.
Since we cannot be too specific due to the limited length of the paper and the large variety of subjects to cover, we will
limit ourselves, in the next sections, to a condensed presentation of the computer programs we use and to brief
descriptions of some elements and techniques of resolution which are more appropriate than others to simulate the
behavior of connections and some which have been developed specifically for that purpose. We will finally present some
applications and point out items requiring further research and development.
STATE OF THE ART
The numerical simulation of steel connections is one of many applications resulting from developments done over the
last twenty years by a group of specialists and several students at Université Laval in Québec and at Université de
Technologie de Compiègne in France. A modular finite element code called MEF [1] and an interactive graphic pre post-
processor called MOSAIC were created and are in continuous development. MEF contains a library of over sixty
standard elements plus as many specialized elements which can be used for two and tridimensional structural and solid
mechanic applications. About twenty other elements are presently under development. The standard and special elements
and modules are used to solve a large variety of problems: linear elastic, nonlinear (plastic, large displacements, large
strains), stability, vibration, dynamic, contact, composite materials, etc… Also, many other specialized elements have
been prepared for specific applications in fluid mechanics, hydraulics, heat transfer, accoustics, form optimization, etc…
We are presently capable of adressing problems like friction, slip, contact, plate stiffening, plate stability and plasticity
which are generally associated with structural steel connections. Bolting and welding can also be reasonably modeled in
2-D and 3-D applications. The types of problems we can analyze are numerous but we do not pretend to be able at the
present time to offer an acceptable solution to every one of them. There is still room for improvement and more research
is required, as shown later.
SPECIALIZED ELEMENTS AND MODULES
Plate and Shell Elements
The library contains a good selection of triangular and rectangular plate and shell elements which can be used for elastic,
elasto-plastic and stability analyses of connections [2–8]. Stiffened plates in two and three dimensional applications can
be properly handled.
Spring Elements
Welded and bolted connections in linear two and tridimensional applications can be grossly but properly represented by
calibrated springs or bar elements [9, 10, 11]. This obviously constitutes the simplest representation.
< previous page page_97 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_97.html07/11/2009 12:06:10 PM
 page_98

< previous page page_98 next page >

Page 98
Bolts can be represented by a combination of bar elements [10, 12, 13]. As shown in Fig. 1a, two bars are used to
simulate the bolt prestress and a third is used to simulate shear. This principle was adapted to two as well as
tridimensional nonlinear elasto-plastic models.

Fig. 1—Bolt modeling.

Contact elements
Contact between two solids can be modeled by a set of nonlinear orthogonal springs calibrated according to the nature of
the contact between the surfaces [10, 14, 15]. These elements can represent separation or firm contact between solids and
perfect slip or friction according to Coulomb’s law.
In order to account for slip and contact in a bolted connection, a contact element is coupled with the bar element which
simulates shear, as shown in Fig. 1b [10].
Weld element
Two welded plates can be roughly modeled considering they behave monolithically. A more refined model is obtained
using springs, as mentioned previously, the spring constants being ultimately determined from experimental tests on the
weld material [16].
For a detailed representation of the behavior of fillet welded connections, we have developed a 15 node tridimensional
prismatic quadratic element which is used in combination with a 20 node tridimensional hexahedric quadratic (Brick)
element to represent the connected plates [11]. These elements allow a representation of the elasto-plastic behavior of the
material considering strain-hardening but were developed assuming small strains and small displacements. The angle of
loading with respect to the orientation of the filled weld is automatically accounted for in the model.
< previous page page_98 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_98.html07/11/2009 12:06:22 PM
 page_99

< previous page page_99 next page >

Page 99
The inherent mechanical properties of the weld material were obtained from recent test results [17].
Each one of these elements has been tested and calibrated against numerous published experimental and numerical
results. The details can be found in the various papers and reports.
Modules
A large amount of sophisticated numerical techniques had to be developed or implemented and tested to simulate
nonlinear plastic and geometric behavior as well as post-buckling behavior and, to some extent, large strains [9, 10, 11,
18]. These are contained in special modules which interact with MEF.
APPLICATIONS
At the present state of development, the MEF and MOSAIC programs and the various specialized modules and elements
can be used to study a large variety of steel connections and problems. Although we are still in the early stages of our
research and development program on modeling of steel connections by finite elements, we have been able to analyze a
number of representative examples [10, 11]. Some of them had already been studied numerically or experimentally and
were used to validate our numerical models.
Due to space limitations, it is not possible to present the results of our analyses; these will be published elsewhere.
Sketches of some of the connections considered, however, are presented in Fig. 2. Problems like friction, slip, contact,
plasticity, prying action and stress distribution, among others, were properly adressed. We are at the present time doing
full numerical and experimental studies of standard girt-to-column connections used in light steel buildings (see Fig. 2d)
and of standard column base plate connections, such as shown in Fig. 3.
FUTURE DEVELOPMENTS
Following is a list of items we have identified for future research and development:
– A major project, already under way, is to create an expert system for nonlinear resolution in which the different
techniques will be used in a simple and efficient manner.
– A routine for graphical color representation of stresses in nonlinear 3-D plastic models is yet to be fully implemented.
– For the simulation of tridimensional welded connections, we shall develop an element which would consider second
order terms since deformations can be quite large in fillet welds. Using proper experimental data, it shall also be
possible to include residual stresses in welded connections. These two additions would certainly improve our model.
Moreover, tridimensional modeling requires the use of powerful computers and the acquisition of a
< previous page page_99 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_99.html07/11/2009 12:06:22 PM
 page_100

< previous page page_100 next page >

Page 100
vectorial processor will improve our ability to perform sophisticated 3-D models.

Fig. 2—Examples of connections studied.

< previous page page_100 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_100.html07/11/2009 12:06:24 PM
 page_101

< previous page page_101 next page >

Page 101

Fig. 3—Typical column base connection

- The bolt model shall be improved since ultimate traction and shear resistances seem to be overevaluated when
compared to available experimental results. This could simply be done by an appropriate calibration of the bar
elements. Besides, plastic failure criteria shall be implemented for welds, bolts and plates.
- What probably constitutes our most challenging project is the development of a global finite element computer
program for frame analysis which would account for partial connection rigidity. This program would make an
extensive use of interactive graphics [22] and would interact with a data base containing pertinent information on
various standard connections. The program would also handle elastic, elasto-plastic, static, dynamic and stability
analyses. A great deal of work is already done.
We have developed a tridimensional beam element which accounts for geometric and material nonlinearities, residual
stresses, initial deformations and post-buckling behavior [9]. The element is based on the theory of thin shells and
considers the progression of plasticity through the thickness of the plates. Connections are simulated by means of 3-D
springs of constant rigidity. The incremental formulation will allow us to consider nonlinear connection behavior without
any great difficulty.
The major task, as one could guess, is to fill the data base with proper information on M-Θ curves for the various types
of connections. Available information would have to be gathered and compiled in an appropriate format. Missing and
new data would have to come from either experimental testing or finite element modeling. This project of academic
character shall keep several people busy for quite some time and is only possible within a framework of international
collaboration between interested researchers.
< previous page page_101 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_101.html07/11/2009 12:06:25 PM
 page_102

< previous page page_102 next page >

Page 102
ACKNOWLEDGEMENTS
The authors wish to acknowledge E.Akoussah, M.Fafard, G.Gendron, Y.Matte and F.Parrot who contributed to the
preparation of this paper.
REFERENCES
1. Dhatt, G. and Touzot, G., The finite element method displayed. J. Wiley and Sons, N.Y. & Toronto, 1984, 510 p.
2. Fafard, M., Beaulieu, D. et Dhatt, G.S., Etude par éléments finis de la stabilité des profilés formés par assemblage
d’éléments plats dans l’espace. Raport no GCT-84–06, Génie civil, Université Laval, Québec, décembre 1984, 168 p.
3. Batoz, J.L. et Ben Tahar, M., Formulation et évaluation d’un nouvel élément quadrilatéral à 12 D.L.pour la flexion des
plaques minces. Université de Technologie de Compiègne, France, 1981.
4. Marcotte, L. et Dhatt, G.S., Instabilité des plaques minces semi-infinies sur fondation élastique par éléments finis.
Rapport GCS-84–02, Génie civil, Université Laval, Québec, 1984.
5. Batoz, J.L. et Dhatt, G., Plaques et coques par éléments finis. Analyse linéaire et non linéaire. Institut pour la
promotion des sciences de l’ingénieur, Paris, juin 1986.
6. Ben Tahar, M., Analyse élasto-plastique des plaques et coques minces par éléments finis. Thèse de doctorat,
Université de Technologie de Compiègne, France, 1981.
7. Matte, Y., Développement de nouveaux éléments triangulaires pour l’analyse linéaire statique et dynamique de
plaques et coques minces. Thèse de maîtrise, Université Laval, Québec, 1986.
8. Talbot, M., Comparaison de deux éléments de coques triangulaires plats utilisant une formulation lagrangienne
actualisée. Thèse de maîtrise, Université Laval, Québec, 1985.
9. Akoussah, E., Beaulieu, D. et Dhatt, G.S., Etude par éléments finis de la stabilité des poteaux dans les bâtiments
industriels. Rapport technique à paraître en 1987, Génie civil, Université Laval, Québec.
10. Gendron, G., Beaulieu, D. et Dhatt, G.S., Modélisation par éléments finis du comportement des assemblages
boulonnés dans les charpentes d’acier. Rapport GCT-87–01, Génie civil, Université Laval, Québec, février 1987, 173 p.
11. Parrot, F., Beaulieu, D. et Dhatt, G.S., Modélisation par éléments finis du comportement des assemblages soudés
dans les charpentes d’acier . Rapport GCT-87–02, Génie civil, Université Laval, Québec, mai 1987.
< previous page page_102 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_102.html07/11/2009 12:06:26 PM
 page_103

< previous page page_103 next page >

Page 103
12. Patel, K.V. and Chen, W.F., Nonlinear analysis of steel moment connections. Journal of the structural division,
ASCE, vol. 118, no 8, August, 1984, pp. 1861–1874.
13. Patel, K.V. and Chen, W.F., Analysis of a fully bolted moment connection using NONSAP. Computers and
Structures, Vol. 21, no 3, 1985, pp. 505–511.
14. Lebouvier, F., Analyse non linéaire de structures planes formées de poutres et d’arcs. Thèse de docteur-ingénieur,
Université de Technologie de Compiègne, Compiègne, France, 1985.
15. Mazurkiewicz, M. and Ostachowicz, W., Theory of finite element method for elastic contact problems of solid
bodies. Computers and Structures. Vol. 17, no 1, 1983, pp. 51–59.
16. Richards, R.M., Rabern, D.A., Hormby, D.E., and Williams, G.C., Analytical models for steel connections.
Proceedings of the W.H.Munse Symposium—Behavior of metal structures, Philadelphia, Penn., May 1983, pp. 128–155.
17. Kennedy, D.J.L. and Miazga, G.S., Behaviour of fillet welds as a function of the angle of loading. Structural
Engineering Report 133, University of Alberta, Edmonton, Alta, 1986.
18. Fafard.M., Algorithmes de calcul automatique des configurations pré et post-flambement de problèmes non linéaires
en calcul des structures. Thèse de doctorat à paraître en 1987, Génie civil, Université Laval, Québec.
19. Sterling, G.H., and Fisher, J.W., A 440 steel joints connected by A490 bolts. Journal of the Structural Division,
American Society of Civil Engineers, Vol. 92, no St3, August 1966, pp. 101–118.
20. Fisher, J.W. and Struik, J.H.A., Guide to design criteria for bolted and riveted joints, John Wiley and Sons, N.Y.,
1974.
21. Standing, K.F., Rentschler, G.P. and Chen, W.F., Tests of bolted beam-to-column moment connections. Welding
Research Council Bulletin, no 218, 1976.
22. Bérubé, C., Lamontagne, R. and Picard, A., CAE System for the Design of Steel Structures. Rapport GCT-87–03,
Génie civil, Université Laval, Québec, 1987.
< previous page page_103 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_103.html07/11/2009 12:06:27 PM
 page_104

< previous page page_104 next page >

Page 104
THREE-DIMENSIONAL PHYSICAL AND MATHEMATICAL MODELING OF CONNECTIONS
A.COLSON
Laboratoire de Mécanique et Technologie
Ecole Normale Supérieure de Cachan
F94230 CACHAN
FRANCE
ABSTRACT
This paper concerns the mathematical modeling of the non linear behaviour of the connections. The interest of a 3D
model rather than a 1D or 2D model is discussed regarding the problems of column stability and spatial structures. The
size of the macro-element which is the physical model of the connection is defined relative to the type of frame analysis.
The main parameters, i.e initial stiffness and ultimate strength, of the modeling are introduced from experimental and
mechanical observations.
Different mathematical expressions, according to the loading path, are proposed, and the hypothesis of reduced isotropy
gives interesting results. Finally, cyclic loading with hardening or softening is presented.
INTRODUCTION
Recent developments in the modeling of the non-linear behaviour of the steel structural connections use a 1D model, i. e.
an curve. However, if we look into current industrial practice there is a lack of information regarding curve
and such modeling. The required level of rigidity regarding the transmissible force at each end of a bar, for a three-
dimensional description, is generally the following:
Normal force : Rigid
Shear force/strong axis : Rigid
Shear force/weak axis : Rigid
Bending moment/strong axis : Rigid, semi-rigid, pinned
Bending moment/weak axis : Rigid, semi-rigid, pinned
Torque : Rigid or pinned, but often ne glected.
< previous page page_104 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_104.html07/11/2009 12:06:28 PM
 page_105

< previous page page_105 next page >

Page 105
Since all the analysis methods do not require a particular model for the rigid connections, it seems that for the plane
structures (normal force, shear force/strong axis, bending moment/strong axis) a 1D solution is adequate. But, in fact,
due to unilateral contact, there is an obvious interaction between normal force and bending moment [1], so the level of
semi-rigidity in bending depends strongly on the level of the normal force. At this stage a 2D model is necessary, more
especially for the in-plane buckling conditions of the columns. Futhermore, even for the plane structures, the out-of-
plane buckling conditions, i.e. for the weak axis, it is necessary to know the behaviour of the column base connection
with respect to the weak axis, and including the effects of the normal force. Then a 3D model is necessary. Furthermore,
such a model allows for considering torsional rigidity and bending about the weak axis for some particular spatial
structures.
PHYSICAL MODELING
For the modeling and the frame analysis the overall connection is replaced by a macro-element (Figure 1), which
contains all the components like bolts, welds, angles, etc. The constitutive equation will give the displacement {D} of
one terminal cross section with respect to the other. Each terminal cross section is supposed to remain plane during the
loading, so classical beam theory is always applicable to all of the bars that connect to the considered mode. The
constitutive equation establishes a relationship between the displacement vector and the force
vector .
At this stage such a macro-element may be geometrically integrated into any finite element program [2] that can handle
non linear analysis.

Figure 1

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_105.html (1 of 2)07/11/2009 12:06:29 PM

page_105

Figure 2
< previous page page_105 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_105.html (2 of 2)07/11/2009 12:06:29 PM

 page_106

< previous page page_106 next page >

Page 106
Size of the macro-element
Generally, regarding the St. Venant Problem, the length of this macro-element should correspond approximatively to the
depth of the bar. In current applications we have chosen to reduce the size to zero (keeping the same constitutive
equation between the two terminal cross sections by lengthening the bar until the neutral point (Figure 2). This is
necessary, particularly for the integration into the P.E.P. program [3], which uses the plastic hinge concept rather than
the progressive plastification of the bar.
The correlation with and interpretation of experimental results must be clearly defined in order to arrive at the proper
comparison between theory and tests.
Main Parameters of the Modeling
Observing the general aspect of the experimental results (Figure 3), at least in the range of the currently required
displacements, it is proposed to use the initial stiffness Ki and the ultimate load Fui as the two main parameters. One or
several other parameters must be introduced in order to describe the non linearity. In a general way, our modeling can be
expressed by the following:

(1)
with [R]: initial stiffress matrix (components Ki)
{Fu}; ultimate load vector (components Fui)
ak: set of parameters inducing the non linearity.

Figure 3

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_106.html (1 of 2)07/11/2009 12:06:30 PM

page_106

Figure 4
< previous page page_106 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_106.html (2 of 2)07/11/2009 12:06:30 PM

 page_107

< previous page page_107 next page >

Page 107
In view of the industrial and practical applications we have tried to minimize the number of ak parameters (1, 2 or 3)
depending on the type of loading and the required precision. After correlating with a number of test data it has been
found that the initial stiffness [R] and the ultimate load {Fu} can be obtained in a forseeable way, through a computation
based on the geometric and the mechanical properties of the components of the connection. This means that for industrial
practice, it is not necessary to do experiments for each kind of connection. Only the ak parameters have to be obtained
from experiments, and it is thought possible to establish a limited number of values for each family of connections.
Otherwise, the partition displacement equation is proposed: where {Dt} is the elastic
component of the displacement and {Dan} represents the inelastic part.
a. Initial Stiffness: [R] matrix, such as
For the simple connections this value can be obtained from a local classical analysis [4] [5]. For the more complex
connections we have to use the finite element method [6]. The calculation cost is limited thince it is the first step of an
elastic analysis. Nevertheless, it would be interesting to developp a specific 3D F.E.M. analysis using automatic mesh
generation, static condensation, substructuring techniques, etc, in order to reduce the calculation costs below those of
NON SAP, ABAQUS and others.
b. Ultimate load (U.L.S.: Ultimate limit surface in the forces space) Limit analysis methods have to be used to find all of
the components of the ultimate load vector (Figure 4). Such methods exist for the most classical connections [5] and also
for tubular connections [7]. For the all-welded connections the ultimate limit surface is generally symmetric, and can be
expressed by the relationship

(2)
For the connections with unilateral contact (bolted, for example) the U.L.S. is unsymmetric. For example, the scheme of
limit analysis (Figure 5), using the Von Mises criterion, can be used to find the U.L.S. in the compression zone due to
concrete fracture [8] [9]. A classical limit analysis method applied to the anchor bolt permits the determination of the U.
L.S. in the tension zone (Figure 6).
< previous page page_107 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_107.html07/11/2009 12:06:31 PM
 page_108

< previous page page_108 next page >

Page 108

Figure 5

Figure 6
MATHEMATICAL EXPRESSIONS
At the moment several mathematical expressions have been developed. The experiments on the column base connections
will provide the remainder, which will then be used to propose the same model for all connections in the structures (for
industrial purpose). Most of them are based on the thermodynamics theory. The first model, only available for a radial
monotonic loading path (Figure 7), is:

(3)
It requires only one parameter (a) to introduce the non linearity. The term ‘‘a” assumes a value between 1.5 to 4.5,
depending on the type and size of the connections (bolted, riveted, welded).
The second model [10] uses an incremental form which can incorporate cyclic and non radial loading:

(4)
The non radial loading path is necessary, particularly in modeling buckling of columns. In the simplified case (2D) used
in the PEP program [3], the expressions are (with and):

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_108.html (1 of 2)07/11/2009 12:06:33 PM

 page_108

(5)

(6)
< previous page page_108 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_108.html (2 of 2)07/11/2009 12:06:33 PM

 page_109

< previous page page_109 next page >

Page 109

Figure 7

Figure 8
The third model [4] utilizes the hypothesis of reduced isotropy which is based on a normality law. This means that in the
space of the reduced forces, and in the space of the reduced displacements, with
the behaviour does not depend on the loading path (Figure 8). The normality law is used with
respect to the first loading surface (F.L.S.), which is homothetic to the U.L.S. (Figure 9), and grows with the loading.
The general mathematical form is:

Under a reasonable set of hypothesis the second and the third model give the same integrated expressions:

(7)
Furthermore, the third model gives . This means that only one parameter is necessary to take into account the
non linearity.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_109.html (1 of 2)07/11/2009 12:06:35 PM

page_109

Figure 9

Figure 10
< previous page page_109 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_109.html (2 of 2)07/11/2009 12:06:35 PM

 page_110

< previous page page_110 next page >

Page 110
CYCLIC LOADING
Inside the F.L.S. the Masing rule, well known in continuum mechanics and extended to structural mechanics, is used to
describe the cyclic loading. At each reversal of loading (order h) a new reference axis system is defined
(Figure 10), with a homothetic ratio of 2 for the variables, thus: and
The softening or hardening during cyclic loading may be modeled by varying the C parameter (equation 7), using, for
example, the expression:
. A comparison between experiment and theoretical results, for a cyclic hardening case
is given in the figure 11.

Figure 11
CONCLUSION
For the modeling of the non linear behaviour we propose physical models based on the geometric and mechanical
properties of the connections, rather than using a curve-fitting technique only. The three dimensional aspect allows be
accounting for of most of the problems, even those associated with planar structures. In their final forms, the
mathematical expressions remain simple enough to use in practical design.
< previous page page_110 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_110.html07/11/2009 12:06:36 PM
 page_111

< previous page page_111 next page >

Page 111
ACKNOWLEDGEMENTS
The author wish to acknowledge P.Penserini who contributed to the preparation of this paper.
REFERENCES
1. Beaulieu, D., Picard, A. “Contribution des assemblages avec plaque d’assise à la stabilité des poteaux”. Construction
Métallique, No2, 1985, pp3–19.
2. Poggi, C., Zandonini, R. “Behaviour and strength of steel frames with semi-rigid connections”. A.S.C.E. Convention,
Detroit, October 1985.
3. Galea, Y., Colson, A., Pilvin, P. “Programme d’analyse de structures planes à barres avec liaisons semi-rigides à
comportement non linéaire”. Construction Métallique, No2, 1986 pp18–25.
4. Louveau, J.M. “Modélisation du comportement cyclique des assemblages de structures métalliques”. Thèse de IIIème
cycle. Université Paris VI 1986.
5. Yee, Y.L., Melchers, R.E. ‘‘Moment-Rotation curves for bolted connections”. Journal of Structural Engineering A.S.C.
E., Vol 112, No3, 1986. pp615–635.
6. Patel, K.V., Chen, W.F. “Analysis of a fully bolted moment connection using NONSAP”. Computers and Structures,
Vol 21, No3, 1985, PP505–511.
7. Mouty, J. “Calcul des charges ultimes des assemblages soudés de profils creux carrés et rectangulaires” . Construction
Métallique No2, 1976, pp37–58.
8. Salençon, J. “Quelques résultats théoriques concernant la butée d’une paroi sur un coin plastique”. Annales I.T.B.T.P.,
No 313, Janvier 1974, pp 187–194.
9. Penserini, P. “Modélisation du comportement des liaisons structures-fondations”. Thèse de Doctorat. Université Paris
VI. To be published.
10. Pilvin, P. “Modélisation du comportement des assemblages de structures à barres”. Thèse de IIIème cycle. Université
Paris VI. 1983.
< previous page page_111 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_111.html07/11/2009 12:06:38 PM
 page_112

< previous page page_112 next page >

Page 112
MATHEMATICAL MODEL FOR SEMI-RIGID JOINTS UNDER CYCLIC LOADS
F.M.MAZZOLANI
Istituto di Tecnica delle Costruzioni
Engineering Faculty
University of Naples (Italy)
ABSTRACT
The main features of a very sophisticated mathematical model for inter preting the actual behaviour of semi-rigid
connection under alternate loading conditions are presented in this paper. After a short analytical presentation, some
possible applications of this model are illustrated. It has been calibrated on the basis of experimental results and starting
from appropriate behavioural parameters ispired to recent ECCS Recommendations, it can be used to extrapolate the
existing results and to interpret the semi-rigid joint performance.
INTRODUCTION
The aim of the present paper is to show some possibilities of application of the mathematical model which has been
proposed in [1] for semirigid connections. The main data of this model are the following:
a) Among the large number of models existing in the literature, the proposed one belongs to the category of the
sophisticated models.
b) The fitness with the actual behaviour of the connection is achieved by means of curve-fitting techniques which utilize
up to 12 numerical coefficients.
c) It has been checked on some existing results of cyclic tests on beam-to-column connections both in steel and in
reinforced concrete with a very satisfactory agreement [2].
d) It has been also used to calibrate the unified testing procedure recently proposed by the task group 1.3 “Seismic
Design” of the European Convention for Constructional Steelwork (ECCS) [3], by means of appropriate behavioural
parameters [4].
< previous page page_112 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_112.html07/11/2009 12:06:39 PM
 page_113

< previous page page_113 next page >

Page 113
e) A recent research program on beam-to-column steel joints [5] provided a lot of experimental results of hysteretic
relationships, which have been strictly interpreted and exploited by means of the proposed model.
THE MATHEMATICAL MODEL
The assumed moment-rotation relationship is subdivided into four branches (see fig. 1a). The increasing branch /a/ (from
O to M+) can be analytically expressed by a three terms relation

(1)
where the first term φ0 is the residual rotation of the previous cycle; the second term is the Ramberg-Osgood type
rotation given by

(2)
the third term is the incremental rotation due to slip between M' and M", given by

(3)
being
M2 the moment which corresponds to a rotation twice the elastic one; α,r,s,K1 numerical coefficients;

The deterioration parameters and are given by

(4)

(5)
depending upon the energy variation (see fig. 1b), being:
the overall energy dissipated at collapse;

the minimum (maximum) value of slip measured at the first cycle and at collapse,
respectively;
< previous page page_113 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_113.html07/11/2009 12:06:40 PM
 page_114

< previous page page_114 next page >

Page 114
∆R the decrease of joint stiffness measured in the cycle before collapse; m,n numerical coefficients.
The decreasing branch /b/ (fron M+ to O) is expressed by

(6)
being

(7)
where ∆'R and ∆"R simulate different unloading conditions (see fig. 1c). Branches /c/ and /d/ are represented by the
same formulation of branches /a/ and/b/.
THE BEHAVIOURAL PARAMETERS
The experiments on structural components under cyclic loads are carried out with particular testing procedures which
usually differ each other. It doesn’t allow to strictly compare the various results coming from different laboratories.
The necessity to unify such experimental procedures is unquestionable and the ECCS tasking group “Seismic Design”
recently decided to work out a proposal for a recommended testing procedure for assessing the behaviour of structural
steel elements under cyclic loads [3].
The main problem was to define the fundamental parameters for the interpretation of the overall behaviour of the
structural components. They have to characterize the degree of ductility, the amount of energy absorption and the
deterioration of rigidity and strenght as far as the number of cycles increases.
The following behavioural parameters have been defined (see fig. 1d,e):
a) partial ductility:

b) full ductility:

c) full ductility ratios:

d) resistance ratios:

e) rigidity ratios:

< previous page page_114 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_114.html07/11/2009 12:06:42 PM
 page_115

< previous page page_115 next page >

Page 115
f) absorbed energy ratios:

The reference parameters Fy, ey represent the upper bound of the quasi-elastic behaviour. They are conventionally
defined by the intersection of the two tangent lines shown in fig. 1f.
INTERPRETATION OF CYCLIC RESULTS
The proposed mathematical model can be used to interpret the results of cyclic tests on structural elements [4]. The
testing procedure given in [3] has been exactly simulated following the same steps. The load versus deflection hysteretic
loops are automatically plotted, as it is shown in fig. 2. As the most significat parameters for comparing the cyclic
behaviour, we have assumed the following:
– peak strenght (Si), corresponding to resistance;
– secant modulus (Ri), corresponding to rigidity;
– energetic efficiency (Ei), corresponding to absorbed energy;
– ductenergy (Di), corresponding to the product of ductility and absorbed energy.
The variations of such parameters as a function of the increasing deflection are also given in fig. 2. These curves are
normalized referring to the limit elastic values (Fy, dy). Their behaviours represent the basic elements when we compare
the performance of different typologies of structural components (i.e. joints, connections,bracings,…) from the point of
view of their reliability under alternate loading conditions. They can be also utilized to define the ultimate limit states
which conventionally correspond to the end of the test. In fact it can occur that the structural component is submitted
during the test to conditions which are out of interest from the physical point of view. Looking at these variations we can
recognize that an ultimate limit state is reached when a significant reduction of the main behavioural parameters arises.
As an example, some ultimate limit states can be identifies as follow:
a) The deterioration of the material due to geometrical or mechanical effects produces a reduction of the peak load
carrying capacity (Si) during the cyclic test.
b) The excess of deformation produces a reduction of the rigidity parameter (Ri), which becomes unacceptable when
leads to unstable effects on the overall structure or introduces too large residual deformations.
c) The structural component is unable to fulfil its function when the energetic efficiency parameter (Ei) is reduc (it can
happen, for instance, in seismic resistant structures).
The interpretation of the experimental cyclic result by means of these
< previous page page_115 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_115.html07/11/2009 12:06:45 PM
 page_116

< previous page page_116 next page >

Page 116
parameters and the definition of their limit values should also require the consideration of their influence on the overall
behaviour of the structure which the component belongs to.
CALIBRATION OF THE MODEL
The cyclic behaviour of 14 beams-to-column specimens has been experimentally analysed in [5]. They have been
designed following the current technological types communly used in rigid and semirigid connections for framed steel
structures. In fig. 3 the specimens are grouped in four main categories (A,B,C,D), which cover the practical range of
behavioural variability of load versus displacement relationships, as it is shown from the corresponding experimental
hysteretic loops.
The proposed model has been applied to such results, giving the simulated curves plotted in the lower part of fig. 4. The
upper diagrams give the ratios between simulated and experimental values for the strenght Fi and the area Ai of a half-
cycle at each value of the imposed displacement ν. The corresponding scatters are usually very satisfactory.
CONCLUSIONS
The above results allow to conclude that the proposed mathematical model is able to fit very closely the experimental
results of beam-to-column joints. It has been satisfactory used in the interpretation, exploitation and extrapolation of
testing results. It seems also a very appropriate model to be introduced in a more sofisticated analysis of the overall
behaviour of semirigid framed structures.
ACKNOWLEDGMENTS
This research is supported by the Italian Ministry of Education. The research team is composed by Edoardo Cosenza,
Antonello De Luca, Attilio De Martino and Ciro Faella.
REFERENCES
1. De Martino A., Faella C. and Mazzolani F.M., Simulation of beam-to-column joint behaviour under cyclic loads.
Costruzioni Metalliche, n. 6 1984
2. Cosenza E., De Luca A., De Martino A., Faella C. and Mazzolani F.M., Inelastic behaviour of framed structures under
cyclic loads. Euromech Colloquium 174, Palermo, October 1983
3. ECCS-CECM-EKS, Recommended Testing Procedure for Assessing the Behaviour of Structural Steel Elements under
Cyclic Loads. Technical Working Group 1.3, Seismic Design, n. 45, 1986
4. De Martino A., Faella C. and Mazzolani F.M., Il comportamento ciclico dei nodi in acciaio: metodologia
interpretative. Giornate Italiane della
< previous page page_116 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_116.html07/11/2009 12:06:46 PM
 page_117

< previous page page_117 next page >

Page 117

Costruzione in Acciaio, 10° Convegno C.T.A., Montecatini, Ottobre 1985

5. Ballio G., Calado L., De Martino A., Faella C. and Mazzolani F.M., Steel beam-to-column joints under cyclic loads:
experimental and numerical approach. Eighth European Conference on Earthquake Engineering, Lisbon, September
1986

Figure 1—The characterization of the hysteretic loops

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_117.html (1 of 2)07/11/2009 12:06:47 PM

page_117

< previous page page_117 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_117.html (2 of 2)07/11/2009 12:06:47 PM

 page_118

< previous page page_118 next page >

Page 118

Figure 2—Simulation of cyclic test results and evaluation of behavioural parameters

< previous page page_118 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_118.html07/11/2009 12:06:49 PM
page_119

< previous page page_119 next page >

Page 119

Figure 3—Tested beam-to-column specimens and experimental hysteretic results

< previous page page_119 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_119.html07/11/2009 12:06:50 PM
 page_120

< previous page page_120 next page >

Page 120

Figure 4—Simulated hysteretic loops and comparison with experimental results for strength (Fi) and energy absorption
(Ai)
< previous page page_120 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_120.html07/11/2009 12:06:51 PM
 page_121

< previous page page_121 next page >

Page 121
Moment-Rotation Relation of Top- and Seat- Angle with Double Web-Angle Connections
by
N.Kishi,1 W.F.Chen,2 K.G.Matsuoka3 and S.G.Nomachi4
ABSTRACT
In this paper, the moment-rotation relationship of top- and seat- angle with double web-angle steel beam-to-column
connections is developed. In this development, the initial elastic stiffness and the ultimate moment carrying capacity of
the connection are determined by using a simple analytical procedure for modeling the top-, seat-, and web-angles of the
semi-rigid connections.
Using the Initial connection stiffness and the ultimate moment carrying capacity so obtained, a three-parameter power
model is found to be adequate for representing the moment-rotation relationship of this type of connections.
1 Associate Professor of Civil Engineering, Muroran Institute of Technology, Muroran, 050 Japan.
2 Professor and Head of Structural Engineering, School of Civil Engineering, Purdue University, West Lafayette, IN
47907 U.S.A.
3 Professor of Civil Engineering, Muroran Institute of Technology, Muroran, 050 Japan.
4 Professor of Civil Engineering, College of Industrial Technology, Nihon University, Izumi, Narashino, 275 Japan.
< previous page page_121 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_121.html07/11/2009 12:06:52 PM
 page_122

< previous page page_122 next page >

Page 122
1. INTRODUCTION
The connection type considered here is the top- and seat-angle with double web-angle connections as shown in Fig. 1.
This connection type has the inherent ductility offered by the flexural deformation of both flange and web angles in the
legs attached to the column. In this paper, an analytical procedure is developed to predict the moment-rotation
characteristics of this type of connections by determining its initial stiffness and ultimate moment capacity. A three-
parameter power model is used here to represent the whole moment-rotation behavior of the connections. The
experimental results reported by Altman et al (1982) and Azizinamini et al (1985) are used here to verify the proposed
procedure.
2. FORMULATION
2.1 Initial Stiffness
Deflected configuration of the top- and seat- angle with double web-angle connections is shown in Fig. 2, based on the
results of Altman et al (1982) and Azizinamini et al (1985). The beam rotated with respect to the stub column, essentially
as a rigid member by pivoting about a point near the surface of the beam compression flange.
To determine the initial elastic connection stiffness Rki, the connection is modeled as follows:
1. The beam and the angle legs adjacent to the beam act as rigid members.
2. Materials of the angle legs adjacent to the column flange are linearly elastic and their displacements are small.
3. The center of rotation for the connection is located at the angle leg adjacent to the compression beam flange.
4. The top angle acts as a cantilever beam in which the fixed support line is assumed to be at the fastener-nut’s edge near
the beam flange in leg adjacent to the column face as shown in Fig. 3.
5. The web angle acts as a cantilever beam similar to the behavior of the top angle shown in Fig. 4.
6. The resisting moment at the center of rotation is neglected.
Based on these assumptions and considering the shear deformation in the leg of the top angle, the horizontal
displacement ∆t of the heel of the top angle corresponding to the beam flange force Pt (Fig. 3) is

(1)
< previous page page_122 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_122.html07/11/2009 12:06:53 PM
 page_123

< previous page page_123 next page >

Page 123
in which
EIt=bending stiffness of the leg adjacent to the column face of top angle,

(2)

(3)
W=nut’s width across flats of bolt,
tt=thickness of the top angle.
Here, the coefficient of shear deformation κ is taken as κ=6/5 (Gere and Timoshenko, 1984).
Following the same procedure for the top angle connection, the mean value ∆a of the horizontal displacement of the
angle’s heel corresponding to total force Pa acting in one web angle (Fig. 4) is:

(4)
in which
EIa=bending stiffness of the leg adjacent to the column face of web angle,

(5)

(6)
ta=thickness of the web angle.
Assuming the Pa acts at the middle height of the web angle and referring to the deflected configuration shown in Fig. 2,
the relationship between the horizontal displacements of each connection (∆t, ∆a) and the connection rotation θr are:

(7)

(8)
in which

(9)

(10)
< previous page page_123 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_123.html07/11/2009 12:06:55 PM
 page_124

< previous page page_124 next page >

Page 124
Noting that the two angles attached to beam web and column flange and taking moment about the center of rotation
(Point C in Fig. 2), the moment has the value

(11)
Substituting Eqs. (1) and (4) into Eqs. (7) and (8) respectively, and then into Eq. (12), the following relationship between
M and θr is obtained:

(12)
Thus, the initial connection stiffness Rki has the value;

(13)
2.2 ultimate Moment
Based on the experimental results reported by Altman et al (1982) and Azizinamini et al (1985), we assume the collapse
mechanisms for the top- and seat- angle connection and of the web-angle connection as shown in Figs. 5 and 6,
respectively.
Since the distance between the two plastic hinges is rather short compared with each angle’s thickness, the effect of
shear force on the yielding of material must therefore be considered. The estimation of the ultimate strength for each
connecting part is based on the analytical procedure proposed previously by Chen and Kishi (1987) or Kishi et al (1987).
For the top- and seat- angle connecting part:
Similar to the procedure used previously by Chen and Kishi (1987), the upper plastic hinge H1 locates at a distance tt/2
from the fixed point, the work equation for the mechanism shown in Fig. 5 is:

(14)
in which Mpt is the plastic moment and Vpt is the shear force acting at the plastic hinges in the top angle leg.
Using the Drucker’s yielding criterion (1956) for the combined bending Mpt and shear Vpt, a 4-th order equation with

respect to is obtained:

(15)
< previous page page_124 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_124.html07/11/2009 12:06:56 PM
 page_125

< previous page page_125 next page >

Page 125
in which
Vot=plastic shear capacity of the top angle leg without coupling,

(16)
σy=yield stress of top angle,
g2=distance between the two plastic hinges,

(17)
kt=distance form the top angle’s heel to the toe of the fillet as shown in Fig. 5.
The ultimate shear strength Vpt can be determined by solving Eq. (15). The ultimate moment at the center of rotation
with no shear action is

(18)
For the web-angle connecting part:
Based on the assumed simple mechanism as shown in Fig. 6, we consider the ultimate strength per unit length at an
arbitrary section first, and then obtain the total ultimate strength by integrating them along the plastic hinge line of the
web angle.
The work equation for the mechanism at an arbitrary section y in Fig. 6 with the plastic moment Mpy and the shear force
Vpy per unit length for the one web angle is given by

(19)
Furthermore, using the Tresca’s yield criterion and the Drucker’s yield criterion (1956) similar to the procedure used
previously for the top- and seat- angle connection, we obtain the 4-th order equation with respect to (Vpy/Voa):

(20)
in which
Voa=plastic shear capacity per unit length of the web angle leg without coupling,

(21)
gy=distance between two plastic hinges at an arbitrary section y,
< previous page page_125 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_125.html07/11/2009 12:06:58 PM
 page_126

< previous page page_126 next page >

Page 126

(22)
ka=distance from the web angle’s heel to the toe of the fillet.
The values of Vpy can be determined by solving Eq. (20). However, the plastic shear force Vpy becomes parabolic
distribution along with angle height y which has the minimum value Vpu at the upper edge of the angle (y=1p) and the
maximum value Vpy= Voa at the lower edge . To simplify the analytical procedure, we assume a linear
distribution for Vpy as shown in Fig. 7. The resultant plastic shear force Vpa has

(23)
The above value is for one angle connected to beam web and column flange. Thus, the value for double web-angle
connection is twice of Eq. (23).
Total ultimate moment capacity:
The ultimate state of the connection is shown in Fig. 8. Using the ultimate strength obtained from Eqs. (14), (15) and
(18) for the top- and seat- angle part of the connection and Eq. (23), for the web-angle part, the ultimate moment
capacity Mu about the center of rotation (Point C) is given by

(24)
in which

(25)

(26)
2.3 Moment-Rotation Relationship
The power model proposed by Richard (1961) and used later by Goldberg and Richard (1963) is adopted here. This
model was used recently by Chen and Kishi (1987) and Kishi et al (1987) to represent the moment-rotation relationship
of top- and seat-angle connections and single/double web-angle connections respectively. This power model is an
effective tool for designer to execute the second-order nonlinear analysis quickly and accurately. This is because the
connection stiffness can be determined directly without iterative procedure and has no negative stiffness.
< previous page page_126 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_126.html07/11/2009 12:06:59 PM
 page_127

< previous page page_127 next page >

Page 127
Using the initial connection stiffness Rki and the ultimate moment capacity Mu of the connection, the moment-rotation
(M- r) relationship can be represented adequately by the power model

(27)
in which
θ0=reference plastic rotation

(28)
The connection stiffness Rk in Eq . (27) is

(29)
and the rotation θr is:

(30)
3. EXPERIMENTAL VERIFICATIONS
To verify the power model proposed here, the tests reported by Altman et al (1982) and Azizinamini et al (1985) are
used. Adding the sub-routine developed in the preceding section to our existing data based program (SCDB) by Kishi-
Chen (1986, 1987), the procedures for the assessment of the M-θr relation can be conducted directly and automatically.
The SCDB program includes the routines for analytical and numerical estimation of M-θr curves with the exception of
those for the control of the data base for semi-rigid steel beam-to-column connections.
A comparison example for M-θr-curves is illustrated in Fig. 9. The experimental results are compared with the analytical
power model proposed here, the modified exponential model as the curve-fitting method introduced previously by Kishi-
Chen (1987) and the polynomial model proposed by Azizinamini et al (1985) as Frye-Morris’s model (1975). The
experimental data are taken from the each level of ultimate moment capacity. Selecting a suitable value for the shape
parameter n, the results obtained by the power model agree rather well with the experimental results similar to that of the
polynomial and modified exponential models. It can therefore be concluded here that the proposed power model
represents adequately the moment-rotation behavior of the top-and seat- angle with double web-angle connections.
< previous page page_127 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_127.html07/11/2009 12:07:00 PM
 page_128

< previous page page_128 next page >

Page 128
4. CONCLUSIONS
In this paper, the moment-rotation relationship of top- and seat- angle with double web-angle connections is developed.
The initial connection stiffness and the ultimate moment capacity of the connection are determined by using simple
analytical procedures and are used as two of the three parameters in the proposed power model. The proposed power
model is found in a good agreement with available results. The power model can be easily implemented in a second-
order analysis.
5. REFERENCES
AISC (1986), Load and Resistance Factor Design Manual of Steel Construction, 1st Ed., AISC, November, Chicago.
Altman, Jr. W.G., Azizinamini, A., Bradburn, J.H. and Radziminski, J.B. (1982) Moment-Rotation Characteristics of
Semi-Rigid Steel Beam-Column Connections, Civil Engineering Department, University of South Carolina, Columbia,
South Carolina.
Azizinamini, A., Bradburn, J.H. and Radziminski, J.B., (1985), Static and Cyclic Behavior of Semi-Rigid Steel Beam-
Column Connections, Civil Engineering Department, University of South Carolina, Columbia, South Carolina.
Chen, W.F. and Kishi, N. (1987), Moment-Rotation Relation of Top-and Seat- Angle Connections, CE-STR-87–4,
School of Civil Engineering, Purdue University, West Lafayette, IN.
Drucker, D.C. (1956), The Effect of Shear on the Plastic Bending of Beams, Journal of Applied Mechanics, Vol. 23, No.
4, December, 509–514.
Frye, M.J. and Morris, G.A. (1975), Analysis of Flexibly Connected Steel Frames, Canadian Journal of Civil
Engineering, Vol. 2, 280–291.
Gere, J.M. and Timoshenko, S.P., (1984), Mechanics of Materials, 2nd Ed., Wadsworth, Belmont, CA, 660–666.
Goldberg, J.E. and Richard, R.M. (1963), Analysis of Nonlinear Structures, Journal of the Structural Division, ASCE,
Vol. 89, No. ST4, 333–351.
Kishi, N. and Chen, W.F. (1986), Data Base of Steel Beam-to-Column Connections, CE-STR-86–26, School of Civil
Engineering, Purdue University, West Lafayette, IN, Two Volumes, 653 pages.
Kishi, N. and Chen, W.F. (1987), Data Base of Steel Beam-to-Column Connections, AISC Engineering Journal, Vol. 24,
Chicago.
Kishi, N., Chen, W.F., Matsuoka, K.G. and Nomachi, S.G. (1987), Moment-Rotation Relation of Single/Double Web-
Angle Connections.
< previous page page_128 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_128.html07/11/2009 12:07:00 PM
 page_129

< previous page page_129 next page >

Page 129
Richard, R.M. (1961), A Study of Structural Systems Having Conservative Nonlinearity, Thesis Presented to Purdue
University, West Lafayette, IN. in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy.

