Printed by Media-Tryck, Lund University 2015

hassan mehri  
Bracing of steel bridges during construction – Theory, full-scale tests, and simulations
Bracing of steel bridges during
construction
Theory, full-scale tests, and simulations
hassan mehri
Faculty of engineering | Lund University

The study as a whole involves investigations of both the temporary and per-
manent bracings that are often required during the construction phase of
steel bridges. Investigations of possible bracing alternatives to the conventional
ones were also of interest. A part of the research included the derivation of
a simplified analytical approach for determining the critical moment of the
laterally braced steel girders at the level of their compression flange. The model
relates the buckling length of the compression flange of a steel girder to its
lateral-torsional critical moment. Moreover, eleven full-scale laboratory tests on
a twin I-girder bridge were performed to evaluate stabilizing requirements of
the commonly used bracings and of the possible alternatives for them during
the construction of composite bridges.

9 789187 993046

Lund University
Faculty of Engineering
Division of Structural Engineering
2015

ISBN 978-91-87993-04-6
ISSN 0349-4969
 Bracing of steel bridges during
construction
Theory, full-scale tests, and simulations

Hassan Mehri

DOCTORAL DISSERTATION
by due permission of the Faculty of Engineering, Lund University, Sweden.
To be defended at lecture hall MA3 at Mathematics Annex building,
Sölvegatan 20, Lund, on 22 January 2016 at 10:15 AM.

Faculty opponent
Professor Emeritus Torsten Höglund
KTH Royal Institute of Technology
Organization: LUND UNIVERSITY Document name: Ph. D. dissertation
Date of issue: December 2015
Author: Hassan Mehri Sponsoring organization: J. Gust Richert Stiftelse-
The Lars Erik Lundbergs Stipendiestiftelse-
Byggrådet- Britek AB- Structural Metal Decks Ltd.
Title and subtitle: Bracing of steel bridges during construction - Theory, full-scale tests, and simulations
Abstract A number of steel bridges have suffered lateral-torsional failure during their construction due to their
lacking adequate lateral and/or rotational stiffness. In most cases, slight bracing can be of great benefit to the main
girders involved through their controlling out-of-plane deformations and enabling the resistance that is needed to
be achieved. The present research concerned the performance of different bracing systems, both those of
commonly used types and pragmatic alternatives. The methods that were employed include the derivation of
analytical solutions, full-scale laboratory testing, and numerical modeling. The results of a part of the study
showed that the load-carrying capacity of The Marcy Bridge that collapsed in 2002 could be improved by adding
top flange plan bracing at 10-20% of its span near the supports. Theoretically, according to Eurocode 3, providing
each bar of an X-type plan bracing having cross-sectional area as small as 8 mm� serves to enhance the load-
carrying capacity of the bridge by a factor of 1.28, which is sufficient to prevent failure of the bridge during the
casting of the deck. The research also included the derivation of a simplified analytical approach for determining
the critical moment of the laterally braced steel girders at the level of their compression flange, which otherwise
can usually not be predicted without the use of finite element program. The model employed related the buckling
length of the compression flange of steel girders in question to their critical moment. An exact solution and a
simplified expression were also derived for dealing with the effect of the rotational restraint of the shorter
segments on the buckling length of the longer segments in beams having unequally spaced lateral bracings. The
effects of this sort are often neglected in practice and the buckling length of compression members in such systems
is commonly assumed to be equal to the largest distance between the bracing points. However, the present study
showed that this assumption can provide an unsafe prediction of buckling length for relatively soft bracings and
can also lead to a significant overdesign in regard to most bracing stiffness values in practice. Full-scale
experimental study on a twin-I girder bridge together with numerical works on different bridge dimensions were
carried out on the stabilizing performance of a type of scaffolding that is frequently used in the construction of
composite bridges. Minor improvements were discussed which found to be needed in the structure of the
scaffoldings that were employed. Findings showed the proposed scaffoldings to have a significant stabilizing
potential when they were installed on bridges of differing lateral-torsional slenderness ratios. Axial strains in the
scaffolding bars were also measured. Indications of the design brace moment involved were also presented which
was approximately between 2 and 4% of the maximum in-plane bending moment in the main girders. Three full-
scale experimental studies were also performed on a twin I-girder bridge in which the location of the cross-beam
across the depth of the main girders was varied. The effects of several different relevant imperfection shapes on
the bracing performance of the cross-beams were of interest. It was found that the design recommendations
currently employed can provide uncertain and incorrect predictions of the brace forces present in the cross-
bracings. Both the tests and FE investigations carried out showed the shape of the geometric imperfections
involved to have a major effect on the distortion that occurred in the braced bridge cross-sections. It was also
found that significant warping stresses could develop in cross-beams having asymmetric cross-sections, the
avoiding of such profiles in the cross-beams being recommended. Finally, seven full-scale laboratory tests of the
end-warping restraints of truss-bracings and of corrugated metal sheets when they were installed on a twin I-girder
bridge were also performed. The load-carrying capacity of the bridge was found to be enhanced by a factor of 2.5-
3.0 when such warping restraints were provided near the support points. Relatively small forces were developed in
the truss-bracing bars in order to such significant improvements in the load-carrying capacity of the bridge to be
achieved. Moreover, bracing the bridge in question by means of the metal sheets that were employed was found to
result in a significantly larger degree of lateral deflection at midspan than use of the utilized truss bracings did.
Key words: brace, stability, steel, bridge girder, construction
Classification system and/or index terms (if any):
Supplementary bibliographical information: ISRN LUTVDG/TVBK-1049/16-SE Language: English
ISSN and key title: 0349-4969, Report TVBK-1049 ISBN 978-91-87993-04-6
Recipient's notes Number of pages 227 Price
Security classification
I, the undersigned, being the copyright owner of the abstract of the above-mentioned dissertation, hereby grant to
all reference sources permission to publish and disseminate the abstract of the above-mentioned dissertation.

Signature Date 4th December 2015

Bracing of steel bridges during
construction
Theory, full-scale tests, and simulations

Hassan Mehri
Copy right © Hassan Mehri

Faculty of Engineering, Division of Structural Engineering

P. O. Box 118, SE-221 00 Lund, Sweden
Report TVBK-1049
ISRN LUTVDG/TVBK-1049/15-SE(185)
ISBN 978-91-87993-04-6
ISSN 0349-4969

Printed in Sweden by Media-Tryck, Lund University

Lund December 2015

En del av Förpacknings- och

Tidningsinsamlingen (FTI)
Preface

One of the major concerns in the design of steel bridges is the global and local
instability of structural members, both during construction and in service.
Catastrophic failures resulting in fatalities have occurred at times when stability
principles have been violated during construction. The stability of steel-bridges
during construction is highly dependent upon the adequacy in terms of both
stiffness and strength requirements of the bracings that are provided. Winter [1]
presented a dual brace criterion, his showing experimentally that the load-carrying
capacity of an approximately 3.5 m long I-shape column (having a depth of 100
mm, a width of 50 mm and a thickness of 1.6 mm) was enhanced by a factor of
fifteen by use of bracings as weak as cardboard strips. The efficiency of such
slight bracings was impressive. It is possible that some of the bridge tragedies that
have occurred could have been avoided by use of very inexpensive bracings. I
decided here to investigate how common bracings function in bridge applications
during what is the most critical stage in terms of possible instability, namely the
construction phase.
The thesis is being submitted for a degree of Doctor of Philosophy at the Division
of Structural Engineering of Lund University. It is based on research carried out
by the author between May 2011 and December 2015. The thesis itself, the
appended papers excluded, is 123 pages in length. No part of the dissertation work
has been submitted for a degree at any other university. The research work was
supervised primarily by Prof. Roberto Crocetti, to whom I am extremely grateful. I
would also like to thank Dr. Eva Frühwald Hansson for her assistance. Special
thanks go as well to Dr. Miklos Molnar, the Head of the Division, for his endless
support and his kindness. I appreciate too the assistance provided by Per-Olof
Rosenkvist (from LTH) and Göran Malmqvist (from SP) during the conducting of
the tests. Jamie Turner (from SMD Ltd in the U.K.) and Thomas Lindin (from
Britek AB) provided the corrugated metal sheets and scaffoldings that were
required during the tests that were employed, I am highly appreciative of their
support. I would also like to thank Fredrik Carlsson (from Reinertsen Sverige AB)
and Ola Bengtsson (from Centerlöf & Holmberg AB) for the consultancy advice I
received from them in our meetings and the discussions we had. I take this
opportunity as well to thank my fellow researchers for the great times and the
discussions we have had.

i
Most importantly, thank you Parvaneh for your patience and encouragement.
Thanks for believing in me more than myself, and being there supporting me
unconditionally.

Hassan Mehri
December 2015

ii
Abstract

A number of steel bridges have suffered lateral-torsional failure during their

construction due to their lacking adequate lateral and/or rotational stiffness. In
most cases, slight bracing can be of great benefit to the main girders involved
through their controlling out-of-plane deformations and enabling the resistance
that is needed to be achieved. The present research concerned the performance of
different bracing systems, both those of commonly used types and pragmatic
alternatives. The methods that were employed include the derivation of analytical
solutions, full-scale laboratory testing, and numerical modeling.
The results of a part of the study showed that the load-carrying capacity of The
Marcy Bridge that collapsed in 2002 could be improved by adding top flange plan
bracing at 10-20% of its span near the supports. Theoretically, according to
Eurocode 3, providing each bar of an X-type plan bracing having cross-sectional
area as small as 8 mm serves to enhance the load-carrying capacity of the bridge
by a factor of 1.28, which is sufficient to prevent failure of the bridge during the
casting of the deck.
The research also included the derivation of a simplified analytical approach for
determining the critical moment of the laterally braced steel girders at the level of
their compression flange, which otherwise can usually not be predicted without the
use of finite element program. The model employed related the buckling length of
the compression flange of steel girders in question to their critical moment. An
exact solution and a simplified expression were also derived for dealing with the
effect of the rotational restraint of the shorter segments on the buckling length of
the longer segments in beams having unequally spaced lateral bracings. The
effects of this sort are often neglected in practice and the buckling length of
compression members in such systems is commonly assumed to be equal to the
largest distance between the bracing points. However, the present study showed
that this assumption can provide an unsafe prediction of buckling length for
relatively soft bracings and can also lead to a significant overdesign in regard to
most bracing stiffness values in practice.
Full-scale experimental study on a twin-I girder bridge together with numerical
works on different bridge dimensions were carried out on the stabilizing
performance of a type of scaffolding that is frequently used in the construction of
composite bridges. Minor improvements were discussed which found to be needed
in the structure of the scaffoldings that were employed. Findings showed the

iii
proposed scaffoldings to have a significant stabilizing potential when they were
installed on bridges of differing lateral-torsional slenderness ratios. Axial strains in
the scaffolding bars were also measured. Indications of the design brace moment
involved were also presented which was approximately between 2 and 4% of the
maximum in-plane bending moment in the main girders.
Three full-scale experimental studies were also performed on a twin I-girder
bridge in which the location of the cross-beam across the depth of the main girders
was varied. The effects of several different relevant imperfection shapes on the
bracing performance of the cross-beams were of interest. It was found that the
design recommendations currently employed can provide uncertain and incorrect
predictions of the brace forces present in the cross-bracings. Both the tests and FE
investigations carried out showed the shape of the geometric imperfections
involved to have a major effect on the distortion that occurred in the braced bridge
cross-sections. It was also found that significant warping stresses could develop in
cross-beams having asymmetric cross-sections, the avoiding of such profiles in the
cross-beams being recommended.
Finally, seven full-scale laboratory tests of the end-warping restraints of truss-
bracings and of corrugated metal sheets when they were installed on a twin I-
girder bridge were also performed. The load-carrying capacity of the bridge was
found to be enhanced by a factor of 2.5-3.0 when such warping restraints were
provided near the support points. Relatively small forces were developed in the
truss-bracing bars in order to such significant improvements in the load-carrying
capacity of the bridge to be achieved. Moreover, bracing the bridge in question by
means of the metal sheets that were employed was found to result in a
significantly larger degree of lateral deflection at midspan than use of the utilized
truss bracings did.

iv
Contents

Preface i 
Abstract iii 
Contents v 
Notations ix 
Publications xiii 
Appended papers xiii 
Paper I) xiii 
Paper II) xiii 
Paper III) xiii 
Paper IV) xiii 
Paper V) xiii 
Contribution of the authors xiv 
Other scientific contributions of the author xiv 
Conference paper xiv 
Supervision of M.Sc. thesis xiv 
1 Introduction 1 
1.1 Objectives 2 
1.2 Limitations 3 
1.3 State-of-the-art 3 
1.4 Terminology 4 
1.5 Outline 6 
2 Examples of bridge failures during construction associated with instability 9 
2.1 Examples of steel-truss bridge failures during construction associated
with problems of instability 10 
2.2 Examples of failures of built-up steel girder bridges during the non-
composite stage associated with their instability 11 
2.2.1 The collapse of Bridge Y1504 in Sweden 13 
2.3 Steel bridge accidents during concreting that were associated with
problems of instability in their timber falseworks 16 
2.4 Conclusions 17 

v
3 Theory of beam stability 19 
3.1 Introduction 19 
3.2 Effects of material inelasticity on bracing requirements 20 
3.3 Effects of residual stresses on buckling load 21 
3.4 Lateral-torsional buckling of doubly-symmetric simply supported beams
subjected to uniform bending 22 
3.5 Modifications required in the basic approach to the critical bending
moment value 23 
3.5.1 Effects of different boundary conditions 24 
3.5.2 Effects of different loading conditions 24 
3.5.3 Effects of lateral restraints (Paper II) 25 
3.5.4 Effects of cross-sectional asymmetry 26 
3.5.5 Effects of inelasticity on lateral-torsional buckling 27 
3.5.6 Effects of variable cross-section on lateral-torsional buckling
(unpublished work) 28 
4 Fundamentals of beam bracing 31 
4.1 Introduction 31 
4.2 Lateral bracing of beams 33 
4.3 Torsional bracing of beams 34 
4.4 “Column-on-elastic-foundation model” for cross-brace stiffness
assessments in steel bridges 36 
5 Lateral-torsional instability concerns during construction of steel bridges 41 
5.1 Bracings required during concreting of the deck 41 
5.2 Bracings required for skewed bridges 43 
5.3 Bracings required for in-plane curved bridges 44 
5.4 Bracing required prior to the concreting stage 45 
6 Bracing options in steel bridges 47 
6.1 Intermediate cross-bracings 47 
6.2 Support bracings 53 
6.3 Bracing of half-through girders 54 
6.4 Full-span plan bracings 55 
6.5 Partial-span plan bracings (Papers I, and V) 57 
6.6 Bracings required in open trapezoidal girders 58 
6.6.1 The equivalent plate concept and forces generated in plan-
bracings by torsion 59 
6.6.2 Forces generated from distortion in the cross-bracings 60 
6.6.3 Forces in the plan bracings due to the web inclinations of
trapezoidal girders 61 

vi
 6.6.4 Forces generated in intermediate external cross-bracings from
torsion 61 
6.6.5 Support cross-diaphragms 62 
6.7 Bracing potential of stay-in-place corrugated metal sheets (Paper V) 63 
6.7.1. Stiffness requirements of the metal sheets 66 
6.7.2. Strength requirements of the metal sheets 66 
6.7.3. Connection requirements 67 
6.7.4. Use of corrugated metal sheets in Twin-I girder bridges 68 
6.8 Scaffolding bracing of steel girders (Paper III) 68 
6.8.1 The common practice in design aimed at providing stability for
the steel girders during construction stage 69 
6.8.2 Shortcomings of the common bridge timber-falseworks 71 
6.8.3. The concept of scaffolding bracing of steel bridges 72 
6.8.4. Effects of the ledgers on the load-carrying capacity of the test
bridge 75 
6.8.5 The effects of the ledgers on the bracing forces 78 
6.9 Bracing potential of precast concrete slabs 81 
7 The effects of initial imperfections on the performance of bracings (Papers III,
IV, & V) 83 
7.1 Effects of the magnitude and the shape of the initial geometric
imperfections on the load-carrying capacity of steel girders 84 
7.2 Effects of the shape and the magnitude of initial geometric imperfections
on brace forces 86 
8 Laboratory tests 89 
8.1 Background 89 
8.2 Tests performed by the author 90 
9 Numerical simulations 95 
10 Conclusions and future research 99 
10.1 Conclusions from the appended papers 99 
Paper I: 99 
Paper II: 99 
Paper III: 100 
Paper IV: 100 
Paper V: 101 
10.2 Future research 102 
Acknowledgements 103 
Appendix I 105 
Details regarding the test setup 105 

vii
Appendix II 111 
Test data not reported directly in the appended articles 111 
Appendix III 113 
AIII.1 Bracing analysis; AASHTO recommendations [82] 113 
Summary of the recommendations regarding the use of cross-
bracings 113 
Summary of the recommendations regarding the use of lateral
bracings 114 
AIII.2 Bracing analysis; Eurocode recommendations [76] 115 
Effects of imperfections in analyzing a bracing system 115 
Lateral-torsional buckling of structural components 116 
Appendix IV 117 
Equivalent plate thickness of typical plan-bracings [2] 117 
References 119 

viii
Notations

The following symbols are used in the present report:

Area of a cross‐section;
The enclosed area defined by the wall‐midline in a thin‐walled
closed section;
, , , , Cross‐sectional area of a single top flange, a bottom flange, and
the web of a steel girder;
, /4 2 ;
, Cross‐sectional area of a diagonal and transversal strut bar;
Distance between the struts in a truss bracing system;
Thickness of a web stiffener;
, , , Moment gradient factors of a beam in its entirety, and of the
unbraced and the braced spans, respectively;
Top flange loading modification factor;
Warping constants; 1 for I sections. For
trapezoidal cross‐sections see 3 ;
Depth of a cross‐section;
Modulus of elasticity;
Tangent and reduced modulus of elasticity;
Distance between the mid‐height of a beam cross‐section and
the plane of a corrugated metal sheet;
Destabilizing force of a compression flange;
Brace design force;
Shear modulus;
Effective shear modulus of a corrugated metal sheet;
, Self‐weight of steel girders and the fresh concrete per unit span
length;
Distance between the centroids of the top and the bottom
flanges of a steel girder;
Depth of a cross‐diaphragm;
Depth of a web stiffener;
Strong axis moment of inertia of a cross‐beam;

ix
 ., . Moment of inertia of the elastic and the inelastic portions of a
cross‐section;
, Moment of inertia of the entire cross‐section, and of the core
portion of a hybrid cross‐section;
Bending stiffness of vertical web stiffeners and a girder's web;
, Moment of inertia of a cross‐section with respect to the " "
strong and the " " weak axes;
, and , Moment of inertia of the compression and the tension flanges,
respectively, with respect to the weak axis of a monosymmetric
section;
, , ⁄ , 4 ,
Torsional constant; ∑ /3 in open cross‐sections and
4 / ∑ / in closed cross‐sections.
Stiffness of a translational spring;
Actual or provided brace stiffness of a lateral brace;
Ideal brace stiffness of a lateral brace;
Design brace stiffness of a lateral brace;
, Effective buckling length factors with respect to the
longitudinal or the weak axis;
Longitudinal span;
Distance between the brace points;
Distance between points of zero twist or lateral
displacement ; the central span length;
, , Effective buckling length of a structural member between
immovable restraints;
, , , Effective buckling length of a system having translational or
rotational bracings between the end‐supports;
, , Critical in‐plane moment of a beam restrained by means of
metal sheets;
, , Critical in‐plane moment of a beam restrained by means of
torsional bracings;
, , Critical in‐plane moment of an unbraced beam;
, , , Critical in‐plane moments corresponding to lateral‐torsional
buckling between the brace points, and to system buckling,
respectively;
Warping brace moment per length of a beam;
Maximum applied in‐plane bending moment;
Theoretical plastic bending capacity of a beam cross‐section;
, , Bending moments with respect to the , , axes in an un‐
deformed configuration;
, , Bending moments with respect to local coordinate axes
x
 , , in a deformed configuration;
Diaphragm effectiveness factor;
Number of intermediate bracings;
Number of half‐sine waves in a buckling mode;
Number of main girders;
Axial compressive point‐load;
, , , Critical load of a braced or an unbraced column;
kN/rad Shear rigidity of a metal sheet equal to / 2 in an
equivalent truss bracing ;
Shear flow;
Equivalent uniform load for geometric imperfections;
Transversal load per unit span length;
Radius of curvature in an in‐plane‐curved bridge;
Radius of gyration;
Transversal span of a bridge;
Tributary width of metal decks per girder 1 /
where ng is the number of girders;
, Maximum torque along the span due to the self‐weight of the
steel girders;
/ Distance ratio of the centroids of the top and bottom flanges
from the neutral‐axis of a beam cross‐section;
Thickness of the plate of a cross‐diaphragm;
Equivalent plate thickness;
Thickness of web‐stiffeners;
Web thickness of a girder;
, Lateral and vertical deformations of a beam cross‐section;
Center‐to‐center distance between the centroids of top flanges
in trapezoidal girders;
Lateral deformation of a column;
, , Longitudinal, horizontal, and vertical coordinates;
Distance of the shear center S.C. from the centroid.
The length of the side segments;
Torsional stiffness of a cross‐beam/‐frame/‐diaphragm;
Girder system stiffness;
Torsional stiffness contribution of web stiffeners;
Effective torsional brace stiffness 4 ;
̅ / ; for bridges with single brace, replace L by 0.75L 4 ,
, Actual provided brace stiffness of a torsional brace;

xi
 , Ideal brace stiffness of a torsional brace;
A cross‐sectional property that takes account of the effects of
cross‐sectional mono‐symmetry;
∆ lateral displacement of a brace point;
∆ Magnitude of initial out‐of‐straightness of a compression
member;
∆ ∆ ∆ total lateral displacement of a brace point;
Sip in a bolted connection;
Normal strain;
twist at a torsional brace point;
Initial twist of a beam cross‐section;
total twist at a torsional brace point;
Shear stress;
Radius of curvature in a bent member;
Relative slenderness ratio;
; total potential energy Internal work‐ external
work ;
, / ;
Normal stress;
∅ Twist of a beam cross‐section;
Skew angle of a cross‐brace with respect to the vertical axis.

xii
Publications

Appended papers

Paper I)

Mehri H. and Crocetti R., "Bracing of steel-concrete composite bridges during

casting of the deck," presented at the Nordic Steel Construction Conference 2012,
Oslo, Norway 2012.

Paper II)

Mehri H., Crocetti R., and Gustafsson P. J., "Unequally spaced lateral bracings on
compression flanges of steel girders," Structures, vol. 3, pp. 236-243, DOI:
10.1016/j.istruc.2015.05.003, 2015.

Paper III)

Mehri H. and Crocetti R., "Scaffolding bracing of composite bridges during

construction," Journal of Bridge Engineering (ASCE), DOI: 10.1061/(ASCE)BE.
1943-5592.0000829, 2015.

Paper IV)

Mehri H., Crocetti R., and Yura J. A., "Effects of geometric imperfections on
bracing performance of cross-beams during construction of composite bridges,"
submitted to Engineering Structures, 19th July 2015.

Paper V)

Mehri H., Crocetti R., "End-warping bracing of steel bridges during construction,"
submitted to Journal of Bridge Engineering, 2nd December 2015.

xiii
Contribution of the authors
The author, Hassan Mehri, planned the test programs, designed the test-setups,
performed the laboratory tests, analyzed the recorded test-data, carried out the
numerical investigations, and wrote the appended papers and the present thesis.
Prof. Roberto Crocetti contributed to the review of the work. Prof. Joseph A. Yura
contributed to both the scientific supervision of Paper (IV) and the review of it.
Prof. Per-Johan Gustafsson helped the author with the derivation of Eq. (8) in
paper (II), and reviewed the analytical solutions of that paper. Dr. Eva Frühwald
Hansson reviewed the text of the thesis.

Other scientific contributions of the author

Conference paper

- Mamazizi S., Crocetti R., and Mehri H., "Numerical and experimental
investigation on the post-buckling behavior of steel plate girders subjected
to shear," presented at the Annual Stability Conference, SSRC 2013, St.
Louis, Missouri, April 16-20, 2013 [5].

Supervision of M.Sc. thesis

- Al-rubaye A., "Theoretical and numerical approach to calculate the shear

stiffness of corrugated metal decks, " Div. of Struct. Eng., Lund Univ. 2014
[6].
- Winge A., "Temporary formworks as torsional bracing system for steel-
concrete composite bridges during concreting the deck, " Div. of Struct.
Eng., Lund Univ. 2014 [7].
- Carlson O. and Jaskiewicz L., "The performance of conventional discrete
torsional bracings in Swedish steel-concrete composite bridges, " Div. of
Struct. Eng., Lund Univ. 2014 [8].
- Ohlin, E., "Metal decks as lateral bracing for composite bridges with
trapezoidal cross-sections, " Div. of Struct. Eng., Lund Univ. 2015 [9].

xiv
1 Introduction

Bridges are an important part of a country's road network. New bridges are often
built over busy roads or railways. Traditionally, bridge construction involves the
in-place casting of concrete, this requiring both time and a large of space, this
often causing serious traffic problems. The economic losses of the traffic delays
thus brought about are very difficult to estimate. These include travel disturbances,
longer travel times, and the unavailability of transportation, all of which lead to
lesser income, and consequently to lesser tax income and social welfare. There are
often major losses as well in the form of indirect effects, and the clear negative
environmental impact of the traffic jams that take place. To reduce disruptions of
this sort, it is extremely important that a bridge's assembly be performed as
quickly, smoothly, and as safely as possible.
Developments in the fabrication of steel girders of high strength steel, weldable,
and able to achieve a high ratio of the moment of inertia to the cross-sectional area
have made the use of steel in the bridge industry attractive. A type of bridge that
makes smooth and relatively quick installations possible and that during the last
few decades has taken over a large part of the bridge market, is that of the so-
called composite steel-concrete bridge. A bridge of this type consists of one or
more steel girders in a composite action with a reinforced concrete deck, this
serving to optimize use of the materials by the steel girders being subjected
predominately to tension, whereas the concrete is subjected mainly to
compression. An overall benefit in the use of steel-concrete composite bridges is
also the fact that during construction of them the steel girders can be erected
rapidly, which ultimately reduces the traffic disturbances brought about. More
importantly, the steel girders bear the construction loads of the scaffoldings while
construction is taking place, transferring these loads to the abutments. This reduces
significantly the amount of scaffoldings that are required.
The installation of steel-concrete composite bridges is, however, a critical matter
in the designing of such bridges, it is often controlling the size of both the steel
girders and the bracings. Altogether 105 out of the 440 cases of bridge failure dealt
with in the report “Failed Bridges, case studies; causes and consequences” [10]
occurred during construction of the bridges. This highlights the importance of
more detailed investigations at this stage. In recent years, a number of accidents
during the construction of steel bridges have occurred due to various instability
phenomena during the lifting, launching, or concreting phases in bridge
construction. An example of such accidents is the collapse of Bridge Y1504 over
1
the Gide River in Sweden that occurred in 2002; see section (2.2.1). The
investigations that took place following the accident required considerable costs
and efforts in themselves, and replacing the bridge cost approximately twice the
original budget for building of the bridge. Although two workers dropped down to
the ground when the bridge collapsed, there were fortunately no fatalities, since
the bridge was not particularly high. There have also been failures in the
construction of composite bridges due to problems caused by the instability of
their falseworks. The falsework failure in connection with the Älandsfjärden
Bridge in Sweden in 2008 is an example of such an accident, five construction
workers there falling 20 meters down to the ground, two of them being killed, and
two severely injured [11]. The author is also aware of the recent bridge collapses
in Norway (Trondheim Bridge on May 8, 2013, two persons killed) and in
Denmark (in Aalborg in June of 2006, one person killed; and in Helsingør in
September of 2014) all of which occurred during concreting of their deck. The
author had no access to the failure reports of these accidents at the time of writing
the thesis.
The failure mode of lateral-torsional buckling plays an important role in
determining the size of the steel bridge girders to employ. The lateral-torsional
instability of steel girders involves the possible twisting of the cross-section of
them and lateral movement of the compression flange. The lateral-torsional
resistance of such slender beams can be improved by either increasing the size of
girders, or providing proper bracings so as to reduce the buckling length of the
compression flanges. In composite bridges, the top flanges are needed mainly to
provide sufficient space for the shear studs when a composite action takes place;
their contribution to resisting out-of-plane deformations being negligible once the
concrete deck has hardened. On the other hand, providing slight bracings can
effectively enhance the resistance of steel girders. Without bracing, too large a
lateral deflection of the compression flange could easily occur. A number of
bracing options for controlling out-of-plane deformations of the main steel girders
are feasible. An effective bracing system should possess adequate stiffness and
strength so as to enhance both the load-carrying capacity of the main girders to a
desired level and to withstand the forces induced in the bracings.
The present study as a whole involves analytical, experimental, and computational
investigations of the bracing requirements of steel bridges during construction.
Current knowledge and related design recommendations concerning bracing
requirements for steel bridges during construction of them are also discussed.

1.1 Objectives
The main objective of the PhD study was to evaluate the stabilizing performance
of the typical bracings (cross-bracings, plan bracings, and corrugated metal sheets)
2
that are commonly used in steel bridge applications during the construction stage.
Investigations of possible brace alternatives such as stabilization by means of
modified scaffoldings were also of interest. The research includes the derivation of
analytical solutions, the carrying out of full-scale laboratory tests, and the
performing numerical simulations.