Fig. 1. Typical Top- and Seat- Angle with Double Web- Angle Connection
< previous page page_129 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_129.html07/11/2009 12:07:01 PM
 page_130

< previous page page_130 next page >

Page 130

Fig. 2. Deflected Configuration of the Angles Connected to Beam and Column at the Elastic Condition

Fig. 3. Cantilever Model of the Top Angle

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_130.html (1 of 2)07/11/2009 12:07:03 PM

page_130

< previous page page_130 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_130.html (2 of 2)07/11/2009 12:07:03 PM

page_131

< previous page page_131 next page >

Page 131

Fig. 4. Deflected Configuration for Cantilever Model of Web Angle

Fig. 5. Mechanism of the Top Angle at the Ultimate Condition

< previous page page_131 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_131.html07/11/2009 12:07:04 PM
page_132

< previous page page_132 next page >

Page 132

Fig. 6. Mechanism of the Web Angle at the Ultimate Condition

< previous page page_132 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_132.html07/11/2009 12:07:04 PM
 page_133

< previous page page_133 next page >

Page 133

Fig. 7. Distribution of Plastic Shear Force along Web Angle Height

Fig. 8. Applied Forces in the Ultimate State of the Connection

< previous page page_133 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_133.html07/11/2009 12:07:07 PM
page_134

< previous page page_134 next page >

Page 134

Fig. 9 Comparison Between Proposed Power Model and Experimental Results

< previous page page_134 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_134.html07/11/2009 12:07:08 PM
 page_135

< previous page page_135 next page >

Page 135
Moment-Rotation Relation of Single/Double Web-Angle Connections
by
N.Kishi1, W.F.Chen2, K.G.Matsuoka3 and S.G.Nomachi4
ABSTRACT
In this paper, the moment-rotation relationship of single web-angle and double web-angle beam-to-column connections
is developed. In this development, the initial elastic connection stiffness is determined by using the simple bending
torsion theory and the ultimate moment capacity by the simple plastic mechanism.
The complete moment-rotation relationship of the connections is represented by a three-parameter power model. The
analytical model is found to be in a good agreement with available experimental results for single and double web-angle
connections.
1. INTRODUCTION
Single and double web-angle steel beam-to-column connections such as those shown in Fig. 1 are widely used in
practice. In this paper, an analytical procedure is developed to predict the moment-rotation characteristics of these
connections by determining the initial stiffness and the ultimate moment capacity of these connections. Power model
used previously by Chen and Kishi (1987) is then adopted to represent the complete moment-rotation behavior of these
connections. The experimental results reported by Lipson et al (1968) on single web-angle connections, and by Bell et al
(1958), Lewitt et al (1966) and Sommer (1969) on double web-angle connections are used here to verify the proposed
procedure.
1 Associate Professor of Civil Engineering, Muroran Institute of Technology, Muroran, 050 Japan.
2 Professor and Head of Structural Engineering, School of Civil Engineering, Purdue University, West Lafayette, IN
47907 U.S.A.
3 Professor of Civil Engineering, Muroran Institute of Technology, Muroran, 050 Japan.
4 Professor of Civil Engineering, College of Industrial Technology, Nihon University, Izumi, Narashino, 275 Japan.
< previous page page_135 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_135.html07/11/2009 12:07:16 PM
 page_136

< previous page page_136 next page >

Page 136
2. FORMULATION
2.1 Initial Stiffness
The general deformation pattern of the web-angle connections is shown in Fig. 2. The experimental results reported by
Bell, Chesson and Munse (1958) and by Lewitt, Chesson and Munse (1966) on double web-angle connections showed
the following behavior:
1. The center of rotation of the connection was near the mid-depth of the beam during the first few increments of
loading.
2. The deformation and subsequent tearing of the connection angles primarily resulted from bending moment, and the
effect of shear deformation on the connection behavior was relatively small.
To determine the initial elastic connection stiffness Rki, we assume the web-angle to behave in the following manner:
1. The effect of shear on connection deformation is ignored.
2. The part of angle connected to column behaves linearly elastic while the part of angle connected to beam behaves as a
rigid body. The connection deformation is small.
3. The part of angle fastened to column flange acts as a moderate thick plate in which the fixed support is assumed to be
at the fastener-nut edge close to the beam web and the concentrated torsional moment is in equilibrium with the
connection moment acting at the free edge. (Figs. 3 and 4).
Based on these assumptions and using the simple bending torsion theory, the governing differential equation with respect
to angle rotation of a moderate thick plate has the form:

(1)
in which

(2)

(3)
mt=distributed torsional moment
E,G=Young’s and shear moduli respectively.
< previous page page_136 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_136.html07/11/2009 12:07:17 PM
 page_137

< previous page page_137 next page >

Page 137
The boundary conditions are

(4)
at the free edge ;

(5)

(6)
and at the fixed edge ;

(7a,b)
in which
g1=gage distance from the fixed support line to the free edge line as shown in Fig. 4.

where
gc=distance form the angle’s heel to the center of bolt holes in leg adjacent to the column face
W=the width of fastener’s nut
k =the distance from the angle’s heel to the toe of fillet.
Solving Eq. (1) together with Eqs. (4) to (7) and setting θr to at the relationship between the connection
moment M and the rotation θr becomes

(8)

in which and putting

(9)
where
Introducing the new parameter β to simplify the equation, the connection moment M is given by

(10)

where the initial connection stiffness is

(11)
< previous page page_137 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_137.html07/11/2009 12:07:19 PM
 page_138

< previous page page_138 next page >

Page 138
and β is defined as

(12)
Equation (11) is the initial connection stiffness of single web-angle connections. Then Rki for double web-angle
connections is twice the value given by Eq. (11). The initial connection stiffness Rki is now expressed in terms of three
parameters t, 1p and g1.
2.2 Ultimate Bending Capacity
Based on the experimental results reported by Bell, Chesson and Munse (1958), we assume the global collapse
mechanism for the web-angle connections as shown in Fig. 5. Since the distance between the two plastic hinges is of the
same order of magnitude when compared with the angle thickness, we must therefore consider the effect of shear force
on the yielding of the material due to bending.
The work equation for the mechanism at an arbitrary section y in Fig. 5 with the plastic moment capacity Mpy and the
shear force Vpy is given by

(13)
The Drucker’s yield criterion (1956) for the combined bending moment Mpy and shear force Vpy, has the form

(14)
in which M0 and V0 are respectively the pure plastic bending moment capacity and shear force capacity per unit length
of web angle.
Using the Tresca’s yielding criterion, we have

(15)

(16)
in which σy is the yield stress of web angles.
Substituting, Eqs. (13), (15) and (16) into Eq. (14) and rearranging the 4-th order equation with respect to (Vpy/V0) is
obtained as

(17)
< previous page page_138 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_138.html07/11/2009 12:07:20 PM
 page_139

< previous page page_139 next page >

Page 139
The value of Vpy can be determined by solving Eq. (17). The plastic shear force Vpy has a parabolic distribution along
the angle height y. It has the minimum value Vpu at the upper edge of the angle (y=1p) and the maximum value Vpy=V0
at the lower edge . To simplify the analytical procedure, the variation of Vpy is assumed to have a linear
distribution as shown in Fig. 6. The resultant plastic shear force F is

(18)
Taking the moment at the lower edge of the angle, the ultimate moment capacity Mu of the connection is obtained as

(19)
Equation (19) is for single web-angle connections. Thus, for double web-angle connections, the value Mu is twice that of
Eq. (19).
2.3 The Moment-Rotation Relationship
The power model, originally proposed by Richard (1961) and later applied by Goldberg and Richard (1963), was used
recently by Chen and Kishi (1987) to represent the complete moment-rotation relationship of top and seat angle
connections. This power model is found to be an effective tool for designer to execute a second-order nonlinear analysis
quickly and accurately. This is because the connection stiffness can be determined directly from the expression without
recourse to an iterative procedure. It has no negative stiffness in the power model. We therefore adopt this model for the
present web-angle connections.
Using the initial connection stiffness Rki and the ultimate moment capacity Mu of the connections, the moment-rotation
relationship can be represented adequately by the power model in the form

(20)
in which
θ0=a reference plastic rotation

(21)
n=shape parameter
The connection stiffness Rk in Eq. (20) is

(22)
< previous page page_139 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_139.html07/11/2009 12:07:22 PM
 page_140

< previous page page_140 next page >

Page 140
and the rotation θr is

(22)
3. EXPERIMENTAL VERIFICATIONS
To verify the power model proposed here, the tests reported by Lipson et al (1968) for single web-angle connections, and
by Bell, Chesson and Munse (1958), Sommer (1969) for the double web-angle connections are used. Adding the
subroutine of the proposed power model to the data base program (SCDB) developed previously by Kishi and Chen
(1986) at Purdue University, the procedures for experimental verifications of the present model can be conducted in a
direct and automatic manner.
Figures 7 and 8 show the comparison of test results with predicted Initial connection stiffness Rki, in which test results
are based on the Kishi-Chen’s modified exponential M-θr curve-fitting expression. The analytical results agree in
general with the experimental values.
The comparison of M-θr curves with tests is illustrated by Figs. 9 and 10. Figure 9 shows the results for a single-web
connection with and 1p=5.5". Figure 10 shows the comparison for a double web-angle connection as
reported by Lewitt et al (1966). In these figures, the experimental results are compared with analytical values based on
the Kishi-Chen’s modified exponential model (1987), polynominal model proposed by Frye-Morris (1978) and the
power model presented here. The analytical results using the power model agree quite well with the experimental results.
It can therefore be concluded here that the proposed power model represents adequately the moment-rotation behavior of
web-angle connections.
4. CONCLUSIONS
In this paper, analytical moment-rotation relationships of single and double web-angle connections are developed. The
initial connection stiffness and the ultimate moment capacity of these connections are first determined by simple
analytical procedures. The three-parameter power model is then used to represent the complete moment-rotation
relationship for these connections. The proposed model is found to be in a good agreement with available experimental
results.
< previous page page_140 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_140.html07/11/2009 12:07:27 PM
 page_141

< previous page page_141 next page >

Page 141
5. REFERENCES
Bell, W.G., Chesson, E., Jr., and Munse, W.H., (1958), Static Tests of Standard Riveted and Bolted Beam-To-Column
Connections, A Progress Report of an Investigation Conducted by the University of Illinois Engineering Experiment
Station.
Chen, W.F. and Kishi, N. (1987), Moment-Rotation Relation of Top-and Seat-Angle Connection, CE-STR-87–4, School
of Civil Engineering, Purdue University, West Lafayette, IN 47907.
Drucker, D.C. (1956), The Effect of Shear on the Plastic Bending of Beams, Journal of Applied Mechanics, December,
509–514.
Frye, M.J. and Morris, G.A. (1975), Analysis of Flexibily Connected Steel Frames, Canadian Journal of Civil
Engineering, Vol. 2, 280–291.
Goldberg, J.E. and Richard, R.M. (1963), Analysis of Nonlinear Structures, Journal of Structural Division, ASCE, Vol.
89, No. ST4, 333–351.
Kishi, N. and Chen, W.F. (1986), Data Base of Steel Beam-to-Column Connections, CE-STR-86–26, School of Civil
Engineering, Purdue University, West Lafayette, IN, 47907, Two Volumes, 653 pages.
Kishi, N. and Chen, W.F. (1987), Data Base of Steel Beam-To-Column Connections, AISC Engineering Journal, Vol.
24, Chicago.
Lewitt, C.W., Chesson, E. Jr. and Munse, W.H. (1966), Restraint Characteristics of Flexible Riveted and Bolted Beam-
to-Column Connections, Department of Civil Engineering, University of Illinois.
Lipson, S.L. (1968), Single-Angle and Single-Plate Beam Framing Connections, Canadian Structural Engineering
Conference, Toronto, Ontario, 141–162.
Sommer, W.H. (1969), Behavior of Welded Header Plate Connections, Master Thesis, University of Toronto, Toronto,
Ontario.
< previous page page_141 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_141.html07/11/2009 12:07:27 PM
 page_142

< previous page page_142 next page >

Page 142

Fig. 1. Typical web-angle connections

< previous page page_142 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_142.html07/11/2009 12:07:29 PM
 page_143

< previous page page_143 next page >

Page 143

Fig. 2. General deformation pattern of the angle of web-angle connections

Fig. 3. Deformation of the angle during the initial loading

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_143.html (1 of 2)07/11/2009 12:07:30 PM

 page_143

Fig. 4. The moderate thick plate modeling of web-angle

< previous page page_143 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_143.html (2 of 2)07/11/2009 12:07:30 PM

 page_144

< previous page page_144 next page >

Page 144

Fig. 5. Mechanism of the part of the angle connected to the column flange at the ultimate condition
< previous page page_144 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_144.html07/11/2009 12:07:31 PM
 page_145

< previous page page_145 next page >

Page 145

Fig. 6. The distribution of plastic shear force

< previous page page_145 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_145.html07/11/2009 12:07:32 PM
 page_146

< previous page page_146 next page >

Page 146

Fig. 7. Initial connection stiffness Rki for single web-angle connections

< previous page page_146 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_146.html07/11/2009 12:07:33 PM
 page_147

< previous page page_147 next page >

Page 147

Fig. 8. Initial connection stiffness Rki for double web-angle connections

< previous page page_147 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_147.html07/11/2009 12:07:34 PM
 page_148

< previous page page_148 next page >

Page 148

Fig. 9. Comparison of single web-angle connections with and

< previous page page_148 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_148.html07/11/2009 12:07:35 PM
page_149

< previous page page_149 next page >

Page 149

Fig. 10 Comparison of a double web-angle connection

< previous page page_149 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_149.html07/11/2009 12:07:37 PM
 page_150

< previous page page_150 next page >

Page 150
MODELLING OF COLUMN-BASE BEHAVIOUR
R.E.Melchers
Department of Civil Engineering and Surveying
The University of Newcastle
Newcastle, N.S.W., 2308
AUSTRALIA
ABSTRACT
An overview of experimental findings for the moment-rotation behaviour of two-bolt, nominally ‘‘pinned”, column-
bases is presented. Monotonic and alternative loading cases as well as the various parameters of interest are described.
Baseplate thickness and bolt capacity were found to be the main factors contributing to connection stiffness and
connection capacity. After the baseplate has been deformed due to high loading, the connection tends to behave “pinned”
as assumed for subsequent lower loadings. Physical parameters were used to develop a mathematical model to describe
the moment-rotation behaviour. For connections governed by the baseplate response, reasonable prediction of stiffness
and capacity up to the maximum was obtained. For cases governed by bolt response, the prediction was less satisfactory.
INTRODUCTION
Rules exist in many design codes and in standard texts for the proportioning of column-bases. Typically the parameters
considered to be of importance are (i) the thickness of the baseplate; (ii) its plan dimensions; (iii) the weld between
column section and baseplate, and (iv) the diameter and strength of the bolts. Generally, the nature of the grout layer
between baseplate and concrete foundation is not considered in design, although the strength and extent of the concrete
foundation itself is considered.
Interestingly, experimental results on which such recommendations are based are not extensive. For realistic column-
bases subject to concentric loading, Bijlaard [1] gave a series of experimental results for connection strength, showing
that the strength of the grout layer was particularly important. It was argued that in combination with rules derived from
beam column connections, the strength of eccentrically loaded column-bases could
< previous page page_150 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_150.html07/11/2009 12:07:38 PM
 page_151

< previous page page_151 next page >

Page 151
also be calculated. However, no experimental verification appears to have been carried out.
Much earlier, Salmon et al [2] had considered eccentrically loaded and nominally “fixed” column-bases. They obtained a
maximum strength for moment capacity as limited by the ultimate strength of the bolts, the baseplate in bending and the
concrete in compression. They used a model based on the behaviour of reinforced concrete sections. The extensive yield-
line patterns given in [1] were not yet in evidence, nor was there consideration of the layer. No experiments appear to
have been carried out.
More recently, Picard and Beaulieu [3] gave a rather similar result based on very similar reasoning. However, they also
reported tests on 15 connections in which column rotations relative to base rotations were recorded. As might be
expected, the nominally “fixed” connections displayed considerable flexibility. This was related to the elastic stiffness of
the connections through socalled “fixidity factors”. It was found that high axial loads increased the stiffness for small
deformations, as also might be expected. Regarding strength prediction, it was found that this was overestimated
(sometimes grossly) relative to experimentally observed capacity.
The stiffness properties of connections in steel frames are of importance for frame deflection prediction, and for stability
calculations [4, 5]. It is therefore somewhat surprising that for beam-column connections, for which quite extensive
experimental and theoretical work has been done [6], there is only limited information about load-deformation behaviour
useful for frame analysis [7, 8]. The aim of the present research activity is to try to provide such information also for
column-bases. Most effort has thus far been reserved for the column-baseplate-concrete foundation interface region;
preliminary studies are in progress for the concrete-soil interface region.
SUMMARY OF TESTS ON “PINNED” BASE CONNECTIONS
The results obtained from full-scale, industry representative column-base connections are described elsewhere in detail
[9]. The bases were nominally “pinned’’, that is, only two holding-down bolts were used, placed symmetrically about the
column-web and along the column minor axis. Every attempt was made to replicate good field practice in such matters
as bolt tightening, packing placement and grouting. The test program consisted of
< previous page page_151 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_151.html07/11/2009 12:07:39 PM
 page_152

< previous page page_152 next page >

Page 152

Figure 1. Schematic of Test Rig

26 tests, including 3 different column sizes (460 UB; 310 UC; 310 UB), baseplate thicknesses of 12, 16, 20, 25 and 30
mm and bolt sizes of M20 and M25. In all cases a 6 mm fillet weld was used to connect the column to the baseplate.
To simulate the shims used to align the base during erection of the frame, four 20 mm steel cubes were placed under the
base, one each at the mid-points of each baseplate edge.
The loading was applied to the connection through a column stub and lever-arm system as shown in Figure 1. This
allowed the ratio moment/axial load to be varied between tests.
It was found that the moment-rotation behaviour of the column-bases depended largely on baseplate thickness. For
relatively thin baseplates, behaviour tended to be initially elastic and then became partly plastic. Eventually, strain
hardening of the baseplate deformation mechanism governed behaviour. This is illustrated in Figure 2. The baseplates
deformed either in the yield-line pattern of Figure 3b or 3c; the pattern of Figure 3a, while theoretically feasible, was not
observed in the present test series.
For thicker baseplates, overall behaviour tended to be governed more by bolt deformation characteristics (see Figure 4);
in some cases the bolts fractured.
< previous page page_152 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_152.html07/11/2009 12:07:40 PM
 page_153

< previous page page_153 next page >

Page 153

Figure 2. Typical Plate-Dominated Response

Figure 3. Baseplate Yield-Line Mechanisms

In summary, the following observations were made [9]:
1. Increasing baseplate thickness markedly increased connection stiffness;
2. Baseplate plan size appeared to have only a marginal effect on stiffness;
3. Increased bolt size led to greater stiffness, and, for sufficiently thick baseplates, to higher maximum capacity;
4. Increased column size for given plate size increased connection stiffness;
5. Increased loading eccentricity led to slight increases in bending stiffness and moment capacities;
< previous page page_153 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_153.html07/11/2009 12:07:41 PM
 page_154

< previous page page_154 next page >

Page 154
6. The location and type of the packing (shims) appeared to be of no importance;
7. There was little evidence of grout failure, presumably because axial compression zone forces were not sufficiently
large;
8. Prying forces were found to occur only at low loads; at high loads sufficient plate-grout separation had occurred to
remove the possibility of prying action.

Figure 4. Typical Bolt-Dominated Response

Figure 5. Behaviour Under Alternating Load

< previous page page_154 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_154.html07/11/2009 12:07:43 PM
 page_155

< previous page page_155 next page >

Page 155
One specimen was subjected to alternate loading, see Figure 5. It will be seen that there is evidence of energy
dissipation, and also, at higher loads, of permanent baseplate deformation such that reversal from positive rotation to
negative rotation can occur under (close to) zero applied moment. This indicates that for bases with sufficiently thin
baseplates the connection will indeed behave as a “pin”, provided sufficiently high previous loading has taken place to
deform the baseplate.
MATHEMATICAL MODELLING
Because it is not feasible to test all possible design combinations for column-base connections, it is appropriate to
develop theoretical relationships between applied moment M, axial force P and connection rotation θ. A previous
attempt [8] to do so for beam-column connections indicated that a relationship between moment capacity and connection
rotation of the form:

(1)
might be appropriate. Here Mp is the plastic moment capacity; as is the initial (elastic) stiffness of
the connection; as is the strain hardening stiffness of the connection and c is a (dimensional)
“constant” found by fitting to the data. It governs the rate of transition from linear elastic behaviour to strain hardening.
Ki may be found from a consideration of the elastic deformation of all components of the connection, in a manner
similar to that described in [8]. In the present case, elastic deformation of the baseplate due to the bolt forces as well as
that due to the outstand of the baseplate on the compression side was modelled using a finite difference program.
Kp was estimated by substituting the strain hardening modulus Est for E in the calculations for Ki relevant for the
baseplate. This method of approach is a well-known technique. Deformation of the bolts was not considered in this way
because bolts exhibit very little, if any, strain hardening. Furthermore, strain hardening was found to be associated with
the plate dominated failure modes (see Figure 3).
Mp was obtained in the usual way from ideal-plastic theory. Yield stresses obtained from samples rather than nominal
values were used.
The results obtained appear to reasonably correspond with the
< previous page page_155 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_155.html07/11/2009 12:07:45 PM
 page_156

< previous page page_156 next page >

Page 156
experimentally obtained values, at least for cases dominated by plate behaviour. For such cases,
. Rather less satisfactory is the correspondence between theory and experiment for
intermediate and for bolt-dominated cases. Full details have been given elsewhere [10] but Figure 6 indicates some
typical results. It is clear that the value of the “constant’’ c appears to depend on the relative contributions of the
baseplate and the bolts to the connection strength.
PRACTICAL IMPLICATIONS
Nominally “pinned” base connections of the type considered herein clearly do possess moment capacity. This remains
the case provided the connection is not deformed beyond elastic limits, which would appear quite likely to be the case
for much of the life of a typical structure. At greater deformation, baseplate deformation may develop, depending on the
thickness of the plate. Thin plates will deform at lower loads and hence will more easily allow the connection to conform
to its assumed behaviour mode. This is illustrated by the cyclic test result of Figure 5.
It is possible, with a sufficiently thick baseplate and high capacity bolts to attain the yield point in the column
compression flange, indicating that in some cases, quite high moment capacity and high stiffness is possible. The
implications for deflection (and also stability) analysis are obvious. What is not so clear is the effect this might have on
existing foundations, and whether this matters.

Figure 6. Correspondence Between Test Results and Predicted M-θ Relationship for Tests 1–9
< previous page page_156 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_156.html07/11/2009 12:07:46 PM
 page_157

< previous page page_157 next page >

Page 157
CONCLUSIONS
An overview has been given of test results on nominally “pinned” base connections and a preliminary attempt at their
mathematical modelling. It was shown that such connections always have strength and stiffness in bending, and, in cases
of thick baseplates and adequate bolts, these factors can be significant. However, deformation of the baseplate at higher
loads will render the “pinned” design assumption true for subsequent relatively low connection rotation. The present
work did not address the possible influence of the behaviour of the soil-foundation interaction, a matter of considerable
importance.
ACKNOWLEDGEMENTS
The author is indebted to Mr. K.K.Hon, Graduate Student, who performed the tests and carried out many of the
calculations.
REFERENCES
1. Bijlaard, F.S.K., Rekenregels voor het ontwerpen van kolomvoetplaten en experimentele verificatie, Report No. BI-
81–51 TNO-IBBC, Delft, 1982.
2. Salmon, C.G., Schenker, L. and Johnston, B.G., Moment-rotation characteristics of column anchorages, J. Struct. Div.,
A.S.C.E., 1955, 81, April.
3. Picard, A. and Beaulieu, D., Behaviour of a simple column-base connection, Can. J. of Civil Engrg., 1984, 12, pp.
125–136.
4. Galambos, T.V., Influence of partial base fixity on frame stability, J. Struct. Div., A.S.C.E., 1960, 86, ST5, pp. 85–
108.
5. Colson, A., Galea, Y. and Pilvin, P., Structures à barres et modélisation du comportement des assemblages, Final
Report, 12th Congress I.A.B.S.E., Vancouver, 1984, pp. 661–667.
6. Jones, S.W., Kirby, P.A. and Nethercot, D.A., The analysis of frames with semi-rigid connections, A State of the Art
Report, Department of Civil and Structural Engineering, University of Sheffield, U.K., 1979.
7. Frye, M.J. and Morris, G.A., Analysis of flexibly connected steel frames, Can. J. Civil Engrg., 1975, 2, pp. 280–291.
8. Yee, Y.L. and Melchers, R.E., Moment-rotation curves for bolted connections, J. Struct. Engrg., 1986, 112, pp. 615–
635.
9. Hon, K.K. and Melchers, R.E., Experimental behaviour of steel column bases, J. Const. Steel Res., (submitted).
10. Hon, K.K. and Melchers, R.E., Moment-rotation curves for “pinned’’ column bases, (submitted for publication).
< previous page page_157 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_157.html07/11/2009 12:07:47 PM
 page_158

< previous page page_158 next page >

Page 158
ON THE NONLINEAR BEHAVIOR OF JOINTS IN STEEL FRAMES
Ferdinand Tschemmernegg
o.Univ.Prof.Dipl.-Ing. Dr. techn.
Head of Institute of Steel and Timber Construction
University of Innsbruck
ABSTRACT
The paper introduces a method to determine analytically the nonlinear moment-rotation-relationship for joints in
structural steel frames. This method was developed on the basis of extensive series performed at the Institute of Steel and
Timber Construction, University Innsbruck, Austria. The field of application includes welded joints as well as bolted
joints.
INTRODUCTION
At the design of structural steel frames problems are mainly encountered during the design of joints, as the elastic-
plastic behavior of the joints is obviously not sufficiently clarified. Furthermore, the influence of joint deflection on the
overall behavior of the frame cannot be neglected, especially in the design of unbraced frames.
This paper reports on the results of tests performed on joints and on a method established on the basis of these test
results which makes it possible to analytically determine the moment-rotation-relationship for structural steel joints.
MATERIALS AND METHODS
Fig. 1 shows a typical welded joint (fig 1a) as well as a typical bolted joint (fig 1b).
< previous page page_158 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_158.html07/11/2009 12:07:48 PM
 page_159

< previous page page_159 next page >

Page 159

Figure 1
Basically there has to be made a distinction between the flexibility of the joint itself and the flexibility of the connection
between joint and the adjacent beam. This distinction together with a macroscopic view of the joint is decisive for the
following approach. By this way, it is possible to investigate the behavior of the very complex mechanism taking place
inside the joint by separating the influence of joint and connection. In order to analyze the behavior of joint and
connections, several different test series were undertaken, as summarized in fig. 2.
Test series 1 served to isolate the forces introduced by the beam flange into the joint and to study this effect separately.
Test series 2, performed on symmetrically loaded joints, allotted for studying the additional effect of beam web on the
load introduction into the joint.
Test series 3—representing an unsymmetrically loaded joint—was performed to study the combined effect of load
introduction and shear acting on the joint.
Test series 4—basically done in the same configuration as series 3—was undertaken to deliver information on the
additional influence on joint behavior of axial forces acting in the column.
< previous page page_159 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_159.html07/11/2009 12:07:49 PM
page_160

< previous page page_160 next page >

Page 160

Figure 2
< previous page page_160 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_160.html07/11/2009 12:07:51 PM
 page_161

< previous page page_161 next page >

Page 161
Refering to measuring methods several publications [1], [2], [3] were issued. In fig. 2 the test specimen and the
measuring equipment are drawn schematically.

Figure 3
The target of the research programm was to analyze separately the elastic-plastic moment-rotation-curves for load
introduction and shear. As represented in fig. 3 the moment-rotation-curve can be determined by the following 4
characteristic values;
– elastic limit moment
– plastic limit moment
– elastic limit of rotation
– plastic limit of rotation
RESULTS
Basing on the measurements and a systematic evaluation, the Joint was modelled in a step-by-step method with the result
that the elastic-plastic behavior of the joint can be described in a pertinent way.
< previous page page_161 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_161.html07/11/2009 12:07:52 PM
page_162

< previous page page_162 next page >

Page 162

Figure 4
< previous page page_162 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_162.html07/11/2009 12:07:53 PM
 page_163

< previous page page_163 next page >

Page 163
Fig. 4 a, b shows a symmetrical and unsymmetrical joint at the plastic limit.
By this way the general spring model given in fig. 5 was developed incorporating the following springs:
– shear spring,
– load introduction springs,
– connection springs.

Figure 5
These springs are assembled together in series.
< previous page page_163 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_163.html07/11/2009 12:07:55 PM
page_164

< previous page page_164 next page >

Page 164

Figure 6
< previous page page_164 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_164.html07/11/2009 12:07:57 PM
 page_165

< previous page page_165 next page >

Page 165
Load Introduction and connection spring are acting in series on the applied moments MA and respectively, while the
shear spring only acts on the difference of applied moments (fig 5).
The case shown in fig, 4 ist the welded connections in this case the springs for the connection can be assumed infinitely
rigid. For other connection types moment-rotation behavior could without problems be represented by an elastic-plastic
spring.
The overall bearing capacity of the joint is determined by the capacity of the weakest spring.
The results of the tests performed made it possible to supply a method for calculating analytically the characteristic
values of the springs for any combination of beams or columns out of the range of European profiles. The moment-
rotation-relationship can be evaluated separately for load introduction+shear. The result of such a calculation is shown
in fig. 6.
The method for calculation is applicable to unstiffened joints as well as stiffened joints.
DISCUSSION
The moment-rotation curves being determined, calculation of frames can be performed now in a simple way with
consideration of elastic-plastic behavior of the joints. This method of frame design is described in detail in [4].
One of the conclusions is that the flexibility of joints is of great influence especially in the design of unbraced frames.
REFERENCES
1. KLEIN,H.: Das elastisch-plastische Last-Verformungsverhalten M- steifenloser, geschweißter Knoten für die
Berechnung von Stahlrahmen mit HEB-Stützen. Diss.Univ. Innsbruck, 1985.
2. HUMER,Ch.: Das Momenten-Rotationsverhalten von steifenlosen Rahmenknoten mit Kopfplattenanschlüssen. Diss.
Univ. Innsbruck, 1987.
3. BRAUN,Ch.: Das Momenten-Rotationsverhalten von steifenlos ausgeführten Rahmenknoten bei Schubbeanspruchung
im überkritischen Bereich. Diss.Univ. Innsbruck, 1987.
4. SZS: Rahmentragwerke in Stahl unterer besonderer Berücksichtigung der steifenlosen Bauweise. Structural Steel
Association Switzerland, to be published in 1987.
< previous page page_165 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_165.html07/11/2009 12:07:58 PM
 page_166

< previous page page_166 next page >

Page 166
SEMI-RIGID CONNECTIONS IN FRAMES, TRUSSES AND GRIDS
Dr.-Ing. Ch. Stutzki, Dr.-Ing. J.Lopetegui
and
Prof. Dr.-Ing. G.Sedlacek
Lehrstuhl fuer Stahlbau, RWTH Aachen, W.-Germany
SUMMARY
Several models for the analysis of non-linear connections are presented. The models are suitable to describe the
nonlinear behaviour of connections according to the problems to be solved. These models are embedded in a general
computer program for the analysis of space frames with geometrical and physical nonlinearity. Three examples are
presented in which the connection behaviour is of great influence on the global structure.
BEAM TO COLUMN CONNECTIONS IN BRACED FRAMES
The characteristic property of a connection in a frame is its moment-rotation behaviour. For some commonly used
connections the -curves are given in fig. 1 in a qualitative way. Not only the linear flexibility, but the whole range of
the -curve is of importance, especially, if an ultimate limit analysis for the frame is performed.
Two influences are neglected in the commonly used static analysis:
1) The size of the connection. Usually the adjacent members are connected to an idealized node which is the point of
intersection of the axes of the members.
2) The flexible, nonlinear -characteristic.
With a simple model both influences are taken into account, (fig.2). The truss members 1 and 2 in fig.2b are assumed to
be rigid, whereas member 3 works as a control member with an axial- force/elongation- behaviour that is controlled by a
given curve (fig.2c). If the axial force in member 3 is N then the theoretical moment in joint is M=N · c, and the rotation
of the joint is
The controlling curve for member 3 is defined as a polygon.
< previous page page_166 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_166.html07/11/2009 12:07:59 PM
 page_167

< previous page page_167 next page >

Page 167

fig. 1: Typical Connections and their moment- rotation-behaviour

The values of the polygon are given as input-data to the program, and may be derived from previous calculations for the
joint behaviour or from tests. The model can be inserted into every plane or spatial framework.

fig. 2: truss model for a beam to column- connection

< previous page page_167 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_167.html07/11/2009 12:08:00 PM
 page_168

< previous page page_168 next page >

Page 168
BEAM-TO-COLUMN CONNECTIONS IN SWAY FRAMES
The connections of sway frames are exposed to large moments due to horizontal forces. In multi-bay frames this leads to
antimetric action of moments from the beams to the inner columns (fig. 3). The main reason for the deformation of the
joint is the shear deformation of the joint panel (fig. 4). For this case a second model has been developed. This model
consists of a truss subassemblage which transfers the differential moment of the beams into the column (fig. 5).

fig. 3: multi-bay sway frame under horizontal load

fig. 4: joint deformation due to shear forces

fig. 5: truss model for a shear connection

< previous page page_168 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_168.html07/11/2009 12:08:03 PM
 page_169

< previous page page_169 next page >

Page 169
Also in this model the axial- force/elongation- behaviour of the control element is controlled by a diagram. The diagram
is described as a polygon, and its values may be taken from test results or are derived from a previous calculation for the
shear panel. The differential moment in the joint is M=h · N where N is the axial force in the control element, and the
angle of the shear deformation is .
The models for the symmetrical loading and for antimetrical loading can easily be combined, (fig. 6) with two control
elements for symmetric moment, according to fig. 2, and one control elements for the differential moment according to
fig. 5.

fig. 6: combined model for symmetric and antimetric moments

EXAMPLE 1
Many calculations and comparisons with other methods and test results have proved that the model is able to simulate
the joint behaviour and its influence on the total structure with good accuracy. It is embedded in a computer program for
space frames with geometrical and physical nonlinear beam elements. The analysis is performed incrementally with an
equilibrium iteration in each step. Only the initial stiffness matrix is used, according to the iteration method of
orthogonal load-deformation-states /2/. It is not necessary to form a tangent stiffness, neither for the beam elements nor
for the joint stiffness. As an example for the application a small study is presented on the influence of different
connection types on the ultimate limit state of a plane frame (fig. 7).
< previous page page_169 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_169.html07/11/2009 12:08:04 PM
 page_170

< previous page page_170 next page >

Page 170

fig. 7: comparison of different ultimate limit loads for a sway frame with different types of connections
BOLTED CONNECTIONS IN TRUSS MEMBERS OF TRANSMISSION TOWERS
Investigating the consequences of irregular foundation settlement of transmission towers, one ends up with axial forces
in some truss diagonals which are much higher than the failure load of these diagonals. This is the result, if a linear
analysis is performed with one of the commonly used computer programmes. However failure has not been observed,
even if settlements of two or some more centimeters occur. The ‘intelligence’ of the real structure and some second order
< previous page page_170 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_170.html07/11/2009 12:08:05 PM
 page_171

< previous page page_171 next page >

Page 171
effects provide some additional safety which is not taken into account in a linear static analysis.
A truss model, taking into account bolt slipping (fig. 8), has been used to investigate the influence of this connection
behaviour on the distribution of internal forces, if in addition to the regular loading, foundation settlement occurs. The
behaviour of the truss element is linear, until the value of friction RH is reached, then slipping may take place, until the
possible sliding distance, due to tolerances between bolt and hole, is reached. From then onwards, linear behaviour is
assumed again. A complete tower (fig. 9 shows the lower part of it) has been analysed, using the described model, with
varying values for friction and sliding distance.

fig. 8: truss model with bolt slipping at the joints

fig. 9
< previous page page_171 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_171.html07/11/2009 12:08:06 PM
 page_172

< previous page page_172 next page >

Page 172

fig. 10
In fig. 10 the axial forces of three diagonal members are plotted against the foundation settlement. The dashed lines are
the result of a linear analysis, the full lines are the result of a nonlinear analysis, using mean values of bolt slipping of 1
mm. This result is only capable of showing a tendency in the distribution of internal forces, not the real forces, because
the exact occurance of bolt slipping and friction in each joint is unknown. However it can be seen clearly that the forces
that are calculated with bolt slipping are much lower than the linearily calculated ones, which gives a good interpretation
of the observed effect (that no failure occurs).

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_172.html (1 of 2)07/11/2009 12:08:07 PM

page_172

< previous page page_172 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_172.html (2 of 2)07/11/2009 12:08:07 PM

 page_173

< previous page page_173 next page >

Page 173
COUPLINGS IN A PONTOON BRIDGE
Pontoon bridges consist of floatable, quasi-rigid boxes of steel that can be connected to each other quickly and without
much effort. This means a kind of construction that is simple to handle even in rough sea conditions. For these reasons
the connection mechanism has large tolerances in order to achieve a quick assemblage. The static analysis of this type of
structure is complicated contrary to the simplicity of its construction, because the deformation behaviour of the couplings
determines the global behaviour of the bridge during the crossing of a vehicle.
The connection behaviour is represented in fig. 11. The moment-rotation behaviour is bilinear: for small relative
rotations of two pontoons the connection behaves like a hinge, and for large rotations the connection is rigid. A simple
model has been developed to simulate this type of coupling (fig. 12). The shear force can be transmitted by two rigid
members, connected by a hinge. The bending moment is transmitted by a control member which behaves according to
the diagram in fig. 11b, depending on the relative rotation

fig. 11: coupling of pontoons

fig. 12: model for the coupling

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_173.html (1 of 2)07/11/2009 12:08:08 PM

 page_173

fig. 13
parameter controlled gap
< previous page page_173 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_173.html (2 of 2)07/11/2009 12:08:08 PM

 page_174

< previous page page_174 next page >

Page 174
The equilibrium iteration for the complete structure is very difficult, especially for low loading, because any foresight of
the deformations may lead to completely wrong results if they are extrapolated from a position with hinges to a final
position where the connections are rigid. In order to overcome this difficulty in the analysis, the incrementation is
performed not in the load, but in the tolerance of the couplings, with a stepwise increase of the factor α (fig.13)
from 0 to 1. With this kind of procedure calculations with very large gaps are possible.
NONLINEAR PROCEDURE
The actual behaviour of a nonlinear system is not described by a tangent stiffness matrix but by so called load-
deformation states /4/. A load-deformation-state consists of a deformation increment vi and of the difference of the
reaction forces before and after the application of the incrementl deformation vi, that is: (vo
is the total deformation of a previous step). The reaction forces can easily be determined applying the previously
described constitutional laws with the deformation as argument.
The complete nonlinear behaviour of the system is described by a set of orthogonal load- deformation- states which are
generated in the course of the iteration. They form a kind of a subspace and have to be free of contradictions in order to
simulate the tangent stiffness. Therefore the iteration is stopped and started again if the system is changing its
characteristics suddenly, because otherwise contradictory load- deformation- states would result.
REFERENCES
/1/ Lopetegui, J., Saleh, A., Stutzki, Ch., Sedlacek, G., Programme for the ultimate limit state analysis of space frames.
User Manual. Report of the Institute for Steel Structures, University of Aachen, 1987.
/2/ Lopetegui, J., Procedure for the solution of nonlinear problems using load—deflection—states. Doctoral dissertation,
University of Aachen, 1983
/3/ Saleh, A., Ultimate limit state analysis of frames with large deflections and plasticity. Doctoral dissertation,
University of Aachen, 1982
/4/ Sedlacek, G., Lopetegui, J., Saleh. A., Stutzki. Ch., Computer-orientated methods for the calculation of space frames
with nonlinear effects.
BAUINGENIEUR, Vol. 60, 1985, pp. 297-305
< previous page page_174 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_174.html07/11/2009 12:08:09 PM
 page_175

< previous page page_175 next page >

Page 175
FLEXIBILITY FORMULAE AND MODELLING OF JOINT BEHAVIOUR IN GIRDERS MADE OF
RECTANGULAR HOLLOW SECTIONS
A.Czechowski, J.Kordjak, J.Bródka
Metal Structures Research
and Design Centre ‘‘Mostostal”
Krucza 16/22, 00–926 Warsaw, Poland
ABSTRACT
This paper deals with the flexibility problem of stepped (unequal width) joints in RHS girders. First of all modes of
flexibility for different joints are defined with relevant elastic formulae being established. Then it is shown how those
linear characteristics can be extended into non-linear range of joint behaviour. Some comparison with test results is also
given. Having known (calculated) joint flexibility one can take it into account in the structural analysis by using so-
called micro-bar model of a joint.
INTRODUCTION
In last 15 years a large amount of research, stimulated and co-ordinated mainly by CIDECT and IIW, has been done in
the domain of RHS joints [1]. Also in Poland, in “MOSTOSTAL” Centre, an extensive research programme has been
carried out, including tests on isolated joints [2–4], as well as on complete trusses [6] and, recently, Vierendeel girders.
Since the joint strength itself is not the only aspect of the structural behaviour, some part of investigations has been
concerned with the problem of joint flexibility and adequate calculation model. It should be mentioned here, that
relatively small number of papers exist in which this particular problem is undertaken, although its implications are
practically
< previous page page_175 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_175.html07/11/2009 12:08:10 PM
 page_176

< previous page page_176 next page >

Page 176
important and in general must not be ignored in design.
The subject of further considerations are unreinforced truss and moment joints (Fig. 1) consisting of RHS members
directly welded together, in which bracing—to—chord for beam -to-column) width ratio is equal or less than 0.8. Such

Figure 1. RHS joints—geometrical parameters

joints have to be treated as flexible ones, otherwise than rigid or semi—rigid connections in conventional steel/work
systems.
ELASTIC FLEXIBILITY FORMULAE
In order to make the term “flexibility” more precisely comprehended, regarding actual deformations of joints, three basic
modes of flexibility are specified (see Fig. 2):
A —translational (typical for axially loaded T and X-joints),
B —antitranslational; push—pull mechanism (typical for truss joints with relatively big gap between bracings),
< previous page page_176 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_176.html07/11/2009 12:08:11 PM
 page_177

< previous page page_177 next page >

Page 177

Figure 2. Modes of flexibility

C —rotational (typical for moment T and X—joints, as well as for overlap truss joints).
The above mentioned flexibility cases have then been solved analytically. The analytical approach accepted for this
purpose can be characterized, as follows:
a) Chord face is considered first as to be a simply supported strip plate, loaded by one or two rigid hollow punches
(bracing members), acting translationally or rotationally respectively;
b) Actual continous plate—punch interaction is replaced by a discrete one, existing in a definite number of points
(supports) along the punch perimeter (see Fig. 2a);
c) Using Green’s function for a strip plate deflection problem a system of linear equations is established, where
unknowns are reactions in points of contact. This procedure is to be repeated for each new set of contact points
coordinates;
d) Solution of the above mentioned equations enables a characteristic deflection due to a unit load to be determined;
e) Since the chord face is not simply supported, but is
< previous page page_177 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_177.html07/11/2009 12:08:14 PM
 page_178

< previous page page_178 next page >

Page 178
TABLE 1
Elastic flexibility formulae for RHS joints

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_178.html (1 of 2)07/11/2009 12:08:16 PM

page_178

< previous page page_178 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_178.html (2 of 2)07/11/2009 12:08:16 PM

 page_179

< previous page page_179 next page >

Page 179
restrained at the edges by adjacent walls, a frame analogy has been used to take this effect into account by introducing a
correction factor.
After having obtained the solutions for different joint parameters approximate flexibility functions of those
dimensionless parameters were established. Final elastic flexibility formulae for truss and moment joints are presented in
Table 1. There is also included a translational—rotational flexibility typical for gap truss joints. This formula consisting
of two flexibility components is based on the assumption, that inner deflection is proportional to antitranslational
flexibility, whereas the outer one is proportional to rotational flexibility of a K—joint treated as a T—moment joint
. Thus both flexibility components can be consequently determined.
NON—LINEAR BEHAVIOUR
This problem is closely associated with elastic and ultimate strength, thus with so—called plastic reserve, which being a
difference between those two limits, is a specific measure of joint flexibility. In this sense rigid connections exhibit
relatively low plastic reserve, whereas joints of low stiffness show a remarkable post—yield behaviour. Many different
geometrical, material and technological factors make the problem of joint behaviour including strength very complex
and extremely difficult to be solved analytically, although some attempts using finite element method are known [5].
Nevertheless even such an apparently accurate approach need to be experimentaly verified.
In this paper a semi—empirical approach is proposed. Elastic (linear) load—deformation relationship can be extended
into non—linear range of behaviour provided that the ultimate strength and the “yield—line” load capacity of a joint are
known. Non—linear secant stiffness formula is given in Fig. 3 and compared with experimentally observed T—joint
behaviour in Vierendeel girder. A fairly good correlation is obtained for exponent n=2. For other joints exponent n may
be
< previous page page_179 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_179.html07/11/2009 12:08:19 PM
 page_180

< previous page page_180 next page >

Page 180

Figure 3. relationship for a moment joint. Non—linear stiffness formula compared with experimental data
different.
MICRO—BAR MODEL OF A JOINT
After having determined joint flexibility the problem is how to take it into account in statical analysis of a whole
structure. For this purpose a micro—bar model of a joint has been devised. The idea of this model is illustrated using a
general example of a gap truss joint with translational—rotational flexibility. Respective components of flexibility are
taken into account by adequate reducing cross—section area of bracing members (see Fig. 4) and by introducing a
system of micro—bars which is “responsible” for rotation only (see Fig. 5).
In contradistinction to conventional systems with pin-ended or rigidly connected members, that one involving micro-bar
models of flexible joints may be called “second order sy-
< previous page page_180 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_180.html07/11/2009 12:08:20 PM
 page_181

< previous page page_181 next page >

Page 181

Figure 4. Equivalent cross—section area of a bracing member reduced due to translational flexibility of Joints

Figure 5. Micro—bar model of a gap truss joint with translational—rotational flexibility

< previous page page_181 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_181.html07/11/2009 12:08:21 PM
 page_182

< previous page page_182 next page >

Page 182
System’’.
A comparison with test results has indicated [6], that second order system is an adequate calculation model for lattice
girders, particularly for statically indeterminant structures.
CONCLUDING REMARKS
Flexibility formulae and calculation model presented in this paper can be directly used in the analysis and design of RHS
girders with flexible joints. This approach enables some specific aspects of the structural behaviour, concerning
deflections, secondary bending moments, stability of bracing members, to be examined and realistically evaluated.
REFERENCES
1. Wardenier, J., Hollow section joints, Delft University Press, Delft, 1982.
2. Czechowski, A. and Bródka, J., Etxide de la résistance statique des assemblages soudés en croix de profilés
rectangulaires, Construction Metallique, no 3, 1977.
3. Zycinski, J., Czechowski, A. and Bródka, J., Static strength of welded K—type joints in rectangular hollow sections,
Archiwum Inzynierii Ladowej. vol. XXXII, no 3, 1986 (in polish).
4. Szlendak, J. and Bródka, J., Yield and buckling strength of T, Y and X joints in rectangular hollow section trusses,
Proc. Instn Civ. Engrs. Part 2. 1985, vol. 79, Mar., pp. 167–180.
5. Korol, R.M. and Mirza, F.A., Finite element analysis of RHS T—joints, J. Struc. Div., ASCE, vol. 108, No ST 9,
Sept., 1982.
6. Czechowski, A., Investigation into the statics and strength of RHS lattice girders, In Welding of Tubular Structures,
Proceedings of the Second International Conference held in Boston, Massachusetts, USA, 16–17 July 1984, Pergamon
Press, 1984.
< previous page page_182 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_182.html07/11/2009 12:08:22 PM
 page_183

< previous page page_183 next page >

Page 183
DISCUSSION OF MATHEMATICAL MODELING PAPERS
The primary thrust of the papers of this session was that complete (=global) connection modeling for behaviour and
strength may be the preferable method, in the long run. Although it presents numerous difficulties, as pointed out by
several participants, especially in the numerical treatment of elasto-plastic phenomena that also may involve strain-
hardening, it will clearly make if easier to apply semi-rigid concepts to frame analyses if the complete moment-rotation
curves for the connections are specified.
Several discussers addressed the problems that are associated with contact problems of the kind that commonly occur in
column footing and end-plate connections. Whether special finite element types were used for this (for example, with
zero stiffness after contact is lost, but regaining it if cyclic loads force a re-contact), or non-linearly elastic springs were
utilized, the end result appeared to be good correlation between tests and theory. These techniques were also applied to
reflect prying action for bolts, as well as friction and the loss of same for slip-resistant connections.
Although it was agreed that it would be preferable to utilize overall connection response characteristics in the frame
response analysis, it would still be necessary to determine the primary response parameters, in the fashion as indicated
by the local analysis session. Specifically, initial stiffness and ultimate strength data must be determined very carefully.
The effects of cyclic loading were discussed by a number of contributors. The problems are particularly complex if full
load reversal occurs, both in terms of the needs of the modeling scheme as well as those of the eventual application to
frame analysis itself.
Several questions were raised regarding the evaluation of column footings and their performance. Thus, structural details
such as base plate to concrete foundation separation were considered, as were anchor bolts, grout and the like. However,
there appeared to be general agreement that this is one area of connection research that is in need of significant
additional studies.
Many of the discussers emphasized the need for concentrating on
< previous page page_183 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_183.html07/11/2009 12:08:23 PM
 page_184

< previous page page_184 next page >

Page 184
connections that are commonly used in industry. For example, North American practice for many types of connections
now involves the use of slotted holes, to facilitate the erection work. It is clear that such will have a significant impact on
the moment-rotation characteristics, at least following the first load reversal or after bolt slip has taken place.
Overall, the many global connection models that have been developed appear to hold excellent promise for more general
applications to connection analysis. The need for being able to assess the importance of the various physical details was
stressed by all presenters as well as discussers.
< previous page page_184 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_184.html07/11/2009 12:08:24 PM
page_185

< previous page page_185 next page >

Page 185
SESSION No 2B
CLASSIFICATION
< previous page page_185 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_185.html07/11/2009 12:08:25 PM
 page_186

< previous page page_186 next page >

Page 186
STRUCTURAL PROPERTIES OF CONNECTIONS IN STEEL FRAMES
J.W.B.Stark
Eindhoven University of Technology,
Den Dolech 2, 5612 AZ Eindhoven,
The Netherlands
TNO Institute for Building Materials and Structures
P.O. Box 49, 2600 AA Delft
The Netherlands
and
F.S.K.Bijlaard
TNO Institute for Building Materials and Structures
P.O. Box 49, 2600 AA Delft
The Netherlands
ABSTRACT
Modern design codes allow the design of structures to be based on the actual load deformation characteristic of
connections. In this paper a classification of connections is proposed. Essential is that the classification depends on the
design method of the connection e.g. elastic method versus plastic method. It is necessary that designers are able to
determine the characteristics of connections. Brief information is given on available connection test data. These
data are essential to develop analytical models to be used in design calculations. Two possible approaches for the
determination of such models are discussed.
INTRODUCTION
In structural steelwork the connections between members play an important role. From an economic point of view the
costs for design and fabrication form a considerable part of the total costs. From a structural point of view the properties
of the connections fundamentally influence the response of the structure to actions. In the past, when ‘working stress’
design was normally used, the connection design was based on rather simple though not necessarily economical
assumptions.
The connections were assumed to behave either as hinges (simple construction) or as infinitely stiff (rigid construction).
The forces on the connections then followed from an elastic analysis of the structure.
< previous page page_186 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_186.html07/11/2009 12:08:26 PM
 page_187

< previous page page_187 next page >

Page 187
The parts of the connections such as end plates and angles, welds and bolts, could subsequently be dimensioned. Even
now this design procedure seems to be used in the majority of cases. However, the introduction of limit state design,
including practical rules for plastic design, requires a more realistic treatment of the connections. When using these
methods, the designer is confronted directly with the fact that for a better insight into topics such as the stability of
columns and frames and for a minimum cost design of members and connections, understanding of the behaviour of
connections is essential.
Another factor is that modern computer programs, now available to the majority of designers, allow a more sophisticated
treatment of connections without an appreciable increase in calculation costs. Finally, also the use of automatic NC
drilling and sawing equipment in the fabricators shop influences the cost relationship between various forms of
connections, leading to a need to minimise the number of welded stiffeners.
Modern design codes e.g. Draft Eurocode 3, allow the design of the structure to be based on the actual load-deformation
characteristics of the connections. It is of course essential that designers are able to determine those characteristics,
preferably in the form of a mathematical model. The available information on this point will be reviewed in an IABSE-
report that is under preparation now [1]. It is drafted in liaison with ECCS-THG 8.2 and should be considered as
complementary to [2] dealing with analysis and design of steel frames.
This paper, to be presented at the “CACHAN-WORKSHOP” is an introduction to [1] and is aimed to collect the latest
information.
DEFINITION OF PROPERTIES OF JOINTS
A steel frame, whether considered as a plane or spatial structural system, essentially consists of linear members joined
together by connections (Fig. 1).

Figure 1. Combination of linear members and connections in a plane structural system.

< previous page page_187 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_187.html07/11/2009 12:08:27 PM
 page_188

< previous page page_188 next page >

Page 188
The response of the system to actions (loads) is in general influenced by the structural properties of both members and
connections. The relevant properties of these elements are strength, stiffness and deformation capacity (ductility).
Assuming for the present that bending is dominant, all three parameters can be presented in a moment-rotation
(curvature) diagram of the type illustrated qualitatively in Fig. 2.

Figure 2. Presentation of structural parameter in a curve.

The properties of a member with parts in compression are dependent upon section geometry and the yield stress of the
steel mainly due to the effect of local buckling.
In most codes the different types of cross-section are classified dependent on their respective moment-rotation
properties. This procedure is used because no practical methods are available to determine the actual relation for
each individual member which can be used in a simplified design.
Depending on the class of the cross-section (form of relation: see Fig. 3) different analysis models may be used.

Figure 3. Load-deflection relationships.

< previous page page_188 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_188.html07/11/2009 12:08:29 PM
 page_189

< previous page page_189 next page >

Page 189
As an example the possibilities for ultimate limit state verification according to the Draft Eurocode 3 are given in Table
1.
TABLE 1
Class of member Capacity of cross-section Analysis of system
Class 1 plastic plastic
Class 2 plastic elastic
Class 3 elastic elastic
Class 4 reduced stress or elastic
effective section
The structural properties of connections can also be presented in a diagram. In Fig. 4 a set of curves for
connections with different types of behaviour is shown.

Figure 4. Moment-rotation curves of beam-to-column connections.

For the use of plastic design the connections can be classified in the following categories:
Nominally Pinned Connections
This type of connections is designed to transfer shear and normal force only. The rotation capacity of the hinge should be
sufficient to enable all the plastic hinges necessary for the collapse mechanism to develop.
< previous page page_189 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_189.html07/11/2009 12:08:30 PM
 page_190

< previous page page_190 next page >

Page 190
Full Strength Connections (connections A and B in Fig. 4)
The moment capacity is greater than that of the member. A plastic hinge will not be formed in the connection but in the
member adjacent to the connection. In theory no rotation capacity is required for the connection. If the connection has
little deformation capacity (connection B) an extra reserve of strength should be required to account for possible
overstrength effects in the member.
Partial Strength Connections (connections C, D and E in Fig. 4))
The moment capacity is less than that of the member. A plastic hinge will be formed in the connection, so sufficient
rotation capacity is required. For this reason connection C is unsuitable. Curve E is typical for a bolted bearing type
connection, showing slip due to the clearance of the holes. For a detailed treatment of the requirements, reference is
made to [3]. This explains how, especially when using partial strength connections in plastic design, information on all
three parameters of the relation is required.
In elastic design traditionally two categories of connections were considered:
Nominally Pinned Connections (Fig. 5)
The connections are assumed to transfer only the end reaction of the beam (vertical shear force and eventually normal
force) to the column.

Figure 5. Nominally pinned connection.

They should be capable of accepting the resulting rotation without developing significant moments, which might
adversely affect the stability of the column. In many countries it is common practice to design structures on a simply
supported basis and then to provide connections which are in effect semi-rigid. A typical example is the flush end plate
detail shown in Fig. 6. This may be unsafe due to insufficient rotation capacity of the connection.
< previous page page_190 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_190.html07/11/2009 12:08:31 PM
 page_191

< previous page page_191 next page >

Page 191

Figure 6. Flush end plate connection.

Rigid connections (Fig. 7)
Rigid connections are used to transfer moments as well as end reactions. Design assumes joint deformation to be
sufficiently small (stiffness large) that any influence on the moment distribution and the structure’s deformation may be
neglected.

Figure 7. Rigid connections

Practical realisation of the assumptions of rigid construction often leads to relatively expensive connection details. Thus
in cases where stiffness is not required for stability reasons (e.g. in braced frames), the use of nominally pinned
connections is usual. However, practical forms of this type will transfer some moments and contribute to the structural
stiffness. On the other hand, rigid connections can be simplified and made less expensive by the omission of stiffeners
and other parts and accepting some flexibility. To fill the gap between pinned and rigid connections, a third category is
defined and accepted in most modern codes.
Semi Rigid (semi flexible) Connections (Fig. 8)
These connections are designed to provide a predictable degree of interaction between members based on actual or
standardized design characteristics of the joints.
< previous page page_191 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_191.html07/11/2009 12:08:32 PM
 page_192

< previous page page_192 next page >

Page 192

Figure 8. Semi rigid (semi flexible) connections.

In conclusion it seems to be evident that there is a need to present information the on relations of joints in a form
which can be readily used by designers.
CONNECTION TEST DATA
Design models for the relation of connections can only be based on a systematic evaluation of test information.
Recently prepared survey’s [4, 5, 6] of available test data for connections between beams and columns permit the
existence of suitable information on particular connection types to be checked.
Of much importance is the Steel Connection Data Bank (SCDB) at Purdue University [7]. Digitised versions of over 300
experimentally obtained curves covering 7 different connection types, are stored in the computer. The tests in this
collection are mainly from sources published in the USA, Canada and the UK. On the European continent also much test
information is available, e.g. [8, 9, 10, 11]. These tests are now collected in the framework of the Eurocode-work and
will be stored in a data base at the University of Aachen. Combination and exchange of these two data bases is
recommended.
ANALYTICAL METHODS FOR THE DESCRIPTION OF THE RELATION
Inclusion of connection flexibility into the analysis of the behaviour of members or complete frames requires that
relations for the connections are available to the designer in a suitable mathematical format.
Two essentially different approaches are possible:
i. An analytical model is developed to represent the relation found by tests on a specific connection as closely as
possible. Strictly speaking
< previous page page_192 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_192.html07/11/2009 12:08:34 PM
 page_193

< previous page page_193 next page >

Page 193
for use of this method test results should be available for every connection to be used in the actual structure.

Figure 9. Analytical model to represent the relation [7]

ii. An approximate method based on linearization of the relation. This approach offers the possibility to develop
semi-emperical models for the calculation of the relevant parameters of the schematized relation. An example of
such a linearisation is given in Fig. 10.

Figure 10. Linearised relation

Formulae for the calculation of the moment capacity are already included in Draft Eurocode 3. Formulae for the
determination of the stiffness are also available for a number of connection types and will be included in an appendix of
the revised EC3. For braced frames and floor beams requirements for the rotation capacity can easily be determined. For
those applications a classification method as used for members can be applied. For unbraced
< previous page page_193 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_193.html07/11/2009 12:08:35 PM
 page_194

< previous page page_194 next page >

Page 194
frames the required rotation capacity can not simply be determined in a general way and needs to be calculated in each
individual case. So for use in unbraced frames formulae for the calculation of the minimum available rotation capacity
should also become available. Work is underway to develop these formulae.
REFERENCES
1. Bijlaard, F.S.K., Nethercot, D.A., Stark, J.W.B., Tschemmernegg, F., Zoetemeijer, P., Structural properties of semi-
rigid joints in steel frames. IABSE paper to be published Dec. 1987.
2. Anderson, D., Bijlaard, F.S.K., Nethercot, D.A., Zandonini, R., Analysis and design of steel frames with semi-rigid
connections. IABSE Survey. Prepared by ECCS-TWG 8.2, Task Group on Stability of semi-rigid frames. To be
published June 1987.
3. Witteveen, J., Stark, J.W.B., Bijlaard, F.S.K., Zoetemeijer, P., Welded and bolted beam-to-column connections.
Journal of the Struct. Div., Proceedings of the ASCE, Vol. 108, no St2, Febr. 1982.
4. Nethercot, D.A., Utilisation of experimentally obtained connection data in assessing the performance of steel frames.
Connection Flexibility and Steel Frames, ASCE, ed W.F.Chen, 1985, pp 13–37.
5. Nethercot, D.A., Steel beam to column connections—A review of test data and their application to the evaluation of
joint behaviour on the performance of steel frames. CIRIA report, RP 338, 1985.
6. Goverdham, A.V., A collection of experimental moment-rotation curves evaluation of predicting equations for semi-
rigid connections. Doctoral dissertation, Vanderbilt University, Nashville, Tennessee, 1984.
7. Kishi, N., Chen, W.F., Data base of steel beam-to-column connections, Structural Engineering report no CE-STR-86–
26, School of Civil Engineering, Purdue University, 1986.
8. Zoetemeijer, P., Bolted connections with flush end plates and haunched beams. Tests and limite state design methods.
Report 6–81–17, Delft University of Technology, Stevin Laboratory, June 1981.
9. Zoetemeijer, P., Bolted beam-to-column knee connections with haunched beams. Tests and computations. Report 8–
81–23, Delft University of Technology, Stevin Laboratory, December 1981.
10. Zoetemeijer, P., Proposal for the standardization of extended end plate connection based on test results. Tests and
analysis. Report 6–83–23, Delft University of Technology, Stevin Laboratory, 1983.
11. Zandonini, R., Zanon, P., Semi-rigid and flexible connections: An experimental investigation. Proceedings of the
International Conference on Steel Structures, Budva, Yugoslavia, 28 Sept.-1 Okt. 1986.
< previous page page_194 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_194.html07/11/2009 12:08:36 PM
 page_195

< previous page page_195 next page >

Page 195
SEMI-RIGID CONNECTIONS IN ISOLATION AND IN FRAMES
J B Davison, P A Kirby and D A Nethercot
Department of Civil and Structural Engineering
University of Sheffield
Sheffield S1 3JD
UK
ABSTRACT
The objective of the present Sheffield programme is the development of a full understanding of the behaviour of steel
frames leading to design approaches that recognise the semi-rigid nature of steelwork connections. It has therefore
embraced in-plane testing of connections, restrained columns in subassemblages and complete frames, as well as
associated theoretical studies. As a result of this co-ordinated approach, a number of points have emerged concerning the
conduct of tests, which could not necessarily have been foreseen, nor would they have shown up in a programme of
joint tests only.
INTRODUCTION
A project has been completed at Sheffield University to provide experimental data on the behaviour of columns and
frames restrained by simple beam to column connections. The work has involved testing a series of joints to determine
their in-plane rotational response, a number of restrained columns and two large three-storey two-bay frames. The action
of the joints in the column and frame tests was monitored and compared with the isolated joint tests. Some interesting
points concerning the action of joints in frames and the implications for conducting joint tests have emerged.
< previous page page_195 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_195.html07/11/2009 12:08:37 PM
 page_196

< previous page page_196 next page >

Page 196
JOINT TESTS
All individual joint tests [1] were conducted with the specific objective of providing data for use later in the
programme. A cruciform arrangement was used with direct loading on the internal column. Rotations were measured by
a set of 3 T-bars attached to the specimen on the column centre line and the beam centre lines as near to the connection
as possible—normally within 100 mm. By monitoring the change in length of the wires attached to these bars, it was
possible to determine the relative rotation within each joint. Such an arrangement could not, of course, measure the
contributions to this rotation due to the different sources of flexibility within the joint, merely the overall effect.
The same light column and beam sections featured in all joint tests were subsequently used in the subassemblage and
frame tests. A total of 25 tests, 8 of which included various types of deliberate lack of fit [2], were conducted. Each test
did, of course, provide two curves, one for each side of the specimen and an illustrative result is given as Fig 1.

Figure 1. Moment-rotation curve from flange cleat joint test.

< previous page page_196 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_196.html07/11/2009 12:08:38 PM
 page_197

< previous page page_197 next page >

Page 197
In some cases a limited number of unloading cycles were included in order to provide data on the unloading stiffness
characteristics of the joints. The main variable in the programme was the type of connection—web cleat, flange cleat,
end plate etc—with several of the joints fixed to the column web. Fig 2 summarises the complete set of data in the
form of average curves for each connection type.

Figure 2. Collection of moment-rotation curves for various connections.

CONNECTION BEHAVIOUR IN FRAME AND SUBASSEMBLAGE TESTS
The performance of the connections in the frame and subassemblage tests was recorded. In the frame tests the bending
produced in small spring steel strips, supporting heavy dumbells and secured to the beams and columns, as a result of
rotation from the vertical was monitored [3]. A complete set of curves for the joints in the frames was thus
obtained. The subassemblage tests [4] were conducted in the horizontal plane and therefore methods of measuring
rotation which rely on the effect of gravity could not be used. Limited access restricted the number of locations which
could be monitored in the same way as the joint tests and a less satisfactory method using two LVDT’s to record the
relative location of the beam and column was adopted. A complete, reliable set of curves for all tests in the
subassemblage series is unfortunately not available. Therefore a complete comparison between joint subassemblage and
frame behaviour is not possible.
< previous page page_197 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_197.html07/11/2009 12:08:40 PM
 page_198

< previous page page_198 next page >

Page 198
Figure 3 compares curves for notionally similar connections—top and bottom cleats with the same beam and
column size, cleat size, bolt size and arrangement and bolt tightness—obtained from isolated joint testing, a
subassemblage test and a frame test. The general form of each curve is very similar with the variation being
approximately the same as that observed between repeat isolated joint tests. This suggests that the characteristics
measured in the basic joint tests provide an acceptable measure of the behaviour of similar connections when functioning
as part of a framework.

Figure 3. Flange cleat moment-rotation behaviour in isolated tests, subassemblages and frames.
Comparisons of the subassemblage and frame tests with finite element frame analysis programmes [5, 6] capable of
including semi-rigid connections showed good correspondence between the predicted and actual response. Fig 4
illustrates the load deflection behaviour of a beam in frame 1 as measured in the test and predicted by the analysis
program. Also shown in Fig 4 are the calculated deflections assuming the beam to be pinned at its supports. A reduction
in deflection at service load of 25% of the simply supported estimate is evident. The analysis was conducted using an
approximate curve consisting of five straight lines closely fitted to the curve obtained from the joint test—see
Fig 5.
< previous page page_198 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_198.html07/11/2009 12:08:41 PM
 page_199

< previous page page_199 next page >

Page 199

Figure 4. Beam deflection for test frame 1.

Figure 5. Approximation of moment-rotation curve for analysis.

< previous page page_199 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_199.html07/11/2009 12:08:42 PM
 page_200

< previous page page_200 next page >

Page 200
It appears that the analysis is not overly sensitive to the exact curve used since some slight variation of the
performance of individual connections around the frame from the joint test behaviour was evident—see figures 6 and 7.
This was also suggested by the authors of the program after undertaking a study of the sensitivity of the computer model
to the number of straight lines required to simulate a moment-rotation curve. In many cases a tri-linear representation
proved to be adequate [6].

Figure 6. Location of joints.

Figure 7. Behaviour of joints in frame test 1.

< previous page page_200 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_200.html07/11/2009 12:08:43 PM
 page_201

< previous page page_201 next page >

Page 201
LESSONS FOR TESTING DRAWN FROM FRAME STUDIES
The existence of a data base of the performance of various forms of structural framing employing semi-rigid connections
permits information on connection behaviour to be extracted that can assist in the conduct of future joint tests.
Probably the most important item is the level of rotation likely to be experienced by the connections as the structure
attains its ultimate load. The experience of the present work, which is limited to non-sway frames under static loading, is
that these are generally far less than the range of rotations normally covered by published data. In this context it is
important to identify the type of component under consideration; axially loaded columns attain their ultimate load with
very little end rotation—typically of the order of 0.005 radians, whereas beams provided with comparatively flexible
connections may require end rotations of 0.02 radians; columns under primary bending e.g. due to unbalanced beam
load, fall somewhere between these two extremes.
If Fig 2 is re-examined in the light of these requirements, it is clear that the most important part of the curve is the
early section, often the initial slope before the region corresponding to joint softening. It is therefore vital that the
measuring systems for rotations are both sufficiently sensitive and are sampled sufficiently frequently during the initial
stages of testing that this region be accurately determined.
A close examination of the behaviour of all the joints in the two frames as columns were brought up to failure showed
that the unloading stiffness of the joints was of more benefit to the column than the loading stiffness of an adjacent joint.
Figure 8 shows the behaviour of joints C and E as the external column was brought up to failure. The much larger
unloading stiffness is readily apparent. It would therefore seem that the investigation of the unloading stiffness of a joint
is equally important as the determination of the initial loading stiffness. The assumption that a connection unloads along
a line parallel to its initial stiffness appears to be reasonable but needs to be conclusively established.
CONCLUSIONS
The work on connection behaviour conducted by Sheffield University has covered the response of joints in isolation,
joints in subassemblages and in full size multi-storey frames. There appears to be no fundamental difference in the action
of similar joints tested in the various environments. Studies of joint behaviour should pay particular attention to their
action in the initial part of the curve and also to their unloading behaviour. The response of connections at large
rotations is likely to be of less importance in interpreting the action of semi-rigid frames.
ACKNOWLEDGEMENTS
The work reported forms part of a continuing investigation into the effects of semi-rigid joint action on the performance
of steel frame structures supported by SERC, CIRIA and Constrado (now subsumed into the SCI) with assistance from
the BRE. The authors are grateful for other help provided by staff of the Department particularly, Mr A M Rifai.
< previous page page_201 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_201.html07/11/2009 12:08:44 PM
page_202

< previous page page_202 next page >

Page 202

Figure 8. Comparison of joints C and E (frame test 2).

REFERENCES
1 Davison, J.B., Kirby, P.A. and Nethercot, D.A. ‘‘Rotational Stiffness Characteristics of Steel Beam to Column
Connections”, Journal of Constructional Steel Research, Vol. 7, 1987. (In Press).
2 Davison, J.B., Kirby, P.A. and Nethercot, D.A. “Effect of lack of Fit on Connection Restraint” Journal of
Constructional Steel Research, Vol. 7, 1987, (In Press).
3 Jennings, D.A., Moore, D.B. and Sims, P.A.C “Instrumenting and Testing Bolted Steel Frame Structures” Proc.,
Determination of Dangerous Stress Levels and Safe Operation Conditions, Edinburgh, Aug. 1986.
4 Davison, J.B., Kirby, P.A. and Nethercot, D.A. “Column Behaviour in PR Construction—Experimental Behaviour”,
ASCE Convention, New Orleans, September 1986 (to be published in Journal of Structural Engineering, ASCE, 1987).
5 Nethercot, D.A., Kirby, P.A. and Rifai, A.M. “Design of Columns in PR Construction—Analytical Studies”, ASCE
Convention, New Orleans, Sept. 1986 (to be published in Canadian Journal of Civil Engineering, 1987).
6 Poggi, C. and Zandonini, R., “Behaviour and Strength of Steel Frames with Semi-Rigid Connections”, Connection
Flexibility and Steel Frames, ed. W.F.Chen, ASCE, 1985, pp. 57–76.
< previous page page_202 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_202.html07/11/2009 12:08:45 PM
 page_203

< previous page page_203 next page >

Page 203
STANDARDISED METHOD FOR MEASURING THREE-DIMENSIONAL RESPONSE OF SEMI-RIGID
JOINTS
M Celikag and P A Kirby
Department of Civil and Structural Engineering
university of Sheffield
Sheffield S1 3JD
UK
ABSTRACT
Whilst procedures for in-plane responses are becoming established little attention has been given to out-of-plane
and torsional characteristics. The findings of exploratory tests are described in this paper which highlight a number of
considerations which need to be resolved before any substantial experimental work is undertaken. There are several ways
in which a beam column connection may be supported and various ways in which the rotation can be computed. It is
suggested that the international research community might give attention to the production of a memorandum to develop
a uniform approach to the problem.
INTRODUCTION
Since the 1930’s many researchers have investigated the behaviour of beam-to-column connections but have mainly
concerned themselves with the in-plane behaviour. Very little experimental information is available on the out-of-
plane and torsional response of semi-rigid connections. A standardised method for in-plane moment-rotation behaviour
of beam-to-column connection has been drafted as a technical memorandum by Task Group 25 of the Structural Stability
Research Council (S.S.R.C.) in 1986 [1].
It is obvious that, in a steel frame, there may be lateral loads acting out-of-plane on beams or unsymmetrical or one-sided
floor loads which cause torsion in beams and therefore on to the beam and column connections. At the same time
eccentric loads may well cause torsional effects. Such restraints are also important in the context of stability, e.g. lateral-
torsional buckling.
This research is aimed at investigating the behaviour of connections due to out-of-plane and torsional loading. Because
cruciform tests rely on symmetry, which can give rise to difficulties, a single connection cantilever arrangement was
chosen (see inset of Fig. 1). The response of the test specimens was found to be more complicated than had been
expected.
< previous page page_203 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_203.html07/11/2009 12:08:46 PM
 page_204

< previous page page_204 next page >

Page 204

Figure 1. The deflection profile of the assemblage for various support conditions.
Some results of the exploratory tests are given and questions regarding the manner in which the relationships
should be determined are raised to initiate an agreed method similar to the S.S.R.C. Memorandum before substantial test
work is carried out.
TEST SPECIMENS, APPARATUS AND INSTRUMENTATION
Flange cleat and web cleat connections were used for exploratory tests. These connections were designed according to an
elastic design method of the Australian Institute of Steel Construction for in-plane loading. Previously similar
connections had been tested for in-plane behaviour and this research was undertaken to find their out-of-plane and
torsional characteristics.
The beam and column elements were the same sections in all tests namely 1000 mm long and
750 mm long. The static yield stresses of the column material and the beam material were 266
N/mm2 and 287 N/mm2 respectively, and were found by testing coupons from their flanges. For the top and bottom
flange cleat connections unequal angles were used; an for the top cleat and a for the
bottom cleat. For the web cleat test again a pair of angle sections was used.
< previous page page_204 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_204.html07/11/2009 12:08:48 PM
 page_205

< previous page page_205 next page >

Page 205
M16 Grade 8.8 bolts with washers were used in drilled 18 mm diameter holes for all connections. Four bolts were used
for the top cleat, six bolts for the bottom cleats and six bolts for the web cleat connection.
A self straining rig was constructed of steel channels and a attachment plate was bolted on the
vertical members to which the column was to be clamped. A hydraulic jack was used to apply the out-of-plane and
torsional loading at the free end of the beam via a load cell. The bolts were tightened up to 160N.m. with a torque
wrench. This torque was adopted as being consistent with the value expected of manual tightening with ordinary tools. A
level was used to ensure the proper alignment of the specimen. For simplicity dial gauges were used for these
exploratory tests which were intended primarily to examine overall response.
TEST PROCEDURES
3.1 Out-of-Plane Response
The single cantilever arrangement of the beam-column connection, as mentioned in section 1, was used. The beam was
connected to one flange of the column with top and bottom flange cleats. For ease of application of load on the minor
axis of the beam, the cantilever arrangement was turned through 90° and then the remote flange of the column was
clamped horizontally against the attachment plate so that the hydraulic jack could be placed vertically to apply a point
load at the free end of the beam web. Four different tests were carried out as listed below. The first, type B, represents
the reference condition; the other types are variations of this situation.

Figure 2. Beam rotations for various support conditions.

< previous page page_205 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_205.html07/11/2009 12:08:49 PM
 page_206

< previous page page_206 next page >

Page 206
Type B: The rear flange of the column was clamped to the attachment plate over its entire length and the beam was
bolted to the front flange of the column by top and bottom flange cleats (see inset of Fig. 3).
Type A: The same connection was used as in type B but an extra plate, equivalent to the depth of beam, was placed
between the rear flange of the column and the attachment plate. The column was clamped in this region only. The aim of
this arrangement was to simulate restraint over the depth of an equivalent beam rather than the clamping action being
present over the full column length (see inset of Fig. 4).
Type C: The test arrangement used was as in type B but with additional stiffeners welded at four positions from the rear
flange and up to the column centreline to substantially prevent the rotation of the rear half of the column web. The
stiffeners were located on the column 125 and 250 mm away from the mid-length of the column in both directions (see
inset of Fig. 5). This was designed to simulate symmetry about the column centreline.
Type D: The test arrangement in type C was used but, in addition, full stiffeners were welded to the column flanges and
to the column web at positions to correspond to the beam flanges. This was intended to substantially eliminate both the
web and flange distortions (see inset of Fig. 6).
Several series of deflection readings were taken along the mid-line of the beam and the corresponding mid-length of the
column to produce a deflection profile of the assemblage for all types as is shown in Figure 1.

Figure 3. curves: unstiffened column web.

< previous page page_206 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_206.html07/11/2009 12:08:50 PM
 page_207

< previous page page_207 next page >

Page 207

Figure 4. curves: rear column flange partially fixed

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_207.html (1 of 3)07/11/2009 12:08:51 PM

 page_207

Figure 5. curves: column web stiffened to centreline.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_207.html (2 of 3)07/11/2009 12:08:51 PM

 page_207

Figure 6. curves: column web fully stiffened.

< previous page page_207 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_207.html (3 of 3)07/11/2009 12:08:51 PM

 page_208

< previous page page_208 next page >

Page 208
3.2 Torsional Response
The rear column flange was clamped vertically on the attachment plate as in type B above. The beam element was
connected to the front column flange by (i) web cleats and (ii) top and bottom flange cleats. Loading was applied by a
hydraulic jack via a plate welded to the free end of the beam which was in turn supported by a bearing thereby imparting
a pure torque to the beam.
Dial gauges were mounted on both sides of the beam web and the column to measure the horizontal deflections due to
the torque, which enabled the beam web rotation and the column web and flange rotations to be determined. Thus the
connection rotation could be calculated. Additionally, attempts were made to measure the free and restrained warping of
the beam flanges.
TEST RESULTS AND DISCUSSION
4.1 Out-of-Plane Tests
Tests were carried out at low moments only, which is thought to be the most important stage, but it is intended to
measure the response over greater ranges in the actual test programme. At these low moments the bending of the beams
can be seen to be relatively unimportant.
The deflection profile of the assemblage for four types of support conditions are given in Figure 1. These profiles
correspond to three levels of moment at the connection which are 0.61, 1.13 and 1.65 kNm. Beam deflections are linear
while those in the column web are not. For different support conditions, when the slope of the beam is extrapolated to the
x-axis, it is found that the beam was rotating about points some distance away from the fixed end of the column. The
distances of these positions, which remain virtually unchanged within each test as moment increases, are 66 mm for type
A, 78 mm for type B, 107 mm for type C and 102 mm for type D. The deflections of the beam and the column decrease
depending on the support conditions in the order of types A, B, C and D.
There are several ways in which the actual connection rotation may be calculated using these deflections. Figure 2 shows
the curves for the four support conditions listed above where is taken as the total beam rotation. Figure 3 shows
the characteristic of figure 2 for support condition type B where the total beam rotation is labelled as rotation type I.
Additionally may be measured in three other ways giving four rotation types:
Type I as total beam rotation
Type II as beam rotation relative to the column centreline
Type III as beam rotation relative to the front edge of the column web
Type IV as beam rotation relative to the front edge of the column flange.
It can readily be seen that these give widely differing connection stiffnesses. Figures 4, 5 and 6 give similar results for
the other three support conditions. There are now 16 possible combinations of support and measurement types with 16
possible characteristics for one connection.
< previous page page_208 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_208.html07/11/2009 12:08:52 PM
 page_209

< previous page page_209 next page >

Page 209
Consider real structures. An exterior column usually does not receive symmetric loads from beams and floors. Type A
and B support conditions could be representative in this case. An interior column often carries symmetrical loads coming
from beams and floors. Type C and D support conditions could be appropriate in this case.
From the authors’ point of view, considering the practical, analytical, and design requirements more than one type of
support condition and type of connection rotation measurement might be needed. A change in the length of the column
specimen used may be another factor which could affect the results. This would suggest that consideration needs to be
given to the way in which such tests are carried out to ensure that enough information is provided to enable the
appropriate curve for any situation to be determined. The authors feel that the most conservative approach namely
the combination of types A and I is perhaps the most appropriate for design. Alternatively it may be necessary to
investigate two different conditions, one for interior columns and one for exterior columns for nominally identical joint
conditions.
4.2 Torsion Tests
The results of torsion tests on two web cleat and one flange cleat connections are shown in Figures 7 and 8, from which
it can be observed that the actual connection rotation is small compared with the twist occurring in the beam. Figure 7,
shows that, at a moment of 2 kNm, the rotation of the web cleat connection (2) is rad, whereas the twist
occurring in a 1 m length of beam for the same condition is rad,

Figure 7. Connection torque-rotation curves.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_209.html (1 of 2)07/11/2009 12:08:54 PM

 page_209

Figure 8. Beam torque-rotation curves

< previous page page_209 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_209.html (2 of 2)07/11/2009 12:08:54 PM

 page_210

< previous page page_210 next page >

Page 210
as can be observed from Figure 8. Thus even this modest connection is torsionally very stiff as compared with the beam
and could be considered as infinitely stiff. However there could be other types of connections where this is not the case.
Attempts were made to measure the free and restrained warping distortions of the top and bottom flanges of the beam for
web cleat and flange cleat connections. For theoretical analyses there is a need to measure the warping bi-moment and
consideration needs to be given as to how this should be done.
4.3 General Observation
The work described above has concentrated on the first cycle of loading applied to a joint. However studies on the in-
plane behaviour of subassemblages and frames [2] have shown, that as column buckling occurs, the direction of the
moment applied at a joint often reverses. It is thus the unloading stiffness of a joint which is frequently most important in
providing restraint to a column as the failure condition is approached.
This occurs even with monotonic frame loading. The question of measuring unloading stiffnesses at various moment
levels is another feature which requires attention.
CONCLUSION
Preliminary out-of-plane tests on beam-to-column connections have shown that there are a variety of ways in which the
column may be supported and a number of ways in which the connection rotation might be measured. The resulting
moment rotation characteristics of the joints vary quite markedly.
The results of torsion tests have demonstrated that even for modest connections e.g. web cleats, the flexibility of the joint
is normally very small when compared with the bare beam. The problem of determining warping stiffness has been
noted. Attention has been drawn to the significance of unloading stiffness.
It is concluded that, with the wide attention being given to semi-rigid action, there is a need to develop a recognised set
of requirements for the conduct of such tests.
REFERENCES
1 Structural Stability Research Council Task Group 25, “Testing of Beam to Column Connections”, Draft Technical
Memorandum, 1986.
2 Kirby, P.A., Davison, J.B. and Nethercot, D.A., “Large Scale Tests on Column Subassemblages and Frames” State of
the Art Workshop, Connections, Strength and Design of Steel Structures, E.N.S.Cachan, May 1987.
< previous page page_210 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_210.html07/11/2009 12:08:54 PM
 page_211

< previous page page_211 next page >

Page 211
DISCUSSION OF CLASSIFICATION PAPERS
There was general consensus among speakers and discussers alike that with the advent of limit states design, it is critical
to arrive at a system of classification of connection response with regard to strength, stiffness and deformation capacity.
In particular, this will facilitate the work of designers, to the effect that it may not be necessary to work with individual
connection M-Φ-curves, for example.
Suggestions for simplified moment-rotation curves were made, to the extent that it would be preferable from an
analytical standpoint to be able to deal with linear relationships rather than highly non-linear ones. Several discussers
expressed concern over this as being a too liberal simplification, although analyses have shown that for most practical
purposes, it is the magnitude of the initial connection stiffness that is the most important to the overall frame response.
This applies to frame drift calculations as well as to stability analyses, although for the latter it is clear that the initial
stiffness is of overriding importance.
Several questions dealt with the initial and subsequent connection stiffness characteristics. Thus, there appeared to be
general agreement that loading and unloading stiffness magnitudes are crucial to the stability of the frame. Test results
for structural subassemblages demonstrated that this was the case as well, and good correlation was found when the
unloading stiffness was set equal to the initial stiffness. It was pointed out that for such problems, the overall rotation
capacity of a connection is of lesser importance; rather, the initial portion of the curve and the shape of the unloading
part are of primary concern.
Several speakers emphasized that although it appears that the unloading stiffness of a connection is the same as the
loading one, this should be examined for a range of connection types and sizes. Similarly, it was pointed out that an
engineering definition of the connection stiffness is required, not the least because many test curves exhibit little or no
initial linear range, or even a much softer connection response than what is eventually determined. This is certainly in
part due to the nature of the physical testing in itself, but further emphasizes the need for a clear definition of the
stiffness.
It was noted that although significant variations can be found in the stiffness values for different types of connections,
their impact on
< previous page page_211 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_211.html07/11/2009 12:08:55 PM
 page_212

< previous page page_212 next page >

Page 212
the response of the frames is not as severe. In other words, with a given type of connection one may use quite different
EI-values, and still arrive at substantially comparable strength and deflection results. This also holds true for stability
analyses of individual columns and columns as part of subassemblages. Again, the importance of the connection
unloading stiffness was stressed by discussers and speakers alike.
< previous page page_212 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_212.html07/11/2009 12:08:56 PM
page_213

< previous page page_213 next page >

Page 213
SESSION No 3
FRAME ANALYSIS
< previous page page_213 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_213.html07/11/2009 12:08:57 PM
 page_214

< previous page page_214 next page >

Page 214
FRAME ANALYSIS INCLUDING SEMI-RIGID CONNECTIONS AND P-DELTA EFFECTS
Socrates A.Ioannides, Ph.d., P.E.
Structural Affiliates International, Inc.
2416 Hillsboro Road
Nashville, Tennessee 37212
USA
ABSTRACT
An analysis technique for Steel Frames that include Semi-Rigid connections is presented. The technique is easy to
implement since it utilizes the stiffness method of analysis of structures. A typical beam element with six degrees of
freedom (d.o.f.) is utilized together with a six d.o.f column element that utilizes stability functions to account for the P-
Delta effect. The connection flexibility is incorporated into the model through a two d.o.f. spring element whose stiffness
is the secant stiffness of the particular connection at the particular load level. Due to the non-linearity of the connection
and the P-Delta effect the solution is iterative and the solution algorithm is presented.
INTRODUCTION
Analysis, design and drift calculations for steel frames are inter-linked processes that jointly dictate the final sizes of the
individual members and connection details of steel frameworks. As a result each one of these has to be compatible in
order that a consistent approach be achieved (i.e. the design approach used depends greatly on the analysis assumptions
and techniques used). For the purposes of this paper only the analysis portion will be examined; design and drift
considerations will be presented in a subsequent paper.
The modeling requirements that will make the analysis, design, and drift calculations consistent will be discussed in this
paper. The use of this consistent approach will result in better understanding of structural behavior by demonstrating the
actual force distribution and by avoiding such ‘‘design approximations” as the “Moment Magnification Factor”.
Specifically the inclusion
< previous page page_214 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_214.html07/11/2009 12:08:58 PM
 page_215

< previous page page_215 next page >

Page 215
of factors such as P-Delta and Connection Flexibility in the analytical model will be discussed and demonstrated.
The A.I.S.C. specification as presented in the Eighth edition of the Manual of Steel Construction[1] describes three types
of construction:
Type I, commonly designated as “Rigid Framing”,
Type II, commonly designated as “Simple Framing”, and
Type III, commonly designated as “Semi-Rigid Framing”.
It further stipulates that “the design of all connections shall be consistent with the assumptions as to type of construction
called for on the design drawings”. Further “type II construction is allowed as long as the wind moments are distributed
among selected joints of the frame and certain provisions are met’’. Since 1964 Disque[2] has promoted the use of type
II construction and has provided certain simplifying design assumptions such as the directional moment connections to
facilitate the design engineer in utilizing this simple type of construction. Work done at the University of Colorado and R.
P.I, by Ackroyd, Gerstle et.al. [3,4] has substantiated the use of type II construction in accordance with the Disque
assumptions with certain restrictions. The purpose of this paper is to deal with type III (Semi-Rigid) not type II.
It can be argued that no connection is either fully rigid or totally pinned or flexible; every connection possesses some
restraint no matter how flexible it is considered and even fully welded connections exhibit some flexibility. The Load
and Resistance Factor Design (L.R.F.D.) [5] by A.I.S.C. recognizes this fact and distinguishes connections in two
categories: Fully Restrained (FR) and Partially Restrained (PR). The technique described herein is suitable for analyzing
frames that include PR connections, provided the Moment-Rotation characteristics of these connections are known.
MODELING
For the purpose of demonstrating the technique and in accordance with common construction practice the columns will
be assumed continuous at the floor lines. However the technique could be extended to utilizing P.R. connections even
between column stacks. The inclusion of Semi-Rigid connections connecting the beams through the column stacks
necessitates a modelling technique that would allow relative rotation between the beam and column at a joint. As shown
in Figure 1 this can be achieved by introducing additional rotational degrees of freedom at each joint; thus exterior joints
will have a total of four degrees of freedom i.e. horizontal and vertical translation, and rotation of the column
< previous page page_215 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_215.html07/11/2009 12:08:59 PM
 page_216

< previous page page_216 next page >

Page 216
stack, and beam-end rotation, whereas interior joints will have five degrees of freedom, namely: the joint translation in
the horizontal and vertical direction, the joint or column stack rotation and the two beam-end rotations. In addition to the
beam and column elements spring elements are introduced connecting the joint rotation to each beam end rotation. In
developing a computer code for analyzing frames such as shown in Figure 1 the following stiffness matrices should be
utilized:

FIGURE 1: TYPICAL FRAME MODEL

A. Beam elements: The stiffness matrix utilized for the beam element should be the standard stiffness matrix found in
basic structures texts[6].
B. Column elements: To account for P-Delta effects in the analysis the column stiffness matrix is formulated utilizing
stability functions. The expressions for the stability functions can be found in basic structural analysis texts[6]. It should
be noted that the stability functions only influence the first 4x4 portion of the stiffness matrix relating the rotational and
shear degrees of freedom. These functions replace the standard stiffness and carry-over functions with ones whose values
depend on the axial load present in a member.
C. Spring elements: The stiffness matrix for the spring element is a simple 2x2 matrix with the stiffness of the spring
(Ks) in the diagonal terms and “-Ks” in the off-diagonal terms.
Frames with semirigid connections are generally more flexible than those designed with fully rigid connections; The P-
Delta effects are, therefore, generally more pronounced and it is the author’s opinion that any analysis that includes
connection flexibility should also account for P-Delta. The column
< previous page page_216 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_216.html07/11/2009 12:09:00 PM
 page_217

< previous page page_217 next page >

Page 217
element described above, formulated with stability functions to account for the loss of stiffness with increasing axial
load, has been found to work consistently well in the presence of semirigid connections with non-linear semirigid
connections.
The proposed method is dependent on the availability of Moment-Rotation curves. In the past this seemed to be the
bottleneck of analysis techniques that considered connection flexibility. Recently, however, efforts from various sources
are beginning to make available such information both in the form of available experimental data and predictor equations
[7,8,9].
A typical moment rotation curve is shown in Figure 2. The non-linear nature of the Moment-Rotation curve shown is
typical of most connections tested in the past [7,8,9]. It should therefore be accounted for in the method of analysis.
Three different stiffness values can be identified with any Moment-Rotation curve:
a) the initial stiffness,
b) the tangent stiffness at any point, and
c) the secant stiffness.