1.2 Limitations
The effects of different loading and boundary conditions, bridge curvature, skewed
supports, and the like, on brace forces and/or on load-carrying capacities have
been investigated by other authors. Studies concerning the effects of such
variations were beyond the scope of the present study. The present study is also
concerned mainly with straight bridges. Although in paper III two-span bridges
were investigated, for the most part of the present study simply supported bridges
were the dominant case studies. However, roughly the same rules as those that
apply for the types of bridges studied here apply to the curved, skewed, and
continuous bridges as well. Although the effects of the concreting sequence and of
launching process on the brace requirements are also relevant to the phenomena
investigated here, they were not examined in the present study. The bridges that
were studied here were subjected to uniform transversal loads applied to the top
flanges, which is a situation very commonly encountered in practice during the
concreting stage. Finally, twin girders and trapezoidal girders were selected for the
case studies, since these are the girders used most commonly in steel bridges in
Sweden.

1.3 State-of-the-art
Although a large number of scientific contributions to study of the stability of steel
beams and columns were reviewed in the literature study presented here, only the
most relevant references are cited in the thesis. Also, the major contributions of
previous research are explained for the most part within the contexts to which they
apply. However, in order to classify them in terms of the method that have been
employed, a brief account of the state-of-the-art within this context is provided in
the following:
Winter [1] developed a simple rigid bar model involving fictitious hinges at the
brace points for determining lower-bound stiffness of the bracings that are
employed in laterally braced columns, their values corresponding to “ideal” brace
stiffness serving as immovable supports. Introducing initial imperfections, Winter
also obtained the magnitude of forces present in the bracings. Introducing a

3
rotational spring representing the flexural stiffness of a given column, Pincus [12]
extended Winter's model in order to determine the bracing requirements of
inelastic columns. Obviously, the main difference between the two models
mentioned above is the load-carrying capacity of the unbraced column, which is
neglected in Winter's model.
Numerous analytical studies of critical load values or of the required stiffness of
bracings in perfect columns or beams under a variety of loading and boundary
conditions and of bracing configurations have been carried out, (e.g. [13-22]).
Some of the studies resulted in closed-form solutions, but most of them made use
of numerical analyses instead. Some of the researches concerned the effects of
imperfections on the load-carrying capacity of simple beams and/or on the
magnitude of brace forces (e.g. [23-25]). There have also been numerous
numerical studies of brace design requirements for imperfect steel bridges (e.g.
[26-35]). Few studies, however, have been concerned with software development
for the analysis of bridges during their construction (e.g. [36]). Various studies
have been carried out on the derivation of simplified solutions for obtaining
critical load values or the magnitudes of bracing forces (e.g. [37, 38]). Very few
studies, however, have dealt with the bracing performance of steel bridges while
taking account of different imperfection shapes that can be involved (e.g. [39]).
Similarly, relatively few full-scale experimental works (e.g. [40-44]) were to be
found on the performance of bracings in steel bridges.

1.4 Terminology
Some important terms will be explained here to assist readers:
- Instability is a condition in which sudden sideway failure occurs in the case
of a member subjected to high compression stresses, typically less than the
ultimate capacity of the material involved.
- Critical load is the load at which a structure passes from a stable to an
unstable state [45]. After this particular load level (bifurcation point) has
been reached, two equilibrium paths are possible. The critical load can be
obtained by examining the equilibrium (either an algebraic equilibrium in
the case of discrete systems having rigid members or a differential
equilibrium in the case of continuous systems having elastic or inelastic
members) or by utilizing the principle of minimum potential energy in the
case of a virtually deformed system. Bifurcation is a branch point in a load-
deflection curve after which when further load is applied two equilibrium
states are possible  the one with zero deflection that can only occur
theoretically in perfect bars, and the other with large deflections.

4
- Post-buckling behavior of structural systems can be studied through
adopting the large-deflection assumption in the case of either perfect or
imperfect members. Three post-buckling situations can theoretically occur:
either a hardening or a softening post buckling, or a transitional case. Plates
and the web of a built-up girder for example, can exhibit considerable
strength enhancement when the critical load has been exceeded. Whereas
shells are imperfection-sensitive, they reach their ultimate resistance values
after partial yielding of the cross-section has occurred, this resulting in a
softening post-buckling. Slender columns reach their critical load value
after only small deformations have occurred.
- Small-displacement (or small-strain) theory: Here the occurrence of
displacements is assumed being very small, this allowing the
approximations of , , and 1.0 to be used to
simplify the mathematical equations involved. Employing small-
displacement theory enables the critical load to be obtained. The assumption
of large-deflections occurring can provide information regarding post-
buckling.
- Principle of minimum potential energy: The total potential energy of an
elastic system consists of the internal work,  i.e. the strain energy
absorbed by the elastic structural members and by the bracings  and the
external work, , i.e. the work performed by the loads applied along the
path traveled from the original un-deformed reference point. According to
the principle of minimum potential energy, the deformations corresponding
to the maxima and the minima of the total potential energy are the
equilibrium positions, the minima corresponding to the stable state.
- Initial imperfections: In practice, all structural members are imperfect in
terms of initial geometry, load eccentricities, and residual stresses. The
large deflection theory of imperfect systems provides a deformation history
including e.g. the loss of stiffness prior to a bifurcation point.
- Falsework or scaffolding: In the thesis, both terms consist of temporary
structures used for construction purposes to mold the concrete deck of steel-
concrete composite bridges and to support fresh concrete until it has
hardened. Two types of such temporary systems tend to be used in practice.
In the present text, the term falsework is used when such temporary
structures are only built up with use of timber-frames/trusses, the term
scaffolding being used when their structure also includes steel-pipe bars.

5
1.5 Outline
The present chapter includes a short description of the problem, the main
objectives of the study, the limitations of the study, and an overview of the
terminology used in the context.
The second chapter reviews a number of steel bridge collapses reported in the
literature associated with problems of instability.
The term lateral-torsional buckling is frequently used in this context. One of the
papers included here, Paper II, involves a number of analytical investigations
concerning the derivation of solutions regarding the buckling capacity of
unequally spaced lateral bracings placed at the level of compression flanges in
steel girders. Accordingly, a brief introduction to the theory of beam instability is
provided in the third chapter. The information contained in Chapter 3
demonstrates the limitations and the difficulties in deriving closed-form solutions
in instability analyses, even in the cases of very simple structural systems.
The fundamentals of beam bracing, lateral-torsional instability considerations
during the construction of steel-concrete composite bridges, as well as the typical
bracing systems commonly used in such bridges are taken up in Chapters 4-6. The
stabilizing potential of corrugated metal sheets and of precast concrete decks here
are also discussed. The bracing performance of scaffolding of types frequently
used in Sweden is considered in Chapter 6. The results of experimental and
numerical investigations of the bracing performance of such scaffoldings are
presented in Paper III.
The shape and the magnitude of initial geometric imperfections have significant
effects on the load-carrying capacity of the steel bridges and on their bracing
forces in particular. This was the main concern in Paper IV. Certain basic
information and discussions that could not be provided in Paper IV are presented
in Chapter 7.
Relatively little information concerning bridge bracing based on laboratory tests is
available. It is highly important that details of the test-setup employed in such
works be available for use by other researchers, for example for the purpose of a
calibration. In Chapter 8, the test setup designed and employed in the present
study is elaborated. The drawing sheets used to build the specimens and the test
setup are also shown in Appendix I.
Commercial programs are valuable tools for research engineers, helping them to
initiate ideas and to expand parametric studies on models that are already
calibrated against test data. In Chapter 9, various concerns regarding the use of
finite element program are discussed. In addition, the techniques used in the
present study to import geometric imperfections into nonlinear analyses are
explained.
6
Finally, in Chapter 10, a summary of the findings and some suggestions for
possible future research are presented. Appendices at the end of the dissertation
provide the opportunity for further test-setup details and test data to be presented.
Appendix III, summarizes the bracing requirements that have been adopted in the
current design specifications both in the U. S. and in Europe.
Five papers written by the author during his Ph.D. study are also appended to the
thesis.

7
2 Examples of bridge failures during
construction associated with
instability

Bridge accidents during construction involving the failure of timber falseworks or

the overall collapse of a bridge often have had tragic consequences. In recent
decades, many bridges have collapsed and many people have lost their lives or
been severely injured as the result. In bridge construction with the huge costs and
the large numbers of workers it involves, safety should be regarded as being more
important than matters of overall construction costs and holding to a time
schedule. The bridge disasters recorded in history should remind engineers of the
consequences of their mistakes; when mistakes are made, the structure tends to
find them. Failure evaluations often report an ignored, underestimated, or unseen
engineering approach one that could easily have been avoided. Valuable lessons
can be learned, however, in reviewing the history of such errors, this enhancing
our understanding of structural responses of this sort under real conditions. In the
present chapter, a number of bridge failures that have occurred during bridge
construction  due either to instability of the bridge itself or to problems in their
falseworks  are summarized and their causes briefly described. There have also
been problems of instability during the construction of bridges having concrete
girders due to the bracings being inadequate. This can be seen for instance, in the
failure of the four-span Souvenir Boulevard Bridge in Laval, Canada, in the year
2000, in which four outer precast pre-tensioned girders slid off their bearings in
each of the interior spans due to inadequate bracing of the precast girders prior to
the concreting of the deck [46]. Emphasis in the present chapter is placed, however
on the instability of steel bridges and their falseworks.
It should be noted that many bridge accidents and the failure reports regarding
them are often not made public in cases in which no fatalities occurred, for fear of
possible legal consequences for the firms involved or harm to their reputations.
The author faced difficulties at times in obtaining basic technical information as
simple as regarding the cause of failure in cases that involved fatalities. Except for
the bridge failure discussed in Section (2.2.1), in which the author had access to
the failure reports, the information regarding other failure cases are from sources
of other types that are referred to.

9
2.1 Examples of steel-truss bridge failures during
construction associated with problems of instability
There have been a number of truss bridge failures due mainly to buckling of their
compression chords or diagonals [10]. These include for example, the following:
- Total collapse of a semi-parabolic truss bridge in Switzerland between
Rykon and Zell, 21 m in overall length that occurred in 1883, a collapse due
to buckling of the upper chord because of inadequate lateral stiffness, one
person being killed;
- Total collapse of a semi-parabolic truss bridge, called the Mountain Bridge,
having a total length of 28 m, that occurred in Austria in 1891, due to the
buckling of compression members because of inadequate lateral stiffness;
- Total collapse of a parallel truss bridge over Cannich in Scotland, 40 m in
total length, that occurred in 1892, due to the buckling of the top chord
because of inadequate lateral stiffness;
- Partial collapse of a semi-parabolic truss bridge, over the Morava River near
Ljubicevo in Serbia, 85 m in total length, that occurred in 1892 due to
buckling of the compression chord;
- Total collapse of a cantilever truss bridge over the St. Lawrence River near
Quebec in Canada, 853 m in total length and having an inner span of 550 m,
that occurred in 1907 due to failure of the under-dimensioned compressed
bottom chord during construction, killing 74 persons. The cross-section of
the chord was built-up of four non-compact web plates.
- Partial collapse of a six-span semi-parabolic truss bridge with a total length
of 554 m near Ohio Falls in Mississippi, U.S., that occurred in 1927, due to
lack of under-water bracings, killing one person.
There were no further reports that the author found of failure of steel-truss bridges
due to problems of instability. This probably indicates that the buckling of
compression chords as well as the lateral bracings they require are now well
understood by engineers, resulting in better production of bridges of this type.

10
2.2 Examples of failures of built-up steel girder bridges
during the non-composite stage associated with their
instability
A number of bridge failures associated with problems of instability in the main
girders are listed below, a brief description of each case being provided. Most of
the failures occurred during erection of the bridges, although in some of the cases
failure occurred during demolition.
- A five-span twin I-girder motorway Bridge with a total length of 272 m
collapsed near Kaiserslautern over the Lauterbach Valley in Germany in
1954. Total collapse of the inner span together with a lateral buckling of the
bottom flange of the side span occurred during erection of the bridge. The
compressive stresses in the bottom flanges of the side span were generated
through deliberate lifting of the internal supports and the applying of an
extra temporary gravity load to the suspended inner span, in order to induce
a pre-stressed condition in the finished concrete deck. The spacing of the
braces was increased from 4 m in the design of the bridge to 8-12 m in the
construction phase without any plan bracings being placed at the level of the
bottom flanges near the supports [10].
- A three-span twin box-girder steel bridge, The Fourth Danube Bridge,
having a total span of 412 m, approximately 32 m width, and 5 m depth
failed in Vienna in November 1969 [47]. The top flange of the final section
was shortened by 15 mm to fit the gap in which the cantilevers met in the
middle. This change was undertaken to adjust the closing section to the
cross-sectional rotations of the cantilevers, due to vertical deflections of the
large cantilevers brought about by their own weight as well as by thermal
elongations of the cantilevers during the day. The drop in temperature in the
evening and reversing of the thermal deformations generated tensile stresses
in the top flange together with compression in the bottom flange. The bridge
experienced major failures due to buckling at both the side-spans and the
inner-span near the regions having a zero bending moment. The width-to-
thickness ratio of the bottom flange stiffeners was also relatively large in
this case.
- A seven-span single trapezoidal girder bridge, The Cleddau Bridge in
Wales, UK, with a total bridge length of 819 m, failed in June 1970 [47].
The huge cantilever arm with a length of 61 m, a width of approximately
20 m, and a depth of 6 m fell to the ground during launching due to
buckling of the cross-girder at a bearing point, killing four people.
- A five-span three-cell trapezoidal steel bridge, cable-stayed in the three
inner spans, The West Gate Bridge over the Yarra River in Melbourne,

11
 Australia, with total bridge length of 848 m, failed in October 1970 [45].
The girders of one span, 112 m in length, buckled after leveling two half-
girders through the self-weight of eleven concrete blocks. Thirty-five
workers perished, some of them while working on the bridge or inside the
boxes, and many while on a lunch break beneath the span, where they were
crushed by the falling span.
- A three-span trapezoidal girder bridge, the Storm Bridge over Rhine, with a
total bridge length of 442 m, collapsed in Germany in 1971 [10]. Because of
buckling of the stiffened bottom flange, the cantilever girder broke off
causing partial collapse of the bridge, killing 13 persons. The longitudinal
stiffeners were not welded to the bottom flange at the joint of two cross-
sections, leaving a 400 mm long section of the bottom flange unstiffened.
- A six-span trapezoidal girder bridge, the Zeulenroda Bridge over the Weida
Reservoir in Germany having a total bridge length of 362 m, failed in 1973
[10]. The 31.5 m cantilever arm of the second span of the bridge collapsed
due to buckling of the stiffened bottom flange (the critical load value,
instead of the design value, had been used to determine the size of the
stiffeners), killing four persons.
- A two-cell steel box girder of a composite bridge at Bramsche over the
Mittelland Canal in Germany having a total bridge length of 60 m, failed in
1974 [10]. The concrete deck had been removed during demolition, leaving
the top flanges without lateral restraint, one person was killed.
- A 97 m long steel girder 2.4-3.65 m in depth collapsed during the erection
of the Syracuse Bridge (with a total bridge length of 670 m) in New York in
1982 [10] due to inadequate bracing, one person being killed.
- Lateral-torsional buckling of a girder weighing 120 tons occurred during the
demolition of a bridge near Dedensen over the Mittelland Canal in Germany
in 1982 [10] after the lateral connections of it had been removed.
- A single steel girder weighing 43 tons fell down during construction of the
Astram Line Metro Railway Bridge in Hiroshima, Japan in 1991 [48] due to
problems of instability, killing 14 persons.
- A three-span triple I-girder bridge, The State Route 69 Bridge over the
Tennessee River, having a total bridge length of 367 m, collapsed in
Tennessee on May 1995 [10]. The bottom flanges were braced substantially
by means of relative lateral bracings, but no lateral bracing was specified
for the top flanges. Cross-frames had been built up by use of double angle
profiles. Total collapse occurred during erection of the bridge when a cross-
frame had been removed in order to fix the connection, one person being
killed.

12
 - A steel girder weighing 50 tons dropped onto the road during a demolition
in Harrisburg, Pennsylvania in 1996 [10]. The failure occurred because of
the girder's flange being cut at two points, causing a reduction in lateral
stiffness, one person being killed.
- Bridge Y1504 in Sweden, with a trapezoidal girder and a total bridge length
of 65 m experienced global lateral-torsional buckling on June 2002 during
concreting of the deck [11]. The stay-in-place metal decks were designated
to also provide lateral bracing of the compression top flanges, yet the type
of corrugated sheets and the number of fasteners had been reduced during
construction. Fortunately, although a few workers fell to the ground below,
the accident had no fatalities because of low height of the bridge.
- The Marcy Bridge in New York with a trapezoidal girder and a total bridge
length of 52 m experienced global lateral-torsional buckling during
concreting of the deck in October 2002, one person being killed [49]. No
plan bracing between the two end supports had been used in this pedestrian
bridge.
- A single steel I-girder (30 m long weighing 40 tons) of the Interstate 70
Bridge in Denver dropped from a freeway bridge into the traffic below in
2004 causing a car to crash, killing three persons [48]. The accident
occurred during the widening of the existing bridge. The girder was
temporarily braced to the existing bridge at five points. However, the
expansion bolts that attached the bracings to the existing concrete deck were
not sufficiently embedded in the deck.
- Three out of seven I-girders of 102nd Avenue over Groat Road Bridge in
Canada buckled in 2015 due to lack of permanent bracings while the girders
were being erected [50]. The subcontractor (the largest private firm in
Canada, with 40 years of experience) misread the specifications regarding
the required bracings. For the one year delay this brought up, the contractor
is required to pay almost $4.2 million in penalties.

2.2.1 The collapse of Bridge Y1504 in Sweden

In this section the original report issued after evaluation of the Bridge Y1504
accident [9] by both the firms involved and the independent parties are reviewed.
Bridge Y1504 over the Gide River in Sweden, located 90 km west of the city of
Umeå, collapsed on June 12, 2002. The bridge had a 65 m long trapezoidal cross-
section provided with nine intermediate cross-diaphragms to control distortion.
Corrugated metal sheets served both as lateral restraints to the top-flanges and as a
stay-in-place formwork for the fresh concrete while the bridge was being built.
The average self-weight of the fresh concrete, including both the reinforcement

13
and the metal sheets was approximately 55 kN/m. The steel girders had an
average weight of 17 kN/m. It was planned the concreting would be carried out in
two steps; first, by covering approximately half of the span symmetrically through
pouring at the mid-span. The cross-section of the bridge suddenly rotated by 90° at
mid-span, after only a quarter of the volume planned for the first step had been
poured. The end-bearings were also severely damaged due to warping of the cross-
section. According to witnesses, the entire accident took place within just several
seconds. Six workers fell into the river, but fortunately were not injured.
A site visit revealed that both the size of fasteners and the type corrugated sheets
employed were altered from the prescribed design in the following ways: The type
of sheets involved changed from TRP45 to PEVA 45 (see Fig. 2.1), the diameter
of the seam fasteners being changed from 6.3 mm to 4.8 mm, the spacing between
the seam fasteners being changed from 150 mm to 300 mm. The diameter of edge
nails placed at each valley was 4.5 mm. Apparently, assuming the stay-in-place
sheets were to function as a formwork only, the constructor replaced the
prescribed sheets by an equivalent or better alternative, which was either available
or cheaper on the market than what had been planned originally. Although such
changes can occur in virtually any project as a routine matter, the failure of the
design and the construction sector to inform each other of the changes made,
created an unsafe working environment, one that could have consequences that
could be tragic.

Figure 2.1
The prescribed (TRP 45) and the corrugated metal sheets (PEVA 45) used for the construction of
Bridge Y1504.

Prior to the concrete pouring commencing, a relative vertical deflections of 18 mm

between the top flanges at mid-span was observed, although this was considered to
be within acceptable tolerance limits. This measurement gave an approximate
initial twist value of 0.005 Rad at mid-span. Calculations regarding system
buckling of the steel girder and global buckling of the metal sheets were missing
in the design documents.
The equations regarding the shear flexibility of corrugated sheets given by
handbooks and guidelines, such as [51, 52], yield considerable discrepancies and
are difficult to validate for a sheet of a particular type [6]. Despite “strong”
contradictory statements regarding the adequacy of the prescribed corrugated
14
sheets in stabilizing the girder, and also regarding the responsibility of the firms
involved, there appeared to be a “general” agreement between the various failure
reports that were issued for the causes of the accident. A review of the failure
reports issued by the independent third parties that were involved showed the
actual shear stress values to probably be greater than the shear resistance of the
critical edge fasteners. This was probably true, even ignoring the torque that could
be generated from the wind load. The folded stiffeners in the type of sheets that
were utilized considerably increases (by a factor of approximately 1.8) the
warping flexibility of the sheets considerably compared with similar sheets that
lack such stiffeners. This increase in flexibility led to strong forces being directed
at the edge fasteners, since the panel was not sufficiently stiff to properly
distribute the shear stresses between the edge fasteners of each panel. Use of two
nails instead of one for each stiffener, one on each side, would have been able to
increase the shear stiffness of the panel appreciably and presumably reduce the
risk of failure.
During concreting, in addition to the in-plane bending stresses (which were
greatest in the mid-span and were zero at the supports), the steel girder with a
semi-closed-section was subjected to shear stresses created by torsion (which was
greatest at the supports and “zero” at mid-span). Assuming the steel girder to have
a thin-walled closed-section, the torque can be assumed to develop shear flows
of ∙ ∑ /2 across the section, being a modification factor to take into
account stress concentration effects due to appreciable changes of thickness in the
cross-section, and being the enclosed area defined by the wall-midline in a
closed section. This shear force should be resisted by both the attachments and the
St-Venant stiffness of the closed cross-section. The torques, , are generated by
the following:
- the self-weight of the steel girder, , due to possible lateral crookedness.
This torque can be approximated by considering, for example, a half-sine-
shape initial crookedness having a maximum value of ∆ /500 at mid-
span, which leads to:

2
2
, ∙ ∆0 2.1
0 500

- an uneven distribution of the fresh concrete across the deck and the
eccentricity of the fresh concrete load from the shear center, both of these
originate from the initial twist of the cross-section.
- the wind load during concreting.
The end panels had to be attached all the way around in order to function properly
as a shear diaphragm. However, the sheets were attached neither to the cross-
diaphragms of the supports nor to the intermediate cross-diaphragms.

15
Consequently, at the support points (i.e. at the location of maximum torque as
explained earlier) the panel had no attachments except a single fastener at each
corner of the free edges (see Fig. 2.2). Thus, the corner fasteners of the free edges
became overloaded, taking shear forces from the sheet. The progressive shear
failure of the edge fasteners at both ends converted the closed-section to an open
cross-section. This significantly lower torsional stiffness may not be adequate to
resist the torque near the support points. Note that the shear center of an open
trapezoidal section is far below the bottom flange and that the rather small
equivalent thickness of typical corrugated sheets can barely move the shear center
towards the centroid. Thus, the difference between the torque arms of an open-
section and a closed-section (utilizing typical thin metal sheets) with respect to the
shear center is negligible.

Figure 2.2
An elevation view illustrating the distribution of the shear forces over the fasteners present at the end
supports, a) representing the case in which the end sheet is properly fastened to the cross-diaphragms
of the supports utilizing an angle profile for example, and b) representing the case in which the end
sheet is not attached to the support cross-diaphragm.

2.3 Steel bridge accidents during concreting that were

associated with problems of instability in their timber
falseworks
A number of accidents have occurred during construction of bridges due to
problems of instability within the falseworks involved, some of these accidents
resulting in tragic events. The failure of the falsework of the Älandsfjärden Bridge
in Sweden in 2008 is an example of such an accident, five construction workers
there felling 20 meters to the ground, two of them being killed and two severely
injured [11].
Among the timber-falsework failures during bridge construction that have
occurred, some 60 cases described in [10], inadequate lateral stiffness or strength
was found to have made a large (20%) contribution to bringing about such
failures. Insufficient bracing or lack of it was also identified as the primary

16
enabling event with respect to the collapsing of bridge falseworks observed in a
survey of bridge falsework failures reported since 1970 [53].

2.4 Conclusions
The author found no information regarding failures of steel-truss bridges being due
to inadequate lateral bracing or to buckling of their compression members during
the last few decades. This indicates clearly that the stability requirements of such
bridges are rather well understood by engineers.
However, built-up bridge girders still fail due to inadequate bracings, or errors in
the proper evaluation of the local and global buckling capacity of the steel girders.
Global lateral-torsional buckling of trapezoidal girders was the failure mode in two
recent bridge accidents.
The stabilizing function of corrugated metal sheets is strongly affected by their
shear stiffness and strength as well as by the shear resistance of their attachments.
Calculations regarding the shear stiffness of metal sheets are rather complex, and
the shear stiffness of them depending upon a number of factors, such as their
geometry, the number of attachments, and warping of the section involved. Further
studies are needed in order to enhance the knowledge for such members when they
function both as a stabilizing system and as a stay-in-place formwork during the
construction of steel bridges.
The failure of Bridge Y1504 taught us that any major change in common design
practice should be highlighted in design documents. Lack of proper
communication between the design and the construction sectors can potentially
create am unsafe working environment in construction operations.
Several accidents have occurred during demolition of steel bridges. This shows
that constructors may well ignore the stability assessments of steel girders during
demolition. There, the steel girders involved can be vulnerable to lateral-torsional
buckling once the lateral support, provided by a concrete deck for example, is
removed. In situations of other types, failures of this sort can occur when the size
of a compression flange is reduced locally during demolition, resulting in a
significant loss in lateral-torsional stiffness.
The recent failures that have occurred during the concreting stage indicate there to
be a need for further investigations concerning the stability of bridge falseworks.
Despite the importance of stability assessments of bridge falseworks, guidelines
regarding the stability requirements of such systems during construction are
ignored in a great extent both in Eurocodes and in other code specifications.
Regardless of having sophisticated plans for management, design, construction,
maintenance, cost analyses, and the like, a simple human error in the design or in
17
the installation of timber falseworks can lead to catastrophic events. Accidents of
this sort very frequently result in fatalities, considerable delays, and significant
extra costs for replacement of what has been destroyed and for failure evaluations.
Despite the elementary techniques used in the structure of bridge falseworks, these
often represent a considerable portion of the construction costs of a composite
bridge. Alternative systems of greater efficiency and safety to the currently used
falseworks can be of great help during construction of steel-concrete composite
bridges.

18
3 Theory of beam stability

A stability criterion represents a limit state such that at a certain load a structure
passes from a stable state, involving a situation in which a small increase in load
generates only a small increase in displacement, to an unstable state one, in which
a small increase in load results in a large change in displacement [45]. Generally,
stability analyses include either studying of the local buckling of a cross-sectional
component in the presence of compressive stresses, or determination of the critical
load of structural members of systems (such as a column, beam, frame, arch, or
truss) that corresponds to their lack of stability. The major concern of the present
research is the lateral-torsional buckling of steel bridge girders. In line with this,
basic concepts concerned with the lateral-torsional instability of beams are taken
up briefly in the present chapter. The text begins with a brief introduction to
matters of elastic and inelastic buckling followed by a discussion of the effects of
residual stresses on the critical load values involved.

3.1 Introduction
In 1729, a Dutch scientist, Pieter van Musschenbroek, performed pioneering work
on the buckling of compression struts, his discovering that the failure load
involved is inversely proportional to the square of the length of the struts [54].
Adopting the assumption of a proportional relationship between the curvature at
any point of a bent member and the resisting moment that develops  a
relationship that Jacob Bernoulli introduced in 1705  the Swiss mathematician
Leonard Euler presented a formula in 1757 for predicting the elastic critical load
value of perfect columns: , / [55]. Euler was uncertain about the
term , “a dimension constant which seems to be proportional to the square or
even cube of thickness and should be obtained experimentally” [55]. In the early
1800's, Euler's formula was widely criticized by engineers and authorities, such as
Coulomb  who believed that the failure load of a column depends only on the
cross-sectional area and not the length of the column. This occurred since the
formula failed in experiments to correctly predict the failure load of columns
composed of materials in use at the time, i.e. masonry and timber. Materials to
which Euler's formula has found to be most applicable, such as structural steel,
first became commercially available some 100 years after Euler's contributions
within this area.
19
More than a century after Euler presented his formula for elastic buckling of
columns, Ludwig von Tetmajer [56, 57] carried out experimental investigations
concerning the causes of the Münchenstein truss railway bridge disaster in June
1891, which killed over 70 persons. The study revealed that the Euler's formula,
that had been used to design such bridges at the time, needed to be modified for
columns having an intermediate slenderness ratio, e.g. one of 0.3 1.4.

3.2 Effects of material inelasticity on bracing

requirements
Euler stated the following:
“The stiffness moment”, / , “is not limited to elastic bodies, and it concerns
a bending by means of which any body resists a change in curvature to
reestablish its original shape” [55].

For columns that experience a bifurcation of equilibrium above the proportional

limit, Euler's formula provides a prediction that overestimates the critical load due
to the regression of strain in the inelastic range, i.e. when / ; where
is the tangent modulus, and and are the normal stress and strain values. To
expand Euler's concept to be applicable within the inelastic range, Engesser
proposed a tangent modulus concept in 1889 [45] from which the critical
slenderness ratio of a column, / , can be calculated using Eq. (3.1), where
is the tangent modulus obtained from the - / curve for a given stress value,
and being the radius of gyration of the cross-section.

/ / 3.1

Performing 32 column tests in 1889, Considere suggested that if buckling occurs

above the proportional limit, the elastic modulus in Euler's formula should be
replaced by an effective modulus, , which is a value between the elastic and the
tangent moduli [45]. This theory is now referred to as the reduced/double modulus
concept, which improves the tangent modulus concept by considering both
material and cross-sectional properties, see Eq. (3.2). The double-modulus concept
makes use of the elastic modulus for the elastic zone of the cross-section, which
has a moment of inertia of . , and a tangent modulus for the inelastic part, which
has a moment of inertia of ..

. .
3.2

20
Pincus [12] stated that since the flexural stiffness of a member decreases in the
inelastic range, its critical load is smaller than the elastic critical load predicted by
the Euler's formula. Thus, to achieve a desired critical load of a column loaded in
its inelastic range and to compensate for the loss in stiffness of the column, braces
stiffer than the “ideal” value for it (= , which serves the brace points similar to
an immovable support), as predicted by Winter's model (see Section 1.3), should
be provided. Gil and Yura [58] reviewed Pincus's claim. In their experimental
studies, artificial inelasticity for a test-column was created through use of high-
strength steel inside of the cross-section and low-strength steel on the outsides. For
the cases studied, in contrast to Pincus's claim, the results showed that the full
bracing requirements were independent of the state of material (i.e. whether it was
elastic or inelastic).

3.3 Effects of residual stresses on buckling load

A large number of tests [56] showed that the load-carrying capacity of columns
having an intermediate slenderness ratio was considerably smaller than the column
strengths predicted on the basis of Eq. (3.2). This occurs due to the detrimental
effects of residual stresses combined with the geometric imperfections and
material nonlinearities that are present. The effects of residual stresses on the
critical load values and on bracing requirements may in practice not be possible to
determine for a given case. A number of studies have discussed the effects of
residual stresses on the strength of very simple column members. For instance,
Galambos [45] investigated analytically the effects of an assumed residual stress
distribution on the load-carrying capacity of a column with a rectangular cross-
section with respect to its either the weak or the strong axis. The residual stresses
were distributed linearly across the cross-section from a tensile stress value
0.3~0.5 times that of the yielding stress at the center of the cross-section to a
compressive stress of the same magnitudes at the edges. Such studies, together
with extensive laboratory tests of varying mechanical properties (such as
slenderness, and cross-sectional shape) and fabrication processes resulted in the
development of column and beam design curves in the current code specifications.
These design curves take account of, for example, the effects of geometric
imperfections, typical load eccentricities, and residual stresses on the design load
value of typical columns and beams used for practical purposes.
In the numerical investigations of the present study, the effects on the performance
of bracings in steel bridges, of the shape and magnitude of geometric
imperfections, along with of the material and geometric nonlinearities involved,
were included in the analyses. Taking account of the effects of residual stresses
was seen, however, as being outside the scope of the present Ph.D. research.

21
3.4 Lateral-torsional buckling of doubly-symmetric
simply supported beams subjected to uniform bending
Lateral-torsional buckling is a failure mode of beams subjected to an in-plane
bending that can occur along the unbraced length causing both lateral movement
of the compression flange and twist of the cross-section, see Fig. (3.1).

Figure 3.1
Lateral-torsional deformation of a doubly symmetric wide-flange beam subjected to a uniform in-
plane bending moment of , where , , and are lateral deflection, vertical deformation, and
twist of the cross-section at the location along the span; , , and are the bending moments
and torque components of in the deformed configuration.

The lateral-torsional buckling of steel beams is of particular importance during the

erection of a bridge before all the bracings have been set in place. There have been
many fatal failures in bridge construction due to lateral-torsional buckling
resulting from improper bracing (see the examples presented in Chapter 2). In this
section, Timoshenko's approach [59] to predicting the critical bending moment of
a doubly symmetric beam subjected to uniform bending moment is first described
briefly. This is followed by various hints of modifications required for the solution
arrived at, enabling different loading and boundary conditions, cross-sectional
asymmetry, and inelasticity to be taken account of. It should be noted that the
lateral-torsional buckling of braced members is rather complex, exact analytical
solutions only being possible for relatively simple situations involving unbraced
beams. However, various conservative recommendations are available on the basis
of numerical studies of different practical demands needing to be met.
Fig. (3.1) illustrates the lateral-torsional buckling deformation of a doubly
symmetric beam, together with a plan view of a deferential length of the beam at
location along the span in both a deformed and a non-deformed situation. The
single span beam in question is subjected to a uniform in-plane bending of .
22
The out-of-plane deformations are represented by centroid lateral and vertical
movements of and , and a cross-sectional twist of ∅. The beam is free to warp
and is restrained against twist at both ends. The material involved is elastic, small
deformation theory being applied ( ∅ ∅ and ∅ 1 , it is being assumed that
local buckling does not occur in any part of the cross-section. In Fig. (3.1), the
bending moments and torsional components of that are applied to the
deformed cross-section are as follows:

; ∅; 3.3

An equilibrium of the external and internal forces requires that the followings
apply [59, 60]:

3.4

3.5

3.6

Only Eqs. (3.5)-(3.6) involve lateral-torsional deformations. Substituting from

Eq. (3.5) into the equation obtained through a differentiation of Eq. (3.6) with
respect to results in the following differential equation:

0 3.7

Substituting the corresponding boundary conditions (where at both ends the lateral
and vertical deflections, the twist, and the bending moment about axis are set to
a value of zero) into the solution of Eq. (3.7) results in the following equation for
the critical moment value:

, , ∙ 1 3.8

3.5 Modifications required in the basic approach to the

critical bending moment value
The basic Timoshenko's approach presented here as Eq. (3.8) was derived on the
basis of the following assumptions: the beam has a doubly-symmetric cross-
section, simply supported boundary conditions, and is subjected to a uniform

23
bending moment. A closed-form solution is rather difficult or even impossible to
obtain for most of the other boundary and loading conditions that exist in practice.
For limited cases, a number of approximate modifications  involving mainly
numerical studies  have been suggested in the literature. Major concerns
regarding possible alternative conditions other than those assumed in derivation of
Eq. (3.8), are explained in the subsections that follow.

3.5.1 Effects of different boundary conditions

A similar mathematical process such as that presented in Section (3.4) can be

carried out so as to obtain predictions of the critical bending moment for different
boundary conditions other than the warping-free and twist-restrained situations
assumed in the derivation of Eq. (3.8). An approximate conservative solution
based on an extension of the effective length method is given in Eq. (3.9) [61].
This equation can be used as a general solution for different boundary conditions.
Under both-end fixed, both-end free, or one-end-free and the-other-end-fixed
conditions, the and values involved would be equal to 1.0, 0.5, or 0.7 [61].

, , ∙ 1 3.9

3.5.2 Effects of different loading conditions

A variety of loading configurations for a beam prone to lateral-torsional buckling

are possible to occur in practice. Transverse loads applied to the top or bottom
flange of the beam shown in Section (3.4), decrease or increase, respectively, the
critical moment value [4]. Unfortunately, for most practical loading conditions,
employing an exact analytical solution for determining the critical moment value
is either impossible or cumbersome. A number of studies (e.g. [62-65]) have
presented expressions that can modify Timoshenko's basic equation so as to
account for the benefits of variable bending moment distributions other than the
uniform bending moment assumed in derivation of Eq. (3.8). Such
recommendations result in critical moment values that are based on conservative
fits to the data involved. For example, the following factor, based on a use of a
so-called three-quarter-point-moment method (see e.g. [4, 66]) is recommended by
AISC [67]:

12.5
3.10
2.5 3 4 3

24
In Eq. (3.10), is the absolute value of the maximum in-plane bending
moment in the unbraced segment, and , . . . , are the absolute in-plane
bending moment values at the quarter, at the center, and at the three-quarter points
of the unbraced segment.
Loads applied above shear center destabilize and those below stabilize the beam as
lateral-torsional buckling develops. To determine the load height effects when
transverse loading is applied to the cross-section at some particular height, Eq.
(3.10) needs to be multiplied by a 1.4 / factor [4], being the location of the
load that is applied with respect to the midheight of the cross-section and is
positive for the values below the midheight of the cross-section. Although the
load-height effect decreases as the length of a member increases [68], the
modification factor mentioned above provides an upper-bound value for the beam
spans typically employed. Eq. (3.10) needs to also be multiplied to a modification
factor of 0.5 2 , / when the beam is subjected to a double-curvature
bending along an unbraced segment; where , is the moment of inertia of the
top flange with respect to the weak axis of the cross-section.
As the result of numerical studies that were performed, Yura et al. [37] showed
that for twin-I girders that were inter-braced by means of typical cross-bracings,
the load height had only minor effects on the critical load values. Park et al. [29]
also developed various modification factors to account for the effects of lateral
restraints. Those critical moment modification factors have been developed for
single beams in connection with numerical studies. While for cases encountered in
practice in which a number of beams involved are inter-connected by means of
cross-bracings, the results arrived at through such various modification factors
might not be truly justified.

3.5.3 Effects of lateral restraints (Paper II)

Fig. (3.2 a) illustrates a situation in which a simply supported beam is braced

laterally at one-third and two-thirds of the span locations. The side segments are
subjected to considerably lesser in-plane bending than the central segment, and
acting as restraining members at both ends of the central segment. Such a case was
studied by Galambos [45], who calculated the effective buckling length of the
central segment as being 0.83 .
Another case of the restraining effects of the side segments on the effective
buckling length of the compression flange of the inner segment is that illustrated
in Fig. (3.2 b). There the side segments, having shorter unbraced lengths provide
restraints on lateral rotation at both ends of the central segment. This case is a part
of an investigation reported by the author in paper II [69]. Although the primary
concern of paper II is the applicability of a simplified method in prediction of the
critical moment value of laterally braced beams, the study also examines the
25
restraining effects of unequally spaced lateral bracings on the effective buckling
length of a critical segment having greater unbraced length. This matter was
investigated analytically by varying the stiffness and the location of the lateral
bracings placed at the level of the compression flange. The results of that part of
the study showed that the side segments of which the unbraced length is shorter
can have a significant impact on reduction of the effective buckling length of the
compression flange of the central segment (down to 0.5 ).

Figure 3.2
End-restraint effects of side segments in laterally restrained beams on the effective buckling length
of the central segment, .

3.5.4 Effects of cross-sectional asymmetry

Additional twisting moments can be developed along the span in beams of

different flange sizes. The twisting moment there, is generated by normal stresses
on each of the differently warped flanges in a mono-symmetric cross-section. Eq.
(3.6) needs to be replaced there by Eq. (3.11) so as to take into account the
additional twist referred above [59]. In Eq. (3.11) is a cross-sectional property
in mono-symmetric sections that is defined by Eq. (3.12), and is the distance
between the centroid and the shear center of the cross-section.

3.11

1
2 3.12

26
Substituting from Eq. (3.5) into the equation obtained through a differentiation
of Eq. (3.11) with respect to results in the following differential equation:

0 3.13

Applying the corresponding boundary conditions in the solution of Eq. (3.13)

yields the following equation as a general solution for obtaining the critical
moment value of beams having mono-symmetric cross-sections. A positive sign
should be employed there when the top flange is in compression, and a negative
sign is when the bottom flange is in compression.

4
, , 1 1 3.14
2

The following approximate formula, obtained on the basis of numerical parametric

studies that have been carried out is also given in the literature [70] for in the
case of wide-flange beams being involved, its providing relatively accurate results
as long as / 0.5, where , / , and is the moment of inertia of the
compression flange with respect to the weak axis of the cross-section.

0.9 2 1 1 3.15

3.5.5 Effects of inelasticity on lateral-torsional buckling

Although a number of parametric studies of the inelastic lateral-torsional buckling

of simple beams have been carried out (e.g. [21, 71, 72]), investigating critical
moment values for the inelasticity of wide-flange members here is not a simple
task. For a beam with a relatively simple cross-section (such as a rectangular
beam) that is subjected to a uniform bending moment the analysis can be divided
into three regions based on yielding of the farthest tension and/or compression
fibers of the cross-section. In approximating the material properties involved
through use of a bi-linear stress-strain model, the elastic modulus of the yielded
regions can be assumed to be “zero”. Thus, the remaining cross-section which
resists the lateral-torsional deformation would only be the elastic core in an
elastoplastic stress distribution across the cross-section. Substituting the cross-
sectional properties obtained on the basis of assumptions referred to above into the
critical moment formula, the critical lateral-torsional bending moment can be
obtained for various lateral-torsional slenderness ratios. However, under non-
uniform-bending conditions in which the maximum bending moment along the

27
unbraced span generates elastoplastic stresses in the cross-section, yielding
commences at the maximum moment location, whereas the cross-section of the
remaining span may be completely within the elastic range.
For design purposes, the design curves given by the code specifications (e.g. [73]),
based on extensive numerical and experimental data are the more practical tools
for examining the effects of nonlinearities.

3.5.6 Effects of variable cross-section on lateral-torsional buckling

(unpublished work)

Most of the theoretical approaches available, including the basic approach

expressed in Eq. (3.8), concern the critical load-capacity of beams of constant
cross-section. However, in order to suit the bending moment distribution, the size
of the cross-section (typically the size of the flanges) is often varied along the
span. Trahair and Kitipornchai [13] discussed the critical moment solutions for a
doubly-symmetric beam having symmetrically stepped flanges. A concentrated
transversal load was applied at mid-span. For the stepped beam, the middle
segment was of greater cross-sectional size. The equations were solved
numerically. The results suggested that a reasonable approximation of the elastic
critical load of such stepped beams can be determined by linear interpolation
between the two critical load values, there calculated for two separate beams of
constant cross-section as the side segments and the middle segment along the
span. Trahair and Kitipornchai claimed this approximation to have the advantage
of being simple and giving either accurate or conservative result.
The author performed investigations concerning the lateral-torsional buckling of
stepped beams having a variety of cross-sectional dimensions and span lengths.
The results of numerical studies here are shown in Fig. (3.3)-(3.6). The depth of
the beams was 1250 mm throughout the span, whereas the span-to-depth ratio
varied as follows: / 3, 5,7,10, and 13. In the doubly symmetric cases, the
flange width to beam depth ratio was approximately 0.25. The ratio of flange
width to flange thickness varied between 16 and 20. It should be noted that the
single beams studied were unbraced along the span between the end-supports.
The vertical axis in the figures represent the normalized critical moment values of
the stepped beams involved with respect to the critical value of a beam with a non-
variable cross-section as the middle segment along the span (i.e. section 1 for each
case, as shown in the figures). The horizontal axis in the figures represents the
middle segment length ratio with respect to the beam span.
The critical moment value of a stepped girder with a given / value in Figs.
(3.3)-(3.6), according to the approximated method suggested by Trahair and

28
Kitipornchai, can be obtained by a linear interpolation between the critical
moments corresponding to / 0 and to / 1.0 values.
For the doubly-symmetric stepped-beams that were studied, the results presented
in Figs. (3.3)-(3.4) show that a linear interpolation between the critical load values
of the side and of the middle segments, all having uniform cross-sections provides
either an accurate or a conservative value for the critical load. However, a linear
interpolation between the critical load values yielded unsafe predictions of the
critical load values of the monosymmetric stepped-beams that were studied; see
Figs. (3.5)-(3.6). Thus, the results provided examples of cases of monosymmetric
beams for which the approximation method suggested by Trahair and Kitipornchai
[13] is incorrect.

Figure 3.3
Critical moment values of doubly symmetric I-girders of varying cross-sectional dimensions. The
beams were subjected to a uniform load applied at the top flange level 0.5 and at the centroid
0.0 ; is the critical moment value of the stepped beam and , is the critical moment
value of a beam of constant cross-section as in section 1.

Figure 3.4
Critical moment values of doubly symmetric I-girders of varying span length. The beams were
subjected to a uniform load applied at the level of the top flange 0.5 .

29
Figure 3.5
Critical moment values of mono symmetric I-girders of varying cross-sectional dimensions. The
beams were subjected to a uniform load applied at the top flange level 0.5 .

Figure 3.6
Critical moment values of mono symmetric I-girders of varying the size of the compression flange.
The beams were subjected to a uniform load applied at the top flange level 0.5 .

30
4 Fundamentals of beam bracing

Bracings are the structural members normally needed to sustain the stability of
load-bearing members in order to reach a desired load-carrying capacity  in
addition to the function they have of resisting the horizontal loads. This takes
place by means of their controlling the out-of-plane deformations of the main
members. The bracings should be adequate in terms of both stiffness and strength.
Since lateral-torsional buckling involves two types of deformation, i.e. lateral
movement and twist, bracings are normally studied in terms of the two separate
categories of bracing, lateral and torsional, that control mainly the lateral
deformation of compression flanges and the relative displacement of the top and
bottom flanges, respectively. However, lateral bracings can also prevent twisting
of a beam cross-section if they are located at a distance from the shear center of
the cross-section. Bracings of these two types can be placed either at discrete
nodal points or continuously along the span involved. In beam analyses, lateral
bracings are normally modeled as linear translational springs, and torsional
bracings as rotational springs. The present chapter introduces the fundamentals of
beam bracing and applications of it within the bridge sector. A brief discussion of
“the columns having elastic supports” is also provided at the end of this chapter,
since this model is currently used by design engineers (see Appendix III) for
estimating the lateral-torsional buckling of the bottom flange of continuous
composite girders braced by means of cross-bracings.

4.1 Introduction
Bracings are generally classified in terms of four separate categories based on their
function they fulfil:
- Discrete or nodal bracing, which controls the corresponding movements of
the braced member at the attachment point; intermediate and support cross-
beams/-diaphragms/or -frames in bridge application are examples of nodal-
torsional bracings.
- Continuous bracing, which resists lateral movement of the compression
flange or twisting of a beam cross-section along the braced span; reinforced
concrete decks are examples of the bracing of the continuous type.

31
 - Relative bracing, which controls the relative movement of two braced
points. Plan bracing is an example of the relative bracing type.
- Lean-on bracing refers to cases in which one member is braced by leaning
on the stiffness of an adjacent member. It can occur in twin- or multi-girder
bridges if at least one of the girders is loaded less or has a larger load-
carrying capacity than the other.
A given bracing type may fit into more than one category as described above. In
strength assessments carried out in a design process, the effective buckling length
of a compression flange can be determined being between two adjacent brace
points if the bracings satisfy both stiffness and strength criteria [1]. The minimum
stiffness value that is required for the bracings of a restrained member to buckle
between restraining points in a discrete bracing system, or to reach a given load
level on a perfectly straight beam in a continuous bracing system, is that the so-
called “ideal stiffness”, . Due to the presence of imperfections, however, a
bracing stiffness greater than the ideal value needs to be provided in order to be
able to rationalize the magnitude of the bracing forces and the deflections involved
[15].
As an example, one can consider here an imperfect column having a hinged
support at the one end and a lateral brace at its top, at which the column is
subjected to a compression point-load (see Fig. 4.1), being the actual stiffness
of the brace, ∆ the initial imperfection, ∆ the lateral deflection, and , /
being obtained on the basis of Winter's model. At the buckling point, , ,
the equilibrium changes from a stable state to an unstable situation, in which one
obtains the following:

∆
, ∆ ∆ ∆ →1 4.1
∆

→ ∆ /∆ 0 ⇒ ∆→ ∞

2 → ∆ /∆ 1 ⇒ ∆ ∆

Figure 4.1
Rigid-bar model concerning the buckling of a column having a lateral brace at its top.
32
Providing the theoretical ideal stiffness value, , / , for the
brace shown in Fig. (4.1), in order to reach a state of buckling between the hinged
support and the lateral brace point, results in infinitively large deflections
occurring at the brace point (see Eq. 4.1), infinite bracing forces thus developing.
If instead twice the ideal stiffness is provided for the brace, the lateral deflection
occurring at the brace point will be equal to the magnitude of the initial deflection
at the brace point, and the brace force being less than one percent of the
compression force in the case of a typical imperfection magnitude of ∆ /500.
Note that the bracing force is a linear function of the lateral deflection that occurs
at the bracing point, i.e. that ∆. For the reasons mentioned, providing at
least twice the ideal stiffness value is recommended [15] in order to appropriately
restrain imperfect members in compression.
Generally speaking, investigations of beam bracing are substantially more
complicated than those of column bracing, since flexural-torsional stability
assessments of restrained beams are computationally more difficult than those of
the flexural buckling of restrained columns. Consequently, current knowledge of
beam bracing is based primarily on numerical and experimental studies rather than
on closed-form solutions.
As explained earlier, the out-of-plane deformation of steel beams can be controlled
by means of lateral bracings applied directly to the compression flange and/or by
use of cross-bracings to resist twisting of the restrained beam. Some types of
bracings, such as a plan bracing or a bracing provided by concrete slab, act both as
lateral and as torsional restraints. Such bracings are more effective than bracing
types in which only lateral or torsional restraints are provided [22]. In the sections
that follow, certain information relevant to the lateral (Section 4.2) and the
torsional (Section 4.3) bracing of beams will be provided.

4.2 Lateral bracing of beams

The stiffness and strength requirements of the lateral bracing provided depends
upon its locations both across the depth of the cross-section and along the beam
span, as well as upon the load height, the number of braces, and the moment
gradient. The most effective position for lateral bracings across the depth of a
cross-section is that of the flange farthest from the center of twist. Lateral bracings
are also more effective when they are attached to the compression flanges (see Fig.
4.2), except for cantilevers, in which bracings are most effective near the tension
flange [74]. In the case of top flange loading of I-girders, the center of twist is
located near the centroid, bracing placed near the centroid thus being ineffective.

33
Figure 4.2
An illustration of load-height and bracing location effects on the critical moment values of laterally
braced beams, , being the critical moment value of the beam having a lateral brace at mid-span
in differenet configurations (1),…, (4), as shown, and , being the critical moment value of the
unbraced beam having a point load at the centroid of the cross-section [4].

For a beam in a reverse-curvature-bending condition along its span, lateral

bracings should be provided for both of the flanges at the inflection point, so as to
prevent twisting of the cross-section at this point [74]. Under such particular
circumstances, a lateral bracing of only the top flange would scarcely be effective.

4.3 Torsional bracing of beams

Torsional bracings are designed to resist the twisting of individual beams (i.e. a
lateral movement relative to each other of the top and the bottom flanges) rather
than to resist only lateral movements of a compression flange. When cross-
bracings are utilized in steel bridge application, this occurs through its limiting the
relative displacements of the top and the bottom flanges of adjacent girders. The
stiffness requirements of torsional bracings are less affected by the location of
loading and bracing across the depth of the cross-section [4], i.e. torsional bracings
of the top flange and of the bottom flange are about equally effective. However,
cross-sectional distortion can strongly affect the effectiveness of torsional
bracings. Accordingly, use of proper web-stiffeners at the site of each cross-
bracing is recommended.
Fig. (4.3) illustrates the effects on the critical load value of the location of
torsional bracings across the depth of a beam for different web-stiffener
arrangements [74]. The most effective location of torsional bracings, such as
cross-beams across the depth of a cross-section, in order for the distortion and the
size of the web-stiffeners involved to be minimized, is that they be located near the

34
mid-height of the cross-section [3]. Reverse curvature has no significant effect on
the torsional brace requirements here.

Figure 4.3
An illustration of the effects of cross-sectional distortion on the critical moment value of a
torsionally braced beam, , being the critical moment value of a beam with a torsional brace at
mid-span in differenet configurations (1),…, (3) as shown, and , being the critical moment value
of the unbraced beam having a point load at the centroid of the cross-section [4].

The overall flexibility of a cross-brace (see 1/ in Fig. 4.4) is affected by the

contributions of the following components [4]: i) the bending flexibility of the
web-stiffeners and of the web of the main girders, 1/ ; ii) the relative in-plane
bending flexibility of the adjacent girders, 1/ ; iii) the torsional flexibility of the
cross-bracings, 1/ ; and iv) the rotational flexibility of the connections between
the transversal beam and the web-stiffener, 1/ .

1 1 1 1 1
4.2

Figure 4.4
A model representing the flexibility of a cross-bracing.

The term 1/ in Eq. (4.2) accounts for a reduction in the overall stiffness of a
cross-bracing occurring due to in-plane relative deflections in the adjacent

35
restrained girders at the location of the cross-bracing in question. The rotational
flexibility of a brace connection can also have a detrimental effect on the overall
flexibility of a bracing system. In practice for the most situations, however, the
connections are assumed to be fairly moment-stiff. More information regarding the
terms used in Eq. (4.2) are given later in Section (6.1).
Substituting ̅ / into Eq. (4.3), the critical moment value of a torsionally
braced doubly symmetric beam can be calculated, where is the number of cross-
bracings between the end supports. Eq. (4.3) was first derived by Taylor and
Ojalvo [75] and was modified later by Yura [4] to account for cross-sectional
distortions, flexibility of the main girders, and asymmetry of the cross-section. For
beams with mono-symmetric sections, should be replaced by , , as
calculated by Eq. (4.4), where and are the distances of the tension and of the
compression flange centroids from the neutral axis of the cross-section, and
, and , are the lateral moment of inertia of the top and the bottom flanges,
respectively.

,
̅
, , , , , 4.3

, , ∙ , 4.4

In the case of laterally braced beams, as the stiffness of lateral bracings increases,
progressive changes in the number of half-sine waves of buckling mode between
the end-supports occur [1, 4, 15]. This commences with a global half-sine
buckling mode occurring between the end-supports and it ends up with a buckling
mode occurring between the brace points. Increasing the stiffness of the torsional
bracings of a beam, however, leads to a half-sine buckling mode appearing
between the end-supports, this dominating the buckling mode shapes of the
compression flange until the stiffness is sufficient to force buckling to occur
between the brace points. Yura [4] suggests that for design purposes, the buckling
mode of torsionally braced beams can be assumed to occur in a half-sine shape
until the bracing stiffness is adequate to force the beam to buckle between the
bracing points.

4.4 “Column-on-elastic-foundation model” for cross-

brace stiffness assessments in steel bridges
In completed state of half-through bridges (see section 6.3), the instability of the
compressive top flanges can be treated as buckling of a laterally restrained column
36
along its length. In addition, Eurocode 3 [76] permits design engineers to use a
simplified method based on a column-on-elastic-foundation model to assess the
effects of initial imperfections and of second-order deformations on the brace
forces generated in the bracings. In the analysis of cross-bracings, this simplified
method can only be used in order to model either the bottom flange of continuous
bridges or the top flange of half-through girders. In both of the systems mentioned
above, the other flange should be braced continuously by means of a concrete deck
or proper plane bracings. Otherwise, for the following reasons, this simplified
method would lead to inaccurate results:
The method models cross-bracings in terms of lateral springs. However, cross-
bracings resist the twist of the cross-section at bracing points, whereas the
translational springs in the model resist lateral displacements. Also, under “full”
bracing conditions, the model predicts there to be “zero” displacement at bracing
points, whereas in reality there are lateral movements at the brace points, since
girders restrained by cross-bracings can deflect laterally. In addition, under full
bracing conditions, the model assumes that the effective buckling length of the
compression flange is equal to the distance between the bracing points. However,
the results of the investigations reported in Paper IV showed this assumption to be
incorrect for bridges braced only by means of cross-bracings, if the effects of
imperfections are also included in the analysis. This will be explained briefly in
Chapter 7.
For a planar beam-column which is restrained laterally by means of continuous
elastic springs having the stiffness value per unit length of the column of ,
subjected to a compressive point load of and an arbitrary transversal load of ,
a general differential equation can be derived studying the equilibrium of a
deformed differential length of the column [45]. If the beam-column has a
prismatic cross-section and elastic material properties, so that , these yield
to:

4.5

Timoshenko and Gere [59] obtained the following solution for the critical load of
the system shown in Fig. (4.5):

,
4.6
, ,

Based on Eq. (4.6), the lower and the upper bounds of the critical loads and the
stiffness values, these corresponding to a particular number of half-sine buckling
waves, , occurring between the end-supports, can be obtained using Eqs. (4.7)-
(4.8).

37
 ,
1 1 4.7
,

1 1 4.8

For a brace stiffness of / 1 , the maximum critical load value

of , / , 1  which corresponds to a buckling mode for which
there are half-sine waves between the end-supports  can be achieved. After
some algebra here, the maximum critical load value can be simplified to the
expression in Eq. (4.9).

, , 2 4.9

Ignoring the first term in Eq. (4.9) and replacing , there by / , gives
the buckling length, , , of the system shown in Fig. (4.5) for the provided brace
stiffness of :
.
, ∙ 4.10
√2

Figs. (4.5)-(4.6) also illustrate the results of Eqs. (4.7)-(4.10):

Figure 4.5
The compression-flange-on-elastic-foundation model used for compression flanges of bridge girders
restrained by means of cross-bracings, where , / , and , are the critical loads
corresponding to a buckling mode having number of half-sine waves between the end-supports,
when the stiffness of is provided for the bracings [45].

38
Figure 4.6
Normalized values of critical loads versus the brace stiffness as obtained using Eqs. (4.6) and (4.9);
, is the critical load of the braced member shown in Fig. (4.5), and , is the critical load of the
unbraced member [45].

Eq. (4.9) can be also used as an approximate solution for a discrete number of
bracings. With only two braces for example (i.e. 2), Eq. (4.8) provides an
equivalent brace stiffness value of 2 / 2 3 / . Substituting
this value into Eq. (4.9) gives , / which is the critical load
corresponding to an effective buckling length equal to the distance between
adjacent brace points.