FIGURE 2: MOMENT ROTATION CURVE

Any of the three may be utilized in the analysis so long as it is consistent with the overall technique used; namely the
tangent stiffness could be used with an incremental load application technique or, as recommended here, a secant
stiffness should be used in an analysis where the full factored load is applied in one step. In the method proposed herein
the secant stiffness is recommended for the following reasons:
a) the factored loads will be applied in one step,
b) the secant stiffness provides an integrated average of how the
< previous page page_217 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_217.html07/11/2009 12:09:01 PM
 page_218

< previous page page_218 next page >

Page 218
connection arrived at the present level of loading,
c) regardless of the mode of loading and unloading at one time in the life of the structure the actual moment-rotation
was followed to arrive at the present state of load.
d) if an initial stiffness is used the deflections derived will be erroneous.
It is further proposed that factored loads be utilized in the analysis, otherwise due to the non-linear nature of Moment-
Rotation curves an incorrect factor of safety may be assumed to exist as shown in Figure 2. Using unfactored loads and
allowable stress design leads to false factor of safety, underestimates total deflections, and thus underestimates the P-
Delta effects.
NUMERICAL ALGORITHM
Having discussed the components of the model and the loads to be applied the numerical algorithm is hereby presented.
First Iteration
The structure stiffness matrix is originally formed by ignoring P-Delta effects and assuming all connections are fully
rigid. This forms the starting point for the iterative process.
Second Iteration
The column axial loads from the first iteration are used to define new column stiffness matrices utilizing the stability
functions.
The spring stiffness for each beam to column semi-rigid connection is obtained from the “available” Moment-Rotation
curves for each type of connection by making an “educated guess” on the relative connection rotation. A method found
by the author to give consistently good results for this initial guess is to superimpose the “beam line” on the moment-
rotation curve (Figure 3). The beam line is a straight line connecting the point of fixed end moment at zero rotation to the
point of zero moment and maximum rotation for the loading condition. This second point is the amount of rotation the
beam would undergo if it was pinned. Although this value is dependent on the type of loading it varies within a narrow
range of 0.016 radians for a point load at mid-span to 0.021 radians for uniformly distributed loads. The algorithm is not
very sensitive to this initial value, so a value of 0.02 radians was found to work well for
< previous page page_218 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_218.html07/11/2009 12:09:02 PM
 page_219

< previous page page_219 next page >

Page 219
commonly encountered loadings. The spring stiffness for each connection is thus determined as the secant stiffness
passing though the intersection of the Moment-Rotation curve and the Beam-Line.

FIGURE 3: BEAM LINE

The structure stiffness matrix is then recalculated utilizing the new column and spring stiffnesses and the problem
resolved.
Typical Subsequent Iteration
The column stiffness matrices are reevaluated based on the axial loads from the previous iteration.
The new spring stiffness for each connection is found by comparing the resulting moment and relative rotation at each
spring location with the Moment-Rotation curve as shown in Figure 4. If the moment-rotation point from the previous
iteration does not fall on the Moment-Rotation curve a vertical line is drawn until the M-Theta curve is intersected. The
new spring stiffness is then defined as the slope of the secant passing through this new point. The convergence has been
found to accelerate if instead of the new stiffness defined above a weighted average of this and the previous iteration is
used as shown in Figure 4.
The new spring stiffness matrices are thus reevaluated and the new structure stiffness matrix formed.
The problem is solved again.
At this point convergence is checked by comparing the results of the present to the previous iteration. Specifically the
relative rotation between the ends of each spring and the column axial loads are compared to the previous iteration. If
convergence is detected the algorithm is stopped; if not the steps described in
< previous page page_219 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_219.html07/11/2009 12:09:03 PM
 page_220

< previous page page_220 next page >

Page 220
this section are repeated.

FIGURE 4: NEW SPRING STIFFENERS

CONCLUSIONS
A method has been presented for the analysis of steel frames utilizing semirigid connections. The basic components of
the model are generally available so implementation of the algorithm should be straight forward for individuals with
basic knowledge of the stiffness method of analysis. P-Delta effects are critical due to the increased flexibility of the
frames and should be included. The resulting iterative problem is solved by iterating on both the P-Delta and connection
flexibility at the same time. Convergence has been achieved in four to six iterations for a variety of frames.
REFERENCES
1. ___, Manual of Steel Construction, American Institute of Steel Construction, Eighth Edition, Chicago, 1980.
2. Disque, R.O., ‘‘Wind Connections with Simple Framing,” Engineering Journal, AISC, Vol. 1, No. 3, July 1964.
3. Ackroyd, M.H., and Gerstle, K., “Behaviour of Type 2 Steel Frames,” Journal of the Structural Division, ASCE,
Vol.108, No. ST7, July 1982, pp 1541–58.
4. Dewkett, K.A., Ackroyd, M.H., “Influence of Connection Behaviour on Steel Frame Designs at Factored Loads,”
paper presented at the ASCE Annual Conference 1984, San Francisco.
5. ___, Manual of Steel Construction, Load & Resistance Factor Design, American Institute of Steel Construction, First
Edition, Chicago, 1986.
< previous page page_220 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_220.html07/11/2009 12:09:04 PM
 page_221

< previous page page_221 next page >

Page 221
6. Beaufait, F.W., Rowan, W.H., Hoadley, P.G., and Hackett, R.M., Computer Methods of Structural Analysis,
Department of Civil Engineering, Vanderbilt University, Nashville, Tennessee, 37235, 1975.
7. Ioannides, S.A., “Flange Behaviour in Bolted End Plate Moment Connections,” Thesis Presented to Vanderbilt
University, Nashville, Tennessee, 1978 in Partial Fulfillment of the Requirements for the Degree of Doctor of
Philosophy.
8. Goverdhan, A.V., “A Collection of Experimental Moment-Rotation Curves and Evaluation of Prediction Equations for
Semi-Rigid Connections,” Thesis Presented to Vanderbilt University, Nashville, Tennessee, 1983 in Partial Fulfillment
of the Requirements for the Degree of Master od Science.
9. Nethercot, D.A., “Steel Beam to Column Connections—A Review of Test Data and Their Applicability to the
Evaluation of the Joint Behaviour of the Performance of Steel Frames,” CIRIA, London, 1985.
< previous page page_221 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_221.html07/11/2009 12:09:04 PM
 page_222

< previous page page_222 next page >

Page 222
NON-LINEAR ANALYSIS OF PLANE FRAME STRUCTURES WITH SEMI-RIGID CONNECTIONS
Yvan GALEA
C.T.I.C.M.—B.P. 1
78470 Saint Rémy lès Chevreuse
FRANCE
1. INTRODUCTION—ABSTRACT
The essential objective of the present paper consists in exposing the procedure of taking into account semi-rigid
connections with nonlinear behaviour, a posteriori introduced into the CTICM-analysis programme of plane frame
structures (PEP).
After a succinct description of the initial state of the PEP-programme, the double spring-ended special member
introduced is presented. The description of the taking into account of nonlinear behaviour of the semi-rigid connections
follows and the paper is completed by results obtained on a two level-three span portal frame.
2. SUCCINCT DESCRIPTION OF THE INITIAL STATE OF THE PEP-PROGRAMME
A detailed description of the initial state of the PEP-programme is given in [1] and [2].
PEP is an analysis programme of plane frame structures the CTICM has developped for research purposes. It allows to
proceed to four types of structural analysis:
- first order elastic analysis,
- second order elastic analysis,
- first order plastic analysis,
- second order plastic analysis.
The main characteristics of the PEP-programme are the following:
a) The plastification in the structural members is based on the concept of plastic hinges with bending moment-axial
force interaction.
b) The determination of the states of equilibrium is based on the “displacement method”.
< previous page page_222 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_222.html07/11/2009 12:09:05 PM
 page_223

< previous page page_223 next page >

increments of the vectors ‘‘nodal loads” |
increments of the vectors ’’nodal loads” |
increments of the vectors nodal loads” |
increments of the vectors nodal loads” |P| and “nodal displacements” |D| during the step.
| and “nodal displacements” |D| during the step. | and “nodal displacements” |D| during the step. | and “nodal
displacements” |D| during the step. OCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "DTD/
xhtml1-transitional.dtd"> Page 223 c) The loads on the structure, which are static loads, maintain their proportionality
during the loading evolution. This leads to the definition of a parameter α, representing a multiplier of the loading
provided by the input data of the programme. d) Any non-linearity in the structural response to the applied loads,
resulting from the occurence of either plastic hinges or second order effects, is dealt with by a step-by-step method
without internal iterations (for example, similar to those of the Newton-Raphson method). The incremental behaviour at
each loading step assumes the formulation: where, - increments of the vectors “nodal loads” |P| and “nodal
displacements” |D| during the step. [Kt] - tangent stiffness matrix of the structure as a function of the plastification state,
of the displacements of the nodes and of the axial forces determined in the structure at the end of the previous step.
The nonlinear behaviour is therefore described by a succession of elementary linear behaviours with Kt being
reactualized at each step. The accuracy is obtained by specifying limited amplitudes for the loading steps. For very
divergent behaviours (close to instability), displacement control is possible. At each step, the condition of the structure is
determined by summing up the variations of condition calculated at the previous steps. Should an instability occur
(collapse mechanism, member buckling, global instability), the detection criterion is the sign of the determinant of the
matrix Kt that changes from positive to negative. e) Prior to the introduction of the present developments, the
connections between the structural members were either perfectly rigid or perfectly hinged. 3. MODELLING OF SEMI-
RIGID CONNECTIONS IN THE PEP-PROGRAMME A detailed description of this modelling is given in [5]. In the
framework of this study, neither the influence of the shear flexibility of the connections upon the global structural
behaviour nor the possible influence of the shear force upon the intrinsic behaviour of the connections are taken into
account. As a consequence, each connection of semi-rigid type is modelled by two equivalent springs: - an axial stiffness
spring Kξ simulating the axial behaviour of the connection when the connected member is strongly compressed or in
tension;
< previous page page_223 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_223.html07/11/2009 12:09:06 PM
 page_224

< previous page page_224 next page >

Page 224
a spiral stiffness spring Kθ simulating the bending behaviour of the connection.
The interaction of these two springs reproduces the global behaviour of the connection. In case of nonlinear behaviour of
the connection, the stiffnesses Kξ and Kθ of the equivalent springs are modified at each step of the “step-by-step”
procedure, according to the evolution of the parameters that have an influence over these stiffnesses (for instance, the
forces upon the connection).
The thus modelled connection is then “attached” to a member whose stiffness matrix integrates Kξ and Kθ. The
modelled connection is therefore considered to have no length.
In this simple way, a special member element is made up of a standard member with the double spring just described at
each extremity (figure 1).
This type of element allows one to treat similarily the members with semirigid connections at one or at both ends. For
the end without semi-rigid connection, an infinite (immeasurable great) stiffness Kξ and an infinite or null stiffness Kθ
according as the connection is perfectly rigid or perfectly hinged respectively, are introduced.
In order to let appear the second order effects and the instability phenomena in strongly compressed members with semi-
rigid connections, the tangent stiffness matrix of the member element represented en figure 1 has been determined.

Figure 1.
This tangent stiffness matrix Rt which relates, during a step, increments of forces at the ends i and j of the member to
increments of displacements of the same ends according to the following relationship:

is expressed as follows in the local system of the member:

< previous page page_224 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_224.html07/11/2009 12:09:07 PM
 page_225

< previous page page_225 next page >

Page 225
Where: N=axial force as obtained at the end of the previous step (negative In compression).
Ω=(vj−vi)/L=global rotation of the member as obtained at the end of the previous step.
L =length of the member between ends i and j.
E =Young’s modulus.
I =inertia of the member cross section.
A=area of the member cross section.
And:

. (compression)

. (tension)

.
The assemblage of the stiffness matrices Rt of all the members, with or without semi-rigid connections, allows one to
obtain the global tangent stiffness matrix Kt of the structure.
< previous page page_225 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_225.html07/11/2009 12:09:09 PM
 page_226

< previous page page_226 next page >

Page 226
4. MODEL OF NON LINEAR BEHAVIOUR OF THE CONNECTION ADOPTED IN THE PEP-PROGRAMME
The model of nonlinear behaviour of the connection adopted at the present time in the PEP-programme is that worked
out by COLSON and PILVIN [3,4]. Hereafter, the basic characteristics of their simplified model are recalled.
The ultimate limit curve of the connection is a bending moment-axial force interaction curve. In a first step, the curve
has been linearized for the sake of simplification (figure 3).
Mu is the ultimate bending resistance of the connection; it is assumed to behave in a similar manner whatever the sign of
the bending may be (symmetrical connection).

are the ultimate axial resistances of the connection in compression and in tension respectively.

Figure 2.

Figure 3.
Figure 2 shows a typical force-displacement curve characteristic of a semi-rigid connection under static loading.
Within the adopted model, the description of the nonlinear behaviour, during the loading phase, is based on knowledge
of experimentally determined values of:
- the initial bending and axial stiffnesses of the connection,
- the ultimate bending and axial resistances,
- other parameters characteristic of the non linear behaviour of the connection,
and of a variable δ depending on the loading level that has been reached and characterizing the distance with regard to
the ultimate limit curve (Figures 2 and 3).
The law of differential behaviour of the connection has been described in form of the following relationships:

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_226.html (1 of 2)07/11/2009 12:09:10 PM

 page_226

(1)
< previous page page_226 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_226.html (2 of 2)07/11/2009 12:09:10 PM

 page_227

< previous page page_227 next page >

Page 227
where:
. dξ, dθ: variations of elongation and rotation within the connection,
. dn, dm: variations of reduced forces

if tension, if compression

. Km, Kn: initial stiffnesses of the connection in bending and elongation if tension, if compression
with
δo=initial value of δ (see figure 3).
. An, Bn, Am, Bm: (experimental) coefficients for characterization of the non linear behaviour
and M-N interaction in the connection.

The tangent stiffnesses of the connection: and are derived from the relationship (1).
Consequently, during a step p of the “step-by-step” procedure of the nonlinear analysis of the structure, the tangent
stiffnesses input for a semi-rigid connection in the tangent stiffness matrix of a member are calculated by:

with:
: variations of reduced stresses during the previous step.
: distance with regard to the ultimate limit curve as determined at the end of
the previous step.
If, during a load step, the ultimate limit curve is transcended, the amplitude of the load step is reduced in order to bring
the point A (figure 3) back on the curve and the connection is replaced by a perfect hinge with regard to the later load
increments. The use of this artifice is simply related to the numerical aspect of the “step-by-step” description of the non
linear behaviour. Strictly speaking, the limit curve should never be reached but only approached asymptotically.

Particular cases: If If
< previous page page_227 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_227.html07/11/2009 12:09:12 PM
 page_228

< previous page page_228 next page >

Page 228
5. EXAMPLE OF APPLICATION
The chosen example of application concerns one of the calibrating frames defined by R.ZANDONINI (Polytechnic
School in Milan), responsible within a subgroup of the ECCS—TC8 “Stability’’ for a study dealing with the stability of
structures with semi-rigid connections.
Both the structure and the loading are defined in figure 4.

Figure 4.
The structure is considered with an initial verticality imperfection . The type of beam-column connection
is depicted in figure 5. The values of the connection parameters have been determined on a supplied bending moment-
rotation curve; they are:

Since the connections are essentially in bending, the following values have been adopted:

(very large)

Figure 5.

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_228.html (1 of 2)07/11/2009 12:09:13 PM

 page_228

Figure 6.
< previous page page_228 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_228.html (2 of 2)07/11/2009 12:09:13 PM

 page_229

< previous page page_229 next page >

Page 229
In figure 6, the evolution of the lateral displacement D at the top of the structure has been drawn as a function of the
loads that increase proportionally with a multiplying factor α, and this for the three following analysis:
- first order plastic analysis+rigid connections (curve A)
- second order plastic analysis+rigid connections (curve B)
- second order plastic analysis+semi-rigid connections (curve C)
The amplitude of the loading step was limited by the most severe condition, i.e. or . As
concerns the curve C, that resulted in 48 steps before reaching the collapse and 4 mn 20 seconds calculation time on a
IBM 4361–3 computer.
In order to take correctly the second order effects into account, the columns have been discretized in 3 sections. Because
the double spring-ended element, described in 2, can not receive loads between its ends, the beams have also been
discretized in 3 sections as shown in figure 7.

Figure 7.
The important divergence of behaviour depending on whether the semi-rigid connections have or have not been taken
into account should be noted. The difference between the collapse load by instability in the elastoplastic range is about
16%. The lack of web stiffeners in the columns at the lower flange of the beams may explain the flexibility of this type
of connection.
6. CONCLUSIONS
The fact of taking into account the semi-rigid nature of the connections in a pre-established programme resulted in
adopting a simplified geometrical model. As a consequence, a certain number of assumptions have been admitted
- connections without their own dimensions,
- shear is not taken into account,
- the ultimate limit curve is linearized.
Some other assumptions belong implicitly to the analysis method adopted for the PEP-programme (plastic hinges, “step-
by-step” method without iterations).
Although a comparison with results obtained by other programmes used abroad has been satisfactory, a further stage is
necessary that will consist in comparing a numerical simulation with large scale tests aimed at confirming the validity of
the adopted assumptions.
< previous page page_229 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_229.html07/11/2009 12:09:14 PM
 page_230

< previous page page_230 next page >

Page 230
7. REFERENCES
1. Lescouarc h Y., Programme de calcul en élastoplastlcité des structures planes. Revue Construction Métallique,
CTICM, no 1, 1976, pp. 40–50.
2. Galéa Y., Programme d’analyse élastoplastique non linéaire de structures planes à barres. Revue Construction
Métallique, CTICM, no 4, 1978, pp. 3–16.
3. Pilvin P., Modèle de comportement statique des liaisons de structures à barres. Revue Construction Métallique,
CTICM, no 2, 1984, pp. 43–50.
4. Colson A., Modélisation des conditions aux limites de liaisons et d’assemblages en mécanique des structures
métalliques. Thèse de Doctorat d’Etat. Paris 6. 18 décembre 1984.
5. Galéa Y., Colson A., Pilvin P., Programme d’analyse de structures planes à barres avec liaisons semi-rigides à
comportement non linéaire. Revue Construction Métallique, CTICM, no 2, 1986, pp 3–16.
< previous page page_230 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_230.html07/11/2009 12:09:15 PM
 page_231

< previous page page_231 next page >

Page 231
NONLINEAR INELASTIC ANALYSIS OF FLEXIBLY-CONNECTED STEEL FRAMES
M.H.Ackroyd
Techtrol, Ltd.
Montreal, Quebec
ABSTRACT
This paper describes a nonlinear computer analysis program that can simulate the behavior of unbraced steel frames that
have flexible beam-to-column connections and/or partial composite girders. It summarizes the modeling capabilities and
cites instances of how the tool has contributed to a better understanding of complex frame behavior and frame design.
INTRODUCTION
This paper describes an analysis program that has been developed for evaluation of structural steel building frames at
their ultimate load level. The program was developed over a period of nine years in a research effort to evaluate the
actual behavior of AISC “Type 2” frames (1), funded primarily by the American Iron and Steel Institute.
SCOPE
The program described herein is a research tool that has been specialized for very precise modeling of a large number of
nonlinear phenomena that arise in the class of planar unbraced sway frames consisting of horizontal beams and vertical
columns. No accounting is made for the possibility of out-of-plane effects such as lateral-torsional buckling of beams,
out-of-plane frame buckling, or torsional buckling of members.
MODELING APPROACHES
In order to accurately predict the behavior of flexibly-connected steel frames, both material and geometric nonlinearities
need to be accounted for in the mathematical models
< previous page page_231 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_231.html07/11/2009 12:09:16 PM
 page_232

< previous page page_232 next page >

Page 232
of the various structural components of the frame. The program described here includes models for inelastic beam-
columns, beam-to-column connections having nonlinear moment-rotation curves (including unloading capability), and
inelastic composite concrete-steel girders with slip.
Beam-columns
The model for steel I-shaped members treats the cross-section as three rectangular elements. All cross-section effects are
computed as analytic expressions relative to these rectangular elements and then in terms of overall section response.
Second order geometric effects are accounted for directly at the member level. Details of the development can be found
in (2).
A. Cross-section Model
Residual Stresses. The residual stresses in the cross-section are represented as initial stress patterns that vary
parabolically and symmetrically along the length of each rectangular component. The patterns in the flanges are taken to
be identical, so that one need only specify the values of the residual stress at the tip of a flange and at the center of the
web: the program automatcally computes the residual stress at the web-flange junction based on equilibrium of the
unloaded member.
Force Resultants. When loads are applied to the section, a linear strain diagram is assumed, elastic-plastic fiber strains
are evaluated as appropriate, and stress resultants are found that equilibrate the applied loads. The portions of the section
that remain elastic are then used to compute the available bending rigidity for use in associated member stiffness
calculations.
B. Spread of Plastification
When loads are applied to a member, the moment diagram is checked to see if any sections along the member length will
experience plastic strains in the presence of axial load and moment. In the regions where plastification is predicted, the
length of the plastic region is computed as the distance from the point where elastic behavior ceases to the point of
maximum moment. At the point of maximum moment, the precise fiber model above is used to compute the bending
rigidity of the cross-section. Then a linear variation of the bending rigidity is assumed from the linear elastic value at the
start of the plastic region to the minimum value just computed. (The accuracy of this approach was corroborated by
ancillary studies comparing finite difference solutions of the continuous problem to the displacements solutions
predicted by the discretized equivalent taper model). A closed form stiffness matrix for the tapered portion of the
member is then combined with the stiffness matrix for the elastic portion of the member, internal
< previous page page_232 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_232.html07/11/2009 12:09:17 PM
 page_233

< previous page page_233 next page >

Page 233
degrees of freedom are condensed out, and the total member stiffness matrix is obtained for use in a frame analysis.
C. Second-Order Geometric Effects
In formulating the stiffness matrix of columns in the frame, the classical stability functions are used for reducing the
flexural stiffness coefficients due to the presence of axial thrust. Overall member P-Delta effects are accounted for by
adding the usual member geometric stiffness matrix to the existing member stiffness matrix.
Nonlinear Beam-to-Column Connections
Flexible beam-to-column connections are modeled as two-degree-of-freedom spring elements having moment-rotation
characteristics appropriate for the given stiffness. The frame analysis program uses a secant stiffness approach, so that
the stiffness is simply the ratio of the current connection moment to the current angular distortion, plus any permanent
angular offsets due to lack of fit or elastic unloading. Details of the development can be found in references 2 and 3.
A. Initial Loading Curves
The initial loading curves for the connection models can be represented as either a Frye-and-Morris polynomial or a
Richard-Abbott exponential. The polynomial forms must be used with care so as not to choose a connection type and
size that is beyond the extrapolation limits of the original Frye-and-Morris data. The Richard-Abbott expression, though,
provides a stable curve-fitting of a wide range of responses, from elastic-plastic to stiffness-degrading.
B. Unloading Behavior
When the moment on a connection decreases, elastic unloading is assumed. In the secant stiffness formulation used here,
this behavior can be handled in the same way that initial lack-of-fit problems are accounted for. The result is that one
need only add an “initial force” vector to the applied load vector based on the permanent angular distortion of the
connection and specify the initial elastic stiffness of the connection in evaluating the two-degree-of-freedom spring
stiffnesses.
C. Finite Joint Size
When the depths of the connected members need to be considered in the frame analysis, the member stiffness matrix can
be automatically adjusted to model rigid panel zones at the ends.
< previous page page_233 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_233.html07/11/2009 12:09:19 PM
 page_234

< previous page page_234 next page >

Page 234
Partial Composite Girders
The model for partial composite girders predicts the nonlinear responses of concrete tensile cracking and compressive
crushing, steel yielding in both compression and tension, and nonlinear slip/deformation of the shear studs bonding the
concrete to the steel W-shape. Details of the development can be found in reference (4).
A. Material Behavior
For each of the three components of the composite girder (steel I-section, concrete, and Nelson studs), realistic material
behavior is assigned according to accepted elemental force-deformation relations. The model for steel is the widely used
elastic-plastic response for individual fibers within a cross-section. The model for concrete is taken as the well-known
Hognestad continuously nonlinear response (5). The load-slip response of the shear stud connectors is that reported in
Olgaard, Slutter, and Fisher (6).
B. Cross-section Model
If one assumes that no transverse separation of the concrete and steel occurs, but that longitudinal slip of the concrete can
occur relative to the steel, then the curvatures of the concrete and steel will be identical. The strains induced by flexure

are taken as linear, but with a possible jump in the value of strain at the concrete-steel interface due to slip, .
The stress fields associated with the above strains will produce a net compressive resultant in the concrete, C, and a
tension resultant, T, in the steel, with a moment arm of z between them. This couple is then the internal reaction to the
applied moment. Compatibility, stress-strain, and equilibrium then lead to a pair of first-order, coupled ordinary
differential equations for the composite girder, as developed in detail by Yam and Chapman (7):

Slip Rate

Shear Flow
where Y is the amount of slip (e.g., inches) and Q is the force in the shear stud.
Solution of these coupled equations must be done by numerical integration techniques for any given girder and loading.
< previous page page_234 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_234.html07/11/2009 12:09:20 PM
 page_235

< previous page page_235 next page >

Page 235
C. Composite Girder Model
The solution technique used for analyzing girders that are parts of frames consists of a preprocessing step of finite
difference solution for isolated girders under gravity loads and a frame analysis that utilizes stiffness data generated in
the preprocessing step.
In the preprocessing step, the desired gravity loads are specified on a simply-supported composite girder with the given
cross-section. This girder is then analyzed for 4 to 6 levels of gravity load factor up to 1.3 or the collapse of the girder.
For each load level, the coupled differential equations are solved numerically using a shooting method of forward
numerical integration using the Euler-Cauchy finite difference equations. Details of this solution method are presented in
Zaremba (4).
Once the differential equations are solved for any given problem, one has obtained the values of curvature at all station
points along the span. These in turn can by integrated to obtain beam slopes and deflections along the span. The resulting
data is then used to establish the load-deflection data to be used in the ensuing frame analysis.
In the frame analysis, an equivalent prismatic model is used for portions of girders in positive bending, where the value
of EI is interpolated from reference responses of the prototypical partial composite girder analyzed in the preprocessing
step.
In general, the response of a composite girder can encompass a range of stiffness regimes. For example, it could start
with a large positive moment at the face of an attached column to result in a tapered end due to ‘‘necking down” of the
effective slab width from the usual effective width along the span to the width of the column flange (under positive
bending), an equivalent prismatic region in the region of positive bending, a region of steel-only in regions of negative
bending, and possibly a region of tapered steel-only when steel yielding occurs under large negative bending moments.
The model used in the frame analysis includes provisions for all of these possibilities, using a substructuring approach
similar to that used for columns under partial plastification.
NONLINEAR FRAME ANALYSIS
All of the individual models described above are organized into a plane frame analysis program that allows combined
gravity and wind loads. The loads are defined independently in terms of generic surface pressures over decks and
cladding areas. Then they are applied to the frame according to tributary areas and user-defined gravity and wind load
multiplier histories. The overall analysis proceeds in
< previous page page_235 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_235.html07/11/2009 12:09:20 PM
 page_236

< previous page page_236 next page >

Page 236
user-prescribed load-step increments, using a secant-stiffness formulation. When convergence is difficult (e.g., when
approaching the collapse load), the program can automatically backstep and re-continue with reduced load step sizes.
Results are played back through high resolution graphics sequences of displaced shapes to facilitate the identification of
modes of collapse and progressive collapse scenarios.
RESULTS
Comparison of Frame Capacity to Test
The above program was used to analyze a frame that was experimentally loaded to collapse at Lehigh University (8) in
order to corroborate its accuracy. It was necessary to include the nonlinear beam-column model and the rigid panel zone
models to simulate the behavior of the rigid frame that was tested. Comparison of results showed that the analysis
predicted a collapse load within 4% of that observed in the test results. Moreover, the sequence and pattern of cross-
section plastifications were well predicted by the analysis.
Comparison of Frame Analysis to ANSYS
At the time of program completion, no experimental results were available for flexibly-connected frames. However a
complete independent analysis check was performed by analyzing a flexibly-connected frame using the ANSYS
nonlinear finite element package (9). Good agreement was observed.
APPLICATIONS
The major use of the above described program has been as a research tool for evaluating the adequacy of frames
designed as Type 2 Construction. The most significant new finding that can be directly attributed to the program is the
unsuspected mode of failure of most Type 2 frames in the so-called “zipper mode” of collapse (10), wherein the leeward
column stack undergoes extensive plastification just prior to the onset of overall frame instability. This mode of frame
failure occurs at load factors below those associated with the classical story shearing modes presumed in “rational”
stability analysis approaches such as used in generating the effective length nomographs. Inadequacies of such design
approaches have been remedied in modified design procedures that have been proposed as a result of studying the
analysis results of both the original designs and modified re-designs (11).
< previous page page_236 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_236.html07/11/2009 12:09:21 PM
 page_237

< previous page page_237 next page >

Page 237
REFERENCES
1. AISC, Specification for the Design, Fabrication, and Erection of Structural Steel for Buildings. Effective November 1.
1978.
2. Ackroyd, M.H., ‘‘Nonlinear Inelastic Stability of Flexibly-Connected Plane Steel Frames”, PhD dissertation,
University of Colorado, Boulder, Colorado, May, 1979.
3. Connolly, J.T., “Shakedown of Flexibly-Connected Plane Steel Frames”, Master of Engineering Thesis, Rensselaer
Polytechnic Institute, Troy, N.Y., May, 1982.
4. Zaremba, C.J., “Limit State Analysis of Partial Composite Girders and Application to Building Frames”, Master of
Engineering Thesis, Rensselaer Polytechnic Institute, August, 1986.
5. Park, and Pauley, Reinforced Concrete Structures.
6. Olgaard, J., Slutter, R., and Fisher, J., “Shear Strength of Stud Connectors in Concrete”, AISC Engineering Journal,
Vol. 8, No. 2, April, 1971.
7. Yam, L.C.P. and Chapman, J.C., “The Inelastic Behavior of Simply-Supported Composite Beams of Steel and
Concrete”, Proceedings of the Institution of Civil Engineers, London, Vol. 41, December, 1968.
8. McNamee, B.M. and Lu, L.W., “Inelastic Multistory Frame Buckling”, Journal of the Structural Division, ASCE, Vol.
98, No. ST7, July, 1972.
9. Edinger, J.A., “Nonlinear Frame Analysis using the ANSYS General Purpose Computer Program”, AISC Research
Report No. 83–1, July, 1983.
10. Ackroyd, M.H., “Limit States of Type 2 Steel Building Frames”, Research Report to Amercian Iron and Steel
Institute, under Project No. 311, “Ultimate Capacity of Flexibly-Connected Steel Building Frames”, Rensselaer
Polytechnic Institute, Troy, N.Y., August, 1981.
11. Cronembold, J., “Evaluation and Design of Type 2 Steel Building Frames”, Master of Architecture Thesis,
Rensselaer Polytechnic Institute, Troy, N.Y., May, 1986.
< previous page page_237 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_237.html07/11/2009 12:09:22 PM
 page_238

< previous page page_238 next page >

Page 238
A FINITE ELEMENT FOR THE ANALYSIS OF SEMI-RIGID FRAMES
Carlo Poggi
Department of Structural Engineering
Politecnico di Milano, Milano, Italy
and
Riccardo Zandonini
Department of Engineering
University of Trento, Trento, Italy
ABSTRACT
A realistic assessment of the frame behaviour requires the influence of joint flexibility to be taken into account in the
analysis. The numerical approach highlighted in this paper makes use of the finite element technique. A beam-joint
element was formulated, in which, both joint deformation and joint finite dimensions are allowed for. The method as an
extension of a highly refined formulation, previously developed for rigidly jointed frames. The validation of the
approach has been obtained against available experimental data.
INTRODUCTION
The significant influence that joint flexibility may have on the performance of framed systems has been recognized since
the early developments of modern structural analysis. Numerical methods which include this parameter in frame analysis
were available more than fifty years ago [1] and approaches more and more refined were then developed [2]. Recent
studies have indicated that design methods recognizing the joint semi-rigid frame action may allow an improvement in
the cost effectiveness of the structure, due to a possible better balance between material and labour costs. Furthermore,
the degree of flexibility of joints, usually modelled as rigid, may in some cases non negligibly affect the frame safety
level. The interest in the semi-rigid frames as a realistic design alternative has then significantly increased. However,
studies are still necessary to thoroughly understand how joint behaviour affects frame response. This knowledge should
enable us to develop sensible design criteria and methods.
Sophisticated numerical approaches need to be set up in order to take into account all the parameters governing the
behaviour of frames with semi-rigid joints. Several methods have been developed very recently [2], which make use of
models at different degrees of refinement for simulating the joint and/or the member response.
This paper intends to highlight the basic aspects of a numerical method, in the frame of the finite element technique,
which adopts a particular formulation [3] allowing for a combination of a refined modelling of the material and
geometrical nonlinearities with computational effectiveness. In particular, the spreading of plasticity through the cross
section and along the member is taken into account. This formulation was proved to be highly accurate in the case of
steel frames with rigid joints [4,5,6].
< previous page page_238 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_238.html07/11/2009 12:09:23 PM
 page_239

< previous page page_239 next page >

Page 239
A joint-beam element was developed, able to take into account both the joint nonlinear behaviour and the joint finite
dimensions. Two different solving strategies were adopted in order to optimize the computer time with reference to the
cases of proportional and nonproportional loading paths.
Several comparisons to available experimental results are also reported and some conclusions on the accuracy of the
proposed method drawn.
JOINT-BEAM ELEMENT MODEL
A joint-beam finite element model was developed, suitable for the analysis of semi-rigid frames (Fig. 1). The main
characteristics of the model are highlighted here; detailed presentation is reported in [7] The model is conceived as an
assemblage of three different parts:
– a central nonlinear beam-column element
– two connection elements
– two rigid bars

Fig. 1 a) Joint-beam element.

b) control points over the central beam element.
Central nonlinear beam element
The material costitutive law, characterized by a bilinear behaviour with cyclic hardening, is described by the following
relations:

(1a,b)

(1c)

(1d-g)
where:

< previous page page_239 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_239.html07/11/2009 12:09:24 PM
 page_240

< previous page page_240 next page >

Page 240
E is the Young’s modulus of the material and σy its yield stress, pl is the plastic part of the total strain (expressed as the
difference of two non-negative and non-decreasing plastic multipliers), εo is the thermal part of the total strain and σrl is
the value of residual stress, hl is an hardening matrix, K is a value related to the slope of the hardening branch and factor
β is a constant to define the type of hardening rule and may vary from 1 (kinematic hardening) to −1 (isotropic
hardening). Vectors rl and nl are constants and λl and Φl are the vectors of plastic multipliers and yield functions. For
simplicity, but without lack of generality, in the following the material is supposed to be elastic perfectly plastic, without
thermal strains and residual stresses (this means ) The central beam element is supposed to be of
constant cross section. Its centroidal axis is denoted by x and symmetry about z axis is assumed (Fig. 1). If the shear
effects are neglected, as in the classical beam theory, the elongation δ(x) and the curvature χ(x) of the beam may be
assumed as generalized strain components governing the deformability of the cross section x. The axial force N(x) and
the bending moment M(x) are the corresponding generalized stresses. Let vectors q(x) and Q(x) be defined as follows:

(2a,b)
where:

(3a,b)
The strain in the x direction, denoted by ε, is assumed to vary linearly with z, so that:

(4)
The model is based on the independent modelling of the displacements (u(x), v(x)) and of the plastic multipliers.

The displacements u(x) and v(x) (collected in vector ub(x)) are modelled as functions of the nodal parameters u by
means of the following relation:

(5)
where Ψ(x) is a matrix of shape functions.
Note that, if all the springs are active, only the first seven of the thirteen degrees of freedom of vector u (Fig. 1a) are
necessary to completely define the displacement field. Otherwise, if several springs are fixed, a slave condition on the
relevant displacements is imposed and the relation is modified by means of a condensation process. This was shown to
be necessary to avoid any ill conditioning of the element stiffness matrix due to the introduction of fictitiously high
spring stiffnesses.
Once eqs. (5) are defined, the strain distribution in the element is given by eqs (4) and (3). A constant distribution of δ(x)
and a parabolic one for the curvature χ(x) is obtained because of the introduction of parameter u7. It is an additional
displacement parameter associated to a fourth order function that permits the use of a limited number of element even in
the presence of geometric nonlinearities [4,5].
< previous page page_240 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_240.html07/11/2009 12:09:26 PM
 page_241

< previous page page_241 next page >

Page 241
The plastic multipliers are modelled, indipendently of the displacements, on the base of their values at an established
number of points (xj,zh) within the element and over the cross sections (Fig. 1b).
To obtain an accurate modelling of the plastic zones it is often necessary that the distribution of plastic strains (p(x,z)) be
controlled at a large number of points. This would produce a lack of compatibility between plastic and total strain
distributions that causes self equilibrated stresses in the isolated element subject to plastic strains only. The procedure
proposed in [3] is adopted to avoid the presence of these spurious terms. This is equivalent to an assumption for the
distribution of N(x) and M(x) along the element that be consistent with the assumed displacement model. For this reason
N(x) is constrained to be constant and M(x) to vary parabolically along x. The coordinates of the control points are those
used to calculate an integral over the element volume by means of Gaussian Quadrature Rules. For doubly symmetric
cross sections the values of zh are reported in [4].
The above described relations together with eqs.1 are combined to obtain the beam elastic-plastic set of relations:

(6a,b)

(6c)

(6d,g)
The expressions of matrices db, nb, hb and rb may be found in [7].
Connection elements
Three springs are located at each end of the beam to simulate the actual behaviour of the joints.
The piecewise-linearized cyclic costitutive laws governing the elastic-plastic spring behaviour are described by the
following relations formally identical to the beam ones:

(7a,b)

(7c)

(7d,g)
Where Qs and qs are vectors collecting the axial, shear and bending stresses and strains of the springs, ds is a symmetric,
positive definite matrix collecting the elastic spring stiffnesses that may also take into account possible coupling of
springs, and hs is a symmetric hardening matrix. Without lack of generality, Koiter’s hardening rule was adopted so that
hs becomes a diagonal matrix.
Rigid bars
Two rigid bars are located at the ends of the beam to account for the actual location of the joints.
< previous page page_241 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_241.html07/11/2009 12:09:27 PM
 page_242

< previous page page_242 next page >

Page 242

Fig. 2 Spring cyclic costitutive law.