39
5 Lateral-torsional instability
concerns during construction of steel
bridges

The construction phase of composite bridges is a delicate operation [49], this stage
often controlling the design of both the steel girders and the bracings. Lateral-
torsional buckling is a failure mode that can occur during lifting, launching of the
girders, or casting of the concrete deck. Examples of some of the well-known
failures that have occurred during construction are: i) the collapse of The Marcy
Bridge in New York city in 2003, ii) the collapse of the Bridge Y1504 over the
Gide River in Sweden in 2002, and iii) the collapse of the State Route 69 Bridge
over the Tennessee River in 1996 (see section 2.2). In the last case, two of the
three spans collapsed into the river once one of the cross-braces was removed to
allow for repair of the brace connections. In this chapter, some of the common
lateral-torsional instability concerns in the design of typical composite bridges
during their construction stage are introduced.

5.1 Bracings required during concreting of the deck

During casting of the deck, the fresh concrete provides no stabilizing support for
the main girders. Proper bracings may be required to control the out-of-plane
deformation of the steel girders under the self-weight of the steel girders and the
fresh concrete; see Fig. (5.1). Tragic events have occurred during concreting the
deck of a number of bridges through the bracings either of the main girders or of
their timber falseworks being inadequate. As mentioned earlier in Chapter 2, the
failures of this type normally occur suddenly.
The concreting sequence along the span, and a symmetric concreting across the
bridge deck are also important concerns during the concreting of the deck. The
sequence of placing of concrete deck can affect the bracing layout. The main
objectives in assessment of the casting sequence are as following: i) minimizing
the installation costs of the bracing system, and ii) reducing the tensile stresses in
the already cast portions of the concrete deck, which can cause early-age cracking
problems.
41
Asymmetric concreting across the width of a bridge cross-section creates torsional
forces, which should be resisted by the cross-section and by the bracings. Such
torques can results in large bracing forces, e.g. in the plane bracing bars.

Figure 5.1
Out-of-plane deformation of steel bridge girders.

During concreting, vertical deflection and twisting of the bridge cross-section at

mid-span should be monitored carefully for each concreting increment. A
softening trend in the load-twist curve can be an indication that instability of the
steel girders can occur for further loading. Possible eccentricity of the self-weight
of the steel girders and/or of the fresh concrete generates torsion with respect to
the shear-center of an initially deformed cross-section (see Fig. 5.2). For instance,
assuming a half-sine wave geometric imperfection, the torque is largest at the end-
supports. Expanding Eq. (3.13) for the concreting condition gives [77]:

0 5.1

Figure 5.2
Torsional forces developed from the self-weight of the steel girder and from the fresh concrete
during the construction phase, where . . is the shear center, . . is the gravity center, and
are the self-weight of the concrete deck and of the steel girder; and and are the distance of the
self-weight loads from the shear center.

42
Here gives the percentage of the fresh concrete loading at which lateral-torsional
buckling can occur. Applying the boundary conditions and solving Eq. (5.1) yields
a value for , where the girder would presumably not fail during concreting
process if 1. Note that the effect of initial twist is not taken into account in the
method described above. In order to take into account the effects of an initial twist,
"∅" in the last term of Eq. (5.1) can be replaced by ∅ ∅ .

5.2 Bracings required for skewed bridges

In skewed bridges, the longitudinal axis of the superstructure, at least at one
intermediate cross-bracing or at one end support is not perpendicular to the major
axis of the substructure; see Fig. (5.3). The vertical reactions at the skewed support
and are not identical, due to the fact that . For this reason, a
torque will be generated at the skewed support, one that should be resisted by the
support bracing. In bridges with skewed supports, the discrepancy between the
vertical displacements, e.g. ∆ and ∆ in Fig. (5.3), of the main girders at a given
cross-section can create considerable torsion in the cross-bracings.

Figure 5.3
A plan view of a twin-girder bridge with a skewed end-support, where and are reactions at
skewed support points, ∆ and ∆ are the vertical deflections of girder 1 and girder 2 at their mid-
span, Ψ is the skew angle, and , and are the diagonal lengths as shown in the figure.

Such torsional forces in the bracings are also generated by traffic load and can
negatively affect the fatigue life of the brace connection. For the construction
stage, the magnitude of the skewed brace forces can be reasonably predicted by a
first-order analysis of the entire bridge, where these forces should be added to the
stability-bracing forces generated by second-order effects and geometric
imperfections. For such bridges, the stability bracing requirements given in the
literature concerned with straight bridges with no skewed supports, and ,
should be modified by a factor of Ψ and 1/ Ψ [35].
Bracing connections having typical web-stiffeners can be problematic in the
installation of skewed braces having a large skew angle. Quadrato et al. [78]
43
proposed the use of half-pipe profiles as web-stiffeners to ease the connection of a
cross-bracing which is not perpendicular to the main girders; see Fig. (5.4). In case
the proposed half-pipe/channel web-stiffeners are welded to the flanges, they can
also enhance the critical load of the main girders by creating nodal warping
restraints.

Figure 5.4
Half-pipe/channel profiles as web-stiffeners applicable at the location of skewed cross-bracings.

5.3 Bracings required for in-plane curved bridges

Providing the cross-bracings and the cross-section of main girders with an
adequate torsional stiffness is more crucial for in-plane curved bridges than for
straight bridges [79, 80]. In addition to the in-plane bending in such bridges, the
girders are also subjected to relatively large torques due to the radial components
of the normal stresses in both flanges, as shown in Fig. (5.5).
The cross-bracings of in-plane curved bridges should be oriented radially. When
the radius of curvature is less than a certain value, resisting the generated torques
often controls the size of the bracings [3]. The second-order effects of such torques
on the load-carrying capacity and on the bracing forces involved can be significant
depending upon the geometry of the bridge; they are of particular importance
during erection when the bracings are not yet all installed. Thus, calculations
based on a first-order analysis of the entire bridge may lead to inaccurate
prediction of the bracing forces generated by the in-plane curvature of the bridge
(which should be added to the stability-bracing forces obtained from the design
specifications for example).

44
Figure 5.5
Torsions generated from the in-plane curvature of a beam, where is the radius of curvature of the
beam, and is the axial force generated in the flanges by .

5.4 Bracing required prior to the concreting stage

A set of girders should be interconnected at several points along their span by
means of a minimum number of bracings prior to the concrete deck being cast, i.e.
under the self-weight of the steel components, when the bracings are not yet all set
in place. This occurs e.g. during lifting, launching, and transportation. These few
bracings are required in order to resist wind load, and to avoid lateral-torsional
buckling of the steel girders. The accidents that occurred during erection of The
State Route 69 Bridge in Tennessee in 1995, and of the 102nd Avenue bridge over
Groat Road in Canada in 2015 (see section 2.2) are examples of lateral-torsional
buckling under the self-weight of the steel portion of the bridge due to inadequate
bracings.

45
Field monitoring of steel bridges has also shown [81] that a proper erecting
sequence, erecting the girders in pairs, providing the steel girders with sufficient
lateral bracings, and providing temporary shoring towers can greatly reduce the
final overall deformations of steel girders which can otherwise lead to undesirable
bracing forces and stresses in the main girders.

46
6 Bracing options in steel bridges

Bracings are essential structural components in steel bridges, both during

construction and service. Some bracings may also be designed to enable the
transfer of loads (such as wind and eccentric live loads) between girders. Slight
bracings can significantly increase the load-carrying capacity of steel girders by
reducing their effective buckling length. Different types and arrangements of
bracings in steel bridges are possible. Utilizing only cross-bracings may not
completely prevent the steel girders from moving laterally. Plan bracings are often
more efficient than cross-bracings since they resist lateral movements of the
compression flanges directly. Yet, they generally conflict with formworks and/or
reinforcing bars, which make the concreting process rather difficult. As a result,
cross-bracings are often preferred in practice.

6.1 Intermediate cross-bracings

Fig. (6.1 a-c) shows the three most common types of the cross-bracings. The
cross-beam type (see Fig. 6.1 a) is preferred for shallow bridge girders, whereas
the cross-diaphragm type (see Fig. 6.1 b) is often used in bridges with trapezoidal
cross-sections. The K cross-frame configuration (see Fig. 6.1 c) is utilized for
relatively deep girders, and is preferred to the X or the Z cross-frame types in
common practice. In cross-diaphragms, large holes of at least 450 mm wide and
600 mm high [82] are needed for inspection and maintenance purposes. When the
top flange is restrained by means of a concrete deck or a plan bracing, the major
function of the cross-diaphragms is to prevent the distortion of a cross-section.
The cross-beams can be constructed by channel profiles due to the ease of the
attachments to the web-stiffeners. However, Mehri et al. [83] (i.e. Paper IV)
recommended considering use of symmetric cross-sections (with respect to their
weak axis) for cross-beams, since relatively large warping stresses can be
generated then in the transversal beam as a result of the eccentricity of brace shear
forces from the shear center of an asymmetric cross-section.
The cross-bracings of a twin I-girder bridge, for example, stabilize the top flange
of the main girders in the sagging regions and the bottom flange near the
intermediate supports. These bracings reduce considerably the effective buckling
length of the individual girders. Also, braced girders of suitable length can be
47
manufactured in pairs prior to transportation to the site, enabling the erection
process to be carried out quicker and more safely. Depending upon the geometry
of a bridge cross-section, cross-bracings are typically used at spacings of 4~8 m
along the span to resist relative lateral movement of the top and bottom flanges of
the individual girders. During service life, a few cross-bracings are required to
transfer possible lateral loads (such as caused by wind or by collisions) between
the main girders. Cross-bracings are also needed in order to control distortion of
the composite cross-section so as to resist torsions generated by eccentric traffic
loads, for example. In bridges of some types such as in-plane-curved bridges or
skewed bridges, the cross-bracings are needed to resist torsions created by the
geometry of the bridge. The maximally 7.5 m requirement for spacing of the cross-
bracings in straight bridges has been eliminated in AASHTO [82] so as to, if
applicable, reduce the number of cross-bracings, which can lead to a reduction in
the number of fatigue-prone attachments [3].

Figure 6.1
Typical cross-bracings in common use in steel bridges: a) cross-beams, b) cross-diaphragms, and c)
cross-frames.

If adequate cross-bracings are provided during construction, the two

interconnected girders need to twist as a single unit to resist the destabilizing
forces along the span. The resisting of such destabilizing forces at the site of each
cross-brace creates uplifting and overturning shear forces in the girders. The brace
shear forces increase locally the bending moment of the one girder and decrease
the bending moment of the adjacent girder. As a result, during the construction of
bridges with twin-I girders one of the girders may reach its critical load earlier
than the other.
In wide bridges, in which more than two main girders are used across the width of
the bridge, cross-bracings may also function in the global action of distributing
traffic loads between the main girders, the details of the brace connections thus
being prone to matters of fatigue. Under such circumstances, cross-bracings are
preferable for connecting the girders in pairs, this creating a number of twin-
girders across the width of the bridge. Lean-on bracings, consisting of parallel
struts at the level of the top and bottom flanges and having hinged connections
with the main girders, should also be used between the girder pairs. The struts are
needed to transfer lateral loads, including the destabilizing forces involved, and to

48
control the transversal spacing between the pairs. These bars should be removed
once the concrete deck has hardened.
In the cross-frames of K , X, or Z types, the contribution of trusses to the torsional
stiffness of the cross-bracing (see in Eq. 4.2) can be obtained by performing an
elastic first-order truss analysis employing horizontal unit forces or displacements
that act at the level of the top chords [4]. However, the contribution of the
transversal beam to the torsional stiffness of cross-beams depends on a situation of
either a single-curvature or a reverse-curvature bending of the transversal beam
(see Fig. 6.2), where is the in-plane moment of inertia of the transversal beam
and is the moment of inertia of a vertical web-stiffener and of a part of the web.
There is the possibility for twin-I girders having a relatively long transversal beam
that the girders will twist in opposite directions, creating a single-curvature-
bending situation in the transversal beam.

Figure 6.2
Single-curvature bending (system i) and reverse-curvature bending (system ii) of a cross-beam
placed near the bottom flanges of the main girders.

The torsional stiffness of the two systems just described is calculated here (see
Eqs. 6.1-6.2). For the single-curvature bending situation (system i):

. .
,
∆/ 0.5
3

49
 1
→ 6.1
, 3 2

For the double-curvature bending situation (system ii):

. .
,
∆/ 2 / 0.5
3 48 / 0.5 3

1
→ 6.2
, 3 12 6

The solutions provided in Eqs. (6.1)-(6.2) are comparable to the general

expression given earlier as Eq. (4.2). Note that is 2 / for a cross-beam in a
single-curvature bending situation, and is 6 / for a cross-beam in a reverse-
curvature bending situation. Also, as can be seen in Fig. (6.2), the brace shear
forces are zero in system (i), there thus being no relative vertical deflections
created in the two main girders. Accordingly, the second term in Eq. (4.2), i.e.
1/ , makes no contribution to the overall flexibility of the cross-beams of which
there is single-curvature bending in the transversal beam. For a situation involving
double-curvature bending, however, such deflections are created by the brace
shear forces at both ends of a cross-brace. This leads to a flexibility of 1/
1/ 12 / . Helwig et al. [26] showed that the effects of this contribution
are significant for twin-I girders for which the solution given for 1/ above can
be employed even when more than one cross-brace is used along the span. If 1/
dominates in Eq. (4.2), system buckling of the entire bridge is a possibility [3].
However, the contribution of web-stiffeners of partial-depth combined with the
contribution of a part of the web of the main girders, i.e. as expressed by the term
1/ in Eq. (4.2), are rather complicated. The distortional flexibility of a cross-
section having a full-depth web-stiffening, , or a partial-depth web-stiffening, ,
can be approximated by use of the following equations [4]:

3.3 1.5
∙ 6.3
12 12

3.3 1.5
∙ ∙ 6.4
12 12

Where and are the width and the thickness, respectively, of the web-
stiffening, is the web thickness of the main girders, and is the contact length
of the cross-bracing.

50
Eq. (6.5) gives the value of minimum transverse span corresponding to a point
from which the buckling mode of a two-dimensional frame shown in Fig. (6.2)
switches from system (ii) to system (i). The accuracy of the solution was verified
by use of a numerical program. However, since the lateral deflection, ∆, is not
constant along the bridge span, the expression is not applicable for determining the
buckling mode of the cross-bracings in a three-dimensional configuration.

1/ 1/

→ 6.5
4

In steel-bridges braced by means of discrete cross-bracings (without this being

combined with any lateral bracing), the buckling length of the compression
flanges, , , will be either:
- Equal to the distance between the intermediate cross-bracings as governed
by Eq. (6.6) [59]. This can occur when a stiffness value greater than that
required for “full/ideal bracing” (i.e. a minimum bracing stiffness value
enabling a brace to function similar to an immovable support) is used for
the cross-bracings. This means that the girders will have to buckle between
the brace points if the cross-bracings are to provide sufficient lateral support
for the compression flanges [1]. It should be noted that Eq. (6.6) is basically
used for predicting the buckling capacity of a simply supported girder
without any intermediate bracing, as explained in Section (3.4).
- Larger than the distance between the intermediate cross-bracings as
governed by Eq. (6.7). This can occur when a lower stiffness value than that
of the full bracing stiffness is provided for the cross-bracings [4, 61].
- Equal to the length of the span, due to system buckling of the entire
structure as a single unit (see Fig. 6.3), which can be estimated by Eq. (6.8)
[37]. This equation was derived on the basis of equation (6.6),
substituting 2 , 2 , and 2 /4 2 /4 for , , and ,
respectively. The resulting expression is applicable to the twin-I girders that
are interconnected by means of a number of cross-beams that are adequately
stiff, and should not be used for laterally braced girders. System buckling
can occur, for example, in bridges having a relatively large span-to-width
ratio. This approximate solution can be also used for girders having
trapezoidal cross-sections [37, 49].

, , , , , ∙ 6.6

51
 ,
̅
, , , , , , 6.7

, ∙ , 6.8

Neglecting the first term of Eq. (6.7) and substituting , , , , , ,

enables a minimum value for the brace stiffness, ̅ / , to be calculated,
this providing a cross-brace function similar to that of an immovable support. An
increase in the brace stiffness greater than the ideal stiffness value would not
enhance the critical moment of girders subject to a system buckling problem.
However, to account for the effects of imperfections, twice the ideal stiffness is to
be recommended as explained in Section (4.1).

Figure (6.3)
System buckling of steel bridge girders; where is the twist of the bridge cross-section at mid-span.

Fig. (6.4) shows a ladder-deck bridge type consisting of two longitudinal main
girders, and transversal cross-girders spaced along the length of the main girders.
In such bridges, both the main girders and the cross-girders are provided with
shear connectors to create a composite action together with the concrete deck. The
connection between the cross-girder and the web stiffeners should be moment-
stiff. For relatively wide bridges, the cross-girder may also require bracings during
the construction stage. In the sagging moment regions, the cross-girder normally
has a constant depth along its span, the top flange restraints being produced by an
inverted U-frame action of the cross-girder and the web stiffeners; see Fig. (6.4 i).
Either near the internal supports or at the support points, three alternatives for
restraining the bottom flanges are possible:

52
 - Through an inverted U-frame action of cross-girders having a deeper
section and of web stiffeners for relatively shallow bridges. This is more
economical than other options are (see Fig. 6.4 ii),
- Hunching the cross-girders near the main girders; see Fig. (6.4 iii),
- Using knee-bracing near the main girder (see Fig. 6.4 iv).

Figure 6.4
Ladder bracing of steel bridges, using system (i) for sagging regions, and systems (ii)-(iv) for
hogging regions or at support points.

For long cantilevers, the cantilever girders having moment continuity with the
cross-girder can be provided for supporting the slab. However, this adds
significantly to the cost. In wide bridges, cross-girders having closer spacing along
the bridge span, e.g. with a spacing of less than 3.5 m between the adjacent cross-
girders, would provide a possibility for using precast decks or corrugated metal
sheets as stay-in-place formworks spanned between the cross-girders.

6.2 Support bracings

Support bracings are necessary to transfer lateral loads from the superstructure to
the substructure. They are also needed to provide the desired boundary conditions
for individual girders and for the bridge as a whole, during construction and while
it is in service. The support bracings also resist and transmit loads at support
points generated through lateral-torsional deformation of the main girders along
the span, as well as loads created by possible eccentricities of the bearing
reactions, possible tilts of the main girders at support points, and from skewed
53
supports. The non-composite stage is normally more critical for the support
bracing than the completed situation is, since the concrete deck also contributes to
resisting torsion at the support points once it has hardened.

6.3 Bracing of half-through girders

In half-through girders, a relatively stiff deck is placed near the bottom flanges, in
which the deck consist of either a concrete slab, or a number of closely spaced
transversal beams having plan bracings; see Fig. (6.5). In such systems, the
compression top flanges can be restrained laterally by the U-frame action of the
laterally stiff deck and the stiffened webs [84]. The connection between the
concrete deck or transversal beams and the main girders should be moment stiff. A
similar bracing system is present for the compression bottom flange in the hogging
regions in the completed condition (i.e. when the concrete deck has hardened). In
both cases that are described above, the lateral bending stiffness of the stiff deck
should be adequate to resist possible lateral deflections of the tension flanges
properly.
In half-through girders having transversal beams and plan bracings, the cross-
beams are often tightened to the patch plates at each end  which are already
welded to the web of the main girders  through use of ordinary preloaded bolts.
In the completed state, the cross beams act compositely with the concrete slab,
which may either be cast over the cross-beams, or encase the cross-beams (partly
or entirely).
Each U-frame is created by the main girders serving as the vertical components,
and a cross-beam or a unit length of the concrete deck as the horizontal
component. The flexibility of the U-frames can be calculated through applying
unit lateral loads at the centroids of the top flanges. The bridges of this type are not
common in highway applications due to the risk of collision. However, a number
of railway bridges are built in the form of half-through girders. At the ultimate-
limit-state, the contribution of the deck to the resistance of the main girders is
usually neglected [84], the main consideration in this design state being the
stability of the top flanges. The buckling length of the top compression flanges can
be predicted approximately by use of the column-on-elastic-foundation model as
explained earlier in Section (4.4). This is illustrated in Fig. (6.5), where, the
stiffness of each U-frame, , can be obtained by a process similar to that
performed in the derivation of Eq. (6.1).

54
Figure 6.5
A schematic illustration of the lateral-torsional buckling of half-through girders.

6.4 Full-span plan bracings

Forming a truss consisting of diagonals and struts, plan bracings reduce the
effective buckling length of the compression flanges by controlling their relative
lateral movements at the ends of each diagonal member; see Fig. (6.6).

Figure 6.6
Two examples of full-span plan bracings, , being the effective buckling length of the compression
flanges of the girders G1-G4.

The unbraced length of a compression flange can be assumed to be the distance

between the struts when the plan bracing is adequately stiff [44]. Most lateral

55
bracings placed at the level of the compression flanges can also provide certain
torsional restraint, the extent of which depending upon their distance from the
shear center. However, this capacity is ignored in most design specifications. The
lateral stiffness of plan-braced steel girders can be assessed through applying
lateral unit loads at the locations of each brace in a two-dimensional structural
model including also the compression flanges.
The three most common types of plan bracings are shown in Fig. (6.7). If
diagonals of a brace type are not designed for tension-only conditions, the
diagonal compression bars should possess sufficient flexural stiffness in two
perpendicular directions. The point of intersection in an X-type brace panel can be
regarded as a brace point for the diagonal compression bar when the other
diagonal bar acts in tension. Note that the unbraced length of these compression
diagonals would be larger than half-length when both diagonals, as parts of the
bridge cross-section, contribute to in-plane flexure.

Figure 6.7
The most common types of plan bracings (plan view).

To assess the stability requirements of plane bracings on the basis of Winter's

approach [1, 3], the compression flanges of the steel girders can be modeled as a
column-on-elastic-supports; see Section (4.4). At each brace bar, these stability
brace forces should be added algebraically to the forces created in the bars by the
torque present there. The torque can be generated by the load eccentricities, the
geometry of the curved bridges (see Section 6.7.1), the web slopes (see Section
6.7.2), and the lateral loads present along the span. The bracing forces generated
by these torsions are largest at the location of maximum torque (i.e. normally near
the supports). Generally speaking, in contrast to straight bridges, the bracing
forces generated by such torsions are dominant in in-plane curved bridges as
compared with the stability bracing forces generated by second-order effects and
geometric imperfections.
Plan bracings are best attached directly to the flanges by bolted connections in
order to reduce the risk of fatigue associated with the use of a gusset-plate
connection to the web. The braces can lead to distortion-induced-fatigue-cracks if
they are attached to the web by means of gusset plates. Plan bracings that are cast
within the deck can conflict with the reinforcing bars of the concrete deck.
However, neither maintenance nor buckling evaluation is required for such
bracings. Plan bracings can be placed at the level of the bottom flanges in I-shaped
56
girders, creating a pseudo-box cross-section with either a concrete deck during the
service period or top flange plan bracings during the construction stage. In
trapezoidal girders, plan bracings placed at the level of the top flanges during
construction also create a pseudo-box cross-section having considerably enhanced
torsional stiffness.
Plan bracings placed near the top flanges but below the formworks can contribute
to resisting in-plane bending of the girders during construction. The contribution
of the top flange bracings to the in-plane flexural stiffness of the main girders is
normally neglected in design. Those bracings are subjected to forces due to strain
compatibility in in-plane bending [38]. Under such circumstances, the risk of
fatigue of the connections should be also considered, or the bracings should be
removed once the concrete deck has hardened. The contributions of the top flange
bracings to the in-plane flexural stiffness of the main girders depend upon the
layout of the trusses, and they can be significant in X-type bracings in particular.
The strut forces in X-type bracings developed from the in-plane flexure of the
main girders can be relatively large, since both diagonals in the adjacent brace
panels are in compression. In contrast, the strut forces of such bracings created by
the torsion involved are relatively slight since one diagonal is in tension and the
other is in compression. Clearly, a combination of the torque and the in-plane
flexure increases the axial force generated in the one diagonal bar in an X-type
panel and decreases the axial force in the other diagonal bar.

6.5 Partial-span plan bracings (Papers I, and V)

In the global lateral-torsional failure mode (see Fig. 6.3) the entire bridge rotates
as a single unit about the shear center of the bridge cross-section between the end-
supports. This type of failure is less sensitive to the number and the size of the
cross-bracings. In open-box girders and in closely spaced twin I-girders without
plan bracings along the span, the St. Venant torsional stiffness is relatively low.
Also, the low level of warping (bending of the flanges in the out-of-plane
direction) stiffness of such cross-sections readily permits considerable twist, which
can lead to an unacceptable degree of rotation at mid-span.
As a solution, the load-carrying capacity of such bridges can be improved by end-
warping restraints being provided near the support points. Eq. (3.9) indicates that
both an end-warping and end-twist restraint one which reduces the corresponding
effective buckling length factor of and , respectively can significantly
increase the critical moment value of a beam. For instance, a half-pipe or channel
web-stiffener (see Fig. 5.4) that is properly welded to both the top and the bottom
flanges of a single I-shape girder, can provide warping restraints to the flanges to
which they are attached. This decreases in Eq. (3.9), resulting in an increase in
the critical load. Similarly, in bridge girders potentially prone to system buckling,
57
plan bracings at support regions (and perhaps at mid-span to prevent further
buckling modes) restrain the lateral rotation of the restrained flanges, this
decreasing the value of in Eq. (3.9). Such modifications increase the warping
stiffness of the cross-section, creating a semi-fixed-end condition, resulting in an
increase in the critical moment value. The corresponding critical moment under
fixed-end conditions can be calculated through replacing by 0.5 0.6 in Eq.
(6.8). Investigations carried out in Paper I [49] concerning the failure of The
Marcy Bridge showed that slight truss-bracings near the supports could prevent
the collapse of the bridge. End-warping restraints of truss-bracings and corrugated
metal sheets were studied experimentally, the results being described in detail in
Paper V [85], which is appended to this thesis.

6.6 Bracings required in open trapezoidal girders

Trapezoidal girders require high fabrication costs due to inclination of the webs. In
a completed state, bridges of this type have particular advantages when compared
with an equivalent twin-I girder bridge, for the following reasons:
i) the bridge possesses greater torsional stiffness, ii) a greater St. Venant stiffness
minimizes the self-weight of the steel girder, and iii) the bridge has greater
durability due to its having fewer edges and no external stiffeners that are exposed
to dirt and to moisture.
However, before the composite action occurs, the open cross-sections are weak
torsionally, and can be susceptible to lateral-torsional buckling of the individual
webs between the cross-bracings, or to system buckling of the entire bridge
between the supports. Increasing the slope of the web and keeping the width of the
bottom flange constant increases the lateral stiffness of a trapezoidal cross-section,
while reducing the critical moment value [86]. The possibility of lateral-torsional
buckling of the trapezoidal cross-sections when they are bent about their weak axis
(i.e. in-plane bending) has been discussed by Attard [87]. The Marcy Bridge
having a trapezoidal cross-section and a lateral-to-in-plane flexural stiffness ratio,
/ , of 1.75, failed due to system buckling [49, 86].

Comparatively large torsional stiffness values for trapezoidal girders can be

achieved converting their cross-section to a pseudo-closed form by use of steel
plates, plan bracings, or corrugated metal sheets attached properly to their top
flanges. In the closed-section configuration, the torsional stiffness is dominated by
the St. Venant component (where 4 / ∑ / whereas in an open cross-
section ∑ /3). Although in the closed-form the warping stiffness is
negligible when compared to the St. Venant stiffness, large warping stresses can
develop due to the cross-sectional distortion that can be present when the bridge
lacks adequate cross-bracings along the span (see Section 6.6.3).
58
Note that, in the completed situation, longitudinal stiffeners are normally needed
in the compressive zones of the inclined webs and of the bottom flange in order to
avoid local buckling.

6.6.1 The equivalent plate concept and forces generated in plan-

bracings by torsion

Kollbrunner and Basler [2] derived expressions enabling typical truss bracings to
be converted to an equivalent plate having a constant thickness of ; see
Appendix IV. The equivalent plate is used to calculate the mechanical properties
of such cross-sections in order to assure that the torsional stiffness and strength are
sufficient to keep rotations and stresses within a magnitude of reasonable size. The
equivalent plate concept is also used in calculating the cross-sectional properties of
trapezoidal girders on the top of which corrugated metal sheets are attached (see
Section 2.2.1). However, the stress concentration factor needs to be taken into
account there, due to large changes that often occur in the thickness of the plates.
Note that relatively little research has been done to assure the accuracy of the
equivalent plate method in such applications. The St. Venant shear flow in the
equivalent plate (see Eq. 6.9) creates axial forces in the brace bars (see Fig. 6.8).

6.9
2

Where is the shear flow, is the enclosed area as defined by the wall-midline
in a thin-walled closed section, and the torque at the middle of each panel, , is
generated by the geometry of curved bridges or by eccentric gravity load across
the width of a bridge cross-section, for example.

Figure 6.8
Forces that develop in plan bracing bars, and , due to the torsion, there, being the shear
flow and being the enclosed area as defined by the wall-midline in a thin-walled closed section.