Frame problem
After a usual assemblage procedure, the entire system relations may be obtained:

(8a)

(8b)

(8c,f)
Where u and F are the assembled vectors of nodal displacements and forces, λ is a vector of plastic multipliers; Φ is a
vector of plastic potentials of the same dimension of λ . R is a vector of positive constants, Kuu is the usual elastic

stiffness matrix while and Kλλ are constant matrices. Such a set of equations is linear in the free parameters
u and λ and must be respective to the sign restrictions and to the complementarity conditions (eqs. 8c,f). The problem
may be solved by means of well known algorithms of mathematical programming or, as an alternative, following an
incremental procedure. Two different incremental solving strategies were adopted. In the case of proportional loading the
step amplitude is automatically calculated by the program while in the presence of non proportional or cyclic loading
paths the step amplitude is imposed by the user.
In the case of non negligible geometrical nonlinearities the stiffness matrix is modified by the use of a geometric
stiffness matrix [5]. The structure of the relations remains the same and the above mentioned solution procedures are still
adequate but some iterations must be introduced in the solution of each step.
< previous page page_242 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_242.html07/11/2009 12:09:28 PM
 page_243

< previous page page_243 next page >

Page 243
COMPARISON WITH EXPERIMENTAL RESULTS
The finite element approach to the nonlinear analysis of semi-rigid frames is an extension of a general formulation
already validated [4,5,6] for the case of steel frames with rigid joints. Its verification against experimental results is
nevertheless necessary in order to check the degree of accuracy of the model adopted for simulating the joint behaviour.
Tests aimed at pointing out the semi-rigid frame action in the inelastic range were conducted recently and represent a
very useful benchmark for the validation of a numerical method. In fact, data related to a series of stability tests carried
out on column subassemblages under nonproportional loading [8], and to experiments conducted on a limited number of
sway frames subject to cyclic lateral forces [9], allow a thorough check of the proposed formulation. The former set of
experiments makes it possible to verify the capability of the method to predict the structural response, with reference to
the complex interaction between column, beams and joints and to its influence on column stability. The latter
experiments enable a verification of the assumed joint cyclic law.
The geometrical configuration and the general loading and restraint conditions of the column subassemblages tested by
Davison and Nethercot are reported in figure 3. The same column and beam sections were used throughout the series.
Tests differed for the type of beam to column connection, the loading and loading history and the axis of column
bending. Instrumentation of the subassemblage allowed to measure, or to determine, the governing mechanical and
geometrical parameters.
COLUMN AXIAL FORCE N AT COLLAPSE (KN)
TEST BEND.* CONN.+ N N ERROR %
AXIS num exp
ST2 S WC 665.10 675.87 −1.59 %
ST3 W WC 556.90 519.91 7.11 %
ST4 S FC 735.80 762.00 −3.44 %
ST7 W FC 481.00 526.40 −8.62 %
ST8 W WSC 491.40 518.25 −5.18 %
ST9 W FEP 484.60 486.49 −0.39 %
*: S=STRONG AXIS W=WEAK AXIS
+: WC=WEB CLEATS FC=FLANGE CLEATS
WSC=WEB AND SEAT CLEATS
FEP=FLUSH END PLATE
fig. 3.a

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_243.html (1 of 2)07/11/2009 12:09:29 PM

 page_243

fig. 3.b
< previous page page_243 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_243.html (2 of 2)07/11/2009 12:09:29 PM

page_244

< previous page page_244 next page >

Page 244

fig. 3.c
COLUMN SUBASSEMBLAGES AFTER [8]

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_244.html (1 of 2)07/11/2009 12:09:30 PM

 page_244

fig. 3.d
< previous page page_244 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_244.html (2 of 2)07/11/2009 12:09:30 PM

page_245

< previous page page_245 next page >

Page 245

TEST No 10 AFTER [9]

Fig. 4

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_245.html (1 of 2)07/11/2009 12:09:31 PM

page_245

< previous page page_245 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_245.html (2 of 2)07/11/2009 12:09:31 PM

 page_246

< previous page page_246 next page >

Page 246
These data were made available to the authors, in the frame of a joint research programme, and a thorough comparison
between numerical and experimental results was carried out and is reported in [10]. Some results are presented in figure
3 where also the mesh adopted in the analysis is plotted. The moment rotation curves determined in Sheffield in a
previous series of connection tests were adopted.
The table of figure 3a gathers the values of the ultimate axial load N in the column. The numerical analysis seems to
predict the ultimate column strength with satisfactory accuracy; the errors are in all cases lower than 9% in absolute
value. The suitability of the simulation method to predict the subassemblage response is confirmed by the comparisons
related to midheight deflection of the column (fig. 3c) and the variation of the moments at the top end of the column (fig.
3d). The good agreement between the numerical and experimental analysis through all the loading process is apparent,
pointing out the capability of the method of evaluating the overall stiffness and the stress state of the system in both the
elastic and inelastic ranges. Uncertainties on actual joint behaviour and on possible eccentricities of the applied axial
load can well explain the differences.
The constitutive law assumed for the connection spring element, based on the Koiter’s hardening rule, do not account for
the possible cyclic degradation of the joint elastic stiffness. A check against the results of a cyclic test by Stelmack [9]
was then made. The connection behaviour was represented by a non symmetrical law, based on the experimental
measured response. Some results are presented in figure 4 related to the lateral frame drift and the performance of the top
right connection. The effect of neglecting the stiffness degradation is apparent in both plots. However, the simulated
frame response shows a satisfactory agreement with the test behaviour, although frame deflections and joint rotation may
be underevaluated also significantly. Anyway, it should be considered that the check is severe because the stiffness
degrading is particularly substantial for the top and seat angle connections used in the test. These results seem to confirm
that the assumed joint law, though simple, may be sufficiently accurate also for cyclic analysis.
CONCLUSIONS
A sophisticated and computationally effective method of analysis was presented and discussed. The approach permits an
inclusion of the joint nonlinear behaviour and the joint finite dimensions in the analysis, besides the spreading of plastic
zones in the members and the geometrical effects.
The validity of the method was checked against experimental results and its adequacy to predict the response of semi-
rigid frames was pointed out. Possible limits to the present formulation with regard to the cyclic modelling of the joint
behaviour are also highlighted.
ACKNOWLEDGEMENTS
This research has been made possible by a grant from the Italian Ministry of Education (M.P.I.).
< previous page page_246 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_246.html07/11/2009 12:09:32 PM
 page_247

< previous page page_247 next page >

Page 247
REFERENCES
1. Pippard A.J.S. and Baker J.F., The Analysis of Engineering Structures, 3rd Edition, Arnold, London, 1936.
2. Anderson D., Bijlaard F.S.K., Nethercot D.A. and Zandonini R., Analysis and Design of Steel Frames with Semi-
Rigid Connections, IABSE Survey, to appear.
3. Corradi L., A Displacement Formulation for the Finite Element Elasto-Plastic Problem, Meccanica, vol. 18, 1983, pp.
77–91
4. Corradi L. and Poggi C., A Refined Finite Element Model for the Analysis of Elastic Plastic Frames, Int. J. of Num.
Methods in Engng., vol. 20, 1984, pp. 2155–2174.
5. Corradi L. and Poggi C., An Analysis Procedure for Nonlinear Elastic-Plastic Frames Accounting for the Spreading of
Local Plasticity, Costruzioni Metalliche, No. 1, 1985, pp. 1–14.
6. Poggi C. and Zandonini R., Behaviour and Strength of Steel Frames with Semi-Rigid Connections, Proc. Session on
Connection Flexibility and Steel Frames, ASCE, Annual Convention, Detroit, October 1985.
7. Poggi C., A Finite Element Model for the Analysis of Flexibly Connected Steel Frames, submitted for publication to
Meccanica.
8. Davison B., Kirby P.A. and Nethercot D.A., Column Behaviour in P.R. Construction—Experimental Behaviour,
ASCE Structures Congress, New Orleans, September 1986.
9. Stelmack T.W., Analytical and Experimental Response of Flexibly Connected Steel Frames, Master Thesis, Report to
AISI, CEAE Dept., University of Colorado, Boulder, 1983.
10. Davison B., Nethercot D.A., Poggi C. and Zandonini R., Analytical and Experimental Studies of Semirigidly
Connected Steel Frames, to appear.
< previous page page_247 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_247.html07/11/2009 12:09:33 PM
 page_248

< previous page page_248 next page >

Page 248
DISCUSSION OF FRAME ANALYSIS PAPERS
It is evident that the methods of frame analysis have advanced to the stage where all significant parameters can be taken
into account, although the computer programs are still mostly research tools. This was addressed by several speakers and
discussers, and it would appear that the development of additional design-oriented programs is critical in the acceptance
and use by the profession of the principles of frames with semi-rigid connections.
Second-order geometric effects (P-Delta) were considered by all of the frame analysis methods that were presented. It
was noted that the magnitude of the P-Delta contribution for many frames can run as high as 20 % when end-plate
connections are used; it may be substantially higher when more flexible connections are used. However, the designers
that have made use of semi-rigid concepts in frame design reported that it was still possible to achieve economical
solutions, but the frame drift is almost always the controlling factor.
A separate report was presented on a comparative study of various frame analysis approaches that are currently in use in
Europe. Despite differences in modeling, such as in the way the global response of the connection was treated, the use of
discrete plastic hinges vs.gradual spread of plastification, and so on, the results of the various programs for several types
of frames correlate very well. This would seem to indicate that as long as first- and second-order material and geometric
effects for the members and the frame are taken into account, along with a connection element that reflects a realistic M-
Φ-relationship, the final results will not differ appreciably. For the moment-rotation relationship it may even be
acceptable to use a very simple bi-linear form, as long as it is not unconservative relative to the real form (for example,
the initial stiffness value should be close to that of the actual connection).
Numerous points were raised as regards the use of composite girders in the frame, and how to incorporate partial
composite action, composite connection effects, reinforcing steel in the slab and around the column, and so on. For
example, it was pointed out that research leading to the development of the composite girder criteria of Eurocode EC-4
had demonstrated that for long spans (up to 20 meters (67 feet)) with a low degree of shear connection, significant
deflection problems could be encountered. However, it was also noted that the increased stiffness of
< previous page page_248 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_248.html07/11/2009 12:09:33 PM
 page_249

< previous page page_249 next page >

Page 249
a composite girder generally leads to the use of a smaller steel shape, with the result that a shallower, more flexible beam-
to-column connection is utilized. Thus, the expected benefits of using the increased stiffness of composite girders may
not be as substantial as originally thought, but it is also clear that further research work needs to address these questions.
Several discussers noted the importance of the increased connection stiffness that results when slab continuity is
achieved (through the use of reinforcing steel bars). In the same way, it was reported that current studies of composite
connection effects have shown that connections with very low stiffness in bare steel frames (double web angles, for
example) are capable of providing significant local restraint when composite action is accounted for. This is clearly an
area where additional research is needed.
< previous page page_249 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_249.html07/11/2009 12:09:34 PM
 page_250

< previous page page_250 next page >

Page 250
This page intentionally left blank.
< previous page page_250 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_250.html07/11/2009 12:09:35 PM
page_251

< previous page page_251 next page >

Page 251
SESSION No 4
FRAME STABILITY AND SIMPLIFIED METHODS
< previous page page_251 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_251.html07/11/2009 12:09:36 PM
 page_252

< previous page page_252 next page >

Page 252
COLUMN BASE PLATE CONNECTIONS
A.Picard and D.Beaulieu
Département de Génie civil
Université Laval, Ste-Foy
Québec, Canada G1K 7P4
ABSTRACT
This paper summarizes the results of experimental investigations on the behaviour of steel column base connections. The
connection is made of a steel base plate welded to the lower end of the column and attached to the footing by means of
two or four anchor bolts. It is generally admitted in analysis that this type of connection behaves like a hinge. However,
two totally different series of tests indicate that the base connection behaves more like a fixed connection.
INTRODUCTION
It is common practice in North America to consider column base connections as either pinned or fixed to the foundation
in calculations. For a perfectly fixed base connection the flexibility ratio at the column base (GL) is equal to 0.0, that is
the concrete footing is infinitely stiff as compared to the column [G=Σ (column/beam) rigidities]. The flexibility ratio is
equal to infinity for a perfect hinged connection. In the Canadian Standard S16.1-M84 [1], the recommended values are
GL=1.0 instead of 0.0 for a fixed connection to account for the difficulty of obtaining perfect fixity, and GL=10.0
instead of infinity for a hinged connection to account for the fact that there always exists a certain amount of rotational
restraint in a hinged connection.
The column effective length factor (K) used in the evaluation of the column capacity is obtained from appropriate
alignment charts. The theoretical relationship between the effective length factor and the nondimensionalized rigidity of
the base connection is shown in Fig. 1 for a column that is pinned at the top. The nondimensionalized rigidity (R) is the
inverse of GL which is a function of the rotational flexibility of the base connection (λ: flexibility factor). It is important
to point out that the curve drops steeply for low values of R, which means that near fixity conditions are obtained at
relatively low values of column base connection rigidity.
< previous page page_252 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_252.html07/11/2009 12:09:37 PM
 page_253

< previous page page_253 next page >

Page 253

Fig. 1—Effective length factor as a function of the rigidity of the column base connection.
To evaluate GL for a partial base restraint the following theoretical equation, which applies to non-sway conditions was
suggested [2]:

(1)
In this equation, θ/M, which is the inverse of the slope of the moment-rotation curve, represents the rotational flexibility
of the base connection (λ) and can be obtained experimentally. The rotation (θ) is the sum of the rotation between the
footing and the surrounding soil and the rotation between the column base and the footing. When the column is free to
sway, the factor 2 is replaced by 6 in Eq. (1). When and the GL value obtained from Eq. (1) relates to
the column weak axis (GLY). The GLX value is computed with and
The objective of this paper is to present the results of an experimental research project dealing with the behaviour of a
simple standard column base connection frequently used in practice. The project was divided in two parts. Buckling tests
were carried out on I section columns to check the adequacy of Eq. (1) and to study the stabilizing action of the base
connection [3]. Flexion tests and compression—flexion tests were performed on column stubs to study the behaviour of
the connection itself and to
< previous page page_253 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_253.html07/11/2009 12:09:38 PM
 page_254

< previous page page_254 next page >

Page 254
determine the influence of the axial compression load on the rigidity of the base connection [4, 5].
BUCKLING TESTS
Experimental Program and Procedure
Nineteen buckling tests were conducted on I section columns of 5.5 m in length, hinged at the top and connected to the
footing as shown in Fig. 2. In two of these tests, a special set-up was designed to simulate the rotation of the footing on
the soil. By means of a centrally located roller and calibrated springs placed at the four corners of the concrete footing, it
was possible to simulate the rotation of the footing on loose sand which is considered as poor soil condition. The main
variable of the tests was the imposed differential displacement of column ends which were applied about the x or y axis.

Fig. 2—Typical column base connection

The rotation of the steel column near the base plate was measured with LVDTs and the bending moment with electrical
strain gauges. Lever arms welded to the lower end of the column were used to amplify the displacements measured by
the LVDTs [3]. The rotation of the column was easily calculated from these measured displacements. In some tests, the
rotation and the bending moment were also measured at the upper end of the column to evaluate the small restraint
offered by the upper pinned connection. This small restraint was taken into account in the analysis of the test results.
Column initial curvatures were carefully measured before each test. Every significant parameter such as applied axial
loads, moments and rotations at column ends, lateral deflections at mid-height, and rotations of the footing were
recorded during the tests. The electronic data acquisition system was capable of recording up to twenty different
measurements in less than forty milliseconds.
Buckling Tests Results
In these tests, all columns buckled about the weak axis. From a practical point of view, the main conclusions that can be
drawn from these tests may be summarized as follows.
(a) For weak axis bending, the measured rotational flexibility at the column base ranges from
to .
< previous page page_254 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_254.html07/11/2009 12:09:39 PM
 page_255

< previous page page_255 next page >

Page 255
With these minimum and maximum values of the flexibility factor, Eq. (1) gives . The
effective length factor is then equal to: 0.71≤Ky≤0.76. As shown in Fig. 1, these values are
lower than the recommanded values in Standard S16.1-M84 for a fixed connection
(b) Since the base connection behaved more like a fixed connection that a hinged connection, the measured buckling
loads largely exceeded the theoretical buckling loads obtained for a pinned base column. The deformed shape of the
buckled columns revealed the presence of a significant rotational restraint at the column base.
(c) For the tests where λ values were measured at both ends of the column, the measured buckling loads were 5.8% to
17.2% larger than the predicted buckling loads based on the measured λ values. It seems therefore that Eq. (1) used to
evaluate GUY and GLY gives acceptable results. It should be noted that in practical cases where the connection at the
top of the column is considered as pinned, the rotational restraint of the top connection is larger than the rotational
restraint of the mechanical device used for the tests.
(d) The moment-rotation curves of the column base plate connections, for weak axis bending, measured during the
buckling tests, show that the connection behaves elastically and the rigidity is constant as long as the base plate remains
in full contact with the footing. A most interesting observation is that separation is initiated only after buckling when the
bending moment becomes large with respect to the axial load. This was the fact in all cases even when large relative
displacements of the column ends were imposed.

Fig. 3—Typical moment-rotation curve measured at the column base.

< previous page page_255 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_255.html07/11/2009 12:09:41 PM
 page_256

< previous page page_256 next page >

Page 256
This conclusion is illustrated in Fig. 3 which shows one of the moment-rotation curves measured during the buckling
tests. The moment developped in the base connection at buckling (Mcr) was smaller than the linear limit (MLin) in all
tests. Consequently the connection was behaving elastically at buckling. Moreover, since the base plate remained in
contact with the footing, the presence of more anchor bolts should have a limited influence on the behaviour of the
connection at buckling.
Another interesting observation can be made from Fig. 3. At ultimate, the base connection developed 78% of the plastic
moment capacity of the column about the weak axis . This is quite a large moment for a
generally considered pinned connection. It seems therefore appropriate to provide a welded connection capable of
developing Mpy between the base plate and the column.
(e) Buckling tests with a special set-up designed to simulate the rotation of the footing on the soil, have shown that no
significant rotation of the footing occurred even though very poor soil conditions were simulated. The moment
developed in the column base connection at buckling and even after was too small to force the footing to rotate.
LOCAL BEHAVIOUR OF THE CONNECTION
Experimental Programs
Two series of tests were carried out on column stubs to study the local behaviour of the base plate connection [4, 5]. In
this short paper, it is not possible to present the full extent of the experimental programs and procedures. The reader is
therefore referred to the published papers. The presentation will be limited to a brief description of some aspects of the
experimental programs.
The main purpose of the first series of tests was to determine the fixity factor of the base connection for strong axis
bending (γx) and to study the influence of the axial compression load on the fixity factor [4]. The measured moment-
rotation curves were used to determine γx.
As explained later, the fixity factor is used for linear analysis of flexibly connected steel frames. The fixity factor of a
connection is given by [6]:

(2)
When considering column base plate connection, λo is the rotational flexibility factor at the base plate level. The fixity
factor varies from zero for a perfectly pinned connection to unity for a perfectly rigid connection
. When and the γ value obtained from Eq. (2) relates to column strong axis (γx). In the
first series of tests, λox was obtained experimentally with and without axial load acting on the column. The experimental
program comprised 15 specimens tested to failure: 7 flexion tests and 8 compression-flexion tests performed on 1220
mm long column stubs.
< previous page page_256 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_256.html07/11/2009 12:09:42 PM
 page_257

< previous page page_257 next page >

Page 257
The main purpose of the second series of tests was to determine the influence of the axial compression load on the value
of GL obtained from Eq. (1). Since the buckling tests have shown that the base connection behaves elastically when
buckling occurs, the moment-rotation curves were measured in the elastic range. Each column was tested at six, seven or
eight axial load levels and was bent successively about both principal axes. More than 150 moment-rotation curves were
obtained from tests performed on 2000 mm long column stubs [5].
Main Conclusions of the Tests
The results of the first series of tests have clearly shown that the flexural stiffness of the base connection is a function of
the applied axial compression load in the column. A significant increase of the flexural stiffness was observed when a
compression force was applied to the column. The test results also indicated that the fixity factor of the base connection
tends towards a limit value when the axial load in the column increases. It was found that for
where C is the axial compression load in the column and Cy is the nominal axial plastic
capacity of the column (Cy=AFy where A is the area of the cross-section and Fy the yield stress).

Fig. 4—Variation of GL with axial load.

< previous page page_257 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_257.html07/11/2009 12:09:43 PM
 page_258

< previous page page_258 next page >

Page 258
The axial compression load has also an effect on the slope of the moment-rotation curve. When the base connections
were not subjected to an axial load, the measured moment-rotation curves were nonlinear almost from the start of
loading. The slope of the M-θ curves decreased continuously with the applied moment. In that case the rotational
flexibility factor (λ=θ/M) and the fixity factor (γ) are not constant. With an axial compression load in the column the
measured moment-rotation curves were linear at least up to 70% of the maximum applied moment. In that case, the
flexibility factor and the fixity factor are constant for a large loading range.
Most of the results of the second series of tests are summarized in Table 1 and in Fig. 4. The mean values and standard
deviations for GL, evaluated at different axial load levels for weak axis or strong axis bending, are given in Table 1.
Each mean value given in that table [GL (50%)] is the average of five measurements and the standard deviations (S)
were computed with the unbiased formula. The GL values having only a 5% chance of being exceeded, assuming a
normal distribution, are also given in the table.
TABLE 1
Statistical parameters of experimental GL values
C/Cy GLX(50%) S GLX(5%) GLY(50%) S GLY(5%)
(%) Test series A (W 150x30 column section)
17.6 1.37 0.19 1.68 0.32 0.07 0.44
22.0 1.14 0.23 1.52 - - -
26.4 0.98 0.21 1.33 0.24 0.05 0.32
30.8 - - - 0.23 0.02 0.26
35.2 0.76 0.11 0.94 0.20 0.03 0.25
39.6 0.70 0.10 0.86 - - -
Test series B (W 200x36 column section)
14.6 2.62 0.39 3.26 0.33 0.09 0.48
21.8 1.80 0.26 2.23 0.25 0.05 0.33
29.1 1.33 0.15 1.58 0.20 0.04 0.27
36.4 1.09 0.19 1.40 0.19 0.02 0.22
40.0 0.94 0.07 1.06 0.18 0.03 0.23
Fig. 4 shows that GL at first decreases rapidly with the axial load but tends towards a limit value. For weak axis bending,
the value of GL is almost constant when the axial load is larger than about 30% of the axial plastic capacity. In practical
situations, the column buckling load is larger than 0.30 Cy.
The base connection of the type studied is generally considered as pinned and the recomemnded GL value is 10.0 [1].
All GL values given in Table 1 are much lower than 10. The test results indicate that for weak
< previous page page_258 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_258.html07/11/2009 12:09:44 PM
 page_259

< previous page page_259 next page >

Page 259
axis buckling a conservative value is and for strong axis buckling if there is no relative
displacement of column ends (sway-prevented condition). In most practical situations, sway is prevented when
considering weak axis buckling, therefore According to Eq. (1), where the factor 2 is replaced by 6 when
the upper end of the column is free to sway, and for the sway-permitted condition.
The effective length of the column is reduced when the flexural stiffness of the base connection is taken into account.
Design examples show that the increase in column strength varies between 10% and 30% in current situations
(slenderness ratios between 60 and 120).
FRAME ANALYSIS
It is well known that connection flexibility affects the overall behaviour of a frame. As the connection flexibility
increases, the frame response changes from shear-type behaviour to flexural cantilever behaviour. Two approaches can
be used to analyse flexibly connected steel frames: the linear and nonlinear approaches.
The linear approach is well suited for current office practice since it requires only minor modifications of the commonly
used displacement or stiffness method. This approach is satisfactory if the rotational flexibility of the connections is
almost constant from the start of loading up to the bending moment which corresponds to factored loads. In our opinion,
the simplest linear method of analysis is obtained by inclusion of fixity factors at the member ends. The member
stiffness matrices are then modified by a correction matrix which includes the fixity factors [6].

Fig. 5—Linearisation of connection behaviour.

< previous page page_259 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_259.html07/11/2009 12:09:46 PM
 page_260

< previous page page_260 next page >

Page 260
The nonlinear approach is used when the behaviour of the connections of a frame is inelastic and nonlinear almost from
the start of loading. It should also be used if the ultimate strength of the frame and the conditions at failure are to be
determined. In this latter case, three different nonlinear effects should be considered: connection nonlinearity, member
plastification and geometric nonlinearity leading to member and frame instability.
Since nonlinear inelastic analyses are too arduous for most designers and unsuitable for design office practice, some
representative linearisation of the nonlinear connection behaviour should be adopted for design calculations. To linearise
the connection behaviour, the secant rotational flexibility can be used, as shown in Fig. 5. The secant is a straight line
from the origin to the point on the curve corresponding to the design moment (kMf).
The following design procedure is suggested. The structure is first analysed assuming perfectly rigid connections to
obtain the maximum moments in the connections. Since the connections are partially rigid, these moments will never be
reached. Then the designer selects a percentage of the maximum moment to design the connections, for instance 75%
. As shown in Fig. 5, the secant rotational flexibility corresponding to the design moment is obtained from
the M-θ curve and is used to compute the fixity factor of the connection. Finally the fixity factors of the connections are
introduced in a computer program, which incorporates the correction matrices [7], and the structure is reanalysed to
obtain new bending moments.

Fig. 6—Linear analysis of a simple frame.

< previous page page_260 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_260.html07/11/2009 12:09:47 PM
 page_261

< previous page page_261 next page >

Page 261
If the bending moment in each individual connection is smaller than the design bending moment (kMf), the design is
satisfactory. If the bending moment is larger, the value of k should be increased. On the other hand, if the bending
moment is much lower than the design moment, the value of k should be decreased. Of course, this design procedure
assumes that sufficient experimental information is available for the type of connection used in the structure.
Fig. 6 shows an example of linear analyses considering fixity factors. The beam-to-column connections are assumed
perfectly rigid but the column base fixity factor varies between 0.0 and 1.0. For the selected frame, the
first-order horizontal displacement (∆) may be expressed in terms of the applied horizontal load (H), taking into account
frame geometry and cross-section properties of beams and columns.
If it is assumed that the column base connection is a perfect hinge, the horizontal displacement is equal to 0.7 times the
horizontal load. If the fixity factor is equal to 0.25, the horizontal displacement is reduced by 43%, when compared to
the displacement obtained neglecting base fixity. Tests have shown that is quite a conservative value for a
simple column base plate connection [4].
It is important to point out that the curve drops steeply for low values of the fixity factor. In other words, the fixity factor
does not have to be high in order to reduce the lateral displacement and the corresponding sway effects in a significant
manner. The advantages of considering the actual flexural stiffness of the column base connection is clearly shown by
the results of the analysis of this simple frame.
CONCLUSION
The test results summarized in this paper indicate that the flexural stiffness about both principal axes of a simple column
base plate connection has a very beneficial effect on column stability and frame behaviour. With these results, the effect
of base restraint can be accounted for in design.
REFERENCES
1. Canadian Standards Association, Steel Structures for Building (Limit States Design). CAN3-S16.1-M84. Rexdale,
Ontario, 1984.
2. Nixon, C.D., The Design of Light Industrial Building. Ph.D. thesis, Dept. of Civil Eng., University of Alberta,
Edmonton, Alberta, 1979.
3. Beaulieu, D. and Picard, A., Contribution des assemblages avec plaque d’assise à la stabilité des poteaux.
Construction Métallique, 1985, no. 2, pp. 3–19.
4. Picard, A. and Beaulieu D., Behaviour of a simple column base connection. Can. Journal of Civil Eng., 1985, vol. 12,
no. 1, pp. 126–136.
< previous page page_261 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_261.html07/11/2009 12:09:48 PM
 page_262

< previous page page_262 next page >

Page 262
5. Picard, A., Beaulieu, D. and Pérusse, B., Rotational Restraint of a Simple Column Base Connection. Can. Journal of
Civil Eng., 1987, vol. 14, no. 1, pp. 49–57.
6. Monforton, G.R. and Wu, T.S., Matrix Analysis of Semi-Rigidly Connected frames. ASCE Journal of the Structural
Division, 1963, vol. 89, no. 6, pp. 13–42.
7. Picard, A. and Beaulieu, D., SUPREM: logiciel d’analyse des structures planes. Rapport GCT-85–09, Génie civil,
Université Laval, Québec, 1985 (rev. 1986).
< previous page page_262 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_262.html07/11/2009 12:09:49 PM
 page_263

< previous page page_263 next page >

Page 263
SIMPLIFIED DESIGN APPROACH FOR END-RESTRAINED COLUMNS IN FRAMES
Reidar Bjorhovde
Professor and Chairman
Department of Civil Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania 15261
U.S.A.
ABSTRACT
Recent studies of the strength, stability and design of columns have focused on the need for giving a more accurate
representation of such members as parts of realistic frames. In particular, the research has addressed the influence of the
connections between the columns and the adjacent parts of the structure, recognizing the stiffness characteristics of the
semi-rigid beam-to-column connections. The paper details one of the approaches that may be used to determine the
buckling load of columns in frames, by taking into account the initial stiffness of the beam-to-column assemblage. It is
demonstrated that this property is crucial to the response and capacity of the member. To facilitate the incorporation of
the method into current design procedures, use is made of the effective length concept, modified to reflect the actual
restraints at the column ends. Finally, it is noted that the solution is equally applicable in allowable stress and limit states
design formulations.
INTRODUCTION
Research and development studies dealing with the strength and behavior of columns have been conducted for many
years. Although the basic problem may appear to be relatively simple, the many factors that influence column stability
make the topic a highly complex one. In addition, solutions that apply to the individual column have limited applicability
to members that are parts of structures. This is due to the interaction of all of the framing members and the overall
response of the frame. For example, second order frame effects can be significant in a number of cases.
The original solutions for compression members dealt with pinned-end, perfectly straight columns with elastic behavior,
which were subsequently modified to take into account material and cross-sectional non-linear effects. Thus, the works
of Euler, Engesser, Considere. Shanley, and Dutheil, among many others [1], are well known, and need not be reiterated
here. Similarly, the advent of the computer made possible the incremental, iterative numerical solutions for the strength
of members with
< previous page page_263 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_263.html07/11/2009 12:09:49 PM
 page_264

< previous page page_264 next page >

Page 264
imperfections in the form of residual stress and initial crookedness. This took place much later than the above
investigations, but the findings have now been incorporated into design standards around the world [1, 2, 3, 4].
Over the past several years it was recognized that although the strength and behavior of the pinned-end column was an
important and convenient reference case, somehow the influence of the actual support conditions of the member should
be accounted for. At the outset it was felt that the restraining effects of the connections between the column and the other
components of the surrounding structure would be likely to raise the actual capacity of the member. Although the
connections would respond in a manner that depended on their rotational stiffness, it was deemed realistic to expect that
buckling could be delayed to a certain degree, even in the case of the most flexible beam-to-column joints.
This paper will provide a brief review of some of the investigations that have examined the capacity of columns with
semi-rigid end connections. The primary strength parameters are identified, along with the development of a design
procedure that allows the engineer to incorporate semi-rigid concepts into current methods of analysis.
INDIVIDUAL COLUMNS WITH END RESTRAINT
Before examining the behavior and strength of end-restrained columns in frames, it is important to understand the
response of the individual member when a measure of restraint is present at the ends. Specifically, this clarifies the
parameters that influence the capacity of the column, as follows:
(1) Restraint characteristics of the column end supports
(2) Column length or slenderness ratio
(3) Material and geometric imperfections
In the following the influence of these parameters will be explained in some detail.
Restraint Characteristics of the Column End Supports
For most practical purposes this identifies the influence of the beam-to-column connection that joins the column to the
adjacent structural elements. The most significant measure of its restraint effect is the moment-rotation or -curve,
and Fig. 1 shows representative samples for some typical structural connections. In particular, the C-measure, as
indicated in the figure, which reflects the initial rotational restraint of the connection, is important. This will be described
in some detail later; at this point it is sufficient to note that the connection stiffness provides the actual column buckling
restraint, along with that of the beam to which it is also connected.
A perfect pin has a C-value of 0 and a rigid connection provides infinite restraint. All real beam-to-column joints lie
somewhere between
< previous page page_264 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_264.html07/11/2009 12:09:50 PM
 page_265

< previous page page_265 next page >

Page 265
these extremes, with the simple or shear-type connections offering a very low moment capacity and stiffness. However,
it has been shown that even such connections can lend significant additional buckling strength to end-restrained columns
[5, 6].

Figure 1 Some Typical Moment-Rotation Curves for Beam-to-Column Connections

Additional information regarding the influence of the connection stiffness on the column buckling strength is provided
by Fig. 2. Using data from a study of Nethercot et al. [7], the figure shows the load-deflection curves that result when the
same column shape is used with three different end connections. Specifically, the curves represent the response of a W-
shape column with pinned ends, double angle connections, and top-and-seat angle connections. It is noted that the
column in question had a slenderness ratio of 120, and an initial out-of-straightness of L/1000, where L is the length of
the member. The large slenderness ratio is the primary cause of the magnitude of the strength increases, as will be
discussed in the following section.
Influence of Column Length
The column strength studies have demonstrated clearly that the longer the member, the larger the relative strength
increase that is created by having a semi-rigid end connection. In other words, for one and the same cross section and
end restraint, the longer the member, the larger will be the percentage increase in the maximum capacity. Examining
various types of beam-to-column connections, researchers have reported strength increases of more than 200 percent for
certain very long columns [5, 8]. However, it is noted that these lengths are well in excess of practical values, as has
been shown by other findings [6, 9]. As a further illustration of this observation, Fig. 3 gives the maximum strength
column curves for members with three types of end connections, and the middle of the SSRC Column Curves [1, 10] is
also included for comparison (it is based on pinned ends).
< previous page page_265 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_265.html07/11/2009 12:09:51 PM
 page_266

< previous page page_266 next page >

Page 266

Figure 2 Load-Deflection Curves for Columns with Various Degrees of End Restraint
The vertical axis in Fig. 3 gives the relative maximum column strength, and the horizontal is based on the non-
dimensional slenderness term. Also shown is the Euler curve, as well as the ranges of practical slenderness values for
two common steel grades (yield stresses of 36 ksi=248 MPa and 50 ksi=345 MPa) that are in extensive use in North
America. Although the strength of the end-restrained columns is still significantly higher than that of the pinned-end
member, it is not on the order of the 200 % that has been indicated. Nevertheless, the benefits are clear, and advantage
should be taken of this contribution to the performance of the structure.
Effects of Residual Stresses
Research has demonstrated that the end restraint effects are not as significant when residual stresses are present in the
cross section [7]. Specifically, the strength increase that is prompted by the end restraint will be less for columns that
buckle in the inelastic range of behavior.
Effects of Initial Out-of-Straightness
Studies have also shown that the end restraint effect is more important when the column is initially perfectly straight [7].
Therefore, for the more realistic maximum column strength models, which incorporate the influence of the crookedness
(as well as the residual stresses), the benefit again will be less than what has been reported, but it still valuable to include
the contributions in some form in design approaches.
END-RESTRAINED COLUMNS IN FRAMES
The preceding discussion has focused on the various parameters that influence the strength of individual end-restrained
columns. The studies
< previous page page_266 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_266.html07/11/2009 12:09:52 PM
page_267

< previous page page_267 next page >

Page 267

Figure 3 Column Curves for Members with Different End Connections

< previous page page_267 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_267.html07/11/2009 12:09:53 PM
 page_268

< previous page page_268 next page >

Page 268
that led to these findings were crucial to the complete understanding of the problem of the stability of such members, but
only limited information was gained regarding the implementation of the concepts into design code formats. Much
debate was conducted regarding the potential for utilizing column curves in the design standards that reflected specific
levels of end restraint, rather than having curves that were based on the traditional pinned-end members. The latter topic
will not be addressed here, short of noting that in North America it appears to have been decided that the pinned-end
column will continue to be the focal point. This has been done on the express premise that although it is recognized that
true pinned-end members do not exist in real structures, such a model reflects accurately the capacity of the member
itself, and is not encumbered by the factors that are related to the overall structure. There are obvious advantages to this
approach, especially when it is borne in mind that frame stability solutions can be provided that reflect the restraint
conditions between the columns and the other structural components. The following section of this paper will detail one
of the potential solutions that has been found to offer significant practical promise [6].
Rather than using complete second order frame stability solutions, it is preferable in many cases to utilize simplified
methods that can account for the column stability through other devices. Among the latter, North American practice for
years has relied on the effective length concept, which essentially transforms a column with real end conditions into one
that is pinned-end, but whose strength is the same as that of the real member. The procedure is well known, and has
served designers adequately for some time. However, the assumptions of the analysis and the manner in which the
effective length factor, K, is determined, have limited the applicability of the concept to structures with well-defined end
conditions. For example, the traditional K-factor solutions are based on rigid beam-to-column connections [1].
Research work by Chen at al. [5, 8] provided a method for finding the effective length of individual columns with
realistic beam-to-column connections at the ends. This solution also examined the effect of the beam stiffness on the
overall column end restraint, and gave data on the K-factors for members with a variety of support conditions. However,
it did not address the larger problem of determining the stability of the column as part of a frame.
The simplified frame stability solution that makes use of the K-factor concept was developed from considering a
subassemblage of an elastic frame, using rigid beam-to-column connections [1], and considering the sway- and non-sway
cases. The characteristic equation for the subassemblage incorporated the K-factor and functions of it, along with the
relative stiffness distribution factor, G, given as

(1)
where the numerator term expresses the combined column stiffness at a joint in the subassemblage, and the denominator
gives the combined beam stiffness at the same joint. Briefly, Eq. (1) reflects the magnitude of the moment that will be
transmitted to the column at the instant of buckling.
Conceptually, Eq. (1) actually shows that the magnitude of the moment
< previous page page_268 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_268.html07/11/2009 12:09:54 PM
 page_269

< previous page page_269 next page >

Page 269
that is transferred to the column is a function of the stiffness of the restraining member or assembly. The higher the value
of G, the more moment the column will be asked to carry. In the most extreme case the columns are infinitely stiffer than
the beams (i.e. G approaches infinity), which means that the beams will transfer no moment to the column. This, in turn,
is an expression of the fact that the beam in this case cannot offer any stabilizing effect for the column. In other words, a
value of G equal to infinity indicates that the column has a pinned end connection.
In real framed structures the beams, which have a certain bending stiffness EI and span L, are attached to the columns of
the frame through beam-to-column connections that have a certain rotational stiffness. The latter is given by the initial
slope of the moment-rotation curve, C, as illustrated in Fig. 1. As shown by Bjorhovde [6], the combined rotational
restraint of the connection and the beam that is attached to the column is given by the effective rotational restraint, C*.
This is exactly the same as the denominator term in Eq. (1), but now revised to take into account actual restraint
characteristics of the assembly.
On the above basis it has been shown [6] that the stiffness distribution factor, G, for a realistic frame should be replaced
by Gr, the end-restraint stiffness distribution factor. Equation (2) therefore takes the place in the effective length solution
procedure, rather than Eq. (1). The former gives the expression for Gr as

(2)
and it is noted that a single C* is used for each column end. The reason for this can be explained as follows:
(a) Behavior of interior columns: In the general frame subassemblage, the beams are loaded and deflected a certain
amount prior to column buckling. The connections are therefore rotated a certain amount. As the column buckles,
two of the connections will unload elastically (stiffness=C) and the other two will continue loading, with a stiffness
equal to the slope of the moment-rotation curve at a point equal to the beam moment before buckling took place. For
the typical interior column, which has two beams framing into each end, this means that if the stiffness of the
loading connection is ignored (which is usually a near-correct assumption, especially for low-moment joints), then
only one connection at each end will provide rotational restraint. Hence, the use of a single C*-value in the Gr-
equation.
(b) Behavior of exterior columns: Following the above evaluation of the behavior of an interior column prior to and
during buckling, it is clear that when an exterior column fails, one end connection will load and the other will
unload. In the solution procedure that was developed [6] it was therefore determined that the values of Gr to use
should be based on one C* at one end, and at the other end a pin could be assumed as the support.
< previous page page_269 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_269.html07/11/2009 12:09:55 PM
 page_270

< previous page page_270 next page >

Page 270
It is noted that in order for the above approach to be utilized, data for the moment-rotation characteristics of typical
beam-to-column connections must be available. Such data collection efforts are currently underway in Britain and the
United States.
At this point in time the effective restraint procedure that has been outlined above should only be applied to the design of
columns in sway-prevented frames. The reason for this is that further research is needed to resolve the questions that
arise with respect to the moment that is transferred into the column from the beams in a sway-permitted structure. The
latter uses moment-resistant connections, and the moments that can develop are substantially larger than those of the
types of joints that have been described in the preceding.
Design examples that have been worked out [6] show column size savings on the order of 5 to 15 percent, depending on
the specifics of the member itself.
SUMMARY
The paper has presented an examination of the strength and behavior of columns with realistic end-restraint conditions,
such as can be encountered in framed steel structures. The individual column strength parameters are given, along with
their relative influences.
A solution procedure is given for finding the effective length factor for end-restrained columns in frames, taking into
account the connection rotational restraint and the bending stiffness of the beams. It is shown that this leads to the
effective rotational restraint, which in turn can be used in a traditional subassemblage K-factor procedure. An expression
is given for the end-restraint stiffness distribution factor, which replaces the well-known G-term of the elastic stability
equations.
It is noted that column economy on the order of 5 to 15 percent can be gained through this solution method. However,
data are needed on the moment-rotation curves and their parameters for many actual beam-to-column connections.
Further research should also be conducted to resolve the applicability of the method to sway-permitted frames.
REFERENCES
1. Johnston, B.G., Editor, ‘‘Guide to Stability Design Criteria for Metal Structures”, 3rd Ed., Wiley-Interscience, New
York. 1976.
2. American Institute of Steel Construction (AISC), “Specification for the Load and Resistance Factor Design,
Fabrication and Erection of Structural Steel Buildings”, AISC, Chicago, Illinois, September 1, 1986.
3. Canadian Standards Association (CSA), “Steel Structures for Buildings-Limit States Design”, CSA Standard No.
CAN3—S16.1–1984, CSA, Rexdale, Ontario, Canada. 1984.
4. European Convention for Constructional Steelworks (ECCS), “Eurocode 3: Recommendations for the Design of Steel
Structures for Buildings”, ECCS, 1986.
< previous page page_270 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_270.html07/11/2009 12:09:56 PM
 page_271

< previous page page_271 next page >

Page 271
5. Sugimoto, H., and Chen, W.F., “Small End Restraint Effects on Strength of H-Columns”, Journal of the Structural
Division, ASCE, Vol. 108, No. ST3, March. 1982 (pp. 661–681).
6. Bjorhovde, Reidar, “Effect of End Restraint on Column Strength-Practical Applications”, AISC Engineering Journal,
Vol. 20, No. 1, First Quarter, 1984 (pp. 1–13).
7. Jones, S.W., Kirby, P.A., and Nethercot, D.A., “Effect of Semi-Rigid Connections on Steel Column Strength’’,
Journal of Constructional Steel Research, Vol. 1, No. 1, September, 1980 (pp. 35–46).
8. Lui, E.M., and Chen, W.F., “End Restraint and Column Design Using LRFD”, AISC Engineering Journal, Vol. 21,
No. 1, First Quarter, 1985 (pp. 29–39).
9. Bjorhovde, Reidar, “Discussion of “End Restraint and Column Stability” by W.F.Chen”, Journal of the Structural
Division, ASCE, Vol. 107, No. ST8, August 1981 (pp. 104–120).
10. Bjorhovde, Reidar, “Deterministic and Probabilistic Approaches to the Strength of Steel Columns”, Ph. D.
Dissertation, Lehigh University, Bethlehem, Pennsylvania, May, 1972.
< previous page page_271 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_271.html07/11/2009 12:09:57 PM
 page_272

< previous page page_272 next page >

Page 272
INFLUENCE OF SEMI-RIGID CONNECTIONS ON THE OVERALL STABILITY OF STEEL FRAMES
F.M.MAZZOLANI
Istituto di Tecnica delle Costruzioni
Engineering Faculty
University of Naples (Italy)
ABSTRACT
This paper summarizes the main results of recent parametrical analysis on the load carrying capacity of semi-rigid
frames. In particular, the effect of partial restraint conditions of sway frames is analyzed with respect to the elastic and
inelastic buckling. In addition to the effect of connections, the influence of geometrical imperfections on the ultimate
behaviour of “industrial frames” is also examined.
INTRODUCTION
The behaviour of semi-rigid frames is influenced by several parameters. The main determinant aspects of the general
problem have been analyzed from different points of view. Starting from the assumption of a mathematical model for
joints under alternate loading conditions [1], an appropriate calculation method has been worked out [2]. The first results
of its application have been mainly devoted to the elastic and inelastic buckling of multi-story frames [3], also including
the effect of geometrical imperfections [4]. This led to a new definition of “industrial frames”, in which the actual
restraint conditions of joints together with the manufacturing tollerances are considered as generalized imperfections [5].
The problem of the overall ductility has been also analyzed in view of the assesment of the structural coefficients (the so-
called q-factors) in the design of seismic resistant structures [6].
ELASTIC BUCKLING
By means of the following parameters:
ακ critical multiplier of the frame with semirigid connections;
α∞ critical multiplier of the frame with perfectly rigid connections;
< previous page page_272 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_272.html07/11/2009 12:09:58 PM
 page_273

< previous page page_273 next page >

Page 273
α0 critical multiplier of the frame with pinned connections;
R ratio of girder-to-column flexural stiffness;
K ratio of joint-to-girder flexural stiffness;
fig. 1 shows some results referred to a portal frame and to a one bay-three story frame.
Curves of fig. 1,a) allow to define three regions which characterize the unstable behaviour of the semi-rigid frames:
- 0<K<3, where even for small values of connection stiffness the critical multiplier
considerably increases and the case of pinned joints can be safely assumed;
- where the actual value of the joint restraint must be introduced in the analysis without
3<K<10, allowing any simplification;
- K>10, where the case of perfectly rigid joints can be assumed provided 80 percent of elastic
critical multiplier is safely adopted.
Fig. 1,b) shows that the obtained results are located with a good approximation on lines passing through the origin,
which are given by the equation:

(1)
Fig. 1,c) provides the value of the costant A for a given R ratio and a given frame typology. These results have been
confirmed by a parametrical analysis on multy-bay (up to three) and multy-story (up to six) frames [3,4] and, therefore,
the linear relation-ship (1) can be proposed for general applications.
INELASTIC BUCKLING
With reference to the M-φ curves of connections and girders, the three cases of fig. 2 can arise: case a) corresponds to a
full strength joint and the global relation-ship can be assumed as elastic-perfectly plastic with a good approximation.
The influence of partial strength joints (case c) produces important alterations in the plastic behaviour of the structure
due to the early formation of plastic hinges (see dotted curves of fig. 2). It causes a strong reduction of the ultimate
carrying capacities together with a change of slope of the decreasing branch of the curves.
Curves of fig. 2 show also the influence of the vertical loads for the given two-story frame with different joint restraint

(K). Going from the case 0.1 to 0.25 (being the critical load of the rigid joint frame) we find that:
– the collapse multiplier is reduced to about a half;
– the decreasing branch increases its slope, leading to a sensible reduction of the global ductility.
In both cases the effect of restraint is sensible in the increasing branch, whereas all curves are connected to only one
decreasing branch, as
< previous page page_273 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_273.html07/11/2009 12:09:59 PM
 page_274

< previous page page_274 next page >

Page 274
it can be aspected because the rigid-plastic mechanism is not influenced by the joint flexibility.
IMPERFECTION SENSITIVITY
The so-called “industrial frames” [5] are today considered in the modern codes which introduce geometrical
imperfections (as out-of-plumb in columns and eccentricities in joints), according to the usual erection tollerances.
The influence of such imperfections on the elastic-plastic behaviour has been analyzed for two portal frames of fig. 3,
respectively fixed and hinged. The different cases are characterized by the column slenderness 100,
160. Collapse multipiers are given in fig. 4. For we observe that all curves of the fixed frame practically reach
95 percent of the squash load Ny, excepted for. In the case of hinged frame, on the contrary, the effect of
semi-rigid connections is more evident and only 60 percent of Ny is reached for (which corresponds in
practice to a rigid joint). As the column slenderness increases, it is evident that also in the case of fixed frame the effect
of semi-rigid connection becomes more and more important. The percentage of collapse load is always decreasing up to
the case of hinged portal with in which only 5 percent of the squash load of column is reached.
CONCLUSIONS
The above results show that it is possible to provide for several structural typologies the values of the elastic critical
multiplier of semi-rigid frames by means a simple linear relation-ship.
The influence of the restraint conditions on the overall behaviour in elastic-plastic range plays a capital rôle on both
strength and ductility of imperfect frames, which must be properly analized according to the definition of “industrial
structures”.
ACKNOWLEDGMENTS
This research is supported by the Italian Ministry of Education. The research team is composed by Edoardo Cosenza,
Antonello De Luca, Attilio De Martino and Ciro Faella.
REFERENCES
1 Cosenza, E., De Luca, A., De Martino, A., Faella, C., and Mazzolani, F.M., Inelastic behaviour of framed structures
under cyclic loads. Euromech Colloquium 174, Palermo, October 1983.
2 Cosenza, E., De Luca, A., Faella, C., Nonlinear behaviour of framed structures with semirigid joints . Costruzioni
Metalliche, n. 4, 1984.
< previous page page_274 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_274.html07/11/2009 12:10:00 PM
 page_275

< previous page page_275 next page >

Page 275
3 Cosenza, E., De Luca, A., Faella, C., Inelastic buckling and post-critical behaviour of steel structures with semirigid
joints. Regional Colloquium on Stability of Steel Structures, Budapest, September 1986.
4 Cosenza, A., De Luca, A., Faella, C., Elastic and inelastic stability of semirigid frames. SSRC Annual Technical
Session on Stability of Plate Structures, Washington, april 1986.
5 Cosenza, E., De Luca, A., Faella, C., and Mazzolani, F.M., Imperfection sensitivity of “industrial’’ steel frames.
International Conference on Steel and Aluminium Structures, Cardiff, july 1987.
6 Cosenza, E., De Luca, A., Faella, C., and Mazzolani, F.M., On a simple evaluation of structural coefficients in steel
structures. Eight European Conference on Earthquake Engineering, Lisboa, September 1986.
< previous page page_275 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_275.html07/11/2009 12:10:00 PM
 page_276

< previous page page_276 next page >

Page 276

Figure 1—Results of the parametrical analysis for the evaluation of the elastic critical multiplier of semi-rigid frames.
< previous page page_276 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_276.html07/11/2009 12:10:02 PM
page_277

< previous page page_277 next page >

Page 277

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_277.html (1 of 2)07/11/2009 12:10:03 PM

 page_277

Figure 2—Influence of the partial strength joints and of the intensity of vertical loads on the load carrying capacity of
semi-rigid frames under horizontal forces.
< previous page page_277 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_277.html (2 of 2)07/11/2009 12:10:03 PM

 page_278

< previous page page_278 next page >

Page 278

Figure 3—Load (N/NY) versus sway (d/dO) relation-ships for imperfect portal frames, fixed and hinged respectively.
< previous page page_278 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_278.html07/11/2009 12:10:04 PM
 page_279

< previous page page_279 next page >

Page 279

Figure 4—Collapse multipliers of semi-rigid portal frames.