59
Wind can also produce significant torque in trapezoidal girders during
construction, since the shear center is far below the bottom flange. In such cases,
the brace forces are larger in the region of maximum torque. In X-type (if not
designed for a tension-only case) and Warren-type plan bracings, one diagonal
would be in tension and the other in compression. In plan bracings of a Pratt-type
in an in-plane curved bridge, the diagonals should be oriented accordingly so that
they are working in tension. The relative lateral displacement of the Pratt-type
brace panels that accumulates, leads to greater lateral deflections occurring at mid-
span than those of Warren- and X-type brace panels. This reduces the lateral
stiffness of such bridges having Pratt-type plan bracings and the efficiency of their
bracings [36]. Also as discussed earlier in Section (6.4), the strut forces that the
torsion results in are considerably greater in Pratt-type bracings than those of
equivalent X- and Warren-types.

6.6.2 Forces generated from distortion in the cross-bracings

Fig. (6.9) presents an example of a trapezoidal cross-section subjected to an

eccentric gravity load that creates both in-plane bending and torsion. The torque
can lead to distortion in a closed cross-section, since shear stresses are not
generated uniformly across the cross-section according to the St. Venant shear
flow.

Figure 6.9
Elevation views illustrating the distortion of a closed cross-section caused by torsional forces during
construction.

60
Proper cross-bracings with a reasonable spacing are required to resist distortion of
the cross-section. The bracing forces in each cross-bracing can be calculated by
studying an equivalent system that separates pure torsion from pure distortion in
the cross-section [79]. Such cross-bracings should limit transversal bending
stresses of distortion of box sections well below the in-plane bending stresses at
the strength limit state [82].
Trapezoidal girders can possess a large degree of the St. Venant torsional stiffness
when their top flanges are provided with proper plan bracings or metal sheets. The
torsional stresses shown in Fig. (6.9) can be resisted by the St. Venant stiffness of
the closed section. An appropriate set of cross-bracings is needed, however, to
avoid additional stresses in the cross-section due to distortion. These additional
stresses (see their distribution in Fig. 6.9) depend upon the magnitude of the
torque and the size of cross-section.

6.6.3 Forces in the plan bracings due to the web inclinations of

trapezoidal girders

A static equilibrium of each top flange of a trapezoidal girder shows that the self-
weight of fresh concrete, , develops a lateral force at each flange as a result of
sloping the web of trapezoidal girders. Such a uniformly distributed force,
0.5 , leads to lateral bending of the two top flanges in opposing directions,
this generating tensile forces in the struts of the plan bracing employed; where is
the web inclination as measured from the vertical axis.

6.6.4 Forces generated in intermediate external cross-bracings from

torsion

External cross-bracings are used to control the relative twist of two adjacent
trapezoidal girders near the mid-span. Such twist can result in an uneven thickness
of the concrete deck involved; see ∆ in Fig. (6.10). The variation of this sort can be
relatively large in curved bridges and in bridges having skewed supports. In
curved bridges, the exterior girder is subjected to greater twist than the interior
girder is, due to the differences in span length. Similarly, different twists of the
adjacent girders having skewed supports lead to a relative rotation of them at mid-
span. A few external cross-bracings near the mid-span can effectively control such
relative twists and vertical deflections of the adjacent girders [3]. Such a function
is required mainly during concreting, thus, the external cross-bracings are often
removed when the concrete deck has hardened. Removal of the external bracings
eliminates the risk of fatigue due to possible contribution of those bracings to the
transfer of traffic loads between the main girders.

61
Figure 6.10
The functio of intermediate external bracing in eliminating the relative displacement, ∆, of the
adjucent girders.

The bracing forces and adequacy of the stiffness of such bracings can be obtained
either by performing a three-dimensional numerical analysis or by assuming a
reasonable value, e.g. 10 mm, for the relative deflection of the two girders so as to
control constructability. Kim and Yoo [88] in considering two different casting
sequences, performed numerical investigations of the performance of external
bracings in twin-trapezoidal curved bridges having either single-span or three-span
features. For the cases that were studied, it was found that the addition of external
bracings having a certain stiffness value beyond one at mid-span had little impact
in terms of controlling the relative displacement of the adjacent girders.

6.6.5 Support cross-diaphragms

Support diaphragms provide twist-restrained boundary conditions at the support

points. In addition, during construction, they transfer lateral loads (such as wind
load) ultimately to the ground. Since the function of the support diaphragms is
governed mainly by their shear action, they need to possess sufficient shear
stiffness / and strength ( / ; see Fig. (6.11), where is the
vertical reaction of the support bearings; and are the depth and the thickness
of the shear diaphragm; and are shear strain and shear stress present in the
diaphragm; and is the shear modulus.

Figure 6.11
The shear flexibility of the support cross-diaphragms; where and are the torsion of the two
adjacent girders at a supprt point, and is the vertical reaction of the support bearings.

62
The shear deformation of the support diaphragms results in a rotation of the entire
bridge. Such flexibility, affecting value in Eq. (3.9), reduces critical moment
value of the bridge against system buckling and can lead to the thickness of the
concrete deck being also uneven. Thus, this flexibility should be kept as slight as
possible. Any reliance on post-buckling resistance of shear diaphragms (including
internal, external, and support types) is not advisable [82].

6.7 Bracing potential of stay-in-place corrugated metal

sheets (Paper V)
Corrugated metal sheets have frequently been used in the U.S. in bridge
applications having trapezoidal or multi-I girder cross-sections to support the
concreting loads. Such members, the ribs of which are oriented perpendicular to
the girders, mainly provide warping restraints for the top flanges rather than lateral
or torsional restraints [3]. The performance of corrugated sheets serving as a
stabilizing system is affected by the following: the shear stiffness and strength of
the sheets, the tearing resistance of the sheets where the edge fasteners are located,
the shear resistance of the edge fasteners (see Fig. 6.12), and the global buckling
of the sheets. Note that second-order effects should be also considered in the
strength calculations carried out.

Figure 6.12
The cross-section of a bridge girder having corrugated metal sheets, a common practice in the U.S..

In multi-I girder bridges (commonly used in the U.S.), the presence of cambers
and having different thicknesses of the top flange of the adjacent girders are
practical challenges in the installation of metal sheets. As a solution to achieving a
uniform concrete deck across the bridges, corrugated metal sheets are commonly
fastened to the angle profiles already welded to the main girders; see the edge
angle profile in Fig. (6.12). The stabilizing performance of such systems is limited
to a considerable extent by the flexibility of their connections [42]. Currently, the
stabilizing potential of such systems should not considered in bridge design,
according to the AASHTO code specifications [82]. However, research performed

63
at University of Texas on the stabilizing potential of metal sheets provided with
improved connections [41, 43, 89] showed that if metal sheets are properly
designed and connected to the main girders, this can enhance considerably the
load-carrying capacity of the steel girders.
In Sweden, the stay-in-place corrugated metal sheets 0.85 mm in thickness were
designated to also act as a stabilizing system for the trapezoidal girder of Bridge
Y1504, which collapsed in 2003. The sheets were attached directly to the top
flanges by nail fasteners, so that in this case, no concerns arose regarding the
flexibility of the edge angle profiles. However, the bracing performance of the
metal sheets employed on Bridge Y1504 was questioned [9] due to the inadequacy
of edge fasteners near the supports; see Section (2.2.1).
In Norway, corrugated sheets with thicker plates 3~5 welded to the top
flanges of the trapezoidal girders have frequently been used as stay-in-place
formworks during construction. Through this eliminating the slip of the
attachments and reducing the warping of the panels having end-closed ribs, a
considerably larger rotational stiffness can be achieved than with use of thinner
sheets and nail or screw attachments.
The bracing performance of metal sheets is strongly affected by the shear rigidity
of the diaphragm, kN/Rad , the value of which depends upon the effective
shear modulus kN/m/Rad and the tributary width per each girder of the
sheets [33]. For design purposes, the effective shear modulus of the sheets (which
depends upon the flexibility of the attachments and of the corrugated sheet) can be
calculated by use of a series of equations such as provided by for example [51] or
[52]. As an alternative, the effective shear modulus can be determined by
performing a laboratory test on a particular case; the effective shear stress value
divided by the shear strain value then giving the effective shear modulus.
Errera and Apparao [90] presented an energy-based solution for determining the
critical moment value of a beam braced by means of shear diaphragms placed on
the top flanges of beams that are subjected to a uniform bending moment.
Lawason et al. (1985) presented the following solution for obtaining the buckling
capacity of a shear-diaphragm-braced beam under transversal loading conditions:

, , 6.10

Where is the shear rigidity of a given metal sheet, is the effective

shear modulus of the metal sheet, is the tributary width per girder of the sheet,
and is the distance of the shear diaphragm from the shear center of the cross-
section. Errera and Apparao [90] also suggested a simplified approximation, as
given in Eq. (6.11), that showed “excellent” agreement with the energy-based
solution given in Eq. (6.10). The first term of Eq. (6.11) is the contribution of the
64
main girders, the last term being the contribution of the shear diaphragms to the
buckling capacity of the braced system.

, , , , 2 6.11

For a monosymmetric cross-section having a span-to-depth ratio of 20, Helwig

and Frank [33] compared the results of FEM with Eq. (6.11) for a variety of shear
rigidities, their defining the value of as being the distance of the shear diaphragm
from the midheight of the girder's cross-section rather than being the distance of it
from the shear center or from the centroid. By substituting for 2 in Eq. (6.11)
and applying the moment gradient factor, ∗ , to the first term, Helwig and Frank
[33] presented a conservative solution (see Eq. 6.12) for predicting the buckling
capacity of a girder restrained by means of a shear diaphragm, under different
loading conditions. The coefficient in Eq. (6.12) represents the efficiency of the
shear diaphragm as a continuous bracing.
∗
, , , , 6.12

As an alternative to Eq. (6.12), the critical moment value of a closed-trapezoidal

cross-section can be calculated by converting the corrugated sheets to an
equivalent plate. Eq. (6.13) gives the equivalent plate thickness value, , for a
specific corrugated sheet type and its attachments. The equivalent thickness is
obtained on the basis of the shear flexibility of a metal sheet, , ,  which
takes account of the followings: slips at the fasteners, warping and shear
deformations of the sheet  and the shear flexibility of an equivalent flat plate,
. The latter can be calculated based on studying the shear deformation of a
cantilever plate (see Eq. 6.14), where is the depth of each plate, and is the
Poisson's ratio.

6.13
∑ ,

2 1 /
6.14
/ ∙

A bending moment gradient (such as one caused by a transversal load) rather than
a uniform bending moment reduces the brace efficiency of the shear diaphragm.
The shear diaphragm is less efficient for cases in which the distance between the
plane of the diaphragm and the center of the twist is small. Note that the distance
in question can vary along the span for most of the transversally loaded girders
[33]. In the experimental work [41], very little displacement of the top flange at
mid-span but significant lateral movement of the bottom flange was observed
before inelastic deformation of the metal sheets occurred. For the cases studied

65
there, this shows clearly that the twist center of the braced cross-section was close
to the top flange at midspan. As a result, in utilizing corrugated sheets, the bottom
flanges may need to be stiffened by means of plan bracings, or their size should be
increased. In the sub-sections that follow, the stiffness, the strength, and the
connection requirements of corrugated sheets in bridge applications are presented
briefly as the results of the research carried out at the University of Texas [31].

6.7.1. Stiffness requirements of the metal sheets

The ideal stiffness of corrugated metal sheets in the case of perfect beams can be
calculated using Eq. (6.12). However, in the study mentioned above [31] it was
recommended that one should provide an effective shear stiffness four times the
ideal value so as to minimize shear deformations of the sheets and the forces in the
fasteners. Taking account of / , this yields the following:

∗
4 , , , ,
′ 6.15

Note that the shear strains along the edges of a brace panel are consistent with the
lateral deformation of the top flanges of the main girders, which are not evenly
distributed along the bridge span. The shear diaphragms attached to the top flanges
are only effective in the positive moment regions. In the negative regions, since
the top flange is in tension and is warped to only a relatively small degree, the
shear diaphragm is ineffective. Under such circumstances, a plan bracing at the
level of the bottom flange can be provided so as to limit the warping of them.

6.7.2. Strength requirements of the metal sheets

A shear diaphragm with ribs oriented perpendicular to the longitudinal axis of the
main girders can be modeled rather well as a plan bracing; see Fig. (6.13). This
enables the stability forces that develop to be analyzed.

Figure (6.13)
The performance of corrugated metal sheets attached to the girders top flanges (top view).
66
The bracing forces there depend upon the magnitude of the initial imperfections
(larger forces being created in the case of larger initial imperfections), the actual
brace stiffness (smaller forces being created by stiffer sheets), the web flexibility,
and the loading conditions. The magnitude of the bracing moment per unit length
of the bridge span (see . / in Fig. 6.14) generated in a shear
diaphragm attached to the top flanges follows the lateral slope of the flanges;
where / ; /2; and are the geometries of each
brace panel as shown in Fig. (6.14); and is the shear force that each panel is
exposed to at its edges. For a simply supported beam under uniform bending
moment conditions, the maximum bracing moment occurs near the supports and
gradually decreases along the span. Whereas under a uniformly distributed load or
a pointed load at midspan, both of them applied at the level of the neutral axis, the
maximum force occurs at a quarter to a third span of the beam [32]. Clearly, the
top flange loading increases the bracing moment and the magnitude of the
maximum twist as compared to loading at the level of the neutral axis.
Assuming a stiffness value of 4 being provided for the corrugated
sheets, Helwig and Yura [31] recommended use of the following bracing moment
value for design purposes:

0.001
6.16

Note that Eq. (6.16) should be modified when loading other than a uniformly
distributed load being applied at the level of the top flanges is present, and if the
shear stiffness of less than 4 is provided.
The center of twist for a braced I-girder with a top flange loading is near to the top
flanges, much of the cross-sectional twist occurring due to the lateral deformation
of the bottom flange. The recommendations above are given for I-girders with no
cross-bracings between the supports. A combination of corrugated sheets with
cross-bracings along the span can make the metal sheets more effective. Utilizing
cross-bracing generally reduces the bracing forces and the required stiffness of the
shear diaphragms [32].

6.7.3. Connection requirements

The shear forces created in the fasteners, and , caused by the brace shear
forces, , can be determined by studying the equilibrium of a given brace panel
[31]. For the case shown in Fig. (6.14), for example, in which there exist five edge
fasteners for each corrugated sheet, the resultant shear force created in the most
loaded fastener is:

67
 2
6.17
5 1.25

Figure 6.14
Shear forces of the edge fasteners of metal sheets acting as a brace on the top flanges of steel girders

Note that a thicker sheet results in low level of ductility in the panel, and thus, in
lower fastener forces.

6.7.4. Use of corrugated metal sheets in Twin-I girder bridges

The stabilizing potential of corrugated metal sheets can be of benefit in multi-

girder bridges having closely spaced longitudinal girders, or in case of single
trapezoidal girder bridges. In contrast to the multi-I girders commonly used in the
U.S., the twin-I girders having a wider transverse span, , are the most common
bridge type in Sweden. In such systems, the large transversal distance between the
main girders requires deeper sheets of sufficient in-plane bending stiffness as
compared with the bridge systems mentioned above. The deeper sheets increase
the thickness of the concrete deck, and thus make the use of metal sheets in Twin-I
girders less attractive in design.

6.8 Scaffolding bracing of steel girders (Paper III)

The recent accidents in the Scandinavian countries of bridges with timber-
falsework showed the methods currently used here to provide support to fresh
concrete suffer from certain shortcomings. A literature study of a large number of
failures of bridges around the world with timber-falseworks revealed a lack of
lateral stiffness and of sufficient bracing to be the major cause of such events; see
section (2.3). Although the problem of instability in the conventional timber

68
falseworks could be easily overcome, relatively elementary methods are being
widely used at this stage of bridge construction. Since a considerable portion of
the time and the budget involved in bridge construction is now allocated to the
installation of such timber falseworks, an alternative for them is of interest. In
Section (6.8.1), the common design practice for providing stability to steel girders
during the construction stage is described, with emphases being placed on the
shortcomings involved. In Section (6.8.2), a typical scaffolding system, the so-
called “Cup-lock scaffolding”, widely used in Sweden is described. In contrast to
timber-falseworks, these scaffoldings are much less vulnerable to problems of
instability. Also, with some minor improvements, these scaffoldings can be
considered to represent a reliable bracing source for bare steel girders, thanks to
their high degree of torsional stiffness. The basic theory involved is dealt with in
Section (6.8.3) and in Paper (III). Additional investigations were also carried out
during the author's Ph.D. study concerning the “lateral” stabilizing potential of the
scaffolding ledgers, the results of which need to be extended; see Sections (6.8.4)-
(6.8.5).

6.8.1 The common practice in design aimed at providing stability for

the steel girders during construction stage

To an extent depending upon the lateral-torsional slenderness ratio of the steel

bridge, , out-of-plane deformations reduce the load-carrying capacity of the
system; see region “C” in Fig. (6.15). During the construction of composite
bridges, under the self-weight of fresh concrete or while different boundary
conditions are present during launching, for example, proper bracings are normally
required so as to limit out-of-plane deformations of the girders. Since slight
bracings can effectively control the out-of-plane deformations, they can
significantly decrease the lateral-torsional slenderness ratio, this resulting in a
considerable increase in the load-carrying capacity. Cross-bracings are the most
common bracing options in steel bridges. Such bracings are conventionally used in
the form of cross-frames/-beams/-diaphragms, normally spaced every 4~8
along the bridge span. Cross-bracings are relatively expensive in terms of
fabrication, erection (fitting problems), and maintenance. Also, cross-bracings are
more efficient when they are placed at the locations of maximum twist and
maximum bending moment. In practice, however, these bracings tend to be spaced
uniformly along the bridge span.
In addition, such permanent bracings can lead to fatigue-sensitive details.
Therefore, reducing the number of cross-bracings, if applicable, and instead
utilizing alternative temporary bracing is of interest. Most bracings are mainly
required during the construction period, and they are normally less important when
the concrete deck has hardened.

69
On the other hand, the location of a critical brace may vary in different
construction stages. Cross-bracings can only resist the twist of an individual girder
at the bracing points, and should be combined with proper plan bracings so as to
control lateral movements both of individual girders and of the entire bridge.
Provisions regarding lateral stiffness are only required during the construction
stages, the concrete deck being able to provide substantial lateral stiffness once it
has hardened. In a completed state, when the concrete deck has hardened, the plan
bracings that are not cast within the concrete deck can attract various forces from
traffic loads. Most plan bracings are relatively soft and could buckle under the
compression forces involved, through the contribution these make to the in-plane
bending of the bridge. Because of the mentioned shortcomings above, and due to
possible conflicts of the lateral bracings with reinforcements, constructors usually
prefer to avoid the use of plan bracings.
In Sweden, a common solution in efforts to enhance the lateral stability of steel
girders during erection is to increase the width of the flanges. This often requires
that the thickness of these flanges also be increased so as to avoid possible local
buckling problems. In composite bridges, however, the top flanges are basically
needed so as to provide sufficient space for the shear-studs once the concrete deck
has hardened. Fig. (6.15) shows in schematic terms a typical design curve, one in
current use by design engineers for sizing the steel girders employed during the
construction phase. Increasing the size of the flanges decreases the lateral-torsional
slenderness of the main girders. Most bridges, in Sweden, are designed with use of
rather stocky girders having a relatively small lateral-torsional slenderness ratio;
see region “D” in Fig. (6.15). Modern fabrication techniques have enabled the use
of high-strength steel from which girders that are more slender  having larger
values and less consumption of steel  can be designed for the composite stage.
However, the current strategy to enhance the lateral stiffness of a bridge by means
of increasing the size of compression flanges is independent of the steel type
involved.

Figure (6.15)
A schematic illustration of a typical design curve for lateral-torsional buckling, being a reduction
factor and being the lateral-torsional slenderness ratio.

70
6.8.2 Shortcomings of the common bridge timber-falseworks

Cup-lock scaffoldings (see Fig. 6.16) and timber-frame falseworks (see Fig. 6.17)
are the most common approaches taken in Sweden for supporting the weight of the
fresh concrete during concreting of the deck in composite bridge. For relatively
shallow girders, the efficiency of steel-truss scaffoldings located between adjacent
girders is reduced since the inclination of the diagonal member is less. In such
circumstances, spacing of timber-frame falseworks every 0.9~1.2 m along the
bridge span are normally preferred. Otherwise, when the depth of bridge is
sufficient, steel -truss scaffoldings (a so-called Cup-lock system) can be an
alternative, a spacing of them at every 1.2~1.3 along the bridge span being
employed. As shown in Fig. (6.16), the top timber chord can be sloped so as to suit
the road transverse profile. Timber planks, 45 45 mm in size, for example, are
normally spanned between the timber chords in both formwork systems. In cases
in which the transversal span between the two adjacent girders is relatively large, a
single longitudinal beam is often placed between the timber frames near the
middle of them so as to avoid possible lateral slide of the bottom chords of the
frames.

Figure 6.16
Steel-truss scaffoldings, a so-called Cup-lock system.

Figure 6.17
A timber-frame falseworks used in a bridge application.

71
Bending moments from construction loads that develop at the cantilevers on both
sides need to be transferred to the main girders. This occurs by means of tension
bars anchored to the web of the main girders near the top flanges plus of inclined
compression bars that are placed on the bottom flanges of the main girder; see Fig.
(6.18). The deeper the girders are, the smaller the compression forces are that
develop in the diagonal bars of the cantilevers. Accordingly, both the distance
between the cantilevers and the length of cantilevers can be increased in the case
of deeper bridge girders. In practice, certain tension rods are also used between the
main girders, located every 2 m for example, along the bridge span so as to
prevent twisting of the individual girders brought up by the bending apart of the
girders at each cantilever; see Fig (6.17). Some bridges have experienced local
web buckling at the anchoring points of the tension bars, however.
The timber-frame falseworks between the main girders transfer the construction
loads, in the direction of gravity to the bottom flanges. In stability analyses of steel
bridges, the construction loads may be assumed to be uniformly distributed on the
top flanges of the main girders. This assumption can be inconsistent, however,
with the actual conditions in which loads are transferred by the scaffoldings and
the falseworks as explained above.

Figure 6.18
The cantilever scaffoldings.

6.8.3. The concept of scaffolding bracing of steel bridges

As an alternative to the current strategies in providing the steel girders with

stability, the utilization of Cup-lock scaffoldings as reliable stabilizing systems (on
the basis of their considerable stiffness and strength) during the construction stage
could be of interest. In view of the fact that such arrangements are already used in
steel bridges during the concreting of the deck, the alternative would not impose
any appreciable additional costs on the overall costs of bridge construction. As

72
shown in Fig. (6.15), this possibility enables design engineers to avoid increasing
the size of the cross-section simply for dealing satisfactorily with the temporary
conditions. The scaffoldings can safely control out-of-plane deformations of the
steel girders before the concrete deck has hardened, so as to be able to provide the
lateral stiffness that is required.
Fig. (6.19) illustrates the torque transferring mechanism that the cup-lock
scaffoldings provide. The scaffoldings normally possess a relatively large degree
of stiffness and strength. By utilizing this potential properly, the steel portion of
the bridge has to deform as a single unit, a matter made possible here by relative
lateral movements of the top and the bottom flanges of adjacent girders being
prevented from occurring, a matter that leads to significant increase in the load-
carrying capacity of the bridge. Currently, cup-lock scaffoldings are not normally
attached to the main girders  such as with the help of mechanical connections, for
example. Since, the scaffolding bars are not designed to transfer tensile forces, the
cup-lock scaffoldings are not able to prevent twisting of the main girders.
However, with some minor improvements, they could be attached to the girders,
even before either launching of the bridge or transportation of components of it to
the construction site. In this way, the stabilizing potential of the improved
scaffoldings could be of benefit not only during concreting of the deck but also
during the launching and lifting stages. This benefit is particularly advantageous
for in-plane-curved bridges and open-trapezoidal girders that are subjected to
considerable torsional stress because of their geometries. This potential can also
reduce the number of cross-bracings needed during the construction stage, and
possibly the size of the main girders.

Figure (6.19)
The torque  generated by destabiling forces, , that develop in the compression flanges 
transfering mechanism of the Cuplock scaffoldings, , and being the brace shear forces, and
being the reaction forces of the cantilevers.

The potential of the scaffoldings for use as a bracing system was first pointed out
by the author. Results presented in an article in the Journal of Bridge Engineering
[11] confirmed the scaffoldings that were studied possessing a substantial bracing
capacity. The research reported on comprised both experimental investigations
concerned with a full-scale test-setup bridge, and numerical studies on several
bridges having differing lateral-torsional slenderness ratios, including the case of a

73
real bridge in Sweden. Fig. (6.20) depicts the test data, both for the conventional
situation in which a cross-beam is placed every 7.5 m (i.e. test TN1), and for a
situation involving attached scaffoldings (i.e. test TN11). When the scaffoldings in
test TN11 were utilized, the load-carrying capacity of the bridge was enhanced by
a factor of 3.7 as compared with the load-carrying capacity achieved in test TN1;
see Fig. (6.20). In addition, the bridge braced by means of the scaffoldings (test
TN11) was able to tolerate considerably larger out-of-plane deformations than the
test TN1 bridge, this generally being regarded as having the advantage of its being
possible to warn the workers involved, of a possible structural collapse. In bridge
collapses due to problems of instability that were reported (see Chapter 2), sudden
failures have often occurred during the construction stage. When the bracing
potential of such scaffoldings is utilized, the enhanced ductility can enable the
bridge to tolerate relatively large out-of-plane deformations. Taking advantage of
the bracing potential of the scaffoldings involved can thus be important in order to
reduce fatalities, this leading to a “safer failure” in the case of a possible collapse.

Figure (6.20)
Test results for the load-carrying capacity of the bridge when braced by scaffoldings (test TN11), and
when braced by conventional cross-beams (test TN1 was performed with a cross-brace stiffness 30
times greater than required for a full bracing condition), standing for scaffolding and for
cross-bracing.

Only slight forces were developed in the scaffolding bars, the load-carrying
capacity of the bridge being increased by some 370% (see / 0.02 at the
ultimate load level in Fig. 6.21).
Another important advantage of the system described is that a bridge restrained by
means of a number of scaffoldings can be considered to represent a robust system.
Also, the scaffoldings stabilize simultaneously both the top and the bottom flanges
in both the sagging and the hogging regions, which for most bracing options is
normally not the case.

74
Figure (6.21)
Test results for the bracing forces generated in the scaffolding bars in test TN11, the system reaching
the ultimate load-carrying capacity there at 0.005 , being the bracing moment
of the scaffoldings at location along the bridge span.

6.8.4. Effects of the ledgers on the load-carrying capacity of the test

bridge

In practice, through certain horizontal bars, “ledgers” as they are called, the
inclined members of a typical Cup-lock scaffolding are interconnected both in-
plane and out-of-plane of the scaffolding trusses; see Fig. (2.22). The ledgers are
used to reduce the buckling length of the inclined scaffolding bars and to provide
the scaffolding trusses lateral stability. The process of assembling the ledgers on
the inclined bars can be carried out quickly in practice through use of the Cup-lock
joints placed at distance of approximately 0.5 m from each other along the length
of the inclined bars.

Figure (6.22)
A schematic illustration of Cup-lock joints and of the ledgers.

75
The Cup-lock joint is fixed by rotating the upper cup around the longitudinal axis
of the inclined bar when a ledger has already been placed in the fixed bottom cup.
In common practice, the ledgers of the two constant lengths of 1200 mm and
1300 mm are available, their cross-sectional size being identical to that of the
inclined bars of the scaffolding. The joint at both ends of a ledger can be
considered as being fairly moment stiff.
In the case of the test-setup bridge, seven brace configurations listed below were
studied numerically:
- System (i), the twin-I girders had an unbraced length of 15 m between the
end-supports.
- System (ii), only one cross-beam was added to System (i) at mid-span.
- System (iii), the typical Cup-lock scaffoldings were installed in System (i)
at spacing of 2500 mm along the bridge span. This represents a situation in
which every second scaffolding is fastened to the main girders, the spacing
of the cross-bracings increasing from 7.5 m to 15.0 m.
- System (iv), one cross-beam was added to System (iii) at mid-span.
- System (v), the typical Cup-lock scaffoldings were installed in System (i) at
spacing of 1250 mm along the bridge span. This represents a situation in
which all of the scaffoldings are attached to the main girders and the
spacing of the cross-bracings increasing from 7.5 m to 15.0 m.
- System (vi), one cross-beam was added to System (v) at mid-span.
- System (vii), the ledgers were placed in the middle of the inclined bars 1, 4,
5, and 8 (see Fig. 6.19), spanned between the scaffolding trusses, there thus
being only four ledgers between any two adjacent trusses.
Figs. (6.23) and (6.24) show the strength ratios versus the twist values at mid-
span, and the strength ratios versus the normalized values of the lateral movements
of the top flange at the mid-span of the test bridge, respectively. An initial twist
imperfection value of /500 was introduced in the main girders at mid-span, i.e.
similar shape to the first buckling mode of the unbraced girders. Obviously, both
the cross-bracings that are conventionally employed and the proposed scaffoldings
resist mainly the twist of each girder. Except for System (vii), Fig. (6.24) shows
that the bridge to still be suffering from a considerable amount of lateral
movement at mid-span in Systems (i)-(vi). However, with use of the typical
dimensions of the scaffolding bars there, the load-carrying capacity of the test
bridge improved by a factor of approximately 4.0, i.e. when comparing Systems
(v)-(vi) with System (ii). In addition, the results show that a combination of the
scaffolding trusses and the cross-bracings commonly employed, i.e. System (vi),
increased the load-carrying capacity of the steel girder slightly as compared with
System (v).