< previous page page_279 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_279.html07/11/2009 12:10:04 PM
 page_280

< previous page page_280 next page >

Page 280
INFLUENCE OF SEMI-RIGID AND PARTIAL STRENGTH JOINTS ON THE BEHAVIOUR OF FRAMES
F.S.K.Bijlaard
and
H.H.Snijder
TNO Institute for Building Materials and Structures
P.O. Box 49, 2600 AA Delft
The Netherlands
ABSTRACT
The behaviour of steel frames is much influenced by the moment-rotation characteristic of the connections between
columns and beams. The use of bolted beam-to-column connections without using stiffeners leading to semi-rigidity and
partial strength, has become more popular. The use of this type of connection gives the opportunity to optimize costs for
connections against costs for beams and columns. The possibility is discussed of using a simplified model for the usually
non-linear moment rotation behaviour of semi-rigid and partial strength connections. Furthermore different design
approaches for the frames are discussed.
INTRODUCTION
In designing steel frames the assuptions made in calculating the force distribution in the frame should be in accordance
with the behaviour of the connections. When calculating the force distribution in a frame in accordance with elastic
theory, the specific joint behaviour with respect to stiffness, such as hinged, rigid or semi-rigid, has to be taken into
account. In all cases, the connections should be able to resist the calculated forces. Hinged connections should be able to
undergo the necessary rotation without producing substantial moments. Rigid connections have to transfer moments and
the corresponding deformation in the connections has to be small, so that they do not influence the calculated force
distribution. Semi-rigid connections must also transfer moments, but the corresponding deformations have an influence
on the force
< previous page page_280 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_280.html07/11/2009 12:10:05 PM
 page_281

< previous page page_281 next page >

Page 281
distribution in the structure.
When the force distribution in a frame is calculated by means of plastic theory, the specific joint behaviour with respect
to strength, such as full-strength or partial-strength, has to be taken into account. Due to the complex behaviour of
connections, it is only possible to calculate their strength and stiffness with relatively modest accuracy. It is, however,
necessary to know the influence of the strength and stiffness of connections on frame behaviour if calculations are to
give a reliable estimate of the actual structural response of the frame.
STABILITY OF FRAMES WITH SEMI-RIGID AND PARTIAL STRENGTH CONNECTIONS
In order to study the influence of the joint characteristic on the structural response of frames, the basic behaviour of a
joint expressed in the moment-rotation characteristic must be known. Three important aspects are the stiffness, the
strength and the rotational capacity of the connection (see Fig. 1). In [1] these aspects are discussed.

Figure 1: Moment-rotation characteristic of connections.

In calculating the stabilility of frames, an important question is whether or not a simplified model for the, in reality, non-
linear connection behaviour can be used. Non-linear connection behaviour, as shown in figure 2(a), can be dealt with in a
geometrical and material non-linear calculation where the load is applied incrementally. However, if a geometrical and
material non-linear calculation is not used to obtain the force distribution In the frame and to verify the stability of the
frame, the non-linear connection characteristic must be schematized. Use of the
< previous page page_281 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_281.html07/11/2009 12:10:06 PM
 page_282

< previous page page_282 next page >

Page 282
initial stiffness Cin leads to overestimating of the stability of the frame. When the bi-linear approximation is used, the
secant stiffness of the connection is taken into account in calculating the frame stability. Suppose, however, that due to

the design load on the frame, the connection is loaded up to a certain moment, say the calculated moment capacity of
the connection. In that case the actual tangent stiffness is far less than the secant stiffness used in the calculation. It can
be shown, however, that using the secant stiffness is a safe approximation to the connection behaviour in calculating the
stability of the frame. Assume a vertical bar with infinite bending stiffness, connected to the ground by a rotational
spring, as shown in figure 2(b). At the top of the bar a horizontal and a vertical load act, producing an overturning
moment:

..... (1)
In the rotational spring a reaction moment:

..... (2)
is activated in the case of a linear rotational spring characteristic (see figure 2(c)). Equilibrium is achieved if:

..... (3)
From figure 2(c) it can be seen that equilibrium is possible if:

..... (4)

The possibility of reaching equilibrium is also dependent on the value of the ultimate moment capacity . In figure 2
(d) the lines a, b and c indicate three different values for H.l. Line b indicates the maximum value for H.l at a certain
value for F.l for which equilibrium is possible. In the case of a non-linear rotational spring characteristic, the same
procedure holds (see figure 2(e)). The maximum load line b does not reach the ultimate moment capacity of the
connection characteristic, but touches that curve. Figure 2(f) shows a non-linear rotational spring characteristic and a bi-
linear approximation to this curve. It can be seen
< previous page page_282 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_282.html07/11/2009 12:10:07 PM
 page_283

< previous page page_283 next page >

Page 283
that the use of the bi-linear approximation leads to a safe calculation of the stability of the frame, because it produces a
lower value for H.l than in the case of the non-linear curve at the same value for F.l. The same holds for the case of a
constant value of H.l (see figure 2(g)). In that case, using the bi-linear approximation leads to a lower value for F.l than
using the non-linear curve.

Figure 2: Influence upon frame stability of a bi-linear approximation to the moment-rotation characteristic of a
connection.
< previous page page_283 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_283.html07/11/2009 12:10:08 PM
 page_284

< previous page page_284 next page >

Page 284
It can be concluded that it is safe to use a bi-linear approximation to the non-linear moment-rotation characteristic for
semi-rigid and partial-strength connections for the purpose of calculating the stability of frames.
ELASTIC DESIGN
Unbraced frames
In calculating the force distribution In the frame, the rigidity of the connections has to be taken into account. In checking
the columns and members under these forces with respect to strength and stability, the strength functions for columns as
well as for beams have to compensate for the non-linear effects which are not yet in the linear force distribution. The
influence of the regidity of the connections is expressed in the calculation of the elastic effective buckling length of the
columns (see Fig. 3).

Figure 3: Elastic effective buckling length of columns in unbraced frames as function of ρ, being the relative stiffness at
the column ends.
Braced frames
Elastic design of braced frames can also be referred to as “weak-column, strong-beam” design [2, 3]. After making a first
order elastic
< previous page page_284 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_284.html07/11/2009 12:10:09 PM
 page_285

< previous page page_285 next page >

Page 285
calculation to obtain the force distribution of the braced frame, the columns are designed first. Then, the beams are
designed in such a way that they do not collapse prior to the columns. The same holds for the semi-rigid connections.
In [4] interaction formulae have been presented for pin-ended beam-colomns under combined axial compression and
bending. To obtain the magnitude of the imperfection e*, the amplification factor n/(n−1) and the buckling factor χ, the
buckling length concept is used. The stability check of the total braced frame is replaced by stability checks of the
individual columns. The columns are ‘‘cut” out of the frame and the resultant pin-ended columns, with the imposed
bending moments and axial forces, are checked for stability. When the individual column approach is consistently
applied, the buckling length should be taken equal to the system length [5]. In actual fact, the columns form part of the
braced frame and it can be shown that the use of the system length as buckling length is conservative in many cases [6].
Therefore, checks on column stability in braced frames are acceptable on the basis of the elastic effective length. This,
however, has consequences for checking of the beams. Use of the elastic effective length as buckling length in the
Interaction formulae, leads to safe estimations for the maximum strength of columns in braces frames, if the beams and
the connections are strong enough and column collapse characterizes the frame behaviour [7,8,9,10]. The moment

distribution at collapse is different from the moment distribution using the first-order elastic analysis. At
collapse, the bending moments on the beam are greater than (Fig. 4). This is caused by the restraining moment Mr that
has to stabilise the column.

Figure 4: Comparison of bending moment diagrams.

In [6] a complete procedure is given to determine the restraining moment Mr.
< previous page page_285 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_285.html07/11/2009 12:10:10 PM
 page_286

< previous page page_286 next page >

Page 286
PLASTIC DESIGN
Unbraced frames
Designing unbraced frames with plastic theory is rather simple when using the formation of a collapse mechanism for
calculating the first order plastic collapse load. The ultimate limit bearing capacity of the frame, which is less than the
plastic collapse load, due to second order effects, can easily be determined by using the Merchant-Rankine formula.
Figure 5(a) shows an unbraced frame with rigid and full-strength connections while Fig. 5(b) depicts the same frame but
with simi-rigid and partial-strength connections. In this figure, the load-deformation curves are plotted

Figure 5: Influence of semi-rigid and partial-strength connections on the calculated bearing capacity of an unbraced
frame: (a) unbraced frame with rigid and full-strength connections; (b) unbraced frame with semi-rigid and partial-
strength connections.
using the various methods for calculating the force distribution. The curves plotted with the thick line indicate the real
behaviour. Due to the flexibility of the connections, the Euler buckling load of the frame will decrease. As a result of the
partial strength of the connection, the bearing capacity, based on the first-order plastic theory, will decrease. Both
phenomena, with respect to the same frame but now with rigid and full-strength connections, will lead to a lower
ultimate bearing capacity of the frame.
It can be seen that the rigidity of the connections has great influence on the stability of unbraced frames.
< previous page page_286 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_286.html07/11/2009 12:10:11 PM
 page_287

< previous page page_287 next page >

Page 287
Braced frames
Plastic design can also be referred to as “strong-column, weak-beam” design [2,3]. The beams are designed first, making
use of the first order plastic force distribution (beam mechanism). Then, the columns are designed in such a way that
they do not collapse prior to formation of a beam mechanism. When the beam falls, a mechanism is formed with a plastic
hinge in the mid-section of the beam and plastic hinges at the supports of the beam. When the moment capacity of the
connections is smaller than that of one of the connecting parts, the plastic hinges will form in the connection itself. If that
is not the case, the plastic hinge will form in the beam just aside the connections. In knee-connections it is also possible
that a plastic hinge is formed in the column when, due to normal forces, the reduced moment capacity of the column
section is smaller than the moment capacity of the beam section [11].
If plastic hinges occur at the top of the column, rotational capacity at the top of the column is necessary [12] if these
plastic hinges are formed prior to the hinge at mid-span.
To reach a beam mechanism a redistribution of moments is necessary. A redistribution of moments can only be attained
when the parts which yield first have sufficient deformation capacity. In many cases this deformation capacity has to be
brought up by the semi-rigid and partial-strength connections [1,6]. The columns are loaded with normal forces and
moments equal to the plastic moment capacities of the connections being part of the beam mechanisms. The bearing
capacity of the beam mechanism is only reached If the columns can withstand the force distribution belonging to the
relevant beam mechanism. Therefore, columns must be checked using interaction formulae [4]. Now the question arrises
which buckling length has to be used. Since the beam can not offer restraint to the column, because of the plastic hinge
near the column end, the system length can be used as buckling length safely. However, this approach is on the
conservative side if used for the frame with rigid supports. In general, the elastic effective length based on the remaining
elastic structure may be used as buckling length in plastically designed braced frames.
< previous page page_287 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_287.html07/11/2009 12:10:12 PM
 page_288

< previous page page_288 next page >

Page 288
DESIGN BASED ON SECOND ORDER THEORY
Using the second order theory, two calculation methods are available:
– geometrical non-linear calculation with linear elastic material behaviour;
– geometrical and material non-linear behaviour.
A geometrical non-linear calculation with a linear elastic material behaviour results in a force distribution which is in
equilibrium in the deformed situation with the factored design load. Imperfections have to be taken into account. The
magnitude of the imperfections is given in [4].
The moment-rotation characteristic can be taken as realistic as possible in the calculations, however, the semi-rigid
connections can also be modelled with the secant stiffness of the bi-linear approximation. This method of analysis results
in the correct force distribution in the whole structure. Correction factors on the moment distribution are superfluous
[13]. Plasticity effects need to be checked separately using interaction formulae for cross-sections loaded with normal
and shear force and bending moment. Formally, because of the assumption of elastic material behaviour in the analysis,
the cross-section should be checked in such a way that the yield stress is reached in the outer fibres of the cross-section
only. However, the use of a plastic interaction formula for checking the cross-section is allowed because this procedure
allows only one plastic hinge to be formed in the structure or more than one but at the same time.
The geometrical and material non-linear behaviour can also be obtained within one calculation [e.g. 14, 15]. In that case
no seperate check on cross-sectional behaviour is necessary.
In most cases, when non-linear theory is used in calculating frames, only the planar behaviour of the members is
considered. Therefore seperate checks on local instability and lateral torsional buckling are still needed to be carried out.
EVALUATION OF DESIGN METHODS
All three design methods discussed in the previous subsections, elastic design, plastic design and design based on second
order theory, are acceptable design methods. Elastic design and plastic design are based on first order force distributions
which are relatively easy to obtain. The post-processing of these results, to take the second order effects and
< previous page page_288 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_288.html07/11/2009 12:10:13 PM
 page_289

< previous page page_289 next page >

Page 289
plasticity Into account, has been formalized in specific design rules. Second order theory results directly into a second
order force distribution, but consequently asks more of the user’s knowledge.
The methods of analysis for evaluating the ultimate limit state of frames is so much refined, that deflections and side
sway become more important and determining. There is a need for criterea for the deflections and side sway dependent
on the purpose of the building and the deformation capacity of the secondary structures attached to the supported frames.
REFERENCES
1. Stark, J.W.B., Bijlaard, F.S.K. “Structural properties of connections in steel frames”, paper to be presented at “State of
the art workshop on connections, strength and design of steel structures”, Cachan, 25–27 May, 1987.
2. Vinnakato, S. “Design of columns in planar frames—A few comments”, Proc. of the National Conference on Tall
Buildings, held at New Delhi, 1973.
3. Vinnakota, S. ‘‘Design of columns as part of frames—Some remarks”, Paper submitted to TG3: SSRC, 1983.
4. Eurocode 3, “Common unified code of practice for steel structures”, Commission of the European Communities,
Brussels 1984, Report EUR 8849 DE, EN, FR.
5. Second International Colloquium on Stability. Introductory Report, Liège, April 13–15, 1977, ECCS
6. Analysis and design of steel frames with semi-rigid joints. ECCS Technical Committee 8 “Structural Stability” (to be
published).
7. Snijder, H.H. and Bijlaard, F.S.K. “The influence of end restraints on the maximum strength of centrically loaded
columns in braced frames”, J. Construct. Steel Research, Vol. 5 (1985) pp. 137–147.
8. Snijder, H.H., Bijlaard, F.S.K. and Stark, J.W.B. “Use of the elastic effective length for stability checks of columns
and consequences for checks on beams in braced frames, Instability and plastic collapse of steel structures”, Proc. of the
Michael R.Horne conference, Edited by L.J.Morris, London, Granada, 1983, pp. 152–163.
9. Snijder, H.H., Bijlaard, F.S.K. and Stark, J.W.B. “Strength requirement for beams in braced frames in relation to
column stability”, Third international colloquium, Stability of metal structures, Final report, Paris, 1983, pp. 247–51.
< previous page page_289 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_289.html07/11/2009 12:10:14 PM
 page_290

< previous page page_290 next page >

Page 290
10. Snijder, H.H., Bijlaard, F.S.K. and Stark, J.W.B. “Design of braced frames using the elastic effective length method”,
Festschrift Roik, Technisch-wissenschaftliche Mitteilungen, Institut für Konstruktiven Ingenieurbau, Ruhr-Universität
Bochum, Mitteilung Nr. 84–3, September 1984, pp. 316–332.
11. Witteveen, J. “Frame analysis including column hinges”, Stability of steel structures, Final report, pp. 309–310,
Second International Colloquium, Liège, April 13–15, 1977, ECCS.
12. Manen, S.E. van “Plastic design of braced frames allowing plastic hinges in the columns”, Heron 27 (1982), No. 2,
pp. 1–20.
13. Vogel, U. “Gedanken zum Sinn und Zuverlässigkeit des Tragsicherheitsnachweises am Ersatzstab bei
Stabsystemen”, Festschrift Roik, Technisch-wissenschaftliche Mitteilungen, Institut für Konstruktiven Ingenieurbau,
Ruhr-Universität Bochum, Mitteilung Nr. 84–3, September 1984, pp. 333–346.
14. Borst, R.de, Kusters, G.M.A., Nauta, P. and Witte, F.C. de ‘‘DIANA—a comprehensive but flexible finite element
system”, Finite element systems, third edn., ed. by C.A.Brebbia, Berlin, Springer-Verlag, 1983.
15. Foeken, R.J.van, Snijder, H.H. “Steel column and frame stability analysis using finite element techniques”, Heron,
Vol. 30 (1985) No. 4, pp. 1–29.
< previous page page_290 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_290.html07/11/2009 12:10:14 PM
 page_291

< previous page page_291 next page >

Page 291
LARGE SCALE TESTS ON COLUMN SUBASSEMBLAGES AND FRAMES
P A Kirby, J B Davison and D A Nethercot
Department of Civil and Structural Engineering
University of Sheffield
Mappin Street
Sheffield S1 3JD
UK
ABSTRACT
The preliminary findings of eight beam and column subassemblage and two three storey two bay frame tests are
reported. It is noted that the behaviour of joints as measured in the frame tests is not appreciably different from that
determined from an individual joint test. It is seen that the semi-rigid joint continuity results in enhanced column
capacity, reduced maximum beam moments and deflections. Existing analysis programs developed at Sheffield and
Milan can accurately simulate this behaviour.
INTRODUCTION
When compared with the normal design assumption of pinned joints, actual semi-rigid joints can be conceived as having
two potential benefits. The first involves restraint provided by the beam to the columns, thus reducing effective lengths
and mid-height deflections and tempering P-∆ effects. The second occurs when considering continuity provided by the
columns at the ends of the beam so as to reduce midspan moments and deflections. Much analytical work [1–3] has
recently been accomplished and computer programs with varying degrees of sophistication exist which can predict the
response of individual columns, beam and column subassemblages and complete structures. The complementary
experimental programme of work needed to verify and validate these capabilities however has lagged far behind. This
paper presents the initial findings of the only known reasonably extensive test programme designed to remedy this
deficiency. As mentioned in a companion paper [4] eight column subassemblage tests and two complete three storey two
bay frame tests with joints nominally identical to those of a specially conducted series of connection tests have been
undertaken. The joint and subassemblage tests were carried out at the University of Sheffield, whilst the frame tests were
conducted collaboratively at the Building Research Establishment (BRE). In addition, two similar frames fabricated with
extended end-plates have been carried out by Hatfield Polytechnic and BRE and a fifth test is yet to be conducted.
< previous page page_291 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_291.html07/11/2009 12:10:15 PM
 page_292

< previous page page_292 next page >

Page 292
TESTS ON SUBASSEMBLAGES
A series of eight column subassemblage tests have been carried out and a full description of the procedures and results
are presented in ref 5. In all cases tests were conducted in the horizontal plane. The columns were 6.5m long with 1.5m
long beams framing in at either end 6.0m apart as indicated in Fig 1. In some cases the beams framed into the column
flanges and in other cases into the column web. A variety of connections were employed ranging from flexible web
cleats to almost rigid extended end plates as listed in Table 1. The ends of the beams remote from the column were
mounted in special bearings which were designed to prevent rotational movement but to freely permit movement parallel
to the direction of the column. To ensure the desired in-plane response the column was restrained from out of plane
movement by bracing in the form of guides, which were provided with sliding bearings so as not to inhibit in-plane
movements.

Figure 1. Subassemblage test arrangement.

< previous page page_292 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_292.html07/11/2009 12:10:16 PM
 page_293

< previous page page_293 next page >

Page 293
TABLE 1
Summary of subassemblage tests
TOTAL COLUMN FAILURE LOAD
TEST CONNECTION TYPE AXIS BENDING Q1 Q2 P ANALYSIS kN EXPERIMENT kN ANAL EXP
kN kN kN
ST2 Web cleats Major 0 73.3 598 671 671 1.00
ST3 Web cleats Minor 41.0 44.0 452 537 520 1.03
ST4 Flange cleats Major 110.0 110.0 463 683* 760 0.90*
ST6 Flange cleats Minor 41.4 43.0 510 594 520 1.14
ST7 Flange cleats Minor 0 40.0 484 524 526 1.00
ST8 Web and seat Minor 67.0 70.0 362 499 518 0.96
ST9 Flush end plate Minor 83.0 85.0 343 511 486 1.05
ST10 Extended end plate Major 93.0 100.6 403 597* 743 0.80*
* Numerical divergence
Loading could be applied independently to the beams and also axially at the head of the column via load cells. The other
end of the column was prevented from translating and a load cell was employed at this location. Correspondence
between the reactive force and the applied loads indicated that the sliding bearings worked well with only minimal
friction losses. Column deflections were measured by LVDT’s and strain gauges were positioned as indicated. On the
beams the gauges were connected to permit bending moments to be monitored whilst on the column both bending
moments and axial loads could be determined.
All columns showed evidence of having been roller straightened, an observation later confirmed by residual stress
measurements, and, in order to induce a representative initial lateral deformation, a small force was applied at midheight.
Loading then occurred in two stages. In the first beam loads Q1 and Q2 were applied up to the levels indicated in Table 1
and these were then held constant whilst axial column load P was applied to failure. After the subassemblage tests,
material yield stresses were obtained from coupons cut from the column flanges. Also shown in Table 1 are failure loads
predicted by the computer program. Apart from the two failure loads indicated by asterisks the correspondence is good
and, in these two cases, it is thought that the analysis was terminated prematurely due to numerical divergence.
A typical load deflection plot is shown in Fig 2 together with the responses predicted by computer programs developed
at the university of Sheffield and reported in ref 6. The close correspondence is apparent.
Table 2 shows the results of some simple ‘design’ calculations which assess the capacity of the column as an axially
loaded member i.e. ignoring the presence of moments transmitted to the column. The capacity is first calculated
assuming that no rotational restraint is provided at the connections giving a pin-ended column. Such a computation
dramatically underestimates the actual capacity even for a column with simple web cleat connections. Also shown is the
capacity determined in a similar manner but with an additional step which utilises the method detailed in ref. 7 to
< previous page page_293 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_293.html07/11/2009 12:10:17 PM
 page_294

< previous page page_294 next page >

Page 294

Figure 2. Load against deflection—test ST8.

TABLE 2
Test and ‘design’ loads for subassemblages
AXIALLY LOADED COLUMN CAPACITY
BASIC PROPERTIES PIN-ENDS MODIFIED CONTINUOUS
TEST Py A Ptest λ Pc Pdes k λ Pc Pdes
(N/mm2) (mm2) (kN) (k=1.0) (N/mm2) (kN) (N/mm2) (kN)

ST2 290 2822 671 92.4 159.5 450.1 1.49 0.67 62.4 225.5 636.4 1.05
ST3 265 2914 520 164.8 57.7 168.1 3.09 0.54 89.0 141.0 410.9 1.26
ST4 273 2983 760 92.3 155.2 463.1 1.64 0.64 59.1 221.1 659.5 1.15
ST6 288 2874 520 164.4 59.1 170.0 3.06 0.53 87.1 151.9 436.6 1.19
ST7 278 2824 526 165.3 58.1 164.1 3.20 0.53 87.6 146.9 414.8 1.27
ST8 279 2837 518 166.2 57.6 163.4 3.17 0.54 89.7 142.9 405.4 1.28
ST9 271 2927 486 166.7 56.9 166.6 2.92 0.53 88.4 145.0 424.4 1.14
ST10 272 2845 743 92.7 154.1 438.3 1.69 0.59 54.7 226.9 645.5 1.15
Py material yield stress from flange coupon tests
A cross-sectional area calculated from measured dimensions
Ptest load at failure in test; Σ (P=Q1=Q2) slenderness (effective length/appropriate radius of gyration)
Pc reduced material compressive strength from Table 27 BS 5950
Pdes A. pc
k effective length ratio
< previous page page_294 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_294.html07/11/2009 12:10:19 PM
 page_295

< previous page page_295 next page >

Page 295
obtain restraint coefficients which are then used in the chart of Fig. 23 of BS 5950 [8] to obtain reduced effective length
ratios. It can be seen that these ratios, taken as unity for the previous calculation, now lie between 0.53 and 0.67. Perhaps
somewhat surprisingly all of the test loads still exceed the enhanced design loads even though no allowance has been
made for the presence of bending moments—indeed the values show a remarkable correspondence. Caution must
however be exercised as there are a number of features which must be borne in mind namely the beams were very short,
the columns were initially very straight and contained minimal residual stresses.
Fig 3 shows the variation in the bending moments observed in the beams and the column ends close to the load
application joint (points A, B and C of Fig 1). In all cases the values have been nondimensionalised by dividing moments
M by the fully plastic moment of the beam Mp and the total axial load in the column P by the column squash load PY.
The values predicted by the computer simulation show reasonable correlation. It is interesting to observe that, as the
beam load is applied, the moment transmitted through to the column attains a value of approximately 0.25MP. As
subsequent axial load is added, this moment does not grow considerably; indeed over the next phase of loading it
actually relaxes. Similar behaviour has consistently been observed with, in some cases, the moment relaxing to zero and
in others actually reversing in sign as the beams are called on to restrain the column as it tries to buckle.

Figure 3. Moments at head of column—test ST8

< previous page page_295 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_295.html07/11/2009 12:10:20 PM
 page_296

< previous page page_296 next page >

Page 296
FRAME TESTS
Two three storey two bay frames have been tested in the Building Research Establishment as part of the Sheffield test
program. One frame involved the column members being bent about their major axes and the other about their minor
axes. In both cases flange cleat connections, which are towards the more flexible end of the range, were employed as two
similar frames with stiff flush and extended end plate connections formed the Hatfield/BRE part of the same sequence of
tests employing a degree of common loading, bracing and monitoring equipment. Fig 4 shows a Sheffield test in which
moments were measured close to column ends, at column midheights, beam ends and adjacent to loading positions.
Deflections were measured at column midheight, at beam load points with rotations also being recorded on the columns
at the beam column centre line intersection points and as close to the ends of the beams as was practicable
(approximately 125mm away from the beam end). Beam loads were applied through saddles suspended from the beams
by hydraulic jacks mounted under the laboratory floor via Macalloy bars. Column loads were similarly applied via
substantial caps at the head of column extensions which protruded above the head of the frame.

Figure 4. Frame test arrangement.

< previous page page_296 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_296.html07/11/2009 12:10:21 PM
page_297

< previous page page_297 next page >

Page 297

Figure 5.
Measured moment distribution around

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_297.html (1 of 2)07/11/2009 12:10:22 PM

page_297

Figure 6.
Measured moment distribution around
< previous page page_297 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_297.html (2 of 2)07/11/2009 12:10:22 PM

 page_298

< previous page page_298 next page >

Page 298
As with the subassemblage tests the loading was undertaken in two stages; load being applied initially to the beams, first
equally up to service load conditions with one beam being loaded more severely to create a form of pattern loading.
Figs 5 and 6 show the moment distribution around frames 1 and 2 after the full design load had been applied to the
beams. These distributions correspond closely to those obtained using a computer program developed at the Polytechnic
of Milan [3]. The beam end restraint offered by the connections can be clearly seen from the moment attracted to the
beam-column intersection. Interestingly the moments attracted to the connections at the extreme top corners of the frame
are not as large as those at the intermediate column. This is due to the flexibility of the column and illustrates that the
moments transmitted by connections depend on the relative balance of connection, column and beam stiffness. Due to
the moments attracted to the column head large deformations in the upper lifts of the external columns caused failure
under applied axial load to occur in these lifts rather than at ground or first floor level as would be expected.
Axial loads in excess of those predicted by design recommendations were sustained throughout the frame, even though
the moments transmitted to the column were greater than the nominal values of the end reaction multiplied by a 100m
eccentricity.
CONCLUSIONS
The in-plane behaviour of joints as determined from individual connection tests and as parts of subassemblages and
complete frames is shown to be consistent. Experimentally determined two dimensional response of the subassemblages
and frames are predicted to an acceptable degree of accuracy by computer programs developed at the University of
Sheffield and the Polytechnic of Milan. The semi-rigid action of conventional beam to column connections has been
shown experimentally to provide restraint against in-plane column failure, to reduce beam deflections and to increase
beam load carrying capacity by reduction of the midspan moment.
ACKNOWLEDGEMENTS
The work reported herein forms part of a continuing investigation into the effects of semi-rigid joint action on the
performance of steel frame structures supported by SERC, CIRIA and CONSTRADO (now subsumed into the SCI). In
particular acknowledgement is made to the BRE who were major partners in the frame tests and in whose laboratories
those tests were conducted. The authors are also graterful for assistance provided by staff of the department and
particularly by Mr A M Rifai.
< previous page page_298 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_298.html07/11/2009 12:10:23 PM
 page_299

< previous page page_299 next page >

Page 299
REFERENCES
1 Jones, S.W., Kirby, P.A. and Nethercot, D.A., “Effect of Semi Rigid Connections on Steel Column Strength” J. of
Construct. Steel Research Vol 1, No 1, 1980, pp. 38–46.
2 Nethercot, D.A., Kirby, P.A. and Rifai, A.M., “Design of Columns in PR Construction Analytical Studies” to be
published in Can. J. Civ. Engg., 1987.
3 Poggi, C. and Zandonni, R., “Behaviour and Strength of Steel Frames with Semi-Rigid Connections” Flexibility and
Steel Frames, ed. W.F.Chen, ASCE, 1985, pp. 57–76.
4 Davison, J.B., Kirby, P.A. and Nethercot, D.A., “Semi-rigid Connections in Isolation and in Frames” State of the Art
Workshop, Connections Strength and Design of Steel Structures, E.N.S., Cachan, May 1987.
5 Davison, J.B., Kirby, P.A. and Nethercot, D.A., “Column Behaviour in PR Construction—Experimental Behaviour’’,
ASCE Convention, New Orleans, September 1986 (to be published in J. of Struct. Engg., ASCE, 1987).
6 Rifai, A.M., “Behaviour of Columns in Sub-Frames with Semi-Rigid Connections” Ph.D. Thesis, university of
Sheffield, 1987.
7 Bjorhovde, R., “Effect of End Restraint on Column Strength—Practical Applications” Engng. J. of AISC. First Quarter
1984, pp. 1–13.
8 British Standard Institution, “Structural Use of Steelwork in Building” BS 5950 Part 1:1985, BSI, London.
< previous page page_299 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_299.html07/11/2009 12:10:24 PM
 page_300

< previous page page_300 next page >

Page 300
SIMPLICITY IN FLEXIBLY-CONNECTED FRAME ANALYSIS
Kurt H.Gerstle
and
Nevis E.Cook, Jr.
Department of Civil, Environmental, and Architectural Engineering
University of Colorado
Boulder, CO 80309
U.S.A.
ABSTRACT
Modeling of flexible connection behavior as elastic-perfectly plastic is proposed for analysis of building frames. Results
based on this idealization are compared with those of fully non-linear analysis for three different examples. It is
concluded that first-order effects of connection flexibility can be captured within engineering accuracy by linearly-elastic
analysis under working loads, and by perfectly-plastic analysis at ultimate.
INTRODUCTION
Recent research [1–3] on flexible connection behavior, and on analysis of flexibly connected frames, has emphasized
non-linear analysis. In general, non-linear analysis is unsuitable for office practice: Non-linear computer analysis is
laborious and expensive, and the results are often so error-prone that their interpretation becomes very complex, more
suited to an academic than a professional environment.
It is the aim of this paper to present a much simpler approach, and to document its suitability for design office practice
and its relation to reality with several examples.
APPROACH
We will represent the moment-rotation relation of a range of flexible beam-column connections in steel building frames
as elastic-perfectly plastic, and incorporate this formulation of the connection behavior in
< previous page page_300 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_300.html07/11/2009 12:10:25 PM
 page_301

< previous page page_301 next page >

Page 301
elastic-plastic frame analysis. With this premise, we will explore the feasibility of representing frame behavior under
working loads by elastic, and the structure strength by plastic analysis. If it can be shown that the frame acts nearly
linearly-elastic at working levels, and possesses sufficient ductility at ultimate, then simple, well-known methods will be
available for the design of flexibly-connected steel frames.
CONNECTION BEHAVIOR
The behavior of typical beam-column connections in steel frames can be represented by moment-rotation curves such as
those shown in Fig. 1 [4]. It should be noted that the range of rotations along the horizontal axis of Fig. 1 vastly exceeds
any values to be expected under conditions prior to collapse. For instance, the end rotation of a simple-supported,
prismatic elastic beam under uniform load causing a midspan deflection equal to 1/360 of its length will be θ=.0089
radians, and the rotation of a restrained beam will be much less than that. Similarly, the connection rotations of a frame
subject to a lateral load causing sway equal to 1/300 of its height will be less than .0033 radians. Connection rotations
under working conditions may thus be well within the near-linear range of the curves of Fig. 1.

Figure 1. Moment-Rotation Relations

Steel frame connections commonly used exhibit great ductility, as shown, for instance, by the curves of Fig. 1. If it can
be shown that the rotation capacity of the connections to be used exceeds the rotation demand necessary for full
redistribution of moments, then it follows that plastic analysis is suitable for determination of the frame strength.
< previous page page_301 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_301.html07/11/2009 12:10:25 PM
 page_302

< previous page page_302 next page >

Page 302
The connection response under load histories such as unloading and load reversal, as shown in Fig. 2, is needed to
predict frame response to more general loadings such as lateral load reversal. We will make the classical elastic-perfectly-
plastic assumptions shown dashed in our analysis.

Figure 2. General Connection Behavior

ANALYSIS
In the following example problems, the “plastic hinge method” [5,6], a piecewise-linear analysis of the structure
sequentially modified by internal hinges to represent local plastification, is applied. The resulting load-deformation
curves will be compared with similar curves from “exact” non-linear analyses, and their behavior at working and
ultimate loads will be studied in order to evaluate the suitability of elastic, and plastic, analysis at these two levels.
Example 1
A uniformly-loaded W18x40 steel beam, 30' (9.15 m) long, of yield strength 36 ksi (247.3 N/mm2), supported by
flexible connections of different types at each end, as shown in Fig. 3, is considered. This size beam was chosen because
a number of connection test results of beams of this size are available.
< previous page page_302 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_302.html07/11/2009 12:10:26 PM
 page_303

< previous page page_303 next page >

Page 303

Figure 3. Structure of Examples 1 and 2

The allowable load of on this beam is governed by the deflection criterion
assuming simple beam action.
Four different connections for which test data were available were considered in this analysis. Their stiffnesses, which
are given in Table 1, cover the range of flexible connections as suggested in Reference 7.
The moment-rotation curves for these four connections, taken from the references listed in Table 1, are plotted in Fig. 4.
The intersection of the beam line with these curves indicates connection moments and rotations under a working load of
1 k/ft (14.6 N/mm). The elastic stiffness k of each of the connections was determined as the secant modulus to this
working point; it, as well as the initial stiffness k0, the assumed plastic connection moment Mcp, and the stiffness ratio
EI/kL are given in Table 1 for the four different connections. They cover the entire previously defined range of flexible
connections.
TABLE 1. Connection Properties
No. Connection Type k0, k-in/rad k, k-in/rad Mcp, k-ft EI/kL
1 3/8" double web angle, 4 bolts [8] 27x103 22x103 16 2.24
2 3/8" double web angle, 5 bolts [8] 96x103 70x103 40 .70
3 Top and seat angles, 150x90x12 mm [9] 144x103 115x103 64 .43
4 Extended end plate to W18x35 Beam, t=1" [9] 642x103 552x103 160 .09
< previous page page_303 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_303.html07/11/2009 12:10:28 PM
 page_304

< previous page page_304 next page >

Page 304

Figure 4. Exact and Elastic-Plastic Moment-Rotation Relations

Using relations of statics, geometry, elastic-plastic beam behavior, and moment-rotation relations shown in Fig. 4, the
beam of Fig. 3 was analyzed for its midpoint deflection A under uniform load w, and the results plotted in Fig. 5. The
solid curves are the results of a fully non-linear analysis based, on the solid lines of. Fig. 4, while the dashed curves
represent the results of a plastic-hinge analysis based on the dashed elastic-plastic curves of Fig. 4. Additionally, the plot
is bounded by dashed lines representing the limiting conditions of simply-supported and fully fixed-ended beam
behavior.
< previous page page_304 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_304.html07/11/2009 12:10:29 PM
 page_305

< previous page page_305 next page >

Page 305

Figure 5. Load-Deflection Relations, Example 1

Example 2
The beam shown in Fig. 3, of length L=30' and loaded with a uniform working load of 1 k/ft (14.6 N/mm), is joined to
columns of story height H=12' of an unbraced frame at both ends. Due to a lateral load V applied to each column, the
columns rotate through equal angles θc, thereby permitting the frame to rack sideways, as shown in Fig. 6. During this
stage, the leeward connection will load further, while the windward connection will unload according to the relations
shown in Fig. 2.

Figure 6. Beam of Frame in Sway

A non-linear analysis based on this connection behavior was devised to determine the connection moments and rotations,
as well as the column
< previous page page_305 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_305.html07/11/2009 12:10:30 PM
 page_306

< previous page page_306 next page >

Page 306
rotation θc due to an applied shear V, and the results are shown solid in Fig. 7. Similarly, these relations were computed
using the plastic-hinge method and the dashed connection curves of Fig. 4, and plotted dashed in Fig. 7
Example 3
Lastly, in order to establish the relation of analysis and experiment, a flexibly-connected frame tested by Stelmack [10]
and shown in Fig. 8 was analyzed by the plastic-hinge method using elastic-plastic connection behavior. This frame,
under sequential gravity and lateral loads, had earlier been analyzed non-linearly by Stelmack, using experimentally
determined connection relations [10]. Fig. 8 shows experimental and analytical plots of sway versus applied lateral load
H. We observe that stiffness and strength of the frame are predicted closely by the simple plastic-hinge method.

Figure 7. Shear-Rotation Relations, Example 2

< previous page page_306 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_306.html07/11/2009 12:10:31 PM
 page_307

< previous page page_307 next page >

Page 307

Figure 8. Lateral Load-Sway Relation, Example 3

CONCLUSIONS
The results of the three example problems indicate the following for flexibly-connected frames:
1. In general, results of elastic-plastic analyses reflect the behavior predicted by non-linear analysis to within reasonable
engineering accuracy.
2. Connection rotations at working levels are within the near-linear range; elastic analysis can be used to predict structure
behavior under working loads.
3. Beyond working levels, the plastic-hinge method will give sufficiently close results. This analysis can be used when
excessive deformations or P-∆ effects need to be considered.
4. Structure strength can be predicted by classical plastic analysis.
Lastly, the need for further testing and analysis of connections to obtain a full range of moment-rotation curves for use in
steel frame design cannot be overemphasized.
< previous page page_307 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_307.html07/11/2009 12:10:32 PM
 page_308

< previous page page_308 next page >

Page 308
REFERENCES
1. Romstad, K.M., and Subramanian, C.V., “Analysis of Frames with Partial Connection Rigidity’’, Jnl. Struct. Div.,
ASCE, Vol. 96, No. ST11, Nov. 1970, pp. 2283–2300.
2. Ackroyd, M.H., and Gerstle, K.H., “Strength of Flexibly-Connected Steel Frames”, Engineering Structures, Vol. 5,
January 1983, pp. 31–37.
3. Edinger, J.A., “Non-Linear Frame Analysis Using the Ansys General Purpose Computer Program”, Technical Report
No. 83–1, A.I.S.C., Chicago, IL, July 1983.
4. Frye, M.J., and Morris, G.A., “Analysis of Flexibly Connected Steel Frames”, Canadian J. of Civil Engrs., Vol. 2, No.
3, Sept. 1975, pp. 280–291.
5. Hodge, P.G., Jr., “Plastic Analysis of Structures”, McGraw-Hill Book Co., New York, 1959.
6. A.S.C.E., “Plastic Design in Steel”, Manual No. 41, 2nd Ed., 1971.
7. Ackroyd, M.H., and Gerstle, K.H., “Behavior of Type 2 Steel Frames”, Jnl. Struct. Div., ASCE, Vol. 108, No. ST7,
July 1982, pp. 1541–1556.
8. Richart, R.M., Private Communication, 1987.
9. Goverdhan, A.V. “A Collection of Experimental Moment-Rotation Curves …”, M.S. Thesis, Vanderbilt University,
Nashville, Tenn. 1983.
10. Stelmack, T.W., Marley, M.J., and Gerstle, K.H., “Analysis and Tests of Flexibly Connected Steel Frames”, Jnl.
Struct. Eng., ASCE, Vol. 112, No. 7, July 1986, pp. 1573–1588.
< previous page page_308 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_308.html07/11/2009 12:10:33 PM
 page_309

< previous page page_309 next page >

Page 309
SIMPLIFIED DESIGN OF FLEXIBLY-CONNECTED BUILDING FRAMES
M.H.Ackroyd
Techtrol, Ltd.
Montreal, Quebec
ABSTRACT
This paper develops a simple design procedure for unbraced building frames that takes advantage of beam-to-column
connection stiffness in reducing girder sizes, and at the same time, avoids premature column plastification typified in
most Type 2 frame designs.
MOTIVATION
Studies of the behavior of Type 2 Designs (1) have shown that exterior columns receive a portion of “fixed-end” moment
from the gravity loaded girders that does not get accounted for in the Type 2 assumptions. Correspondingly, the girders
tend to be overdesigned, because Type 2 assumptions do not take advantage of the rotational restraint actually provided
by the “simple” connections. In some frames above nine stories tall, these analysis approximations can result in the onset
of frame instability at load factors below 1.3, but for all frames, these approximations will result in oversized girders.
Therefore, one would like to modify the prediction model to account for the beneficial effects of connection stiffness on
girder forces and its detrimental effects on exterior columns. Also, one would like to avoid having to rely on
sophisticated computational tools, as that would defeat the original spirit of simplicity of Type 2 design.
APPROACH
A thorough study of the analysis results for the frames studied above showed that the deformed shapes of the frames
were similar to the shape one would obtain by superposing two shapes: (1) the portal shear model associated with wind
loading, and (2) a gravity model having inflection points at midheights of exterior columns and essentially no flexure of
the interior columns.
< previous page page_309 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_309.html07/11/2009 12:10:34 PM
 page_310

< previous page page_310 next page >

Page 310
Based on these observations, it appears that wind moments are predicted well by the standard portal method of
approximate analysis. However, a new model for gravity loads needs to account for the transfer of moment from girders
into the exterior columns, as well as to account for beneficial “flexible-end” moments on girder ends. This paper
proposes a model for gravity loading of Type 2 frames that can satisfy these criteria as well as provide the basis for
simple design aids, suitable for manual computation.
Two subassemblages have boundary conditions that kinematically map onto the suggested gravity model, above. For
interior bays of a frame, it seems appropriate to neglect the column rotations and consider only the gravity-loaded girder
with flexible rotational springs connecting it to rigid supports. For exterior bays of a frame, it seems appropriate to
assume inflection points at midheights of columns and hence to model only half the column above and half the column
below.
The notation used herein for frame components is:
Ig = moment of inertia of girder
Ic, a = moment of inertia of column above
Ic, b = moment of inertia of column below
ha = story height of column above
hb = story height of column below
1 = span of girder
k = representative stiffness of flexible connections (both connections identical)
It should be noted that the exterior bay model subsumes the interior bay model in light of the fact that, as moments of
inertia of the columns tend toward infinity, the right end of the model tends toward a rigid support as in the interior bay
model. Therefore, only the exterior bay model needs to be analyzed as the general model of behavior.
The analysis used here is the widely used “direct stiffness’’ approach, where individual elemental models are combined
to give the needed structural model. For this structure, only three elemental models are necessary: a standard prismatic
beam element represents the girder, rotational springs represent the flexible connections, and prismatic beams with far-
end-pinned represent the half-columns. The elements and associated stiffness matrices are shown below.
< previous page page_310 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_310.html07/11/2009 12:10:34 PM
page_311

< previous page page_311 next page >

Page 311

GRAVITY MODEL FOR INTERIOR BAYS

GRAVITY MODEL FOR EXTERIOR BAYS

< previous page page_311 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_311.html07/11/2009 12:10:36 PM
page_312

< previous page page_312 next page >

Page 312

ELEMENTAL MODEL FOR GIRDER

ELEMENTAL MODEL FOR FLEXIBLE CONNECTION

< previous page page_312 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_312.html07/11/2009 12:10:37 PM
 page_313

< previous page page_313 next page >

Page 313

ELEMENTAL MODEL FOR HALF COLUMNS

Using the direct stiffness method, the assembled structural stiffness equations become:

Now, if one assumes that only symmetric gravity loads are applied to the girder, then the loading to be applied to the
structural model is simply the conventional fixed-end beam moments, Mf, specified in degrees of freedom 2 and 3:

< previous page page_313 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_313.html07/11/2009 12:10:38 PM
 page_314

< previous page page_314 next page >

Page 314
Next, this 3-by-3 system is readily inverted in algebraic form to give expressions for the three rotations, 01, 02, 03.
These rotations can then be back-substituted into the rotational spring models at each end of the girder to obtain
expressions for the “flexible end moments” (as contrasted to “fixed end moments”) to be used in the frame design. After
much algebraic simplification, these expressions become:

where Ci and Ce are coefficients which convert the standard fixed end moments to the proper “flexible end” moments
for the flexibly-connected frame being designed. They are defined as

and the following nondimensional terms are defined:

Connection Stiffness Parameter, a=EIg/k1

Relative Flexibility Factor,

Note that the “Relative Flexibility Factor’’ is simply the inverse of the well-known relative rigidity factor, G, used for
estimating effective lengths of columns using the Jackson/Moreland nomographs.
Note that as connection stiffness, k, and column stiffness, Ic/h, become indefinitely large, then both “a” and “g”
approach zero, the coefficients, Ci and Ce approach 1, and the “flexible end moments” become the usual fixed end
moments, Mf.
A final point of corroboration is obtained by letting “g” become infinite (far-end pinned condition) and “a” become zero
(fully continuous beam). It can be seen that the right moment, Me, goes to zero, and the left moment, Mi, goes to 1.5Mf.
This would also be given by superposing an equal—and-opposite “Mf” at the right end of the girder to nullify the fixed-
end “Mf”: the left moment becomes 1.5Mf with the 0.5 carry-over factor.
< previous page page_314 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_314.html07/11/2009 12:10:39 PM
 page_315

< previous page page_315 next page >

Page 315
The above expressions are easily automated on a programmable calculator. Alternatively, a plot of realistic ranges of the
Connection Flexibility Parameter and Relative Flexibility Factor can provide a simple design aid for obtaining flexible
end moments on girders, as illustrated below.
PROPOSED DESIGN PROCEDURE
The above development for predicting flexible end moments can be readily incorporated into the usual Type 2 design
procedure, by appending refinement design calculations to the Type 2 results. This section summarizes the proposed
Modified Type 2 Design procedure. Verification of its safety and economies that can be achieved in its use are described
in reference 1.
MODIFIED TYPE 2 DESIGN PROCEDURE
STEP 1: FRAME DEFINITION (same as Type 2)
STEP 2: LOADING DEFINITION (same as Type 2)
STEP 3: ASSUMPTIONS
Assume column effective length factor for in-plane buckling and moment amplification, Kx=1.5
STEP 4: ANALYSIS (same as Type 2) Wind loading—use the Portal Method.
Gravity loading—assume girders are pinned.
STEP 5: DESIGN (same as Type 2)
Use allowable stress design provisions, Part 1 of AISC. Use either extended flange plate connections or Type 2 wind
connections (top-and-seat angles with web clips).
DESIGN CONNECTIONS FOR WIND MOMENT ONLY.
STEP 6: GRAVITY RE-ANALYSIS
a. Calculate initial elastic stiffness of connection, ki (e.g., inch-kips/radian).
b. Calculate representative connection stiffness:
Use k=ki for flange plate connections.
Also compute plastic moment capacity of flange plates, My (e.g., inch-kips).
Use k=0.5*ki for other types of connections.
c. Calculate Connection Flexibility Parameters for girders
a=EIg/k1
d. Calculate Relative Flexibility Factors at exterior joints

< previous page page_315 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_315.html07/11/2009 12:10:40 PM
 page_316

< previous page page_316 next page >

Page 316
e. Determine flexible end moment coefficients for girders

f. Calculate flexible end moments of all girders.

Mi=Mf/Ci (not to exceed My, if flange plates)
Me=Mf/Ce (not to exceed My, if flange plates)
g. Adjust frame moment diagrams for new flexible end moments.
Girder midspan moments Exterior column axials

Interior column axials

Column moments

REPEAT STEPS 5 AND 6 UNTIL NO CHANGE

REFERENCES
1. Ackroyd, M.H., “Design of Flexibly-Connected Steel Building Frames”, Final Report to American Iron and Steel
Institute, Project No. 333, Rensselaer Polytechnic Institute, Troy, N.Y., November, 1985.
< previous page page_316 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_316.html07/11/2009 12:10:41 PM
page_317

< previous page page_317 next page >

Page 317

DESIGN AID FOR FLEXIBLE END MOMENT CALCULATIONS

< previous page page_317 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_317.html07/11/2009 12:10:42 PM
 page_318

< previous page page_318 next page >

Page 318
DISCUSSION OF PAPERS ON FRAME STABILITY AND SIMPLIFIED METHODS
The papers that were presented in this session gave significant information on the development of practical approaches to
column and frame stability, as well as the development of methods that may be feasible for design office usage. The
correlation between theory and tests has been found to be very good for the advanced solution techniques; it was agreed
that a major effort should be undertaken to bring semi-rigid concepts into practice, and the simplified method approach
may be the preferred way.
Several speakers and discussers affirmed earlier comments regarding the influence of the initial and unloading stiffnesses
of the beam-to-column connections. This would also apply to column footings, although research should be conducted to
establish the full range of moment-axial load interaction, and the effects of this on the M-Φ-relationship for the footing.
However, it was indicated that the column footings generally provide much more restraint than what has been assumed
in the past, especially for longer columns. The need for full-scale tests was emphasized, to provide further correlation
studies with the theoretical findings. Long-term effects and cyclic loads were also mentioned; it was pointed out that
both may lead to a deterioration of the stiffness.
Good correlation was found between the stability studies of individual columns and columns in subassemblages and
frames, especially insofar as the influence of the connection stiffness is concerned. The application of the column
stability solution to unbraced frames was debated; originally it was thought that the effect of the transfer of the beam
moment into the column at the instant of buckling would make for an unconservative application. However, the frame
subassemblage and full-scale tests had demonstrated that the buckling capacity and behaviour were much less sensitive
to the moment transfer, which would appear to indicate that the column stability solution may have wider applicability.
This is a subject in need of further research.
There was general agreement that the development of simplified frame analysis and design methods had the potential for
offering economy in design as well as construction. For example, taking advantage of the end restraint would lead to
lower girder stress levels. However, several discussers noted that serviceability checks would be critical. It was also
pointed out that revised drift limits would have to be developed
< previous page page_318 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_318.html07/11/2009 12:10:42 PM
 page_319

< previous page page_319 next page >

Page 319
for frames with semi-rigid connections, recognizing that the commonly used (average) limit of 0.0025 for the ratio of
building drift to building height had been arrived at for structures with rigid connections. In other words, if the
computations of the drift values are more accurate, then the drift limits will have to be revised. This would appear to be a
suitable topic for a major research effort.
Summarizing, there seems to be substantial agreement on the buckling characteristics and controlling parameters for
members and frames. Ultimate limit states are well defined, the solutions are less sensitive to connection stiffness
variations than previously thought, and tests and theory compare favorably.
< previous page page_319 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_319.html07/11/2009 12:10:43 PM
 page_320

< previous page page_320 next page >

Page 320
This page intentionally left blank.
< previous page page_320 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_320.html07/11/2009 12:10:44 PM
page_321

< previous page page_321 next page >

Page 321
SESSION No 5
DESIGN REQUIREMENTS
< previous page page_321 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_321.html07/11/2009 12:10:45 PM
 page_322

< previous page page_322 next page >

Page 322
LRFD CONNECTION DESIGN IN USA
Robert O.Disque
Director of Building Design Technology
American Institute of Steel Construction
ABSTRACT
The American Institute of Steel Construction has been researching a limit state design procedure for structural steel since
1969. The principle investigator has been Dr. T.V.Galambos, then of Washington University and now at the University
of Minnesota. The design specification and Manual of Steel Construction was completed in 1986. The USA is now in the
process of converting from the traditional allowable stress design (ASD) to a limit state procedure, termed Load and
Resistance Factor Design (LRFD). This paper presents the basics of how connections are designed using LRFD.
Basic Behavior
All steel connections have a unique Moment-Rotation Relationship. See Figure 1. The characteristics of the curve
depend on the thickness of the material, size and number of bolts, configuration of detail material, etc. The state of the
art does not permit a designer to predict the behavior of the connection, except in a few well defined cases. LRFD
classifies frames according to the type of connection—partially restrained (PR) or fully restrained (FR). The designer
must decide which of these he/she has before analizing the frame.
Fig. 2 shows examples of FR connections. Fig. 3 shows examples of PR connections.
< previous page page_322 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_322.html07/11/2009 12:10:46 PM
page_323

< previous page page_323 next page >

Page 323

FIG. 1

FIG. 2

FIG, 3
< previous page page_323 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_323.html07/11/2009 12:10:47 PM
 page_324

< previous page page_324 next page >

Page 324
Bolts in Tension
The design load in tension is specified as (.75) Fu(Ag), where is the resistance factor and in this case equals 0.75. Fu
is the ultimate strength of the bolt material and Ag is the gross area of the bolt. In contrast to what may be permitted in
some European specifications, the bolts must be pretensioned. Prying action may be a factor in bolt size and material
thickness.
Bolts in Shear—Strength
Bolts in bearing type connections transfer shear load through the shank of the bolt. Pretensioning does not affect this
capacity. In many applications, therefore, pretensioning is not required. In these cases, the connection is termed ‘‘snug-
tight”. In cases where the performance of the connection is critical, pretensioning is required. The specification defines
the two cases. If the threads are excluded from the shear plane, the design values are higher than when they are included.
This philosophy is the same as ASD.
Bolts in Shear—Serviceability
In cases where frequent load changes are contemplated which might result in a fatigue failure, bolted connections should
be designed as slip-critical, formally called friction-type. Since this type of loading only occurs at working loads the
connection is designed at service loads rather then at factored loads as is the usual case with LRFD. The specified
resistance is appropriately reduced.
Bolts—Material in Bearing
There are two conditions for the material bearing value.
Condition 1—Edge distance is more than 1.5 times the bolt diameter.
Condition 2—Edge distance is less than 1.5 times the bolt diameter.
Condition 1 contemplates a failure by gross distortion in the material. Condition 2 protects failure by tearing out a “slug”
of material between the bolt and the end of the material.
For condition 1 the design load is specified as 1.8xFux bolt diameterxt (material thickness). For Condition 2,
< previous page page_324 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_324.html07/11/2009 12:10:47 PM
 page_325

< previous page page_325 next page >

Page 325
the design load is .
Because of the two conditions, connection design may require that the bolt loaded next to an edge may have a different
bearing value than the other bolts.
For Condition 1, the specified value is with . If the designer is not
concerned with hole distortion, 3.0 may be substituted for 2.4.
Bolts—Block Shear
Block shear is a failure mode first codified in the AISC Manual in 1980. The LRFD procedure is different but more
logical. There is no longer an edge distance requirement normal to the load but block shear must be checked for either of
two possible failure modes: tension yielding and shear fracture or tension fracture and shear yielding. The larger of the
two governs because block shear is a tearing phenomenon and the fracture term is the determining one and it is not
obvious which is that term. Tension yield is taken as 0.75 Fy; shear fracture= 0.45Fu. Tension fracture=0.75xFu; shear
yield= 0.45Fy.
Bolts—Interaction
Interaction equations are specified for both strength at factored loads and slip at service loads. For strength, the design
tension is,

The slip critical allowable shear is,

where tb is the pretension
Bolts—Eccentric Loads
When a bolt group is loaded eccentrically (Fig. 4), rotation occurs about an “instantaneous center”. The procedure for
determining the design load is as follows.
1. Assume location of instantaneous center (i.c.).
2. Determine Ri for the heaviest load bolt (farthest from the center of gravity) from,

< previous page page_325 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_325.html07/11/2009 12:10:49 PM
 page_326

< previous page page_326 next page >

Page 326
3. Ri for all the bolts is calculated based on determined from the linear ratio of the bolt distance from the i.c.
4. R and the moments for all bolts are summed and the statics for the group is checked.
5. The location of the i.c. is iterated until equilibrium is satisfied.

FIG. 4
Bolts—Prying Action
For certain devices where a bolt is in tension (Fig. 5) the bolt is subjected to additional load. This, of course, adds to the
bending in the plate. Although, the phenomenon is complex, AISC has simplified it by assuming that the bolt receives
additional tension. In reality, the additional load is probably a combination of tension and bending. Fig. 6 diagrams the
procedure as formulated by AISC.
The first extreme is when the plate is very thin and in double curvature. In this case the bolt is loaded the
most. The second extreme is with a thick plate in single curvature. Here the bolt receives only the
calculated applied load. It is necessary that α be assumed to be within 0.0 and 1.0.
< previous page page_326 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_326.html07/11/2009 12:10:49 PM
 page_327

< previous page page_327 next page >

Page 327

FIG. 5

FIG. 6
The dimensions used in the formulas are shown in Fig. 7
The formulation follows:

< previous page page_327 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_327.html07/11/2009 12:10:50 PM
 page_328

< previous page page_328 next page >

Page 328

FIG. 7
The plate thickness is calculated to resist bending and is,

The design procedure shown here is for a strength limit state and as such, used with factored loads. For a serviceability
check, as in the case of fatigue loading, a different formulation is required. In this case, the designer is referred to the 8th
Edition of the AISC Manual of Steel Construction.
Welds—Fillet
The LRFD design value for fillet welds is exactly 1.5 times that for allowable stress design. Identical designs result when
the live load—dead load ratio is 3.0. This design value is,

where,

FEXX=Classified strength of weld metal

A study of tests on fillet welds indicates that the LRFD specified value is very conservative—especially when the load is
transverse to the longitudinal axis of the weld.
< previous page page_328 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_328.html07/11/2009 12:10:51 PM
 page_329

< previous page page_329 next page >

Page 329
Welds—Eccentric Loads
An instantaneous center procedure, similar to that used for bolts, has been developed for eccentrically loaded weld
groups. However, the welded case is very much more complicated because the strength deformation curve for fillet
welds depends on the angle to which the weld is loaded. As seen in Fig. 8 it was assumed that the ultimate strength value
was a straight line and equaled the specified value of 0.6xFEXX. When the deformation on a particular weld segment
was small the computer program followed the load—deformation curve down the real curve as indicated in Fig. 8. When
the strength of the weld group was determined, a of 0.75 was applied.
Groove Welds
The design strength of groove welds is essentially the same in LRFD as in ASD, that is, it matches the strength of the
material it joins. The effective throat dimension depends on the welding process, welding position and the included angle
at the root of the groove.

FIG. 8
< previous page page_329 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_329.html07/11/2009 12:10:55 PM
 page_330

< previous page page_330 next page >

Page 330
STRUCTURAL STEEL CONNECTIONS IN CANADA CURRENT REQUIREMENTS AND FUTURE TRENDS
Michael I.Gilmor
Canadian Institute of Steel Construction
201 Consumers Road, Suite 300
Willowdale, Ontario, M2J 4G8
CANADA
ABSTRACT
The design of structural steel connections in Canada as covered under CSA Standard CAN3-S16.1-M84 is discussed,
particularly as related to the ultimate strength of bolts and welds, and to the serviceability of slip-resistant connections.
The results of recent research on partial joint penetration groove welds, longitudinally and transversely loaded fillet
welds, gusset plate connections, and the strength of fillet welds in HSS truss joints are presented, as well as some of the
conclusions drawn to date.
INTRODUCTION
In 1974 Canada adopted its first structural steel design code based on limit states design, CSA Standard S16.1, “Steel
Structures for Buildings-Limit States Design”, as an alternative design philosophy to the traditional allowable stress
methodology. Since then, an ever increasing number of steel structures have been designed with this new standard,
ranging from shopping centres to a 40 000 tonne nuclear power plant. With the introduction of the 1985 edition of the
National Building Code of Canada, the older allowable stress design standard, the 1969 edition, was no longer
referenced, leaving the 1984 edition of the limit states design standard [1] as the only structural steel design standard in
Canada.
Although the formats differ from their allowable stress counterparts, the clauses related to connections covered similar
areas but have been expanded where new knowledge was available. In S16.1, strength requirements for bolts in tension,
in shear, and in combined shear and tension are given. The
< previous page page_330 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_330.html07/11/2009 12:10:56 PM
 page_331

< previous page page_331 next page >

Page 331
strength of the plate elements within a bolted joint are written as bearing requirements. Serviceability requirements of
bolted connections, where slip at service level loads is critical to the usefulness of the structure, are written in terms of
the total slip resistance of the joint.
For welds, resistances are given for complete and partial joint penetration groove welds, fillet welds, as well as plug and
slot welds, in terms of the strength of the weld metal and the base metal, at the joint.
S16.1 permits simple construction, special simple construction, and continuous construction. Special simple construction
permits the use of connections to resist gravity loads on the basis of simple construction and lateral loads by distributing
the moments due to lateral loads among selected joints of the frame. However the Standard qualifies this method and its
use seems limited to smaller structures, as there are no rules set down in S16.1 for the method of analysis nor are the
behavioural assumptions given for various connection geometries.
For continuous construction, the Standard does provide the requirements for rigid joints similar to those in AISC’s
Specification.
CURRENT REQUIREMENTS
Bolted Connections
S16.1 is concerned primarily with high strength bolted connections and describes their tensile and shear resistances in
terms of the ultimate tensile strength of the bolt, Fu, the cross-sectional area of the bolt based on the nominal diameter,
and the following ratios, as appropriate, the tensile stress area to nominal area, the root area to nominal area, shear stress
to tensile stress. The resistance factor, the number of bolts, and the number of shear planes complete the specification of
the joint’s resistance.
For serviceability, the slip resistance is stated in terms of the ultimate tensile strength of the bolt, the nominal area of the
bolt, and a factor relating slip probability for the surface type and condition, tensile stress area to nominal area, and
preload to ultimate tensile load.
< previous page page_331 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_331.html07/11/2009 12:10:57 PM
 page_332

< previous page page_332 next page >

Page 332
The bearing resistance of the plate material within a bolted joint is stated in terms of the ultimate tensile strength of the
plate, its thickness and the end distance of the bolt.
Bolted joints in shear must meet the shear and bearing requirements and may also meet the slip requirements. However,
slip-resistant joints are considered to be the exception rather than the rule. Installation of high strength bolts is restricted
to turn-of-nut method or the use of direct tension indicators.
For inspection, the emphasis is on the observation of the actual bolt installation. For bearing-type connections subject to
shear, visual inspection, that the bolt has been tightened, is all that is necessary.
Welds
Under limit states design the technical requirements for welds are virtually unchanged from allowable stress standards.
Depending upon the type of load and its orientation, the weld resistance is taken as
i) the same as the base metal,
ii) the lesser of the shear of the base metal on the fusion face, or
iii) shear on the effective throat of the weld as a function of the area of weld, the ultimate tensile strength of the weld
metal, based on its electrode classification, and the ratio of shear to tensile strength of the weld metal.
The Standard does permit the designer, for the weld group in a connection, to use an ultimate strength analysis method to
determine the resistance, instead of the traditional elastic analysis method, but no method is given.
Restrained Members
Members retrained either due to full or partial end fixity must be designed for shear and moment. S16.1 provides
requirements for column web
< previous page page_332 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_332.html07/11/2009 12:10:58 PM
 page_333

< previous page page_333 next page >

Page 333
stiffeners when beams frame into H-type column members based on the 1959 work of Graham, Sherbourne, Khabbaz
and Jensen.
For corner and “T” type joints, the designer is cautioned to detail the joint to minimize the risk of lamellar tearing.
Bolts and Welds in Combination
While bolts and welds are permitted in the same shear plane, the number of bolts are to be determined based on slip
resistance at service loads, and the ultimate resistance of the connection is limited to the greater of that of the weld or
bolt group. This recognizes that the load-deformation response of bolts differs from that of welds. Other than performing
a load-deformation analysis of the connection as a whole, this conservative approach, to take the larger of the two
individual resistances, is used.
RECENT RESEARCH RESULTS
As the cost of fabricated, erected structural steel is influenced heavily by the man-hours required for fabrication, it is
essential to ensure that connections are both economical and reliable. In the past few years a number of research projects
have been undertaken in Canada which deepen our knowledge on connection behaviour and reliability. This portion of
the paper will touch on some of their results.
Partial Joint Penetration Groove (PJPG) Welds
According to any standard in North America for welding, AISC [2], AWS [3], or CSA, a PJPG weld is permitted to
carry, in tension perpendicular to the weld axis, only two-thirds of the load of a complete joint penetration groove
(CJPG) weld of similar weld size and electrode strength. This produces considerable fabrication difficulties when
backing bars cannot be complete or when the root pass cannot be backgouged and rewelded, in order to qualify as a
CJPG weld. Thus a joint with as much as 99.9% penetration can, by specification, carry only 66% of the load in the
parent plates.
In 1977, Popov and Stephen [4], in a limited series of tests, investigated the strength of pairs of PJPG welds, in static and
cyclic behaviour, in splices of W14 columns. Welds from complete to only 23% penetration were tested.
< previous page page_333 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_333.html07/11/2009 12:10:59 PM
 page_334

< previous page page_334 next page >

Page 334
Recently, Kennedy and Gagnon [5] tested 75 specimens with groove welds of 5 different penetrations, ranging from 20%
to 100% of the plate thickness, in plates of two steel grades, some singly and some in pairs. The PJPG welds tested
singly were intended to demonstrate any influence of eccentricity of load with respect to the weld axis. All welds were
single V 45° bevel SMAW welds made with 3.2 mm diameter E48018 electrodes.
For the specimens tested singly, the mean failure stress of the weld was 1.15 times the ultimate stress of full penetration
welds with a CoV of 10.2%. For welds tested in pairs the ratio was 1.17 with a CoV of 11.2%.
Kennedy and Gagnon noted that their results were consistent with those of Popov and Stephen, as well as Lawrence and
Cox, and drew the following conclusions:
(a) ‘‘…such welds on a unit area basis are as strong as full penetration welds…”;
(b) “Welds tested singly (and therefore eccentrically) were as strong as double welds tested concentrically. The
specimens with eccentrically loaded welds deflected sideways reducing the eccentricity on the welds as well as the
ductile straining produced a mean uniform stress distribution across the weld throat.”; and,
(c) “As the percent penetration decreases the restraint to lateral deformation increases with the result that the weld
strength increases over that proportionate to the weld area.”
For design, Kennedy and Gagnon recommended that “PJPG welds loaded in tension be designed on the basis of the
percent penetration multiplied by the ultimate tensile resistance of the plate. Root passes must be properly made to
achieve uniformity of depth and the desired penetration.”
Fillet Welds
Current weld design standards base the design for fillet welds on the strength of longitudinal fillet welds regardless of the
direction of loading [1– 3]. Although CSA Standards S16.1 and W59 [6] permit an ultimate strength analysis none is
given. In 1971, Butler and Kulak [7] proposed that ultimate
< previous page page_334 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_334.html07/11/2009 12:11:00 PM
 page_335

< previous page page_335 next page >

Page 335
strength be calculated from the load-deformation response of the weld as a function of the angle of load using the
instantaneous centre of rotation model.
Although strengths of fillet welds have been studied since the early 1930’s, few tests have been done on fillet welds
loaded at intermediate angles. Butler and Kulak, Clark [8], Holtz and Harre, Swannell and Skewes [9], and Biggs et al
[10] have contributed to the experimental data while, Neis [11], and Marsh [12] have attempted to develop mathematical
models.
In a carefully designed and instrumented test programme involving 42 specimens (seven different loading angles, two
weld sizes with three specimens for each combination), Miazga and Kennedy [13] showed that the strength of fillet weld,
determined by the vector sum of the longitudinal and transverse stresses, is overly conservative.
Miazga and Kennedy demonstrated that an analysis method based on the maximum shear stress failure criteria is in good
agreement with the test data, with a test-to-predicted ratio of 1.01. Further, the maximum shear stress failure criteria is
better than methods based on either maximum normal stress or von Mises strain energy of distortion failure criteria.
In their tests, all specimens failed in the weld while the plates were not strained beyond the yield strain. Miazga and
Kennedy also examined the role, if any, of frictional forces, as proposed by the International Institute of Welding (IIW),
in the analysis of the results and concluded that they “do not appear to be a significant factor in the analysis”.
In an examination of the measured weld deformations, they stated: “By normalizing the deformations of the welds,
measured in the direction of loading, to the gauge lengths used to obtain the deformation, it is concluded that the fillet
weld ductility is essentially independent of the loading angle, and that the deformations are proportional to the gauge
length”.
Gusset Plate Connections
While gusset plates have been used for decades, a rational design method has not been fully developed. Since
Whitmore’s [14] effective width concept not much has been reported until work by Bjorhovde [15], and Richard [16].
< previous page page_335 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_335.html07/11/2009 12:11:01 PM
 page_336

< previous page page_336 next page >

Page 336
These recent programmes have combined physical tests with finite element models to better define tensile behaviour.
However, in certain circumstances concerns have been raised when a single gusset plate lapped to a connection plate
were subject to compressive loads. Cheng [17] has shown, based on limited tests, that the effective width approach can
grossly over-estimate the collapse load of the connection while ASSHTO’s limit of 930/√Fy for an unsupported edge is a
conservative limit in such situations.
An obvious way to avoid the problem is to detail the connection to ensure that a stiffener, even to the connection plate, is
provided. Certain shapes can be coped or blocked to ensure a flange or leg is retained to act as a stiffener on one side of
the connection. Alternatively, an angle can be attached to the connection plate to provide the necessary stiffness.
HSS Truss Connections
While much research has been done on the joint strength of HSS members connected in a truss, less attention has been
paid to the actual performance of the connectors, usually the welds. When fillet welds are used, a weld leg size of 1.2
times the wall thickness has been recommended. However, in many situations, chords and webs are sized for more than
strength. Often the joint strength rules ensure that the web members are larger than required solely to carry calculated
loads. Thus a design methodology relating the size of the fillet weld to the calculated force to be carried by the truss
member would be more satisfactory and more economical.
Work is currently underway at the University of Toronto under Jeff Packer to investigate the strength and behaviour of
fillet welded HSS truss connections. Tests on isolated joints have been completed [18] and full scale truss tests are now
being planned.
It is hoped that this research will develop rational design rules for fillet welded HSS truss joints to complement the IIW
joint strength rules.
CONCLUSIONS
While adopting a limit states design standard has helped focus both the code writers and the designers attention on the
actual failure mechanisms of connections, research continues to point to areas where the margins of
< previous page page_336 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_336.html07/11/2009 12:11:01 PM
 page_337

< previous page page_337 next page >

Page 337
safety are inconsistent and improvements can be made with significant cost savings. Controlling weld volume in any
structural steel joint results, not only in reduced fabrication costs and improved ability to compete with concrete
structures, but also in reduced shrinkage and distortions in the work piece, and more favourable residual stresses.
REFERENCES
1. Canadian Standards Association, CAN3-S16.1-M84, Steel Structures For Buildings (Limit States Design), Canadian
Standards Association, Rexdale, 1984
2. American Institute of Steel Construction, Load and Resistance Factor Design Specification for Structural Steel
Buildings, American Institute of Steel Construction, Chicago, 1986
3. American Welding Society, Structural Welding Code—Steel, American Welding Society Inc., Miami, 1986
4. Popov, E.P. and Stephen, R.M., Tensile capacity of partial penetration groove welds, Jour. Struc. Div., American
Society of Civil Engineers (ASCE), New York, 1977, 103, ST9, pp.1721–1729
5. Gagnon, D.P. and Kennedy, D.J.L., The ultimate strength of partial joint penetration groove welds, paper to be
delivered at the Canadian Engineering Centennial Convention, Canadian Society of Civil Engineering, Montreal, 1987
6. Canadian Standards Association, Welded Steel Construction (Metal-Arc Welding), W59, Canadian Standards
Association, Rexdale, 1984
7. Butler, L.J. and Kulak, G.L., Strength of fillet welds as a function of direction of load, Welding Research Supplement,
Welding Reasearch Council, May, 1971, pp. 231s-234s.
8. Clark, P.J., Basis of design for fillet-welded joints under static loading, Proc. Conf. on Improving Welded Product
Design, The Welding Institute, Cambridge, England, Vol 1, pp. 88–96
9. Swannell, P. and Skewes, I.C., The design of welded brackets loaded in plane: Elastic and ultimate load techniques,
Civil Engineering Report, University of Queensland, Brisbane, Australia, Feb., 1977
10. Biggs, M.S., Crofts, M.R., Higgs, J.D., Martin, L.H., and Tzogius, A, Failure of fillet weld connections subject to
static loading, Proc. Conf. Joints in Structural Steelwork, Pentech Press, London, England, 1981,pp. 1.92–1.109
11. Neis, V.V, New constitutive law for equal leg fillet welds, Jour. Struc. Div., ASCE, Vol 111, No. 8, pp. 1747–1759
12. Marsh. C., Strength of aluminum fillet welds, Welding Journal, Welding Research Council, Vol. 64, No. 12, pp. 335s-
338s
< previous page page_337 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_337.html07/11/2009 12:11:05 PM
 page_338

< previous page page_338 next page >

Page 338
13. Miazga, G.S. and Kennedy, D.J.L., Behaviour of fillet welds as a function of the loading angle, Structural
Engineering Report, University of Alberta, No. 133, Mar., 1986
14. Whitmore, R.E., Experimental investigation of stresses in gusset plates, Engineering Experimental Station Bulletin,
University of Tennessee, No. 16, Knoxville, 1952
15. Bjorhovde, R. and Charkrabarti, S.K., Tests on full-sized gusset plate connections, Jour. Struc. Div., ASCE, Vol,
111, No.3, Mar., 1985, pp. 667–684
16. Richard, R.M., Rabern, D.A., Hormby, D.E., and Williams, G.G., Analytical models for steel connections, Proc.
Behaviour of Metal Structures W.H. Munse Symposium, ASCE, May 17, 1983, pp. 128–155
17. Cheng, J.J. R., and Hu, S.Z., Compressive tests of gusset plate connections, paper presented to Annual Meeting,
Structural Stability Research Council, Mar. 25, 1987
18. Frater, G.S., Weldment design for hollow structural sections, M.A.Sc. Thesis, Dept. Civil Engineering, University of
Toronto, Dec., 1986
< previous page page_338 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_338.html07/11/2009 12:11:05 PM
 page_339

< previous page page_339 next page >

Page 339
PRACTICAL CONSIDERATIONS IN THE DESIGN OF FRAMES WITH PR CONNECTIONS
Stanley D.Lindsey, Ph.D., S.E.
Stanley D. Lindsey & Associates, Ltd.
1906 West End Avenue
Nashville, TN 37203 USA
ABSTRACT
The codification process that will allow structural engineers to design buildings using PR connections has been
completed in the United States with the publication of the LRFD Specification. The design of such frames requires a
different design approach due to the non-linear response of the frames. The writer’s firm has had experience in the
practical application of the frames using PR connections. The purpose of this paper is to emphasize some of the most
important factors for the practicing engineer to evaluate when designing buildings with PR connections.
One of the most promising structural concepts now receiving a great amount of attention in both research and in practical
applications is the use of PR connections in frames. In the United States with the recent publication of the current limit
design specifications[1]—LRFD—use of actual connection restraint has been recognized. Through the LRFD
specifications, the designer has the guidelines to produce safe designs that employ PR connections. The field is now
open. While the design rules are there through LRFD, the actual experience and recorded performance of frames in real
buildings with PR connections are far behind the technical design aspects. It is true that frames using other assumptions
of connection behavior, such as the Type 3 semi-rigid of the ASIC specifications, have been used for many years.
However, frames using the actual moment rotation curves for the connections have not yet been designed in great
numbers in the United States. Thus, a limited performance base is available to the profession. This paper will describe
some of the important design aspects that have been learned through the writer’s actual design of buildings with PR
connections.
Very importantly, a word of caution is given. The design of buildings with PR connections requires a special level of
design consideration to produce a safe and economical end result. The design
< previous page page_339 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_339.html07/11/2009 12:11:07 PM
 page_340

< previous page page_340 next page >

Page 340
problem is non-linear and that one fact alone sets the design apart from the normal design process used by the majority
of today’s engineers. The entire design process for a system that responds in a non-linear fashion as compared to one that
has a linear response is different. Unless done properly, a non-linear system will not always have a sufficient degree of
safety and may well be unsafe. Earlier papers[2,3] by the writer have given some specifics concerning inelastic vs.
elastic designs.
One starts any design with PR connections with moment-rotation curves for the connections. An immediate observation
is made. The prediction of the behavior of the system will only be as accurate as the connection moment rotation curves
that are used. The designer must know the limits of his curves. He must not extrapolate beyond known limits and he
must not change the connection from the tested connection that generated the moment rotation curve. While these points
seem obvious when mentioned, in the real world of the practitioner, it is very easy to allow a slight modification to a
connection detail, say, for the convenience of the fabricator. Even the simplest change can substantially affect behavior.
So practically speaking, the designer must watch very, very carefully to see that his design in the field is the same as his
design in his computer. Whereas in buildings with other types, conformance of final design to intended design may not
be as important, but in buildings with PR connections, it is of paramount importance. If what is used is not the exact
same as what was designed, the result could mean disaster.
The analysis routines used for frames with PR connections are extremely important. Assuming valid moment-rotation
curves, the predicted behavior used for design must accurately portray actual behavior in a fashion sufficiently accurate
for design. The writer has found that an elastic second order analysis using non-linear moment rotation curves and
factored loads will produce sufficient results for design. The second order analysis should include, as a minimum, frame
P-delta, loss of column stiffness due to axial load on columns, individual member P-delta, and inelastic moment rotation
curve capabilities. In addition, it should have the computation routines in the program for column K-factors for each
loading. Programs of this type have worked well in the writer’s design practice. More elaborate analysis routines, for
instance, to include partial plastifications of column cross sections has not been found to be necessary for design.
Sufficient accuracy for design of buildings does seem to require such elaborate programs. For research applications;
< previous page page_340 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_340.html07/11/2009 12:11:08 PM
 page_341

< previous page page_341 next page >

Page 341
however, it is a different matter and for these instances it is wise to use the most accurate tool available to verify
research results.
Another important design aspect is to match the proper type PR connection to the frame that is to be designed. As
an example, a very low building several bays deep in the direction of light lateral wind loading does not require a
connection that will develop a large moment in the connection. Therefore, a practical design will use the
connection with just the right amount of restraint, such as a pair of web angles or a very thin end plate. Top and
bottom seat angles or thick end plates are not needed. The writer has found the best economies come from using
the connection that furnishes the least restraint but yet yields a safe structure. The more the restraint, the more
costly the connection so, therefore, it is wise to use the least possible restraint in the connection to achieve the least
expensive connection cost. As lateral loads increase in relation to gravity loads, as the building becomes less and
less deep in the direction of loading and as the building becomes taller, more and more connection restraint is
demanded. The art of matching available connection restraints to required demand is presently being learned by
designers. As the experience is gained, it will be possible to produce a reference of past successes and failures for
the profession to use. The PR connection selection will become an integral part of the design process.
A very important design consideration in frames with PR connections is the drift of these frames. The less the
connection restraint, the more the drift. Much research has been focused on the ultimate loads needed on frames
with PR connections for collapse and shake down. From a practical designer’s view, allowable drift limits for
damage or occupant comfort seem to always come before a frame is even close to collapse. The fact is especially
true if one adopts the philosophy of using the minimum connection restraint required for the actual demands.
The writer’s experience has been that almost without fail, drift requires the most connection restraint, not gravity
loads. In the writer’s designs, drift is controlled by selectively adjusting the available connection restraint to the
demand. If one follows that procedure, a safe building design controlled by allowable drift will result.
One final consideration must be mentioned and that is individual and overall stability. For overall stability, any
analysis routine for frames with PR connections should indicate that unstability exists when the
< previous page page_341 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_341.html07/11/2009 12:11:09 PM
 page_342

< previous page page_342 next page >

Page 342
convergence criteria of the routine is not met. If a valid solution does not exist, then the design is not valid.
Individual member restraint; i.e., effective length, is needed to design compression members. One must be able to
accurately compute the effective restraint offered to a column by the beams connecting with PR connections to
that column. Column K factors are directly influenced by the available connection restraint. The smaller the
available restraint, the larger the K factor. Since available connection restraint is variable under different loads,
K factors will vary with loading. This variance is due generally to the non-linear shape of most moment rotation
curves. In practical terms, the designer must be very careful to use the proper procedures to determine his K
factors, lest he can have a very unconservative design. The less stiff the connection selected by the designer, the
greater this problem becomes. Therefore, practical designs must always be precise in this area of computation.
It is possible to design frames with PR connections and produce safe economical structures. Many factors must be
considered and must be carefully designed to avoid having a structure that does not perform as intended.
Hopefully, this paper has presented some practical aspects for consideration by designers when confronted with
the task of designing frames with PR connections.
REFERENCES
1. Manual of Steel Construction, Load & Resistance Factor Design, American Institute of Steel Construction,
Chicago, Illinois, 1986.
2. Lindsey, S.D., Ioannides, S.A., and Goverdhan, A.V., ‘‘An Analysis and Design of Beams Partially Restrained
Connections,” American Institute of Steel Construction, Engineering Journal, Volume 22, No. 4., 1985.
3. Lindsey, S.D., “Design of Frames with PR Connections,” to be published in the Constructional Steel Research
(Special Issue).
< previous page page_342 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_342.html07/11/2009 12:11:09 PM
 page_343

< previous page page_343 next page >

Page 343
OBSERVATIONS ON THE SEISMIC BEHAVIOUR OF STEEL CONNECTIONS AFTER THE MEXICO
EARTHQUAKES OF 1985.
Enrique Martinez-Romero
Consulting Structural Engineer
Av. Nuevo Leon 54–2nd. floor
Mexico, D.F. 06140 MEXICO
ABSTRACT
The Mexico City Earthquakes of September 19 and 20, 1985 affected about 3000 buildings, 330 of them suffered either
collapsed or severe damage. Since there are about 1,200,000 buildings in Mexico City, while the number of affected
number is large, its percentage is very small. Only 12 Steel buildings received important damage. This paper describes
the principal types of steel construction existing in Mexico City at the time of the Earthquakes, particularly as regards to
its connections, and relates them with its construction time. A recopilation of the Seismic Codes development is given
and the connection’s seismic behaviour is discussed. Proposed implementations of the new Code to incorporate the use
of Flexible Connections are discussed, based on the evidences left by the earthquakes.
INTRODUCTION
Seismic provisions for the design of buildings in Mexico City, have—been incorporated into the Design Codes since
1942. Several revisions of them have taken place since then, based principally on the observations of building damages
and frame behaviour under extreme earthquake loading. The major earthquakes occurred in 1943, 1957, 1962, 1979 and
1985, served to calibrate previous provisions and to incorporate changes in the newest—versions (1).
< previous page page_343 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_343.html07/11/2009 12:11:10 PM
 page_344

< previous page page_344 next page >

Page 344
The construction of steel-framed buildings in Mexico City started back in 1910 with the execution of Museums,
Theatres, Railroad Stations and Monuments. The first multistory-steel-framed office and hotel buildings appeared in the
late 20’s, all of them built with ASTM A-7 type of structural-steel, using built-up-columns and I-beam sections in simple
construction. Lateral load resistance was accomplished either thru shear action of the multiple masonry filler walls or by
elementary double-angled X-bracing systems.
By the mid 30’s some steel framed buildings incorporated knee-braces to the simple beam-to-column connections, in
order to eliminate the need of some of the interior filler walls, without losing its lateral load resistance.
By the late 40’s and early 50’s most of steel framed multistory buildings were incorporating moment-frame action by the
use of moment beam-to-column connections in riveted construction. Welded-moment connection appeared in the 60’s
and high-strength bolded connections in the 70’s.
No specific provisions were made in the Building Codes to account for the flexibility of semi-rigid connections in
particular. However, begining with the Emergency Regulations issued on 1957, the recognition of the inelastic frame
action in the seismic behaviour of building frames, originated a series of different approaches to building seismic design.
The observations of the damage produced by the September 1985 earthquake on several steel-framed buildings, most of
them built before 1957,—contributed in a very significant manner to understand the behaviour of—different types of
beam-to-column connections. Even though most of them—were not specifically designed bearing in mind its amount of
flexibility—and rotation capacity, the assessment of their seismic behaviour displayed interesting parameters, some of
which are discussed in the present paper.
TYPES OF BEAM TO COLUMN CONNECTIONS.
Figure 1 shows some of the most common beam-to-column connections used in riveted construction in the 30’s decade.
In some cases, encasement in
< previous page page_344 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_344.html07/11/2009 12:11:11 PM
 page_345

< previous page page_345 next page >

Page 345
concrete of both beams and columns for fire protection purposes, served—also as a joining material with the filler
masonry walls. This fire proofing added significant strength to both the lateral and vertical loading—capacity of the steel
frame. Standard diagional angle-X-bracing, normally with high slenderness ratio, substituted the wall in some instances,
and fire proteciton was not always supplied.

Figure 1. Riveted beam-to-column connection in Simple-Construction.

Some buildings like the Hotel Del Prado, in Mexico City, had been retrofitted in the late 50’s, after being moderately
damaged by the 1957—earthquake. The steel frame received welded knee braces under and above the beams, to improve
the lateral load capacity of the bare frame. Such a reinforcement proved to be satisfactory inspite of being executed in a
very—crude way. (see elements marked with * in Fig. 1)

Figure 2. Typical beam-to-column connection with flange brace

< previous page page_345 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_345.html07/11/2009 12:11:12 PM
 page_346

< previous page page_346 next page >

Page 346
Figure 2 represents the first serious intent to attain lateral load resistance of the steel buildings in the 40’s without the
need of filler masonry walls. They also used riveted construction and A-7 Steel. This type of construction displayed large
drifts during the macro-earthquake of September 1985, as a result of large joint rotations. It “was often found the knee
brace either buckled or failed in tension, and some of the buildings which survived, suffered permanent lateral
deflections which condemned them for demolition. Surprisingly, some others like the City Hall building complex, at the
Central Square in Mexico City, fared very well showing up some damage only, and are being repaired now.
Figure 3 shows the most typical type of connections found in the steel construction of the 50’s decade. Riveted moment-
connections of this types in moment-resistant-frame construction displayed and outstanding seismic behaviour in most
cases, like the famous 44-story Latino Americana Tower and several other tall buildings, which suffered no damage at
all, nor—given evident signs of distress on past major earthquakes.