76
Figure (6.23)
The normalized load-carrying capacities versus the twist values at mid-span for systems (i-vii).

Figure (6.24)
The normalized load carrying capacities versus the normalized lateral deflection of the compression
flanges at mid-span for systems (i-vii).

A combination of the ledgers and scaffolding trusses controlled to a great extent

both the lateral deflection and the twist of the girders. Obviously, the scaffolding
trusses acted as torsional bracings thanks to their high degree of torsional stiffness.
One matter connected with this that was studied here was the effects that the
ledgers have on the bracing forces created in the scaffolding and ledger bars, a
matter taken up in the next section. However, the lateral bracing capacity of the
ledgers needs to be more thoroughly investigated.

77
6.8.5 The effects of the ledgers on the bracing forces

Numerical studies were performed to investigate the axial forces that developed in
the scaffolding bars and in the ledgers when uniformly distributed loads were
applied to the top chord of the scaffolding trusses, these loads representing the
construction loads generally encountered in practice. The forces that developed
represented the vector sum of the stabilizing forces and the forces created to
support the self-weight of the fresh concrete. Fig. (6.25) shows the normalized
magnitude of the forces created in the inclined bars along the bridge span; see
members 1-8 in Fig. (6.19). An imperfection shape similar to the first eigenmode
of the unbraced girders, its having a maximum twist of /500 at mid-span, was
introduced in the girders. The results obtained for both systems (v) and (vi)
showed that among the bras 1-8, the inclined bars 4 and 5 (see Fig. 6.19) provided
the major contribution to stabilization of the girders whereas the other bars (1, 2, 3,
6, 7, and 8) contributed mainly to carrying the construction loads. The inclined
bars 4 and 5 were found to act in tension and compression, respectively; see Fig.
(6.25).

Figure (6.25)
A comparison of the bracing forces induced in scaffolding bars 1- 8 of Systems (v) and (vi).

Fig. (6.26) depicts the normalized values of the axial forces created in the inclined
members 4 and 5 along the bridge span when different values for the initial
imperfections were introduced in System (v). Clearly, higher values for the initial
twist led to higher levels of the stabilizing forces in the scaffolding bars.

78
Figure (6.26)
The bracing forces, , , generated in scaffolding bars 4 and 5 of System (v) when initial
imperfections of differing magnitude were introduced.

The effects of the ledgers on the magnitude of the bracing forces in the scaffolding
bars were also investigated. A maximum initial twist value of /500 for the
girders was introduced. Fig. (6.27) depicts the normalized axial force values that
developed in the scaffolding bars of System (vii). Uniform loading of the upper
chords of the scaffoldings involved loads representing the construction loads
employed in practice. It was found that in utilizing the ledgers, the resultant forces
in the inclined bars were all in compression in system (vii), the magnitude of the
forces being almost constant. This occurred because of the fact that the ledgers
increased the lateral stiffness of the system, so that lesser stabilizing forces were
generated in the inclined bars than the forces generated by the self-weight of the
fresh concrete (compared the results obtained here with the data shown in Fig.
6.25).

Figure (6.27)
Bracing forces, , , generated in the inclined bars (1-8) of System (vii), the ledgers being placed at
the mid-length of scaffolding bars 1, 4, 5, and 8 between the adjacent trusses.
79
Fig. (6.28) depicts distribution of the axial forces that developed in the ledgers of
System (vii). The ledgers acted mainly in tension and in compression. The
distribution of the forces along the span shows the ledgers to be generating a
bending stiffness with respect to the z-z axis, as shown in Figs. (6.27)-(6.28).

Figure (6.28)
The bracing forces, , generated in the ledgers of system (vii), the location of the ledgers being
marked as C1, C2, C3, and C4.

Thus, the ledgers, firmly connected to the scaffolding trusses, can be regarded as a
sort of “Vierendeel beams” that having long lever arm, take the loads that acting in
the horizontal plane, see Fig. (6.29).

Figure (6.29)
The scaffolding's Vierendeel effects on forces generated in ledgers C1-C4.

It should be noted that the length of each ledger is relatively short (either
1200 mm or 1300 mm in line with current practice), and that the bracing forces
that are generated in the ledgers tend to be relatively small. Accordingly, the
second-order effects of the axial bracing forces on the ledgers were negligible. The
load-carrying capacities presented earlier for systems (vi) and (vii) represented
two extreme cases for the connection rigidity of the Cup-lock joints (i.e. pure
80
hinges in system (vi) and infinitely moment-stiff in system (vii)). In reality, the
level of performance is between these two extremes, as is the case for all of
connection types.
Although the ledgers strongly enhanced the lateral stiffness of the test bridge, they
may well not represent suitable lateral bracings for controlling the lateral
deformations of real steel bridges by creating Vierendeel beams. The
investigations discussed in this section did confirm, however, that a combination
of the bracing potential of the scaffolding trusses with an appropriate lateral
bracing system could well be more beneficial. Such lateral bracings can be
achieved by modifying the placement of the ledgers so as to achieve, for example,
a Pratt-truss form along the bridge span. It should be noted, however, that even
without the ledgers the scaffoldings can still increase considerably the load-
carrying capacity of steel bridges, a capacity level that may well be sufficient for
most of the situations encountered in practice.

6.9 Bracing potential of precast concrete slabs

The research group at the University of Texas under the supervision of Dr. Todd
Helwig is currently investigating the bracing potential of precast concrete slabs for
in-plane-curved bridges. Fig. (6.30) shows a photo taken by the author during his
visit to Ferguson Laboratory of Structural Engineering at the University of Texas.

Figure (6.30)
An experimental study of the stabilizing potential of precast decks presently being performed in the
Ferguson Laboratory of Structural Engineering at the University of Texas.

As shown in Fig. (6.31), in the situation currently encountered in practice, the

precast slabs employed have no mechanical connections with the shear studs
placed on the main girders or on the cross-girders. Clearly, both the cross-girders
and the main girders can be assumed to be restrained continuously at the top
81
flange level when the joints are grouted during construction. Developing a
practical mechanical connection between the concrete slabs and the main girders is
an important challenge in connection with such bracing alternatives. In a manner
similar to what applies to corrugated metal sheets, such systems have limited
efficiency in sagging regions and have no stabilizing effects on the compression
flanges in hogging regions of I-girder bridges.

Figure (6.31)
The most common approach to installing the precast concrete decks of steel bridges.

82
7 The effects of initial imperfections
on the performance of bracings
(Papers III, IV, & V)

Deciding during the computational analyses of different possible imperfection

shapes, which of these would provide the maximum level of bracing forces can be
difficult. The bracing forces are affected to a significant extent by a number of
factors including in particular the magnitude, the shape, and the distribution of the
initial imperfections along the span, and the load gradient at the brace point [39].
Various code specifications, such as AASHTO [82] and AISC [67], recommend
that a relative initial crookedness value of /500 be considered for the
compression member located between the brace points, being the distance
between the adjacent bracings. Eurocode, in contrast, suggests the use in stability
analyses of a half-sine-shape initial imperfection of /500, being the bridge
span. Regardless of the obvious differences between these recommendations, a
literature study revealed that relatively few investigations have been carried out on
the effects of imperfections on bracing forces that are present particularly in steel
bridge applications.
In building applications, such as in simply supported truss roofs, the imperfections
that are present can be approximated in two ways:
- i) by an equivalent lateral force, e.g. 8 ∆ ∆ / , applied
uniformly to the compression flange, where ∆ is the lateral deflection of the
restrained flange under plus external lateral loads, ∆ is the magnitude of
the geometric imperfections, is the beam span, / is the axial force
present in the compression flange under a uniform in-plane bending of ,
and is the distance between the centroids of the top and the bottom flanges
[73]; see also Appendix (AIII.2);
- ii) by performing a second-order large-displacement analysis of the initially
deformed compression chords [68].
In the former method, the second-order effects should be also included in the
calculation of ∆. If the brace is sufficiently stiff to act in a manner similar to an
immovable support, the bracing forces will be of reasonable magnitude.

83
As an example, if one assumes, for a simply supported beam with a sufficiently
stiff lateral brace placed at mid-span at the level of the compression flange that
∆ ∆ /500, the bracing force then would be approximately: 0.04
1.25 8 2 /500 / .

7.1 Effects of the magnitude and the shape of the initial

geometric imperfections on the load-carrying capacity of
steel girders
Fig. (7.1) shows the normalized load-carrying capacity versus the twist at mid-
span of the test setup having the bracing configuration of system (v) as described
earlier in Section (6.8.4), different initial twists of /500 , /1000 , and /
2000 being introduced to the model. The stiffness of the system was
found to decrease dramatically when the magnitude of the initial twist was
increased. The initial twist value of /500 considered here is on the basis of the
Eurocode 3 recommendation [76] for initial imperfection of steel bridge girders.

Figure 7.1
Effects of imperfections on the load-carrying capacity of System (v) as described in Section (6.8.4),
scaffoldings being present every 1250 mm and there being no cross-beam at mid-span.

Despite the clear effects that could be noted, there is relatively little information
available regarding the effects of the shape of the initial imperfections on the
performance of typical bracings in steel bridges [39]. Mehri et al. [83] showed that
the shape of the initial imperfections can strongly affect both the load-carrying
capacity of the girders and the bracing forces. Depending upon the initial shape
and the magnitude of the geometric imperfections involved, the girders of a twin I-
girder bridge braced by means of cross-beams can sway either in the same
direction or in the opposite direction, this creating either single- or double-
curvature-bending of the cross-beam [83].
84
It is common practice among design engineers to assume that the compression
flange of steel girders would buckle between the cross-bracings when the bracings
are provided with sufficient stiffness . . , , ; see Fig. (7.2 i). The
load-carrying capacity of the system there can be calculated then on the basis of
the buckling capacity of each girder spanned between the bracing points. The
effects of the load gradient and of the shape of the initial imperfections that are
present are ignored, however, in this statement of the matter. For instance, an
initial imperfection of the compression flanges in the shape of a half-sine-wave
between the end-supports may result in a buckling mode of the compression flange
in the manner shown in Fig. (7.2 ii) rather than buckling between the brace points
in the manner shown in Fig. (7.2 i). The critical load corresponding to the buckling
mode (ii) can be much greater than that corresponding to the buckling mode (i);
see the appended Papers (IV and V) [83, 85].

Figure 7.2
Effects of imperfections on the performance of cross-bracings

The effects of different imperfection shapes on the critical load value can be
explained easier in terms of some particular examples of column bracing. For
instance, assuming , / and 0.0 conditions to apply,
Eq.(4.5) in Section (4.4) yields the results given in Eqs.(7.1)-(7.2) for the total
end-slope values at the support points when the geometric imperfection shapes of
, sin / and , sin 2 / , respectively, are initially
introduced for a simply supported column [45].

,
, 0 / , , 7.1
1
,

,
, 0 2 / , , 7.2
1
4 ,

85
In practice, although a column having a full-sine-shaped imperfection, i.e.
, sin 2 / , might never reach a critical load value of 4 ,
corresponding to an S-shape buckling mode, the column might not buckle at a
, load level either. Depending on the magnitude of an imperfection, snapping
through from an initial S-shape to a bow shape deformation, and in general a
substantial change in the out-of-plane deformation profile, might require much
additional strain energy to be consumed. Under such circumstances, a column may
reach a load-carrying capacity in excess of the minimum critical load value
corresponding to its lowest eigenmode. Of the infinite number of imperfection
shapes that are possible, only one can be the same as that of the first eigenmode of
a braced system. Although this worst scenario might be relevant in some cases and
might possibly occur in practice, its occurrence is very unlikely in most practical
situations. For instance in a twin-I girder bridge having a number of intermediate
cross-bracings, a sine-wave imperfection shape (a shape similar to the first
buckling mode shape) having inflections at the locations of the cross-bracings is
very unlikely to occur.

7.2 Effects of the shape and the magnitude of initial

geometric imperfections on brace forces
The magnitude of the initial imperfections affects the bracing strength
requirements directly. Eq. (7.3) presents a general expression for the relationship
between the initial imperfections, ∆ , and the bracing forces, , [1].

/
∆ ∆0 7.3
1 /

Where ∆ is the deformation of the compression member at the brace point, and
and are the actual (provided value) and the ideal (minimum required
value to act similar to immovable support) stiffness values of the bracing in
question.
The bracing force obtained on the basis of Eq. (7.3) can be adjusted by a
modification factor of 1 /∆ if any slipping, , also occurs in bracing
connections in cases of oversized holes [3].
Investigations reported in Paper (IV) showed initial imperfections to also have a
strong effect on the magnitude of bracing forces that develop in the cross-bracings,
and often on the load-carrying capacity of slender girders as well. Both tests and
FE investigations that were conducted showed the shape of geometric
imperfections to have a dominant effect on the distortion of the cross-section of a
bridge resulting in either a single-curvature or a reverse-curvature bending of the
86
cross-beam members, regardless of the location of the cross-beam member across
the depth of main girders. In investigations of brace forces in the cross-beams of
steel bridges, in addition to the half-sine shape between the end-supports currently
recommended by Eurocode 3, the study recommends that a similar imperfection
shape, but one having an eccentric distribution with its maximum magnitude at a
quarter of the bridge span, for example, also be considered in the bracing analyses.

87
8 Laboratory tests

8.1 Background
Bracing forces obtained in experimental investigations are highly sensitive to
friction at the support points, as well as between the loading apparatus and the
main girders. Creating “roughly” ideal boundary conditions is crucial in
performing various tests carried out in stability investigations. In order to
minimize lateral restraints, Helwig et al. [41] used a truss-form gravity load
simulator such as that shown in Fig. (8.1), one anchored to the reinforced concrete
floor. Two loading simulators were positioned at the one-third points along the
bridge span. Tensile axial forces generated by 900 kN capacity hydraulic actuators
were used to apply compression forces at the level of the top flanges of both
girders. In order to eliminate the tipping restraint of the loading beam, knife-edges
were welded to the ends of each loading beam. The objective was to study the
stabilizing potential of corrugated metal sheets when they are attached to the top
flanges of twin I-shape steel girders.

Figure (8.1)
The gravity load simulator used at the University of Texas for studying the stabilizing performance
of corrugated metal sheets on twin I-girder bridges.

89
8.2 Tests performed by the author
An alternative loading apparatus to the system shown in Fig. (8.1) was designed
by the author for performing the laboratory tests (see Fig. 8.2), this being done for
following reasons: i) building two truss-simulators would increase the expenses
and require considerable time, ii) some friction at the hinged points of the loading
apparatus was found to still exist in the system shown in Fig. (8.1); and iii) it was
desired that each girder be able to deform independently of the other in the lateral
direction. The system shown in Fig. (8.1) needed to be modified for this purpose.

Figure (8.2)
The following details of the test setup are shown: a) the loading apparatus, b) the boundary
conditions for the top flanges, c) the knife-edge boundary conditions of the bottom flanges at the one
end, d) the slide free boundary conditions of the bottom flanges at the one end, e) the four roller
bearings cut by means of CNC to be placed between the top flanges and the loading apparatus but
replaced with the shaf-beraings, f) the four shaft-bearings used between the top flanges and the
loading apparatus.

90
Fig. (8.2) shows the employed loading apparatus, the boundary conditions, and the
bearings placed between the loading apparatus and the main girders. Fig. (8.2 e)
shows the original plan (roller bearings cut by CNC) for the bearings designed so
as to minimize the friction between the loading apparatus and the main girders. A
preliminary test revealed, however, that there was considerable friction at those
points that resisted the lateral movement of the main girders. Alternatively, four
high quality SY510M ∅50 mm shaft-bearings having a load capacity of 100kN
were utilized at the points shown in Fig. (8.2 f). Pre-test investigations that were
carried out verified the shaft-bearings being sufficient for the purpose of the study.
Due to a limited budget, the test girders and their connections needed to be used
again and again in eleven planned tests involving different bracing configurations;
see TN1-TN11 in Fig. (8.3). In order to avoid inelasticity, the maximum stresses
needed to be kept well below the yielding stress of the steel girders (e.g. at less
than 50% of the yielding stress of them) in accounting for the combined effects of
residual and of nominal stresses.

Figure (8.3)
Full-scale tests performed by the author, TN1-3 reported in Paper IV, TN4-10 reported in Paper V,
and TN11 reported in Paper III.

Before the test setup was ordered, a number of numerical and analytical analyses
were carried out to ensure that elastic buckling would occur well below the
91
yielding point, before the last test TN11, which involved loading in the inelastic
region. The measured initial out-of-straightness values of the flanges prior to each
of the tests that were carried out confirmed the permanent deformations being
negligible.
Fig. (8.4) shows the location of the potentiometers, the LVDTs, and the strain
gauges used during tests TN1-TN11. LVDTs were used instead of potentiometers
at points where the deformations were expected to be very small, e.g. at S/N-1/2 or
at S/N-9/10 points during the tests TN4-TN10.

Figure (8.4)
The location of potentiometers, LVDTs, and strain gauges for the setups of the tests TN1-TN11
performed by the author.

92
The potentiometers and the LVDTs employed were "Celesco Stingpot 300 mm
0.2 mm", "Novotechnik TR 100mm, TR 50 mm 0.05 mm", and "Tex 0200
0.05 mm". Strain gauges were attached to the cross-beams at mid-span (8 gauges:
SG/NG1-4), to the truss-bracings (20 gauges: F1-F10), and to the scaffolding bars
(32 gauges: FR1, FR3, FR5, FR7) all being the single-direction "Kyowa 15 mm"
type. The loading piston and its maximum capacity was "Enerpac 600kN". The
loading piston was placed at the center of the H-shape loading apparatus, which
was built-up using European standard I beam profiles. The longitudinal beam of
the loading apparatus was a 5000mm long IPB330 standard profile type
reinforced by 15mm thick steel plates welded onto the topmost surface of the top
flange and onto the bottommost surface of the bottom flange. The transversal
beams were IPE330. Web stiffeners were welded at the loading point and the
shear reaction points.
Further details concerning the test setup are provided in Appendix I.

93
9 Numerical simulations

Performing full-scale experimental investigations can be very expensive. Efforts to

create semi-ideal boundary conditions and care in eliminating undesired friction
are essential to ensure the usefulness of the structural responses such as in the case
of the bracing forces. At the same time, analytical solutions are only possible for
relatively simple and limited cases for which a number of assumptions may be
required to simplify the problem. The derivation of closed-form solutions in
bracing assessments of bridges taking into account various conditions such as
those of load-height, asymmetry, cross-sectional variations, imperfections, and the
like are very difficult if not impossible to carry out. The simplified methods that
are available such as the Grillage analogy for the analysis of bridges require
considerable simplifications, and must be repeated for a number of possible
buckling mode cases. Human errors can also occur in arriving at assumptions and
in the superposition of principals. Simplified methods using the analogy of line-
beams on springs can also be valuable and are often suited to particularly common
situations. The line-beam models, however, do not consider directly possible
transversal effects such as generated by skewed supports, or by radial flange forces
in curved bridges. Also, the effects of load-height, and of bracing location across
the depth of girders, are difficult to deal with by use of line-beam models.
Although initial imperfections can have a substantial effect on the magnitude of
bracing forces, the stability analysis of bridges is highly complex when the initial
imperfections are included in the analyses. At times, therefore, there is a need of
advanced numerical tools for carrying out sophisticated analyses.
Numerical programs are widely available and are relatively easy to use. However,
they have the following potential disadvantages:
- The chance of errors is higher,
- It may take longer to set up the model and extract the results,
- The results can only be relied on with greater confidence if the program is
used with sufficient practice and with adequate understanding of the
modelling assumptions.
- Outputs are highly dependent upon the accuracy of the input data and the
mechanical properties that are defined such as the element type and the
meshing size,
- More checking is needed.

95
Numerical programs are more vulnerable to human errors if one lacks sufficient
practice and understanding. It should be noted, however, that human errors are
inevitable regardless of the method employed if personnel involved lack
competences. Despite the inherent disadvantages noted above, numerical programs
are highly valuable tools when used to initiate ideas or to perform parametric
investigations or sensitivity analyses after the model is verified against simplified
approaches, experimental data, and analytical solutions. Without doubt, valuable
knowledge can be obtained in research with greater confidence in the results by
utilizing a combination of the various tools available, including analytical
approaches, numerical programs, and laboratory tests.
The commercial software Abaqus [91] was used here mainly for the numerical
investigations. Shell elements S4R, were used for modeling webs, flanges, web-
stiffeners, and cross-beams since the major portion of the strain energy involved is
due to in-plane deformation of the elements. In the case of shell elements,
sufficient mesh densities with the aspect ratios close to unity were employed in
three directions. The effects of material nonlinearities and large deformations were
also taken into account in third-order incremental analyses. Multi-linear material
properties were used for steel grade S355 having stress values of 355, 360, 539,
and 571 MPa, corresponding to strain values of 0.2, 1.3, 5.5, and 11.3%, and for
the S460 steel type with stress values of 460, 466, 561, and 605 MPa,
corresponding to strain values of 0.2, 1.2, 4.0, and 11.2%. Top flange uniformly
distributed loading was considered in the analyses where applicable, since this is
the most common loading condition in practice. However, four-point-flexure
loading was considered for modeling when a calibration against test data was
being carried out.
Two methods for conducting buckling analyses are available in numerical
programs: eigenvalue analysis and incremental analysis. An eigenvalue analysis
requires considerably less computational effort since in an incremental analysis the
global stiffness matrix becomes updated at each increment. Three different
methods for introducing initial geometric imperfections are available: i)
performing a linear superposition of buckling mode shapes, ii) using the
displacements that take place in a static analysis, and iii) applying displacements
to nodes directly. Defining the precise shape of imperfections was not the aim of
the study. Imperfections consisting of multiple superimposed buckling-mode
shapes regarded as perturbation of the geometry involved were considered in the
nonlinear analyses. In use of this approach, two analyses were performed on the
same model: a linear buckling analysis of a perfect form of the bridge to determine
the buckling modes, and an incremental third-order nonlinear analysis. In the latter
analysis, the scaled initial imperfections obtained from the former analysis, were
added to the perfect geometry to create a perturbed mesh. Abaqus imports
imperfection data in terms of node geometries. Although during the two separate
analyses that were performed no compatibility check of the two models carries out

96
by Abaqus, the original model and the subsequent one should be consistent in
terms of the assembly, and the node and element numberings, and the like.
To extract nodal information concerning the buckling mode shapes, the following
text was added to the input file of the first analysis, the “last mod” being the
desired number of eigenmodes to be extracted.
---------------------------------------------------
*Output, field, variable=PRESELECT
*NODE FILE, Global=yes, last mod=1
U
-----------------------------------------------------
To introduce the geometric perturbation to the incremental analyses, a linear
superposition of scaled buckling mode shapes was defined. A modification using
the text that follows was carried out in the input file of the subsequent analyses.
The second and third lines of that text (see below) consist of two values, the first
one being the eigenmode number, and the second one the scale factor.
----------------------------------------------------
*Imperfection, File=first job file name, Step=1
1, 15
2,-10
-----------------------------------------------------

97
10 Conclusions and future research

10.1 Conclusions from the appended papers

Paper I:

Paper I evaluated the collapse of The Marcy Bridge that collapsed in New York
City in 2002. It was found that the load height effects need to be considered in the
design of both the braced and the unbraced trapezoidal girders. For trapezoidal
girders that are sufficiently braced by cross-bracings, the load-height modification
factor has a constant value. However, the value decreases when the web-slope of
the unbraced girder is increased with respect to the vertical axis. Although the
stiffness of the cross-bracing employed in The Marcy Bridge was several times
greater than that required for a full bracing condition, the bridge failed,
nevertheless, due to system buckling. The results of the study showed that the
load-carrying capacity of The Marcy Bridge could have been improved by adding
top flange bracing at 10-20% of the span length near the end-supports. Providing
X-type plan bracing with relatively small cross-sectional area, as little as 8 mm
for each bar, could have enhanced the load-carrying capacity of the bridge,
according to Eurocode 3, by a factor of 1.28 which would have been sufficient to
prevent the failure of the bridge during casting of the deck.

Paper II:

In Paper II, a simplified model for determining the critical moment value of
laterally braced steel girders placed at the level of their compression flange was
introduced. This is often difficult or cumbersome to deal with, without the use of
Finite Element programs. The simplified solution that the model provides can be
of considerable help to design engineers, either in preliminary decisions
concerning the size of different bridge girders and their bracings, or in checking
the numerical results. The model relates the buckling length of the compression
flange of a steel girder to its lateral-torsional critical moment. Both an exact
solution and a simplified expression were derived to take account of the rotational
restraining effects of unequally spaced lateral bracings on the effective buckling
length of the compression flange in question. This is a matter that has been
neglected in practice, its tending to be assumed there that the buckling length of a
99
compression member is equal to the largest distance between the bracing points.
The paper showed that this assumption can give inaccurately unsafe results for
very low bracing stiffness values and can at the same time lead to significant
overestimation problems for relatively high bracing stiffness values. Typical
design curves are presented that can help design engineers considerably in
choosing appropriate brace locations and stiffness values without the need of using
commercial finite element software for this purpose. The approach can also help
one better understand both column and beam bracing and theories concerning
them.

Paper III:

Paper III, presents the results of the original study  no previous investigation of
which appears to have been carried out  of the stabilizing potential of one type of
scaffolding, so-called “Cup-lock scaffolding”, widely used in the construction of
composite bridges in Sweden. Currently, scaffoldings are not considered for use as
stabilizing systems in the design of composite bridges, despite the relatively high
degree of torsional stiffness they possess. The paper presents the results of full-
scale experimental and numerical investigations concerning the torsional bracing
performance of such scaffoldings. Minor structural improvements in the
scaffoldings were first needed, as discussed in the paper. There was found to be
significant bracing potential in the structurally improved scaffoldings when they
were installed on bridges of differing dimensions and lateral-torsional slenderness
ratios in the numerical investigations. The bracing moments generated in the bars
of such scaffoldings were measured. Indications of the bracing forces involved,
expressed as a percentage of the maximum in-plane bending moment measured in
the main girders, were also provided for design purposes. Since utilizing the
bracing potential of such scaffoldings does not substantially increase construction
costs, the bridge industry can benefit considerably from taking advantage of the
bracing potential of such scaffoldings so as to enhance both the load-carrying
capacity and the ductility of bridges during construction.

Paper IV:

Paper IV presents the results of experimental and numerical investigations of the

effects of the shape of imperfections on the bracing performance of the typical
cross-beams in steel bridges. Three large-scale experimental studies of a twin I-
girder bridge were carried out in which the location of the cross-beam across the
depth of the main girders was varied at mid-span. In numerical investigations,
three bridge structures of differing lateral-torsional slenderness ratio were studied
for determining the effects of several different relevant imperfection shapes on the
bracing performance of the cross-beams in straight steel bridges. The results

100
obtained showed that the design recommendations currently used can yield unsafe
and incorrect predictions of the bracing forces involved. In addition, it was found
that slender bridge girders can reach a load-carrying capacity significantly greater
than what is predicted theoretically for the critical load value of the braced girders,
the extent of this difference depending upon the shape of geometric imperfections
present initially in the main girders. Both tests that were carried out and the FE
investigations showed the shape of imperfections to also have significant effects
on the distortion of the cross-section of the braced bridges, its resulting in either a
single-curvature or a reverse-curvature bending of the cross-beams involved
regardless of the location of the transversal beam across the depth of the main
girders. It was also found that both lateral bending and torsional moment could
develop in the cross-beams, their magnitudes being affected in a great extent by
the shape of initial imperfections. Since significant warping stresses can develop
in the cross-beams of asymmetric cross-section, avoiding such beam profiles in
cross-beams is suggested.

Paper V:

Paper V presents the results of seven full-scale laboratory tests of the restraining
effects on end-warping of plan bracings and corrugated metal sheets when they
were installed near the supports of a twin I-girder bridge. Four tests were
performed when single-panel Z-type or double-panel Warren-type bracings were
attached on the top flanges near each support. In addition, three tests were
performed when typical metal sheets were attached by use of nail fasteners of
different arrangements to the top flanges. The bracing forces generated in the truss
bars were also measured using a pair of strain gauges attached on each bar. The
load-carrying capacity of the bridge was found to be enhanced by a factor of
2.0~3.0 when the girders were provided with such warping restraints of both
types. The metal sheets employed there, however, showed significantly larger
lateral deflections than the utilized plan bracings did. Relatively small forces were
generated in the truss bars in order to achieve these marked improvements in the
load-carrying capacity. The brace forces in the employed truss bars were
compared with the ones obtained through an approximate method being used
frequently by the design engineers for the analysis of such bracings. The results
obtained by the approximate method  which is suggested also by Eurocode 3 
agreed fairly well with some of the brace forces that measured during the tests
whereas, the predication gave unsafe results with very large discrepancy for few of
bars.
The study showed that such inexpensive modifications could be of valuable
benefit to narrow bridges that are prone to system buckling.

101
10.2 Future research
Despite the fact that bracings are generally highly effective in controlling the
lateral-torsional deformation of steel bridges, relatively few guidelines and design
recommendations are available in the code specifications regarding this matter.
The use of proper bracings is often crucial for avoiding certain possible failure
modes during construction. The size, layout, and type of bracings employed can
directly affect the initial stresses that develop in steel girders due to out-of-plane
deformations during construction, especially in the case of in-plane curved
bridges. There is much yet to learn regarding the effects of bracings during
construction. Also, effects of the concreting sequence and of the shape of the
initial imperfections have been little investigated. The effects of load height,
imperfections, variations in cross-sections, asymmetries, eccentricities, and the
like make the predication of bracing requirements rather complicated. Further
investigations are needed to obtain a better understanding of how design engineers
can size bracings as satisfactory as possible.
Scaffoldings can serve as a reliable stabilizing source during construction of
bridges. They can also serve as a temporary bracing system during the widening of
existing bridges or the repairing the bracings (the repair of which led to the
collapse of the three-span Tennessee bridge in the U.S.; see Section 2.2). The
present Ph.D. work has shown that such scaffoldings can serve to substantially
enhance both the ductility and the robustness of steel bridges during construction.
This should be studied in greater detail for more practical purposes. Such
scaffoldings can also further the industrialization of bridge construction and the
shortening of construction time. Attaching scaffoldings to the main girders can
enable the development and also the use of “smart” scaffoldings in the future,
together with the automatic monitoring of deformations and stresses during
construction on the basis of brace forces generated in the scaffolding bars. The
data collected can considerably benefit understanding of the behavior of typical
steel bridges during different stages of construction.

102
Acknowledgements

The financial support this Ph.D. study received from the “J. Gust Richert Stiftelse”
Dnr 2012/05, and from “The Lars Erik Lundbergs Stipendiestiftelse” Dnr 2013/07
and Dnr 2014/05, as well as a research grant from Byggrådet to support the
laboratory testing is gratefully acknowledged.
The material support that Britek AB provided for the scaffoldings required and
that Structural Metal Decks (SMD) Ltd provided for the corrugated metal sheets
that were needed during the laboratory tests is likewise greatly acknowledged.

103
Appendix I

Details regarding the test setup

Figure (AI.1)
details regarding the end-supports.

Figure (AI.2)
End-support components.

105
Figure (AI.3)
End-support components.

Figure (AI.4)
The dimensions of the cross-beam and of its web-stiffeners at mid-span used in tests TN1-TN11.

106
Figure (AI.5)
Elevation views of the two identical main girders, and plan views of their flanges.

Figure (AI.6)
A plan view of the loading apparatus.

107
Figure (AI.7)
Details of tests TN4, TN5, TN6, and TN7; TN4: a Z-type single panel brace with a cross-beam at
mid-span, TN5: a Warren-type brace with a cross-beam at mid-span, TN6: a Z-type single panel
brace without a cross-beam at mid-span, TN7: a Warren-type brace without a cross-beam at mid-
span.

Figure (AI.8)
Details of tests TN8, TN9, and TN10; where corrugated metal sheets were attached to the top flanges
near the supports, nail fasteners being located either at every second valley (in TN8) at each valley
(in TN9) or at each valley provided with stiffening angle profiles at its free edges (in TN10).

108
Figure (AI.9)
Details of test TN11: the scaffoldings were attached to the main girders here at each 1200mm or
1300mm along the bridge span.

109
Appendix II

Test data not reported directly in the appended articles

Figure (AII.1)
Bracing forces measured in the truss bars during test TN6.

111
Figure (AII.2)
Bracing forces measured in the truss bars during test TN7.

112
Appendix III

AIII.1 Bracing analysis; AASHTO recommendations

[82]

Summary of the recommendations regarding the use of cross-bracings

The need for cross-bracings shall be investigated both for all stages of the
construction procedures and the final conditions that are present, this including but
not limiting to the following:
i) the transfer of wind loads between the girders and to the bearings, ii) the
stability of the bottom flange in compression, iii) the stability of the top flange in
compression prior to the concreting of the deck, and iv) the distribution of the
vertical and the live loads that are applied to the bridge.
The cross-bracings for rolled beams and for built-up girders should be as deep as
practicable, for rolled beams their being at least half of the beam depth and for
built-up girders their being at least 0.75 of the girder depth. Cross-diaphragms
having a span-to-depth ratio of greater than 4.0 should be designed as beams. For
bridge sections consisting of two or more boxes, external cross-bracings shall be
used at end supports. For such systems, the external cross-bracing at the interior
supports can only be eliminated if an analysis indicates that the individual boxes
are torsionally stable. At the location of external cross-bracings, internal cross-
bracings should be also provided.
The cross-bracings should be placed normal to the main girders if the supports are
not skewed. They can be placed parallel to the skewed support lines if these are
not skewed from normal more than 20 degrees. In such situations, the web-
stiffeners should be oriented within the plane of the skewed cross-bracings. When
the support lines are skewed from the normal by more than 20 degrees,
intermediate cross-bracings shall be placed normal to the girders. The effects of
the tangential components of the force transmitted by the skewed end-cross-
bracings shall be taken into account when the cross-bracing is skewed. The
removal of highly stressed cross-bracings in the vicinity of interior supports that
are shewed by more than 20 degrees from the normal, for example, may be
beneficial as long as out-of-plane deformation of the girder is not excessive.

113
Where end-supports are skewed by more than 20 degrees from the normal, the first
intermediate cross-bracing adjacent to a skewed end-support should be offset by a
minimum of 0.4 or 1.5 , being the cross-brace spacing and the girder web
depth, this reducing the presence of a stiff-load-path that can attract and transfer
large transverse forces to the skewed support.
Cross-bracings shall be spaced along the bridge span in nearly uniform fashion so
as to promote constructability and to allow the use of simplified methods of
analysis for calculating flange-lateral bending stresses. The requirements for cross-
bracings spaced at less than 7.5 m from one another given in previous versions of
the AASHTO specifications are replaced here by requirements based on a rational
analysis aimed at reducing the presence of fatigue-prone attachment details. The
spacing of the intermediate cross-bracings, , in an in-plane curved I-girder
bridge shall not exceed /10 , being the girder radius and the limiting
unbraced length, which can be determined from / , where is the
effective radius of gyration for lateral torsional buckling and the compression-
flange stress within the cross-section at the onset of nominal yielding. At a cross-
brace spacing greater than , significant lateral flange bending is likely to occur.
Intermediate cross-bracings should be placed at or near the in-plane maximum
moment in the main girders and near to both sides of the field splices that are
present. The need for additional temporary or permanent cross-bracings should be
also taken into account in connection with the transportation and the erection of
each shipping piece at the lifting points involved.

Summary of the recommendations regarding the use of lateral bracings

Corrugated metal sheets should not be assumed to provide the top flange in
compression with adequate stability during concreting of the deck.
Lateral bracings that are not required for the completed state should not be
considered as being primary members of the girders in question, and can thus be
removed after construction of the bridge is completed. Lateral bracings, if
required, should be placed either in or near to the flange that is being restrained. In
I-girder bridges having lateral bracings at the level of the bottom flange, a pseudo-
closed section can be created together with the hardened concrete deck, this
increasing the forces in the cross-beams as they serve to restrain the shape of the
closed cross-section.
To reduce the occurrence of large relative lateral movements during construction
in the case of I-girder bridges having spans greater than 60 m, either temporary or
permanent lateral bracings in one or two panels close to the end-supports  and
near the internal supports when continuous bridges are involved  can be
considered in the design stage. Although such bracings, when placed at the level of
the bottom flanges are about equally effective as compared with those placed at
114
the level of top flanges, top lateral bracing tends to be more preferred, since in the
completed state these are subjected to live-load forces to a lesser degree. For
horizontally curved bridges having a relatively sharp curvature, it may be more
beneficial to provide both top and bottom flanges with lateral bracing of this type.
The shear center of an open trapezoidal section is located below the bottom flange.
The use of top lateral bracing raises the shear center to a position closer to the
center of the pseudo-closed section, this increasing appreciably the torsional
stiffness of the cross-section. If such bracings are attached to the webs, the cross-
sectional area of the bridge should be reduced in calculating the shear flow
involved, and a means for the transfer of forces from the lateral bracings to the top
flanges should be provided. For straight trapezoidal girders having spans greater
than 45 m in length, full-span lateral bracings should be provided. The cross-
sectional area of diagonal members of top lateral bracings in both straight and
horizontally curved trapezoidal girders should be greater than 0.03 , being the
center-to-center distance between the top flanges. This criterion is recommended
so as to ensure that the warping stresses of the top flanges will be less than ten
percent of their in-plane bending stresses.

AIII.2 Bracing analysis; Eurocode recommendations

[76]

Effects of imperfections in analyzing a bracing system

The effects of imperfections can be included in the analysis of a bracing system

through specifying an equivalent geometric imperfection in the form of a bow-
shape having a maximum magnitude of ∆ :

1
∆ 0.5 1 ∙ .1
500

The effects of an initial bow-shape imperfection can be replaced by an equivalent

uniform destabilizing force of when a bracing system is needed in order to
stabilize the compression flange of a beam of constant depth:

∆ ∆
8 .2

Where / , and ∆ is the in-plane deflection of the bracing system due to

both and any external loads.

115
Lateral-torsional buckling of structural components

As a simplified approach in analyzing stability of the bottom flanges of continuous

girders located between supports, the second-order effects and the effects of
geometric imperfections on supporting springs can be taken into account by
specifying an additional lateral force at the connection to the spring:

, 1.2
100

1
, 1.2 .3
, 80 1 / ,

Where , / , , and is the distance between the adjacent bracings.

If the compression force of the flange of a multi-span bridge (in a completed

state after the concrete deck has hardened) is constant over the span as a whole,
the critical load capacity of the bottom flanges , can be calculated as follows:

2
, , .4

The lateral supports having a stiffness value of , can be assumed to be rigid if

4 , / . A safety verification can be carried out using
̅ ∙ / , , where , /3; and , is the area of the
compression zone of the web.

116
Appendix IV

Equivalent plate thickness of typical plan-bracings [2]

- For Warren-type bracing:

2 2 1.5
/ ∙ / / /3 ∙ 2/

- For K-type bracing:

2 2 1.5
/ ∙ / 2 0.25 / / 4 /12 ∙ 2/

- For X-type bracing:

2 2 1.5
/ ∙ / / 2 /12 ∙ 2/

- For Pratt-type bracing:

2 2 1.5
/ ∙ / / / /12 ∙ 2/

117
References

The main references used in the context of the thesis are listed below:

[1] G. Winter, "Lateral bracing of columns and beams", Vol. 125, No. 1, ed: ASCE,
1960, pp. 807-826.
[2] C. F. Kolbrunner and K. Basler, Torsion in structures an engineering approach,
Springer-Verlag, New York, 1969.
[3] T. A. Helwig and J. A. Yura, "Steel bridge design handbook: bracing system
design", U.S. Department of Transportation, Federal Highway Adminstration,
Washington, D.C. FHWA-IF-12-052, 2012.
[4] J. A. Yura, "Fundamentals of beam bracing", Engineering Journal, pp. 11-26,
2001.
[5] S. Mamazizi, R. Crocetti, and H. Mehri, "Numerical and experimental
investigation on the post-buckling behavior of steel plate girders subjected to
shear", presented at the Annual Stability Conference, SSRC 2013, St. Louis,
Missouri, 2013.
[6] A. Al-rubaye, "Theoretical and numerical approach to calculate the shear stiffness
of corrugated metal decks", M.Sc. Thesis TVBK-5229, Structural Engineering,
Lund University, 2014.
[7] A. Winge, "Temporary formworks as torsional bracing system for steel-concrete
composite bridges during concreting of the deck", M.Sc. Thesis TVBK-5231,
Structural Engineering, Lund University, 2014.
[8] O. Carlson and L. Jaskiewicz, "The performance of conventional discrete
torsional bracings in steel-concrete composite bridges: a survey of Swedish
bridges", M.Sc. Thesis TVBK-5241, Structural Engineering, Lund University,
2015.
[9] E. Ohlin, "Metal deck forms as lateral bracing of composite bridges with
trapezoidal cross-sections", M.Sc. Thesis, Structural Engineering, Lund
University, 2015.
[10] J. Scheer and L. Wilharm, Failed Bridges: Case Studies, Causes and
Consequences: Wiley, 2011.
[11] H. Mehri and R. Crocetti, "Scaffolding bracing of composite bridges during
construction", Journal of bridge Engineering (ASCE),DOI:10.1061/(ASCE)BE.
1943-5592.0000829, 2015.
[12] G. Pincus, "On the lateral support of inelastic columns", AISC Engineering
Journal, vol. 1, pp. 113-115, 1964.
[13] N. S. Trahair and S. Kitipornchai, "Elastic lateral buckling of stepped I-beams",
ASCE J. Struct. Div., vol. 97, pp. 2535-48, 1971.
119
[14] M. A. Serna, A. López, I. Puente, and D. J. Yong, "Equivalent uniform moment
factors for lateral–torsional buckling of steel members", Journal of Constructional
Steel Research, vol. 62, pp. 566-580, 2006.
[15] J. A. Yura, "Winter's bracing approach revisited", Engineering Structures, vol. 18,
pp. 821-825, 1996.
[16] C. M. Wang, S. Kitipornchai, and V. Thevendran, "Buckling of braced
monosymmetric cantilevers", International Journal of Mechanical Sciences, vol.
29, pp. 321-337, 1987.
[17] C. M. Wang, C. J. Goh, and E. P. Chew, "An energy approach to elastic stability
analysis of multiply braced monosymmetric I-beams", Mechanics of Structures
and Machines, vol. 17, pp. 415-429, 1989.
[18] C. M. Wang, K. K. Ang, and L. Wang, "Optimization of bracing and internal
support locations for beams against lateral buckling", Structural and
Multidisciplinary Optimization, vol. 9, pp. 12-17, 1995.
[19] Z. Vrcelj and M. A. Bradford, "Elastic distortional buckling of continuously
restrained I-section beam–columns", Journal of Constructional Steel Research,
vol. 62, pp. 223-230, 2006.
[20] J. Valentino and N. S. Trahair, "Torsional restraint against elastic lateral
buckling," Journal of Structural Engineering, vol. 124, pp. 1217-1225, Oct. 1998.
[21] J. Valentino, Y. L. Pi, and N. S. Trahair, "Inelastic buckling of steel beams with
central torsional restraints", Journal of Structural Engineering, vol. 123, pp. 1180-
1186, 1997.
[22] G. S. Tong and S. F. Chen, "Buckling of laterally and torsionally braced beams",
Journal of Constructional Steel Research, vol. 11, pp. 41-55, 1988/01/01 1988.
[23] Y. L. Pi and N. S. Trahair, "Nonlinear inelastic analysis of steel beam-columns,
II: Applications", Journal of Structural Engineering, vol. 120, pp. 2062-2085, Jul.
1994.
[24] J. A. Yura and G. Li, "Bracing requirements for inelastic beams", presented at
2002 SSRC Annual Stability Conference, 2002.
[25] Y. C. Wang and D. A. Nethercot, "Bracing requirements for laterally unrestrained
beams", Journal of Constructional Steel Research, vol. 17, pp. 305-315, 1990.
[26] T. A. Helwig, J. A. Yura, and K. H. Frank, "Bracing forces in diaphragms and
cross frames", presented at Structural Stability Research Council Conference, pp.
6-7, April 1993.
[27] Q. H. Zhao, B. L. Yu, and E. G. Burdette, "Effects of cross-frame on stability of
double I-girder system under erection", Transportation Research Record, pp. 57-
62, 2010.
[28] Q. Zhao, B. Yu, E. G. Burdette, and J. S. Hastings, "Monitoring steel girder
stability for safer bridge erection", J. of Performance of Constructed Facilities,
vol. 23, pp. 391-398, 2009.
[29] J. S. Park, J. M. Stallings, and Y. J. Kang, "Lateral torsional buckling of prismatic
beams with continuous top-flange bracing", Journal of Constructional Steel
Research, vol. 60, pp. 147-160, 2004.
[30] B. Kozy and S. Tunstall, "Stability analysis and bracing for system buckling in
twin I-girder bridges", Bridge Structures-Assessment, Design & Construction,
vol. 3, pp. 149-163, 2007.

120
[31] T. A. Helwig and J. A. Yura, "Shear diaphragm bracing of beams, II: design
requirements", Journal of Structural Engineering, vol. 134, pp. 357-363, 2008.
[32] T. A. Helwig and J. A. Yura, "Shear diaphragm bracing of beams, I: stiffness and
strength behavior", Journal of Structural Engineering, vol. 134, pp. 348-356,
2008.
[33] T. A. Helwig and K. Frank, "Stiffness requirements for diaphragm bracing of
beams", Journal of Structural Engineering, vol. 125, pp. 1249-1256, 1999.
[34] J. S. Davidson, M. A. Keller, and C. H. Yoo, "Cross-frame spacing and
parametric effects in horizontally curved I-girder bridges", Journal of Structural
Engineering, vol. 122, pp. 1089-1096, Sep. 1996.
[35] L. Q. Wang and T. Helwig, "Stability bracing requirements for steel bridge
girders with skewed supports", Journal of Bridge Engineering, vol. 13, pp. 149-
157, Mar.-Apr. 2008.
[36] C. Topkaya, Widianto, and E. B. Williamson, "Evaluation of top flange bracing
systems for curved box girders", Journal of Bridge Engineering, vol. 10, pp. 693-
703, 2005.
[37] J. A. Yura, T. A. Helwig, R. Herman, and C. Zhou, "Global lateral buckling of I-
shaped girder systems", Journal of Structural Engineering, vol. 134, pp. 1487-
1494, 2008.
[38] Z. Fan and T. A. Helwig, "Behavior of steel box girders with top flange bracing",
Journal of Structural Engineering, vol. 125, pp. 829-837, 1999.
[39] L. Wang and T. A. Helwig, "Critical imperfections for beam bracing systems",
Journal of Structural Engineering, vol. 131, pp. 933-940, 2005.
[40] J. A. Yura, B. Phillips, S. Raju, and S. Webb, Bracing of steel beams in bridges,
Center for Transportation Research, University of Texas at Austin, 1992.
[41] O. Egilmez, T. A. Helwig, and R. Herman, "Buckling behavior of steel bridge I-
girders braced by permanent metal deck forms", J. of Bridge Eng., vol. 17, pp.
624-633, 2012.
[42] O. Egilmez, R. Herman, and T. A. Helwig, "Lateral stiffness of steel bridge I-
girders braced by metal deck forms", Journal of Bridge Engineering, vol. 14, pp.
17-25, 2009.
[43] O. Egilmez, T. A. Helwig, C. Jetann, and R. Lowery, "Stiffness and strength of
metal bridge deck forms", Journal of Bridge Engineering, vol. 12, pp. 429-437,
2007.
[44] B. H. Choi, Y. S. Park, and T. Y. Yoon, "Experimental study on the ultimate
bending resistance of steel tube girders with top lateral bracing", Engineering
Structures, vol. 30, pp. 3095-3104, 2008.
[45] T. V. Galambos and A. E. Surovek, Structural stability of steel: concepts and
applications for structural engineers, John Wiley & Sons, 2008.
[46] R. Tremblay and D. Mitchell, "Collapse during construction of a precast girder
bridge", Journal of Performance of Constructed Facilities, 2006.
[47] B. Åkesson, Plate buckling in bridges and other structures, CRC Press, 2007.
[48] J. Leib, (2006, 21 March 2006), A series of deadly errors. Available:
http://www.denverpost.com/news/ci_3622757.
[49] H. Mehri and R. Crocetti, "Bracing of steel-concrete composite bridge during
casting of the deck", presented at the Nordic Steel Construction Conference 2012,
Oslo, Norway 2012.
121
[50] Edmonton Journal, (2015, 19 July 2015), Lack of bracing caused 102nd Avenue
Bridge girders to buckle, Available: http://www.edmontonjournal.com
[51] L. D. Luttrell, "diaphragm design manual", Steel Deck Institute, 1981.
[52] ECCS, "European recommendations for the application of metal sheeting acting
as a diaphragm", in ECCS publication, 1995.
[53] J. André, R. Beale, and A. M. Baptista, "A survey of failures of bridge falsework
systems since 1970", Proceedings of the ICE-Forensic Engineering, vol. 165, pp.
161-172, 2012.
[54] J. F. Bell, The experimental foundations of solid mechanics, Springer, 1984.
[55] B. G. Johnston, "Column buckling theory: historic highlights", Journal of
Structural Engineering, vol. 109, pp. 2086-2096, 1983.
[56] J. P. Den Hartog, Strength of Materials, Dover publications Incorporated, 1949.
[57] S. Timoshenko and G. H. MacCullough, Elements of strength of materials, New
York, D. Van Nostrand Company Inc., 1935.
[58] H. Gil and J. A. Yura, "Bracing requirements of inelastic columns", Journal of
Constructional Steel Research, vol. 51, pp. 1-19, 1999.
[59] S. Timoshenko and J. M. Gere, Theory of elastic stability, New York, McGraw-
Hill, 1961.
[60] V. Z. Vlasov, Thin-walled elastic beams, Jerusalem, Israel Program for Scientific
Translations, 1961.
[61] R. D. Ziemian, Guide to stability design criteria for metal structures, 6th ed.
Hoboken, New Jersey, John Wiley & Sons, 2010.
[62] M. G. Salvadori, "Lateral buckling of I-beams", Transactions of the American
Society of Civil Engineers, vol. 120, pp. 1165-1177, 1955.
[63] P. Kirby and D. A. Nethercot, Design for structural stability, Halsted Press, 1979.
[64] D. A. Nethercot and K. C. Rockey, "Unified approach to elastic lateral buckling
of beams", Engineering Journal, American Institute of Steel Construction Inc, vol.
9, pp. 96-107, 1972.
[65] G. Powell and R. Klingner, "Elastic lateral buckling of steel beams", Journal of
Structural Division, 1970.
[66] T. V. Galambos, Guide to stability design criteria for metal structures, 5th ed.,
John Wiley & Sons, 1998.
[67] AISC, "Specification for structural steel buildings", vol. ANSI/ AISC 360-10,
Chicago, Illinois, 2010.
[68] SCI, Stability of steel beams and columns, in accordance with Eurocodes and the
UK National Annexes, SCI P360. UK, 2011.
[69] H. Mehri, R. Crocetti, and P. J. Gustafsson, "Unequally spaced lateral bracings on
compression flanges of steel girders", Structures, vol. 3, DOI:10.1016/j.istruc.
2015.05.003, pp. 236-243, 2015.
[70] S. Kitipornchai and N. S. Trahair, "Buckling properties of monosymmetric I-
beams", Department of Civil Engineering, University of Queensland,1979.
[71] N. S. Trahair, "Inelastic buckling design of monosymmetric I-beams",
Engineering Structures, vol. 34, pp. 564-571, 2012.
[72] Y. L. Pi and N. S. Trahair, "Nonlinear inelastic analysis of steel beam-columns,
I:Theory", Journal of Structural Engineering, vol. 120, pp. 2041-2061, Jul 1994.
[73] EC3, "Design of steel structures –Part 1-1: General rules and rules for buildings",
EN 1993-1-1:2005, ed. Stockholm, Swedish Standards Institute, Sweden, 2005.
122
[74] J. A. Yura, "Bracing for stability, state-of-the-art", in Proceedings of the 13th
Structures Congress. Part 1 (of 2), 3-5 April, Boston, USA, pp. 88-103, 1995.
[75] A. C. Taylor and M. Ojalvo, "Torsional restraint of lateral buckling", J. of the
struct. Div., vol. 92, pp. 115-130, 1966.
[76] EC3, "Design of steel structures–Part 2: Steel bridges", SS-EN 1993-2:2006,
Swedish Standards Institute, Stockholm, Sweden, 2009.
[77] P.-O. Thomasson, "Lateral-torsional instability of box girder bridges at erection",
International Association for Bridge and Structural Engineering (IABSE), pp.
414-415, 2008.
[78] C. Quadrato, W. Wang, A. Battistini, A. Wahr, T. A. Helwig, K. Frank, and M.
Engelhardt, "Cross-frame connection details for skewed steel bridges", The
University of Texas at Austin, FHWA/TX-11/0-5701-1, 2010.
[79] Z. Fan and T. A. Helwig, "Distortional loads and brace forces in steel box
girders", J. of Struct. Eng., vol. 128, pp. 710-718, 2002.
[80] T. Helwig, J. A. Yura, R. Herman, E. Williamson, and D. Li, "Design guidelines
for steel trapezoidal box girder systems", Univ. of Texas at Austin, Center for
Transportation Research, FHWA/TX-07/0-4307-1, 2007.
[81] B. J. Bell and D. G. Linzell, "Erection procedure effects on deformations and
stresses in a large-radius, horizontally curved, I-girder bridge", Journal of Bridge
Engineering, vol. 12, pp. 467-476, 2007.
[82] AASHTO, "AASHTO LRFD Bridge Design Specifications", 7th ed., Washington
D. C., American Association of State Highway and Transportation Officials,
2014.
[83] H. Mehri, R. Crocetti, and J. A. Yura, "Effects of geometric imperfections on
bracing performance of cross-beams during construction of composite bridges",
Engineering Structures, Submitted on 19 July 2015.
[84] SCI, Steel bridge group: guidance notes on best practice in steel bridge
construction, 5th issue, The Steel Construction Instutute, UK, 2010.
[85] H. Mehri and R. Crocetti, "End-warping bracing of steel bridges during
construction", Journal of bridge Engineering (ASCE), submitted on 2nd Dec 2015.
[86] J. A. Yura and J. A. Widianto, "Lateral buckling and bracing of beams, a re-
evaluation after the Marcy Bridge collapse", Structural Stability Research Council
2005 Annual Stability Conference, 6-9th Apr 2005, Montreal, QB, Canada, pp.
277-294, 2005.
[87] M. M. Attard, "General non-dimensional equation for lateral buckling", Thin-
Walled Structures, vol. 9, pp. 417-435, 1990.
[88] K. Kim and C. H. Yoo, "Effects of external bracing on horizontally curved box
girder bridges during construction", Engineering Structures, vol. 28, pp. 1650-
1657, 2006.
[89] T. A. Helwig, O. Egilmez, and C. Jetann, "Lateral bracing of bridge girders by
permanent metal deck forms", Texas Department of Transportation, Austin, 2005.
[90] S. J. Errera and T. V. S. R. Apparao, "Design of I-shaped beams with diaphragm
bracing", vol. 102, pp. 769-781, 1976.
[91] ABAQUS/CAE, V 6.13 [computer software], RI, Simulia, 2013.

123
Paper I
 Nordic Steel Construction Conference 2012
Hotel Bristol, Oslo, Norway
5-7 September 2012

BRACING OF STEEL-CONCRETE COMPOSITE BRIDGE DURING

CASTING OF THE DECK

Hassan Mehri, Roberto Crocetti

Division of Structural Engineering, Lund University, Sweden

Abstract: Trapezoidal cross sections are often used as main longitudinal load-bearing systems in
steel-concrete composite bridges. A critical design stage for these girders occurs during casting of
the bridge deck, when the non-composite steel section must support the entire construction load,
including the wet concrete. A research work was undertaken to study lateral torsional buckling
(LTB) capacity and stiffness requirements of U-shaped girders with focus on discrete torsional
bracings and top lateral truss bracing, under uniform loading condition due to self-weight of
structural system and wet concrete. Findings are then compared with the results obtained by
previous research works and common code specifications.

Keywords: bracing, lateral torsional buckling, stiffness, composite bridges.

1 Introduction
Generally the top flanges of the girders are connected to the concrete deck as a continuous lateral
bracing and the finished work has high torsional stiffness but during erection and construction
period when the deck has not hardened or been attached, compression flanges of built-up steel
girders are susceptible to instability as the girder is as open cross section and therefore relatively
flexible in torsion. Different types of bracing systems are normally used to stabilize bridge girders
against LTB. Typical bracing systems are i) discrete torsional bracing systems and ii) lateral
bracing, in the form of a horizontal truss system. Minimizing the numbers of bracing used, will
lead to a more efficient design since this bracing makes up a significant amount of the total costs.
Moreover, top flange bracing has generally no function once the concrete deck has cured.

2 Background and literature review

Lateral torsional buckling capacity [1] of a simply supported beam subjected to uniform
bending moment about the strong axis is:
π π2 E 2 Iz Cw
M0 = �EIz GJ + (1)
Lb L2b
2 Nordic Steel Construction Conference 2012

Where 𝐿𝑏 =the distance between lateral twisting supports; 𝐸=modulus of elasticity; 𝐺=shear
modulus; 𝐽=torsional constant; 𝐼𝑧 =lateral moment of inertia; 𝐶𝑤 =warping constant. It should
be noted that the Eq. (1) was derived assuming the mentioned load and boundary conditions
for a single, doubly-symmetric beam which is restricted to twist and free to warp at both ends.
To account for the effect of variable moment gradients and load conditions, a modification
factor, 𝐶𝑏 , is typically applied to Eq. (1). When the lateral and the warping end restraints are
unequal or are not free, the following general approximate method [2] may be used:
Cb π�EIz GJ π2 ECw
Mcr = Cb Mo = � 1+ 2 (2)
K z Lb GJ�K L � y b
Where 𝐾𝑧 and 𝐾𝑦 are buckling length coefficients about weak axis, z, and strong axis, y,
respectively. Decreasing 𝐿𝑏 and providing appropriate bracing system, 𝑀𝑐𝑐 will increase. Thus
yielding or local buckling-not global buckling- will control the strength of the girder.
Different types of cross frames or diaphragms are generally used to appropriately decrease the
distortion of the girders by preventing relative twist between the two wings (Fig. 1). For both
twin I-shaped and trapezoidal girders, with an adequate lateral cross-bracing system the
warping stiffness and lateral moment of inertia of the cross section will be high enough to
force two “wings” work together against torsion and lateral bending (Fig. 1a), otherwise the
buckling capacity of the system will be dropped and global lateral buckling of half of the
system about its own warping center is a possibility.