Figure 3. Moment-resistant beam-to-column connection typical of the 50’s decade

The very high ductility levels exhibited by these connections overcame the large demands of overloading produced by
the extreme earthquake suffered in 1957, 1962 and 1985, as it will be explained later.
Figures 4 A and 4 B give an indication of the most recent types of beam-to-column connections in moment resistant-steel
framed buildings.
< previous page page_346 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_346.html07/11/2009 12:11:13 PM
 page_347

< previous page page_347 next page >

Page 347

Figure 4. Welded beam-to-column connection (2) and H.S. bolted connection used in moment-resistant.
With the exception of the five-building “Pino Suarez” complex, built—with moment resistant space frame in which one
of the three 23-story towers collapsed upon one of the two 14-story buildings (remaining the other three in the range of
severely damaged to moderately damaged), after the September 1985 earthquake, all other buildings having beam-to-
column connections similar to the one shown in Fig. 4-A, displayed a very satisfactory ductile behaviour (2, 3).
In the case of this particular building complex, an overloading of some floors, the proximity of the building’s natural
period with the ground motion, and the very long duration of the almost constant pulses of the—ground, produced
successive deterioration of the joints in the inelastic—range, causing either buckling or fracture of some beam-truss
diagonal—which propicitated a collapse mechanism through large interstory drifts, enhanced P-∆ effect of columns,
local buckling of the box-column flange—plates, and rupture of the joining weldements.
Practically all the rest of buildings with these types of connections, had a very satisfactory ductile behaviour, although in
some instances, large lateral deflections and insuficcient clear distances between adjacent—buildings, propiciated further
damage by pounding. Buildings with connections of the type shown in Fig. 4-B, displayed a remarkably good seismic
behaviour with no damage suffered. This was also the result of a better—
< previous page page_347 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_347.html07/11/2009 12:11:14 PM
 page_348

< previous page page_348 next page >

Page 348
analysis, design and construction practices developed in the 70’s decade.
DUCTILITY DEMAND
It is convenient to compare first the strengths required by the seismic codes used in the design of the structures inspected
after the earthquake (collapsed, damaged or unaffected) with those demanded by the measured or estimated ground
motions during the 1985 Earthquake and with the supplied strengths (i.e. actual strengths of the constructed buildings at
the time that such an earthquake occurred), in order to better understand the actual behaviour of the various types of
connections included in this document.
The damage assessment carried out on the steel buildings located in the most severely affected area of Mexico City,
identified 102 engineered steel buildings (4). The most common type of framing found in this survey was the MRFO
(Moment-Resistant-Frame-Only), being the second most common the MRFD (Moment-Resistant-Frame with Diagional
Bracing). Therefore, the following discussion concerns mostly to these types of structures.
Based on the seismic coefficients specified by the various codes and on some simplifying but consistent assumptions, the
demanded yielding seismic coefficient Cy, corresponding to the demanded nominal strengths (Vy/W) were estimated on
its upper and lower bounds (5). Figures 5-A and 5-B show these values of Cy compared with the linear elastic demand
strength by the earthquake motion, with 5% damping.

Figure 5. Comparison of Linear elastic demanded strength (LEDS) demanded by the 5% damped Linear elastic response-
spectra (LERS) of the recorded E-W component at SCT station (5).
< previous page page_348 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_348.html07/11/2009 12:11:15 PM
 page_349

< previous page page_349 next page >

Page 349
It can be noted that the strength demanded by the earthquake overexceeded the upper bounds of the minimum nominal
yielding strengths required by the seismic codes since 1942. However inspite of this clear evidence, a lot of buildings
fared quite well. The most satisfactory explanation of it lies in two main reasons: a) The strength supplied by the filler
walls, facades and non-structural elements, normally not considered in designs, and b) a large amount of inelastic
(plastic) redistribution of the internal—forces and moments due to the ductile behaviour of the connections.
The first point is a well recognized effect in which the actual buildings will always supply strengths 30 to 180% larger
than the theoretical design strength of the bare framing only (6). The second one is of interest to us now since it
recognizes the ductile behaviour of the structure. This mere fact moved the Mexican Government to incorporate the
inelastic design spectra into the Building Codes of 1976. The observation of the damage to buildings in September 1985
produced a substantial increase of the seismic coefficients and a more stringent requirements for the ductile behaviour—
of frames. Thus the inelastic response spectra (IRS) obtained by dividing the linear elastic response spectra (LERS) by
the so-called ductility coefficient Q, given by the newest codes, presents a more realistic situation which explains better
the survival of so many buildings in Mexico City. (Fig. 6). According to the 1987 Code, (7) the recomended ductility
values for steel moment-resistant-frames were reduced from the 6 to 2 range specified in 1976 to the range of 4 to 1.
Similarly, the seismic coefficient C corresponding to the soft-grounded area was increased from 0.24g given by the 1976
Code to 0.40g.

Figure 6. Inelastic response spectra (IRS) derived from the 5% damped LERS of the E-W component at SCT, divided by
Q=4.
We could well summarize that the New 1987 Seismic Code, the demanded strength for ductile-moment-resistant frames,
mandate a 167% increase in the linear elastic demand response—spectra, but also a decrease—in the value of the
ductility ratio Q from the 6–4 range for the Same type of frames, to the 4-3 range, for the moment-resis-
< previous page page_349 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_349.html07/11/2009 12:11:16 PM
 page_350

< previous page page_350 next page >

Page 350
tant-frames. All of this represent an increase of 2.5 and 2.2 times the demanded strength required by the 1976 Code, for
the DMRF and the MRF respectively. In addition a further increase of the importance factor from 1.3 to 1.5 essential
buildings, allows an additional margin to the upgraded demanded strength.
FUTURE DESIGN TRENDS AND RESEARCH NEEDED
At this point, the seismic behaviour of steel framed buildings has been superficially identified with the type of
construction. Its supplied strength has been approximately compared with the evolution of the Code’s strength demands
and the 1985 earthquake’s demand, suggesting that the enormous difference between the supplied elastic strength and the
demanded elastic strength was furnished in great part by the many incursions of the steel—frame into its inelastic
behaviour. It was also pointed out that in many—instances, the beam-to-column behaviour displayed large ductility
values—(4 or more) during the long duration of the seismic event, by means of—several stable hysteresis cycles.
However, it is difficult to assess the elastic rotations of the connections and consequently the deformational behaviour of
such connections.—This of course, deals both with the connection flexibility and the panel-zone-deformations, both of
which were indeed responsible in a good part of the large drifts observed in some frames, but scarce energy dissipation.
A revision of the 1987 Building Code is now under consideration, to incorporate sound basis for the limit-state-design
philosophy which accounts for the flexibility of the beam-to-column connection, in order to rely better in the actual
reserve of inelastic strength of the steel frame.
The elastic strength of the frame should be sufficient to resist the design earthquake (10-year recurrence period) without
significant non-structural damage, whereas the ductile inelastic behaviour of the frame, should act as a second-defense
line, to resist extraordinary earthquake (50-year recurrence period), with some structural damage, but without collapse.
Here it will be helpful to evaluate the highly redundant type of cons-
< previous page page_350 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_350.html07/11/2009 12:11:18 PM
 page_351

< previous page page_351 next page >

Page 351
truction represented by the Moment-Resistant-Space-Frames, with boxed columns and trussed girders… How much
ductility does this type of construction provides when subjected to several cycles of extreme earthquake loading?… How
does the collapse mechanism of this type of construction develops in—sucssessive pulses of lateral loading?… What
should the required strenght of the side fillet welds be in a boxed columns to avoid local buckling of the flange plates, as
in the case of the Pino Suarez building complex?…—and finally, what b/t ratios should be recommended for the boxed-
columns plates, to allow a satisfactory hysteretical behaviour during long-duration earthquakes?…
The questions above stated represent only a few of the many unknows—presently intriquing our understanding of the
exceptionally good performance of many engineered steel buildings which fared quite well during the—September 1985
earthquake in Mexico City, like the new Loteria Nacional—building which had every argument to have been damaged
(irregular shape,—corner building, first story soft, about two seconds natural period, etc.), and had not a single window
broken. And by the same talken, the required englightment of our knowledge to ascertain the causes of the damage and
collapse suffered by similar types of buildings like the Pino Suarez Complex, in the same event.
Even though this topic has been partially focussed by some researchers (8), and numerous literature is already available
on connection flexibility in Steel Frames (9), future research is urged on this matter.
REFERENCES
1. Departamento del Distrito Federal. “Normas de Emergencia para las Construcciones en el Distrito Federal’’. Octubre
1985.
2. Hanson R.D., Martin H.W., and Martinez-Romero, E., “Performance of Steel Structures in the September 19, and 20,
1985 Mexico City—Earthquakes”. Proceedings fo the AISC’s National Engineering Conference, Nashville, Tenn. June
1986.
3. Martinez-Romero E. “The behaviour, overview of damage and Retrofit of Steel Buildings after the Earthquakes of
September, 1985 in Mexico City”. Proceedings of the Pacific Structural Steel Conference Auckland, N.Z. August 1986.
< previous page page_351 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_351.html07/11/2009 12:11:19 PM
 page_352

< previous page page_352 next page >

Page 352
4. Martinez-Romero, E. “Damage assessment of Steel Buildings in Mexico City, after the earthquakes of 1985”.
Proceedings of the ASCE’s International Conference: The September 1985 Earthquakes of Mexico City Lessons
Learned and Factors involved”. Mexico City, September 1986.
5. Bertero V.V. “The Mexico Earthquake of September 19, 1985 Performance of Building Structures”. 2nd. International
Conference of Engine ering and Technology. Asociación Mexicana de Ingenieros Mecánicos y Electricistas and Society
of Hispanic Professional Engineers . Mexico City, August 1986.
6. Brokken, S.T. and Bertero V.V., ‘‘Studies on the effects of Infills in Seismic Resistant Construction”., Report No.
UCB/EERC-83/11. Earthquake Engineering Research Center, Univ. of California, Berke ley, CA. October 1981
7. Departamento del Distrito FEderal. “Reglamento para las Construciones en el Distrito Federal, Edición 1987”. (to be
released in june 1987).
8. Rosenbaum, D.B. and Richard, R.M., “Flange Stress and Strains in W-sections welded to Box Columns”. The
Departament of Civil Engine ering and Engineering Mechanics. University of Arizona. Tucson, Arizona. April 1986.
9. Chen, W.F., editor “Connection Flexibility and Steel Frames”. Proceedings of a session sponsored by the Structural
Div. of the ASCE. Detroit Michigan, October 1985.
< previous page page_352 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_352.html07/11/2009 12:11:20 PM
 page_353

< previous page page_353 next page >

Page 353
MODIFIED EUROCODE 3 DESIGN RECOMMENDATIONS FOR HOLLOW SECTION LATTICE GIRDER
JOINTS
J.Wardenier
Department of Civil Engineering
Delft University of Technology
Delft, The Netherlands
ABSTRACT
The Eurocode 3 recommendations have been extensively discussed in all international committees dealing with the
design of hollow section joints. Based on these discussions and additional research, the IIW Subcie XV-E, the Cidect
Joint Working Croup and ECCS-TC-10 have proposed a common modification. The modified recommendations are
presented in this paper.
INTRODUCTION
The design recommendations for hollow section joints in Eurocode 3 have been based on the IIW-XV-E
recommendations published in 1981. Based on the results of new research and re-analyses carried out in the last years,
the IIW-XV-E recommendations for hollow section joints have been revised.
The various international committees (IIW-XV-E, Cidect Joint Working Group, ECCS-TC-10) working in this field have
produced a common document with proposed modifications for Eurocode 3. These modified recommendations are given
in-this paper. Although the IIW/ Cidect recommendations (to be published) cover a larger range of joints for simplicity
the modified Eurocode 3 recommendations cover the same types as previously.
< previous page page_353 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_353.html07/11/2009 12:11:23 PM
 page_354

< previous page page_354 next page >

Page 354
1. Hollow section lattice girder joints
1.1 Scope
In this section design rules are given to determine the static strength of single plane joints in lattice structures composed
of hollow sections (being round, square or rectangular in shape) or combinations of these with open sections.
The general provisions for the scope given in clause… apply, except that the wall thickness is limited to a minimum of
2.5 mm.
1.2 Definition
Single plane joints in lattice structures signify, in this context, connections between members situated in a single plane
which transmit primarily axial loads.
The axial load distribution in a lattice girder may be determined on the assumption that the members are connected by
pin-joints. Secondary moments in the joints, caused by the actual bending stiffness of the joints, and eccentricities within
the limits given in subclause 1.4 can be neglected.
The gap g is defined as the distance measured along the length of the connecting face of the chord between the toes of
the adjacent brace members. The overlap ov is expressed as q/px100% as shown in figure 1.

Figure 1. Gap and overlap of joints

< previous page page_354 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_354.html07/11/2009 12:11:24 PM
 page_355

< previous page page_355 next page >

Page 355
1.3 Welds
In welded joints, the connection should be established over the entire perimeter of the hollow section by means of a butt
weld, a fillet weld, or a combination of the two.
Fillet welds should satisfy the following conditions:

where a=throat thickness of the fillet weld to be determined in accordance with subclause…
This requirement may be waived where departure from them can be justified in regard to strength and to deformation
and/or rotational capacity.
1.4 Field of application
The calculation rules given in this section can be applied provided the following conditions are observed:
i. The angle between the chord and the brace member or between brace members should be at least 30°.
ii. Moments resulting from eccentricities may be neglected in calculating joint strength if the eccentricities are within
the following limits:

where e=eccentricity
do=diameter of the chord
ho=depth of the chord

Figure 2. Eccentricity of joints

< previous page page_355 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_355.html07/11/2009 12:11:25 PM
 page_356

< previous page page_356 next page >

Page 356
The members that meet at a joint must have their ends prepared in such a way that their cross-sectional shape is not
modified.
1.5 Welded joints “between circular hollow sections
The axial force(s) in the trace member(s) due to the design ultimate load should not exceed the design capacity according
to the formulae given in Table 1.
1.6 Welded joints with square or circular hollow section brace mem bers and a square hollow section
chord.
The axial force(s) in the brace member(s) due to the design ultimate load should not exceed the design capacity
according to the formulae given in table 2. Table 2a defines the range of validity of these formulae.
1.7 Welded joints between, rectangular hollow sections
For joints between rectangular hollow sections and joints between square hollow sections outside the range of validity
for 1.6 the design strength should be based on the following criteria where applicable:
i. chord face failure
ii. chord wall failure (yielding or instability)
iii. chord shear failure
iv. chord punching shear failure
v. brace failure with reduced effective width
The design strength may be determined according to: “Design recommendations for hollow section joints”, IIW-
document XV-…-87.
1.8 Welded joints with hollow section brace members and an I-section chord
The axial force(s) in the member(s) due to the design ultimate load should not exceed the design capacity according to
the formulae given in table 3. The range of validity is given in table 3a.
< previous page page_356 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_356.html07/11/2009 12:11:26 PM
 page_357

< previous page page_357 next page >

Page 357
TABLE 1
Design strength capacity of axially loaded welded joints between circular hollow sections

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_357.html (1 of 2)07/11/2009 12:11:27 PM

page_357

< previous page page_357 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_357.html (2 of 2)07/11/2009 12:11:27 PM

 page_358

< previous page page_358 next page >

Page 358
TABLE 2
Design strength capacity of axially loaded welded joints with square or circular brace members and a square hollow
section chord

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_358.html (1 of 2)07/11/2009 12:11:28 PM

page_358

< previous page page_358 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_358.html (2 of 2)07/11/2009 12:11:28 PM

 page_359

< previous page page_359 next page >

Page 359
TABLE 3
Design strength capacity of axially loaded welded joints between hollow section braces and an I or H section chord

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_359.html (1 of 2)07/11/2009 12:11:30 PM

page_359

< previous page page_359 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_359.html (2 of 2)07/11/2009 12:11:30 PM

 page_360

< previous page page_360 next page >

Page 360
TABLE 2a
Range of validity of Table 2

TABLE 3a
Range of validity of Table 3

< previous page page_360 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_360.html07/11/2009 12:11:31 PM
 page_361

< previous page page_361 next page >

Page 361
SYMBOLS
Ai : cross sectional area of member i (i=0, 1, 2)
AQ : effective cross sectional area of the chord for shear
E : elastic modulus of steel

: axial load in the cross section of the chord at the gap

: joint design strength capacity based on load in member i (i=1 or 2)
Ov : overlap
Q : shear force (Q=Ni · sin θi)
Qp : shear yield capacity of a section
a : throat thickness of a fillet weld
bi : external width of a square or rectangular hollov section (RHS) for member i (i=0, 1, 2)
be : effective width for brace to chord connection
: effective punching shear width
be(ov): effective width for overlapping brace to overlapped brace connection
bm : effective width for the web of the chord
c : coefficient
di : external diameter of a circular hollow section (CHS) for member i (i=0, 1, 2)
e : eccentricity of a joint
fo : maximum compressive stress in the chord due to axial force and bending moment
fop : maximum compressive stress in the chord excluding the stress due to the horizontal
brace load components
fyi : design value for the yield stress in member i (i=0, 1, 2)
g : gap between the braces of a K or N joint
g' : gap g divided by the wall thickness of the chord
hi : external depth of a section, for member i (i=0, 1, 2)
hw : depth of the web of an I or H section chord
i : integer used to denote member of joint, i=0 designates chord and i=1–2 the brace
members. Normally i=1 refers to the strut and i=2 to the tie.

n, n' :
(ov) : index used for overlapped brace
ro : radius of an I- or H-section
ti : wall thickness of member i (i=0, 1, 2)
tw : web thickness of an I- or H-section
α : factor giving the effectiveness of the chord flange for shear

β : (average) brace to chord diameter ratio or width ratio

γ : half chord width, or half chord diameter to wall thickness ratio

γm : joint coefficient
θi : included angle between a brace member (i=1 or 2) and the chord
< previous page page_361 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_361.html07/11/2009 12:11:32 PM
 page_362

< previous page page_362 next page >

Page 362
THE RESPONSE OF STEEL STRUCTURES WITH SEMI-RIGID CONNECTIONS TO SEISMIC ACTIONS
Prof. Dr.-Ing. G.Sedlacek, Dipl.-Ing. M-S Koo
Lehrstuhl für Stahlbau, RWTH-Aachen
W. -Germany
and
Prof. Dr. G.Ballio
Department of Structural Engineering,
Politecnico of Milan
Italy
1. Method for the aseismic design in Eurocode 8
The Eurocode 8 provides unified rules for the aseismic design of buildings, which base on the response spectrum
method. Steel structures according to these rules are designed as follows:
A structure is modelled as a linear elastic dynamic system, along which the masses due to dead weight and cooperating
loads are distributed.
The natural vibration periods Ti are calculated and the response values R(Ti) are taken from the response spectrum for
the relevant soil condition.
For a system as in fig. 1 a static horizontal force H, which acts parallel to the assumed seismic direction and is applied at
the centres of the masses, may be determined with equation (1)

(1)
and a safety assessment in view of the limit of the assumed elastic model, i.e. the first attainment of yield stress in the
extreme fibres of the cross-sections, may be conducted. In equation (1) A is the maximum ground acceleration applicable
to the relevant seismic zone and q is the behaviour factor.
This behaviour factor takes account of the energy dissipation due to the real nonlinear structural behaviour, which
depends on the topology and loading of the structural system, the second order effects and the plastic rotations of the
members and their connections in case of local ductility.
< previous page page_362 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_362.html07/11/2009 12:11:34 PM
 page_363

< previous page page_363 next page >

Page 363
The behaviour factor also allows for the real dynamic limit states in relation to the fictitious limit state of initial yielding
assumed in the safety assessment. Hence the q-factor performs a correction factor for the linear dynamic model and its
limit states.
In this paper the influence of semi-rigid connections to the q-factor will be treated.
2. Method for the calculative determination of the q-factors
A rational procedure for the determination of q-factors has been proposed by Ballio, Perotti, Rampazzo, Setti /2/.
According to this method the q-factor is determined by its definition in Eurocode 8, which is represented by the
equations (2-a, 2-b, 2-c):

(2-a)

(2-b)

(2-c)
In equ. (2-a) the qa-factor describes the relation between the maximum ground acceleration a at the real limit state and A
as applied in the safety assessment with the linear response method; in equ. (2-b) the qδ-factor is used to derive the
deflections δ at the real limit state from the deflections δel calculated for A with the linear response method. By the
definition in Eurocode 8 both the qa and the qδ-factors are equal, equ (2-c).
This definition results in a linear dependence between the applied maximum ground acceleration and the maximum
deformation of the structure, or between qa and qδ as demonstrated in fig. 2.
This linear dependence can be confirmed by calculating the qa-qδ graph for several qa-values with the time history
method the accelerogram being defined such that it fills the assumed response spectrum, see fig. 3, and an unlimited
elastic constitutive material law and no second order effects being taken into account.
In applying a more realistic elasto-plastic material model and a second order theory the time-history method produces a
nonlinear relationship between qa and qδ with values for small and moderate qa-values and for
greater qa-vaiues.
The cross point of the nonlinear graph obtained by the time-history method with the straight line gives maximum q-
values that fulfill the definition of equ. (2-a), (2-b), (2-c) on the safe side.
< previous page page_363 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_363.html07/11/2009 12:11:35 PM
 page_364

< previous page page_364 next page >

Page 364
Though this definition of the q-values is rather artificial, the q-values so obtained normaly are close to the physical limits
caused by sway down effects and represent safe side values in view of other limit state as lowcycle-fatigue or fracture of
connections, fig. 4.
These limits of course have to be checked separately, unless their attainment is prevented by special detailing rules, e.g.
fracture strength of the net section > yield strength of the gross section and maximum b/t—ratios.
In the table 1 some numerical values of the q-values for single and double storey-frames, calculated for different natural
periods T(sec) of the structures and different strong motion periods t(sec) of the applied accelerograms are indicated and
compared with the q-values as proposed in the first draft of Eurocode 8.
The probleme yet to be solved is how these q-values are controlled by the behaviour of semi-rigid connections.
3. Modelling of semi-rigid connections in view of their behaviour in structual systems excited by
earthquakes.
The hysteretical behaviour of bolted end-plate connections in terms of moment-rotation dependences has been
experimentally determined /4/. In assuming stable loops the hysteretical behaviour can be idealized according to fig. 6
and be modelled by the truss model of Stutzki, see fig. 7.
In applying this truss model to frames with semi-rigid connections the q-values can be determined in the way described
obove.
In fig. 8 for a single storey frame the δ-a—dependence is calculated for rigid connections and semi-rigid connections.
This example demonstrates that semi-rigid connections cause a greater deflections and hence higher natural vibration
periods Ti which may result in a smaller response R(Ti).
In addition the q-values are greater (by about 20 %) than those for rigid connections.
A third advantage may be realised by the fact that the ultimate moment capacities of rigid connections are smaller than
those of the connected beams and columns, which makes them passiv controlling devices that protect the columns and
the beams from being overstressed.
This may be the philosophy for developing specific semirigid connections for the aseismic design, e.g. friction controlled
connections with slotted holes etc, that behave rigidly due to friction grip for earthquake events in the serviceablity range
and assure energy dissipation by slip-effects without yielding in the ultimate range, with background safety due to
‘‘strain-hardning effects” by the bearing forces at the ends of the slotted holes, fig. 9 /6/, /7/.
< previous page page_364 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_364.html07/11/2009 12:11:36 PM
 page_365

< previous page page_365 next page >

Page 365
References
/1/ EuroCode 8: common unified rules for structures in seismic regions, Commission of the European Communities, Eur.
8850, 1983
/2/ Ballio, G. /Pertti, F. /Rampazzo, L. /Setti, P.: Determinazione del coefficiente di struttura per costruzioni metalliche
sogette a carichi assiali. 2. convegno nazionale l’ingegneria sismica in italia, rapallo 1984
/3/ Koo, M-S: Untersuchung zum Einfluss der Bebendauer,Strukturausbildung und des Verhaltens von Verbindungen auf
die Sicherheit von Stahlbauten bei Erdbebenbeanspruchungen, Diss. RWTH-Aachen 1987
/4/ Indagina sperimentale sulla resistenza e duttilita di collegamenti strutturadi, institute of construction theory of the
university of Pisa, Monografia 5
/5/ Stutzki, Ch.: Traglastberechnung räumlicher Stabwerke unter Beräcksichtigung verformbarer Anschlüsse, Diss.
RWTH-Aachen, 1982
/6/ Roik, K. /Dorka, U.: Vibration Control of structures under earthquake loading by friction-grip elements, Part 1: How
to put brakes on a building. Earthquake eng. and struc. dyn. (demnächst)
/7/ Roik, K. /Dorka, U.: Optimales Systemverhalten unter Erdbebenbelastung durch Reibelemente mit
“dimensionierbarer Dissipation”, Ruhr-Uni Bochum SPD 151-Berichte Nr. 6, März 1987
< previous page page_365 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_365.html07/11/2009 12:11:36 PM
 page_366

< previous page page_366 next page >

Page 366
Table 1: Comparison of q-values according to the draft EC 8/1984 and as numerically determinated

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_366.html (1 of 2)07/11/2009 12:11:38 PM

page_366

< previous page page_366 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_366.html (2 of 2)07/11/2009 12:11:38 PM

page_367

< previous page page_367 next page >

Page 367

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_367.html (1 of 2)07/11/2009 12:11:39 PM

 page_367

fig 1: The response spectrum method and its elements

< previous page page_367 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_367.html (2 of 2)07/11/2009 12:11:39 PM

 page_368

< previous page page_368 next page >

Page 368

fig 2: Definition of the behaviour factor q

< previous page page_368 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_368.html07/11/2009 12:11:40 PM
 page_369

< previous page page_369 next page >

Page 369

fig 3: Example for an artificial accelerogram with a strong motion period of 7sec that fills the response spectrum
< previous page page_369 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_369.html07/11/2009 12:11:41 PM
page_370

< previous page page_370 next page >

Page 370

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_370.html (1 of 2)07/11/2009 12:11:42 PM

 page_370

fig 4 Comparison between the q-value as defined in Eurocode 8 and the real physical limits
< previous page page_370 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_370.html (2 of 2)07/11/2009 12:11:42 PM

 page_371

< previous page page_371 next page >

Page 371

fig 5: Moment-rotation hysteresis for bolted end—plate connections, from /4/

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_371.html (1 of 2)07/11/2009 12:11:43 PM

 page_371

fig 6: Idealisation of the hysteresis in fig. 5

< previous page page_371 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_371.html (2 of 2)07/11/2009 12:11:43 PM

 page_372

< previous page page_372 next page >

Page 372

fig 7: Modelling of semi-rigid connections by substitute trusses according to Stutzki /5/

< previous page page_372 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_372.html07/11/2009 12:11:44 PM
page_373

< previous page page_373 next page >

Page 373

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_373.html (1 of 2)07/11/2009 12:11:45 PM

 page_373

fig 8: Comparison of the δ-a graphs for a frame with rigid and semi-rigid connections
< previous page page_373 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_373.html (2 of 2)07/11/2009 12:11:45 PM

 page_374

< previous page page_374 next page >

Page 374

fig 9: Moment-rotation behaviour due to friction slip effects in semi-rigid connection" /6/, /7/
< previous page page_374 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_374.html07/11/2009 12:11:46 PM
 page_375

< previous page page_375 next page >

Page 375
DISCUSSION OF DESIGN REQUIREMENTS PAPERS
Numerous points were raised regarding individual connector capacities and the corresponding deformations that can be
anticipated. For example, the American design criteria for bolts in bearing imply an ultimate deformation of
approximately 12 mm (0,5 inches). This is not a serviceability limit, since it simply happens to be the deformation that
will have taken place by the time the bearing failure occurs. However, it was agreed that various ultimate as well as
serviceability limit states should be examined.
Special note was made of the block shear failure mode; this is apparently only given in North American standards, and
not in Europe. This concept has now been extended to gusset plate and other tension-type connections, both for bolted
and welded joints.
Many discussers felt that the only realistic way in which semirigid design concepts could be entered into the codes would
be with the use of limit states principles. The detailed structural design example that was presented illustrated this very
clearly, not the least because serviceability criteria are currently very poorly defined for flexibly-connected frames.
Nevertheless, serviceability in some form must be addressed, although a number of discussers questioned the wisdom of
doing so through design code requirements.
Many comments were made regarding the introduction of new steel types and grades on the market, and how current
design standards are utilized in such cases. Some concern was expressed that past research, which was mostly based on
mild structural steels, had been used to prepare the design criteria that now are being applied for materials with
significantly different properties. Although there was general consensus that this type of recognition is warranted, it was
indicated that as long as the ratio of yield stress to ultimate stress is adequate (to facilitate stress redistribution), and the
material ductility and weldability are satisfactory, it should be acceptable to use the current codes with new steel grades.
However, some form of occasional verification may be warranted. By the same token, it was also suggested that design
standards should establish scopes and extent of coverage, to add to their background information. This is particularly
important for codes that address seismic design requirements essential, whose attendant material and structural ductility
needs are essential to the adequate performance of the structure.
< previous page page_375 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_375.html07/11/2009 12:11:47 PM
 page_376

< previous page page_376 next page >

Page 376
This page intentionally left blank.
< previous page page_376 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_376.html07/11/2009 12:11:48 PM
page_377

< previous page page_377 next page >

Page 377
SESSION No 6
DATA BASE ORGANIZATION
< previous page page_377 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_377.html07/11/2009 12:11:48 PM
 page_378

< previous page page_378 next page >

Page 378
NEEDS, CONCEPTS AND ORGANIZATION: A DISCUSSION
As outlined in the Introductory Notes of these Proceedings, the increasing amount of information on the response of
various types of beam-to-column and column footing connections makes it imperative to provide for some organization
and channeling of the findings, to maximize usage and minimize overlap. The need for harmonizing the data gathering
was stressed by all participants, to the effect that the manner of presentation, the interpretation of the results and the
views towards their eventual application should be kept in mind. To that end it was felt that accessing the data base
through a computer network such as E.A.R.N., for example, would be of significant potential.
It was observed that one of the problems of using current data is the highly variable quality of the manner of reporting.
Thus, too many reports do not offer sufficient details for the results that are presented. For example, connection
fabrication details are often omitted. Similarly, although a large amount of data is available, important gaps do exist,
especially when it comes to three-dimensional behaviour and strength. With a measure of data collection coordination,
especially in the form of standardized data sheets (as discussed in the following), the efforts of disseminating the
available and developing information can be vastly enhanced. Several discussers noted the data collection efforts that
had already been started, but indicated that these were strictly just that, with relatively few attempts at analyzing or
unifying the findings.
It was agreed that the eventual aim of any data collection organization should be to assist in the development of design
requirements. Without such, it cannot be expected that the advanced concepts of frames with semi-rigid connections can
get much beyond being the province of the very sophisticated designer.
Several points were made regarding the need to avoid any form of duplication of research efforts, and how an effective
and easily accessible data base can serve this purpose. This should also be extended to inform potential users about on-
going research work.
Summarizing, it was agreed that without a well-organized and-maintained data base, including clearly established
limitations/extent of the collection effort, significant advances could not be expected to occur at a reasonable pace.
The data sheet format and possible connection classification
< previous page page_378 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_378.html07/11/2009 12:11:49 PM
 page_379

< previous page page_379 next page >

Page 379
approach that are presented in the following address these needs in detail. It is noted that it was agreed to restrict the data
collection effort to beam-to-column connections and column footing connections under monotonic loading conditions.
Data on fasteners would not be a part of the collection, nor would fatigue characteristics. Although both of the latter are
needed, the limitations are necessary to make the task manageable.
< previous page page_379 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_379.html07/11/2009 12:11:50 PM
 page_380

< previous page page_380 next page >

Page 380
DATA SHEET FORMAT PRESENTATION
Reidar Bjorhovde
Department of Civil Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania, U.S.A.
and
Jacques Brozzetti
C.T.I.C.M.
Domaine de Saint Paul
Saint Rémy lès Chevreuse, France
ABSTRACT
This brief paper provides a discussion of the information that should be provided by a beam-to-column connection data
sheet. It is emphasized that such sheets may be put to use by other researchers in similar or related studies, but also by
designers for actual construction projects. Details of the connection that has been tested are given, along with material
properties (test values) of members and connectors, drawing(s) of the connection test setup and loading procedure, and
clear indications of how loads, moments, deflections and rotations were obtained. The data sheet should also give the
complete moment-rotation (or load-deflection) curve, indicating service and ultimate loads and deformation magnitudes.
Finally, it is suggested that the initial stiffness of the connection be provided, preferably on the basis of a partial
unloading curve. A sample filled-out data sheet is given at the end, using the format that has been adopted by C.T.I.C.M.
INTRODUCTION
On the basis of the papers that have been presented at this conference, as well as the research data collection efforts that
are
< previous page page_380 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_380.html07/11/2009 12:11:51 PM
 page_381

< previous page page_381 next page >

Page 381
underway in several locales around the world, it is clear that much information is currently available regarding the
strength and behaviour of a number of types of beam-to-column connections. It is important to gather these data into a
coherent and unified basic format, in order that several primary objectives can be attained, as follows:
(1) On a world-wide basis, to obtain a data base of experimental results for beam-to-column and column bases
connections.
(2) With a view to the development of strength and design criteria and equations, to promote the mutual recognition of
test data and the techniques for evaluation and modeling, in order that code writers and designers around the world
can have access to the largest possible assembly of findings.
(3) In order to establish a reliable data base as well as procedures for the exchange of experimental results, to arrive at a
format of data presentation that will facilitate their interpretation and use.
(4) On an ongoing basis, to evaluate research and development needs and to update the requirements as additional data
become available.
(5) By noting the data that are required for subsequent usage, to promote improved methods of testing and reporting.
DATA SHEET PRESENTATION
The attached Data Sheet is described in the following. It is emphasized that it represents one possible format for such
sheets; however, the clarity of the presentation of the data makes it easy to use.
The first page contains three primary types of information:
(1) The source of the test data that are presented, in particular, the reference document that gives additional information
about the test. It is critical that sufficient retrieval data are given, in order that the researcher/designer can obtain
further information. Thus, authors, report or paper title, location, date, and where a copy may be obtained should be
indicated.
(2) The steel grades of the different components of the connection (i.e. beam, column, connection plates and/or angles,
bolts and welds). If at all possible, the material properties that are shown should be based on tests, and the applicable
material standard (e.g., ASTM, CSA, NF, DIN, BS, etc.) and number should be given.
(3) Detailed measurements and geometric characteristics of the
< previous page page_381 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_381.html07/11/2009 12:11:52 PM
 page_382

< previous page page_382 next page >

Page 382
connection and its elements. If at all possible, these data should be based on measurements, rather than referring to
nominal sizes. Shape designations (e.g. HEA 200) should be noted wherever applicable. If measured sizes are not
available, nominal figures must be given, and it should be noted that such is the case.
The second Data Sheet page summarizes the details of the test setup and the connection itself. It is important to indicate
the following:
(1) Exact load application and measurement points, including relative distances.
(2) Form of load application (static or dynamic; monotonic or cyclic) and measurement (load cell or otherwise).
(3) Forms and instruments of strain and displacement measurement and corresponding directions.
(4) If rotations are given, indicate how they were obtained, including any computational details.
(5) Manner of test control (i.e. load, stroke or displacement control).
(6) Detailed testing procedure.
The third Data Sheet page gives the load-deformation curve, including pertinent coordinates for points such as the end of
the linear range, the ultimate load-carrying capacity, and the initial stiffness of the connection, and the failure mode. The
latter should preferably be based on an unloading curve, for example, by unloading from one half of the projected
ultimate strength of the connection to one tenth of this value before resuming the original loading procedure. The use of
unloading curve data will likely eliminate the types of erractic behaviour that are sometimes encountered during the first
application of load in connection tests.
The load-deformation curve may be given in the form of applied load vs. measured deflection, as is shown in the
example. It may also be given as a moment-rotation curve, but it is critical that in this case the Data Sheet must also
provide information on how the moment and rotation magnitudes were arrived at. To assure sufficient detail for
subsequent uses, the load-deformation curve should be based on a minimum of 35 test points. If available, digitized test
curve data should also be provided.
The last Data Sheet page provides additional observations or remarks which may be of value to the user. For example, it
may be valuable to have some comments regarding connection ductility, second order (local deformation) effects, etc.
Data on the response characteristics of column footing connections should follow the general outline of the preceding. In
this case it is also necessary to include information on the concrete and anchor bolt materials, the length and type of
anchoring of the bolts, the use of
< previous page page_382 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_382.html07/11/2009 12:11:52 PM
page_383

< previous page page_383 next page >

Page 383

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_383.html (1 of 2)07/11/2009 12:11:54 PM

page_383

< previous page page_383 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_383.html (2 of 2)07/11/2009 12:11:54 PM

page_384

< previous page page_384 next page >

Page 384

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_384.html (1 of 2)07/11/2009 12:11:54 PM

page_384

< previous page page_384 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_384.html (2 of 2)07/11/2009 12:11:54 PM

page_385

< previous page page_385 next page >

Page 385

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_385.html (1 of 2)07/11/2009 12:11:56 PM

page_385

< previous page page_385 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_385.html (2 of 2)07/11/2009 12:11:56 PM

 page_386

< previous page page_386 next page >

Page 386
OBSERVATIONS:
The horizontal stiffener in the column web was welded at shop at the wrong position. See the figure.

fourth page
< previous page page_386 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_386.html07/11/2009 12:11:57 PM
 page_387

< previous page page_387 next page >

Page 387
grout, and so on. These additional data can be included on Data Sheet pages 1 and 2.
SUMMARY OF WORKSHOP PROPOSAL
It is emphasized that the Connection Data Sheet should not give any form of analytical details. It is intended only as a
vehicle for standardization of connection test data presentation and to facilitate the eventual work of the data user.
Contact individuals for the Data Sheet collection effort were identified during the workshop, as follows:
A. Beam-to-Column Connections:
For Europe: J.Brozzetti
C.T.I.C.M.
Domaine de Saint Paul
F-78470 Saint Rémy lès Chevreuse. France
For North America and Australia: R.Bjorhovde
Department of Civil Engineering
University of Pittsburgh
949 Benedum Hall
Pittsburgh, PA 15261. U.S.A.
B. Column Footing Connections:
For Europe: F.S.K.Bijlaard
TNO Institute for Building Materials and
Structures
P.O. Box 49
NL—2600 AA Delft. The Netherlands
For North America and Australia: D.Beaulieu
Département Génie Civil
Université Laval
Ste. Foy, Quebec G1K 7P4. Canada
Contact individuals for Japan will be determined at a later date. The assignment of the contact individuals is to collect
the Data Sheets with the aim of eventually publishing a Data Sheet Collection Report.
Information regarding the collection effort at any one time can be obtained by contacting the individuals noted in the
preceding.
< previous page page_387 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_387.html07/11/2009 12:11:58 PM
 page_388

< previous page page_388 next page >

Page 388
CLASSIFICATION OF CONNECTIONS
Reidar Bjorhovde
Department of Civil Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania, U.S.A.
Jacques Brozzetti
C.T.I.C.M.
Saint Rémy lès Chevreuse, France
and
André Colson
Laboratoire de Mécanique et Technologie
Ecole Normale Supérieure de Cachan
Université Paris VI
Cachan, France
INTRODUCTION
The following represents a presentation of connection classification concepts only, and is not intended for direct use at
this time. However, it is felt that the principles of categorization are rational and relatively easy to use for practical
purposes. Moreover, it will facilitate the classification of new connections or for connections where M-Φ-curves are not
currently available.
BASIC PRINCIPLES
A non-dimensional format is preferred, where the vertical axis gives the ratio M/Mp and the horizontal axis gives Φ/Φp.
These terms designate the following:
M = connection moment
Mp = fully plastic moment of beam
Φ = connection rotation
< previous page page_388 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_388.html07/11/2009 12:11:59 PM
 page_389

< previous page page_389 next page >

Page 389
Φp = beam rotation at beam moment=Mp (see Figure 1)

Figure 1 Definition of Beam M-Φ-curve

It is noted that the initial slope of the beam M-Φ-curve, as shown in Figure 1, is a function of the length of the beam
element. Figure 2 demonstrates this further, for a range of beam lengths as a function of the beam depth.

Figure 2 Beam M-Φ-curve as a Function of Beam Length

< previous page page_389 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_389.html07/11/2009 12:12:00 PM
 page_390

< previous page page_390 next page >

Page 390
In this figure h indicates the depth of the beam; Cbeam is the initial slope of the beam M-Φ-curve; CR, CSR and CF
indicate the stiffness of rigid, semi-rigid and flexible beam-to-column connections, respectively.
For the above it is necessary to define and choose a length of beam that is regarded as the most suitable. It is noted that
stiffer connections require a shorter beam length, in order to arrive at compatibility between the beam that is used to non-
dimensionalize the connection M-Φ-curve.
The connection M-Φ-diagram is plotted dimensionally first, to get the clear relationship between moment and rotation. It
can assume the form as illustrated in Figure 3.

Figure 3 Connection M-Φ-Relationship

In this figure the term Mult indicates the ultimate connection moment capacity (in case of the lack of a clear plateau, it is
necessary to define the value of Mult as a function of Φmax, the required maximum rotation). The term Φcp is the
rotation for a point as shown, and C is the initial stiffness of the connection.
CLASSIFICATION APPROACH
Figure 4 gives the schematic, non-dimensional connection M-Φ-classification diagram. Ranges of response are
indicated, and it is evident that it will be easy to categorize new connections according to this scheme.
< previous page page_390 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_390.html07/11/2009 12:12:01 PM
 page_391

< previous page page_391 next page >

Page 391
It is emphasized that this classification system is conceptual only at this stage. Further studies are expected to lead to
quantified data for the classification diagram.

Figure 4 Non-Dimensional Connection Classification Diagram System

< previous page page_391 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_391.html07/11/2009 12:12:02 PM
 page_392

< previous page page_392 next page >

Page 392
This page intentionally left blank.
< previous page page_392 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_392.html07/11/2009 12:12:03 PM
page_393

< previous page page_393 next page >

Page 393
RESEARCH AND DEVELOPMENT NEEDS
< previous page page_393 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_393.html07/11/2009 12:12:04 PM
 page_394

< previous page page_394 next page >

Page 394
INTRODUCTORY COMMENTS
The research and development needs that are given in the following were extracted from the observations of the speakers
and the discussions that took place during the workshop. They represent the essence of the work of the research
reporters, and although no explanatory notes are provided along with each subject, it is felt the topics give a realistic
impression of the work that needs to be undertaken.
The research topics are listed in the order of the technical sessions of the workshop, and have not been prioritized in any
form. Some of the subjects are somewhat broad in scope, and may in fact entail multiple research projects.
RESEARCH AND DEVELOPMENT NEEDS
1. Local Analysis of Joints
(a) Ultimate limit states for connections with monotonically increasing moment-rotation relationship.
(b) Three-dimensional connection analysis and tests.
(c) Loading and unloading behaviour of typical building beam-to-column connections.
(d) Development of standardized methods of connection testing and reporting.
(e) Clarification of the prying phenomenon for bolted connections.
2. Mathematical Modeling
(a) Two-and three-dimensional modeling of connections through the full range of behaviour.
(b) Incorporation of strain-hardening and fracture considerations into connection modeling.
(c) Finite element modeling of bolts in tension, including bolt pretension effects and contact considerations.
< previous page page_394 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_394.html07/11/2009 12:12:05 PM
 page_395

< previous page page_395 next page >

Page 395
(d) Development of pre—and post—processing tools for computer analyses of connections.
3. Classification
(a) Data base organization, including fasteners, beam-to-column connections, tubular connections, composite
connections, and fatigue data.
(b) Development of computerized data bank for ease of access and transfer.
(c) Integration of current data banks into one cohesive assembly.
(d) Preparation of connection data sheet
4. Frame Analysis
(a) Verification of theoretical results through full-scale subassemblage and frame tests.
(b) Serviceability and ultimate limit states analysis criteria for frames and their connections.
(c) Cyclic load behaviour of semi-rigid connections.
(d) Definition of bolt pretension to produce specific forms of connection slip behaviour.
5. Frame Stability and Simplified Methods
(a) Stiffness characteristics of column footing connections, including axial load and column effects.
(b) Verification of column footing analytical models.
(c) Initial and unloading stiffness properties for beam-to-column connections.
(d) Application of end-restrained column buckling theories to columns in unbraced frames.
(e) Moment transfer from beam to column during buckling of end-restrained subassemblages.
(f) Influence of axial load on subassemblage and frame stability.
(g) Drift characteristics of flexibly connected frames.
(h) Development of serviceability criteria for flexibly connected frames.
(i) Interaction of connection and bracing stiffness in braced frames.
< previous page page_395 next page >

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_395.html07/11/2009 12:12:06 PM
 page_396

< previous page page_396

Page 396
6. Design Requirements and Codes
(a) Investigate and settle differences between the standards of various countries in their treatment of fastener strength
and behaviour.
(b) Ductility criteria for high-strength bolts.
(c) Tension-shear interaction relationship for bolts.
(d) Clarify design criteria and approaches for prying.
(e) Classification system for connections according to M-Φ-relationship or similar.
(f) Effective width criteria for connections between tubes and other types of cross sections.
(g) Design criteria for use with high strength steels: Ductility, weldability, applicability of current requirements.
(h) Ductility requirements for connections in the presence of axial loads.
(i) Rotation capacity of connections under cyclic loading.
(j) Behaviour and strength characteristics for composite connections.
< previous page page_396

file:///P|/9780203497890__gigle.ws/_84129__9780203497890__9780585461199__/files/page_396.html07/11/2009 12:12:07 PM