Fig. 1: effectiveness of intermediate cross-bracing stiffness during casting the deck.

The warping constant, 𝐶𝑤 , for a doubly symmetric I-shaped cross section is 𝐼𝑧 (ℎ/2)2 and for
an open top box girder with sloping or vertical webs can be calculated by equations given in
handbooks or other approaches [3]. For both cross sections, the torsion constant is
𝐽 = Σ(𝑏𝑖 𝑡𝑖3 /3) where h, 𝑏𝑖 and 𝑡𝑖 are distance between top and bottom flange centroids, width
and thickness of plates that make up the girder cross section, respectively. Assuming that the
connections between the cross frames and the girder are hinged (so that there is no
“Vierendeel effect”) and the cross frames are stiff enough in their plane, the total torsional
rigidity of twin doubly symmetric, simply supported twin I girders is:
h2 S2
K T = K T,st + K T,w = 2GJ1 + 2E � Iz,1 + Iy,1 � (3)
4 4
Where 𝐾𝑇,𝑠𝑠 =St. Venant rigidity, 𝐾𝑇,𝑤 = total warping rigidity, S= distance between web
centroids of I girders and index 1 indicates related parameter to one girder about the centroid
of each wing, 𝐶1. Substituting 𝐾𝑇,𝑠𝑠 , 𝐾𝑇,𝑤 and 2𝐸𝐼𝑧,1 for the 𝐺𝐺, 𝐸𝐶𝑤 and 𝐸𝐼𝑧 respectively, in
Eq. (2) gives the global buckling moment of doubly symmetric twin I girders [4], 𝑀𝑔 :
2
2πCb πE
Mg = �EIz,1 GJ1 + � 2
� (Iz,1 h2 + Iz,1 Iy,1 S 2 ) (4)
K z Lg 2K y Lb
Nordic Steel Construction Conference 2012 3

Yura [4] has presented a simplified equation retaining only the term 𝐼𝑦,1 𝑆 2 , which is the
dominating term of the Eq. (4). Thus the global lateral buckling capacity is directly
proportional to the girder spacing “S”. In this study, Yura also mentions that a substitution of
𝐼𝑧,1 with effective moment of inertia, 𝐼𝑧1,𝑒𝑒𝑒 , could be directly applied to a mono-symmetric
cross sections and open box girder system with reasonable accuracy:
π2 SE
Mg = Cb 2 �Iy,1 Iz1,eff (5)
Lg
t
Iz1,eff = Iz1,c + Iz1,t (6)
c
where 𝑡/𝑐 is the distance ratio of tension and compression flanges from the centroid of cross;
Iz1,c and Iz1,t are the lateral moment of inertia of the compression and the tension flanges
about weakest axis, respectively. In the current work, the validity of Eq. (5) for LTB of Marcy
Bridge will be also investigated.

3 Case study
The Marcy Bridge in New York was a pedestrian bridge and collapsed during construction,
resulted in nine severely injured workers and one fatality. The bridge, which spanned 51m,
consisted of a trapezoidal girder with concrete deck acting compositely as a top flange. The
concrete deck was 4200mm wide and approximately 200mm thick. The bridge was straight in
plan and arched in elevation and collapsed during casting of the concrete deck. Casting begun at
the abutments and moved to the mid-span. As the pouring reached the mid-span, global lateral
buckling occurred and the girder collapsed twisting off of its supports at the ends, Fig. 2. The
cross sectional dimensions at mid-span and near the supports are shown in Fig. 3. The web
thickness is constant over the entire length, while bottom flange and top flanges thicknesses
varies between 22mm and 28mm. Eight internal k-frames were used in the Marcy Bridge with
no top lateral truss bracing. All cross frame bracing members were L3 × 3 × 3⁄8 (in) angles
with a cross sectional area of 1361𝑚𝑚2 .

Fig. 2: The Marcy bridge after collapse [5]. Fig. 3: Steel cross sectional dimensions of Marcy
Bridge (CS-e: at the ends, CS-m: at the mid span)

4 Mono-symmetric cross sections

The effect of having unsymmetrical cross section (i.e. with different flanges) should be taken
into account in stability analysis especially when the compression flange is the smaller flange.
Eq. (1), (2) and (4) are derived assuming doubly symmetric cross section. The lateral torsional
buckling of mono-symmetric beams can have complicated behaviour depending on boundary
4 Nordic Steel Construction Conference 2012

conditions, type of loading and degree of asymmetry. It must be also noted that for mono-
symmetric beams, where bending is in plane of symmetry, the shear center and the centroid
do not coincide. Most of previous approaches for buckling of mono-symmetric cross sections
can lead to incorrect results for particular cases [6].
An exact formula was presented by Vlasov [7] expressing the elastic buckling moment of
simply supported mono-symmetric beams under uniform moment. Since then, equations for
the elastic critical moment have been developed by various authors [6] with slight
modifications on coefficients and expressions depending on various boundary and loading
conditions. A general formula for the critical bending moment is given by Galambos [2]:
Cb π2 EIz βy 4 Cw K 2z GJ(K z Lb )2
Mcr = �1 ± �1 + � + �� (7)
2(K z Lb )2 β2y Iz K 2y π2 EIz
Where 𝛽𝑦 =the coefficient of asymmetry can be approximated [8] as follow:
2
Iz
βy = 0.9d(2ρ − 1) �1 − � � � (8)
Iy
Where 𝜌 = 𝐼𝑧𝑧 /𝐼𝑧 and d is the depth of cross section. To account the effects of mono-
symmetry of the cross section, other approach given in AASHTO [9] is substituting 2𝐼𝑧𝑧,1 for
𝐼𝑧,1 in which Eq. (1) is used.
The result of an exact solution [8] with approximate coefficient of asymmetry and also an
approximate equation given in AISC LRFD [10] are used in current work to verify the model.

5 Linear eigenvalue buckling analyses

Generally, two types of analysis can be performed to study buckling problems using the finite
element method: 1) Linear eigen-value buckling analysis, and 2) non-linear incremental
buckling analysis. The eigen-value buckling analysis is used in the current study which is
limited to problems where the pre-buckling displacements are relatively small and any
changes in material properties do not significantly affect the assumption of linearity.

5.1 Finite element modelling

The commercial finite element software, SAP 2000 [11], was used for the current research.
Four-node shell elements with sufficient fine meshing were utilized for modelling the cross
section. Shell elements were utilized since a significant portion of the total strain energy of
the deformed state of these structural systems is due to in-plane behaviour of the elements.
Beam elements with 6 degree of freedoms were used to simulate lateral top flange and cross-
frame bracing throughout the girder. In order to avoid “Vierendeel behaviour”, beam elements
were considered as hinged at the both ends.
At the ends of the unbraced length of the girder, the beam is free to warp but prevented to
twist laterally. The girder is free to slide in longitudinal direction at only one end.
A linear elastic isotropic material with E=200GPa and n=0.3 is used for all elements. Linear
eigenvalue buckling analysis has been carried out and the material non-linearity is neglected.
Marcy Bridge was designed as a vertically curved beam with a camber at the mid-span. In this
work the model is assumed initially straight and the benefits of presence of deck formwork
close to the top flanges are also neglected conservatively.
Nordic Steel Construction Conference 2012 5

5.2 Model verification

The results obtained by numerical analysis should always be checked by means of traditional
approaches, engineering judgments and laboratory tests. Eq. (1) for LTB capacity is
conservative in predicting the buckling strength for the unbraced case due to the more
favourable moment gradient, load and support height in practice; factors that are all neglected
by the equation. Other factors which are not considered in general equations for LTB capacity
of a beam are the magnitude and distribution residual stresses and effect of initial
imperfections (loading and geometry), discontinuities in the cross section and slope of the
webs in trapezoidal cross sections. The analytical aspects of determining the critical moment
strength of a beam are quite complex and closed-form solutions exist only for most simple
cases. Global LTB values of Marcy Bridge, assuming constant cross section CS-e, (Fig. 3), is
quantitatively presented with exact and approximate solutions in Table 1.

Table 1: Result of FEM analysis and theoretical solutions (Loads at N.A.)

Analysis type Mcr (MNm) Diff. (%)
FEM 10.4 -
Exact solution, with approximate 𝛽𝑧 [8] 10.0 -3.9
Approximate, AISC LRFD [10] 8.7 -16.9
Eq. (5) 10.3 -1.3

The theoretical solutions are derived for loading with uniform distributed moment while FEM
simulations are derived for girders subjected to uniform distributed load. Therefore theoretical
equations should be adjusted [4] by gradient moment factor, 𝐶𝑏 = 1.12, which is suggested
by other approaches such as [8]. Comparisons between theoretical and FEM results show
good agreement despite that LTB of trapezoidal cross sections is more complicate than
buckling of twin I girders as the shear center is outside of the cross section and the effects of
inclined webs in LTB capacity is ignored in all theoretical solutions. The similar way used for
twin I girders showed more consistent results with general solutions. Supports height in
theoretical formulas is assumed at the Neutral axis level (N.A.) but in reality and in FEM
simulations, supports are placed at bottom layer of the cross section and global buckling
capacity of the girder is slightly increased in this case.

5.3 Non-prismatic cross sections

The flange thickness of the cross-section of the Marcy Bridge varies between 22mm and
28mm at the top flange and between 20mm and 22.5mm at the bottom flange. The effect of
changing the cross section along the girder is normally not considered in traditional lateral
torsional buckling solutions.

Table 2: Effect of Non-prismatic cross sections (Load at N.A.)

Section type Mcr (MNm) Diff. (%)
Constant Cs-e (Fig. 3) 10.4 -0.1
Non-prismatic (Marcy Bridge) 10.4 -
Constant Cs-m (Fig. 3) 10.7 +3.1
6 Nordic Steel Construction Conference 2012

Although, there is no high change in cross sectional dimensions in Marcy Bridge, the results
shown in Table 2 indicate that this effect should be considered in buckling capacity of the
girders when cross-sectional properties are significantly changed. Next investigations are
based on real cross sections of Marcy Bridge.

5.4 Load and support height

The effect of load height must be taken into account for determining the LTB capacity of a
beam. The case where the load acts on the compression flange of simply supported beam is
the most detrimental as it increases the torque arm. Top flange loading reduces buckling
capacity of a single I-shaped beam for mid-height loading by an approximate factor of 1/1.4
whereas tension flange loading improves the buckling capacity by a 1.4 factor. Yura [4] has
achieved that the load height effect can be ignored for doubly symmetric twin I girders, with
single-curvature bending moment gradient and adequate intermediate torsional bracing. The
effect of load height on buckling capacity with respect to different cross frame stiffness for
Marcy Bridge is shown in Fig. 4. It is evident that load height effect cannot be ignored for
trapezoidal girders even though the girder is torsionally braced by means of cross frames with
���b (= Nβ⁄Lb ), where N=number of intermediate cross
large equivalent torsional stiffness, β
frames and βb =torsional stiffness of each cross frame. Top flange loading reduces LTB
capacity of the Marcy Bridge by 0.806 compared to loading at neutral axis (N.A.).

Fig. 4: Effect of load height for Marcy Bridge where Mcr,br and MN.A.,ubr are global LTB capacity
for braced and unbraced case (Load at N.A.); bracing by means of intermediate cross frames

The effect of load height for different web slopes is also investigated where the vertical depth
was held constant while top width was varied and the results are shown in Table 3.

Table 3: Effect of Load height for different web slopes (cross frames as Marcy Bridge, M.B.)
Mcr,ubr (MNm) Mcr,br (MNm) Mcr,Top /Mcr,N.A.
Slope of webs
Top N.A. Top N.A. Unbraced M.B. braced
0° (𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝑤𝑤𝑤𝑤) 6.1 7.6 12.3 15.2 0.806 0.806
6.5° 5.7 7.1 10.0 12.5 0.795 0.806
13° (𝑀𝑀𝑀𝑀𝑀 𝐵𝐵𝐵𝐵𝐵𝐵) 5.1 6.5 8.4 10.4 0.779 0.806
Nordic Steel Construction Conference 2012 7

The results indicate that for trapezoidal girders interconnected with cross frames the load
height effect is not affected with increasing slope of webs if the depth of girder is kept
constant while top flange loading effect increases with sloping of the web for trapezoidal
girders which are not interconnected by means of cross frames. It is also shown that critical
bending moment of these girders decreases with increasing slope of the webs for both cases:
unbraced, Mcr,unbr, and braced , Mcr,br even though 𝐼𝑧 is increasing. Bracing is achieved by
means of intermediate cross frames.

5.5 Equivalent stiffness of internal cross frames as torsional bracing system

For low stiffness of internal cross frames, the wings of the girder system (Fig. 1b) behave
independently, governed by general LTB equations considering half part of the girder.
Increasing the cross-frames stiffness, the girder will still buckle in a single half wave until
buckling occurs between the cross frames. Yielding can also control the capacity of the girder
[12]. It can be derived from Fig. 4 or Fig. 5 that for torsional bracing stiffness higher than the
ideal stiffness, the effectiveness of the cross-frames stiffness is dramatically reduced. The
global buckling of a girder system can also be predicted by considering the cross frames as
continuous torsional bracing along the length of the girder [12]:
2
Cbr EIz β�T
Mg = �Cubr
2 2
Mubr + ≤ Min(Ms , Mp ) (9)
CT
Where 𝛽𝑇̅ = 𝑁𝛽𝑇 /𝐿𝑔 , N=number of intermediate cross frames; 𝑀𝑢𝑢𝑢 =bucking capacity of
unbraced beam; 𝐶𝑢𝑢𝑢 , 𝐶𝑏𝑏 are limiting factors corresponding to an unbraced and an effectively
braced beam, respectively; 𝐶𝑇 is top flange loading modification factor; 𝑀𝑠 and 𝑀𝑝 are the
critical moment corresponding to buckling between the brace points and the plastic strength of
the section, respectively; and 𝛽𝑇 =the effective torsional brace stiffness including the stiffness
of the cross frames, 𝛽𝑏 , the web stiffness, 𝛽𝑠𝑠𝑠 , and girder system stiffness, 𝛽𝑔 , as follow [12]:
1 1 1 1
= + + (10)
βT βb βsec βg
For k-brace systems [12] such as cross-frames which are used in Marcy Bridge, 𝛽𝑏 is:
2Eb2bot h20
βb = 3 Ab (11)
8Lc + b3b
Where bbot =width of bottom flange; h0 =height of cross frame; Lc =length of diagonal
members; and Ab =cross area section of cross-frame members. Although it is shown that the
sum of intermediate cross frames stiffness is the dominating variable and not the actual
number or spacing of the cross frames [4], the moment capacity of girder system drops when
the cross-frame spacing exceeds 0.25Lb [13]. It should be considered that in some cases,
fewer cross-frames are normally used under erection rather than service conditions. The GLB
capacity of Marcy bridge with different number of intermediate cross frames, Mcr,N−br,
compared to unbraced buckling capacity, Mcr,ubr , with respect to equivalent torsional
���b , is shown in Fig. 5. The result indicates the cross frame bracing system used in
stiffness, β
Marcy Bridge has a stiffness which is several times larger than the required torsional stiffness
corresponding to full bracing. It is also shown that providing only one intermediate cross
frame at mid span is more efficient than two cross frames, each at one third of the span, due to
maximum lateral deformation at the mid span. For Marcy Bridge, Iy ⁄Iz > 1.0 where Iz and Iy
are respectively in-plane and out-of-plane moments of inertia, if only the steel cross section is
present. Marcy Bridge is one example proving that, the concept “lateral-torsional buckling of
8 Nordic Steel Construction Conference 2012

the girder cannot occur if the beam is bending about the weaker bending axis” is not valid for
trapezoidal girders without top flange bracing system.

Fig. 5: Equivalent cross frame stiffness of M.B. (Load at N.A.). N=Number of cross frames.

5.6 Top flange bracing of partial or entire length of the girder

Top flange bracing of the U-shaped girder is significantly effective in increasing GLB
strength as it creates a ‘pseudo box’ cross section which has much higher torsional stiffness to
resist system buckling. LTB can be avoided by properly positioned and designed lateral truss
bracing, before dropping off due to combination of lateral deflection and twisting. Partial or entire
top flange lateral bracing is commonly used to ensure that the design moment does not exceed the
LTB capacity. It should be noted that setting a top lateral bracing system for the entire length
of the girder may be too expensive. It has been shown that installing partial top flange bracing
near the supports is more efficient than bracing at the middle of the girder [14] as it converts
lateral supports from hinged to semi-clamped. In this study it is also indicated that using
single diagonal bracing system (Fig. 6) improves GLB strength of the girder in the same
extent as using X-type system if the cross section area of the diagonal member of the single
brace system is equal to the sum of the cross section areas of the X-type bracing. Having high
amount of Lbr /Lb at the ends will increase the GLB of the girder, where Lbr =top flange
braced length of the girder. For three different trapezoidal cross sections with different length
between 45m to 62.5m it is shown [14] that top bracing of those bridges with more than 20%
of the entire length at each end does not have a large impact on the GLB strength. Local
buckling of compression flange or webs can occur rather than global lateral torsional buckling
[15], as the unbraced length decreases.

Fig. 6: Common top flange bracings Fig. 7: Decreasing the unbraced global length of the
girder by providing partial bracing at the ends
Nordic Steel Construction Conference 2012 9

The effect of entire X-type top flange bracing for Marcy Bridge is shown in Fig. 8.

Fig. 8: Effect of entire top flange bracing on GLB capacity of Marcy Bridge, 𝑀𝑐𝑐,𝑏𝑏 .
𝑀𝑐𝑐,𝑢𝑢𝑢 =GLB capacity of Marcy Bridge without top flange bracing; 𝑀𝐸𝐸 =Maximum bending
moment of M.B. due to self-weight of the steel girder and wet concrete deck; and 𝑀𝑅𝑅,𝑏𝑏 =design
moment resistance of M.B. with top flange bracing, with respect to 𝐴𝑏 , According to Eurocode 3.

With such a bracing, the moment capacity of top flange braced girder, Mcr,br , increases
linearly with respect to increasing cross area of top flange bracing bars, 𝐴𝑏 . It can also be
derived that for Marcy Bridge providing top flange truss bracing was a necessity to carry
applied moment, MED , caused by self-weight of steel, formworks and wet concrete. It must be
noted that the effect of imperfections should also be taken into account and more bracing are
needed to reach the resistance moment, MRd,br , according to Eurocode 3 or other codes. It is
shown in Fig. 9 that with a partial top flange bracing, applied over a relatively small part of
span length at each end, LTB capacity of Marcy Bridge significantly increases due to
providing warping end restraint similar to the laterally fixed ends. Maximum global LTB
capacity of the adequately partial top flange braced bridge at the ends can be predicted
considering K z ≈ 0.5 in theoretical equations such as Eq. (7). It is also shown that partial top
flange bracing at the ends is more efficient than at the mid-span.

Fig. 9: comparison between entire, partial ends and partial mid-span top flange bracing.
𝑀𝑐𝑐,𝑡𝑡𝑡−𝑏𝑏 =GLB capacity of Marcy Bridge with top flange bracing; 𝑀𝑐𝑐,𝑡𝑡𝑡−𝑢𝑢𝑢 =GLB capacity of
Marcy Bridge without top flange bracing
10 Nordic Steel Construction Conference 2012

6 Conclusions
The main conclusions are:
1- Load height effect must be considered in design for both braced and unbraced cases.
This effect is constant for efficiently braced trapezoidal cross sections by intermediate
cross frames and decreases with increasing the slopes for unbraced girders. For
trapezoidal girders subjected to uniform vertical load at the top flanges, the
modification factor, 𝐶𝑏 , is suggested as 0.9 (= 0.806 × 1.12) to take the effect of load
height and moment gradient into consideration.
2- Buckling capacity of non-prismatic cross sections cannot be predicted by current
general solutions and the effect of changing properties of the cross section should be
properly taken into account.
3- Intermediate cross frames stiffness for Marcy Bridge was several times higher than
required for full bracing against distortion. It should be noted that these values for
stiffness and partial bracing length are achieved conducting linear buckling analysis of
Marcy Bridge and must be adequately enhanced to account the effects of
imperfections.
4- If the results of an analysis indicate inadequate bending moment capacity, the strength
of the girder can be improved by adding top flange bracing at the 10-20% of the span
near the supports. Providing X-type bracing with relatively small area cross section
(≈ 8mm2 ) moment capacity of Marcy Bridge increases about 28% which was needed
to prevent the girder against lateral torsional buckling during casting the deck.

References
[1] S. Timoshenko and J. M. Gere, Theory of elastic stability. New York: McGraw-Hill, 1961.
[2] T. V. Galambos, Guide to stability design criteria for metal structures, 5th ed.: John Wiley & Sons,
Inc., 1998.
[3] M. Khorasani and S. F. Stiemer, "Design of mono-symmetirc plate and box girders," The university of
British Columbia (Vancouver)2010.
[4] J. Yura, T. Helwig, R. Herman, and C. Zhou, "Global lateral buckling of I-shaped girder systems,"
Journal of Structural Engineering, vol. 134, pp. 1487-1494, 2008.
[5] Pedestrian Bridge Collapse. Available: http://www.exponent.com/pedestrian_bridge_collapse/
[6] I. Balaz and Y. Kolekova, Stability of monosymmetric beams. Amsterdam: Elsevier Science Bv, 1999.
[7] V. Z. Vlasov, Thin-walled elastic beams 2. ed., rev. and augmen. ed. Jerusalem: Israel Program for
Scientific Translations, 1961
[8] J. A. Yura, "Bracing for stability - state-of-the-art," in Proceedings of the 13th Structures Congress.
Part 1 (of 2), April 3, 1995 - April 5, 1995, Boston, MA, USA, 1995, pp. 88-103.
[9] "Standard Specifications for Highway Bridges (17th Edition)," ed: American Association of State
Highway and Transportation Officials (AASHTO), 2002.
[10] AISC, "Specification for structural steel buildings," ed. Chicago, Illinois: American Institute of Steel
Constructions,Chicago, Illinois, 2005.
[11] CSI, "CSI analysis reference manual," ed: Computers and Structures Inc., Berkeley, California, 1995.
[12] J. A. Yura, "Fundamentals of beam bracing," Engineering Journal, vol. 38, pp. 11-26, 2001.
[13] Q. H. Zhao, B. L. Yu, and E. G. Burdette, "Effects of cross-frame on stability of double I-girder system
under erection," Transportation Research Record, pp. 57-62, 2010.
[14] J. A. Yura and J. A. Widianto, "Lateral buckling and bracing of beams - a re-evaluation after the marcy
bridge collapse," in Structural Stability Research Council - 2005 Annual Stability Conference, Apr 6 - 9
2005, Montreal, QB, Canada, 2005, pp. 277-294.
[15] B. H. Choi, Y.-S. Park, and T.-y. Yoon, "Experimental study on the ultimate bending resistance of steel
tub girders with top lateral bracing," Engineering Structures, vol. 30, pp. 3095-3104, 2008.
Paper II
 Structures 3 (2015) 236–243

Contents lists available at ScienceDirect

Structures

journal homepage: http://www.elsevier.com/locate/structures

Unequally spaced lateral bracings on compression flanges of steel girders

Hassan Mehri a,⁎, Roberto Crocetti a, Per Johan Gustafsson b
a
Div. of Structural Engineering, Lund Univ., Box 118, Lund 22100, Sweden
b
Div. of Structural Mechanics, Lund Univ., Box 118, Lund 22100, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: In the bridge sector, lateral bracings can be provided e.g. in the form of metal decks or horizontal truss bracings.
Received 26 February 2015 Those bracings are more efficient at the regions of maximum lateral shear deformations generated from
Received in revised form 6 May 2015 destabilizing forces of compression flanges e.g. near to the twisting supports. A model is presented in this
Accepted 17 May 2015
paper, which relates the lateral buckling length of compression flange of steel girders to their lateral torsional
Available online 2 July 2015
buckling moment and can be used to investigate stiffness requirement of lateral bracings applied on the compres-
Keywords:
sion flanges between the twisting restraints. Analytical solutions were derived for the effects of bracing locations
Lateral bracing and bracing stiffness values on buckling length of compression flanges. Moreover, an exact and a simplified
Stiffness requirement solution for the effect of rotational restraint of shorter-spans on critical load value of the compression members
Steel girder with unequally spanned lateral bracings were derived. The model can be suitable for design engineers to prelim-
inary size the cross-section of beams and lateral bracings.
© 2015 The Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction and background
Relatively little lateral bracings can greatly enhance load carrying ca-
pacity of slender steel columns and beams by limiting their out-of-plane
deformations [15]. However, improper restraint against lateral torsional
buckling can be detrimental. A number of bridge failures have occurred
due to improper lateral bracings. Two examples are: the collapse of The
Marcy Bridge [7] in New York, and Y1504 Bridge [2] in Sweden. These
bridges were designed with trapezoidal cross-sections and both col-
lapsed due to global lateral torsional buckling during concreting of the
deck. No lateral bracings were used in the first case, while stay-in-
place corrugated metal sheets were designated to act as lateral stabiliz-
ing system in the latter bridge. However, for steel bridge applications,
Egilmez et al. [4] showed, both analytically and experimentally, that
corrugated metal decks if properly designed and connected to the
girders, can significantly reduce lateral deformation of steel girders.
Yura et al. [16] derived a simplified expression for global buckling of
steel bridge girders which corresponded to the failure mode for the two
mentioned bridges. Lateral bracing at partial span near to the abutments
can enhance the load carrying capacity of the girders which are prone to
global buckling by creating a semi-clamp condition at the supports [16].
Mehri and Crocetti [7] showed that providing relatively “soft” truss brac-
ings along a partial length of bridge span near to the abutments (e.g. α ¼
0:1 in Fig. 1), the global buckling of The Marcy Bridge could have been
avoided; where α is the partial bridge span from the twisting supports
at both ends which are laterally braced by means of e.g. either truss
⁎ Corresponding author. Tel.: +46 46 222 7397.
E-mail addresses: Hassan.Mehri@kstr.lth.se (H. Mehri), Roberto.Crocetti@kstr.lth.se
(R. Crocetti), per-johan.gustafsson@construction.lth.se (P.J. Gustafsson).
bracings or corrugated metal decks. An example for the application of
the present study can be to estimate critical bending moment value of
such bridges by studying required lateral bracing stiffness near to the
abutments, see Fig. 1, in order to create semi-clamped boundary
conditions.
The exact solutions for bracing requirements and load carrying
capacity of compression members are only possible for simple cases,
with certain boundary and loading conditions. There have been numer-
ous previous studies on bracing requirements of simple beams and
columns [17]. Among these studies, there are investigations on critical
moment value for a variety of loading and bracing conditions, and dif-
ferent cross-sectional properties. In a number of studies, Timoshenko's
energy approach was used to find the optimal locations for bracings of
simple beam structures, see e.g. [11,13,14]. Some of the studies have
led to predicting conservative values for critical loads and stiffness of
bracings that are already included in some code specifications. Finian
et al. [6] studied the stability of imperfect steel beams which were
restrained by means of a number of discrete elastic bracings, and gave
expressions to estimate the magnitudes of bracing forces.
Recommendations for critical moment values basically consist of
applying a number of coefficients that statistically give lower bound
results. For beams for instance, the coefficients are applied to critical
moment, obtained from e.g. Timoshenko's approach [10] to consider
the effects of different conditions such as moment gradient, load height,
cross-sectional symmetries and boundary conditions. Generally, those
recommendations have been presented for a limited number of simple
cases and can also lead to high discrepancies in the results, especially
when a combination of the effects is considered.
Winter [15] presented a model with rigid bars, which are hinged at
the locations of equally spaced transitional springs, to study the lateral

http://dx.doi.org/10.1016/j.istruc.2015.05.003
2352-0124/© 2015 The Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
 H. Mehri et al. / Structures 3 (2015) 236–243
Fig. 1. Global buckling of “narrow” steel girders and the performance of partial-span lateral
bracings.
bracing requirements of columns. Winter's model predicts a minimum
required (or “ideal”) stiffness value for those particular cases, which is
considered to serve equivalent to immovable lateral support. Beyond
this threshold stiffness, any increase in brace stiffness will not enhance
the critical load of the columns. Galambos [5] discussed column cases
with unequal spans. Plaut [8] studied the bracing requirements of col-
umns with single lateral brace at an internal arbitrary point between
the pinned or elastic supports. For the mentioned cases, he also showed
that the “ideal” brace stiffness only exists when the bracing is at the cen-
ter of a uniform column at mid-span. Plaut and Yang [9] also studied the
behavior of pinned-end columns with two intermediate lateral bracings
for two cases: equally spaced bracings, and a case with unequally spaced
bracings at a specific location. For a column with three-span lateral brac-
ings, he stated that “The optimal locations of the internal braces are not ob-
vious unless full bracing is possible”. The present paper discusses the
indications concerning the optimal locations for unequally spanned later-
al bracings which gives the largest critical load for a given brace stiffness
value. The brace in this study were placed symmetrically with respect to
the midspan of the girders which is normally the case in practice.
However, the main purpose of this paper is to analytically investi-
gate the applicability of a proposed simplified model to predict critical
moment of laterally braced girders varying the stiffness and location
of bracings. This approach can greatly benefit design engineers either
before or after performing the final design in order to size the beams
and bracings or to check the results obtained from using commercial
software. For this purpose, exact and approximate expressions were de-
rived to examine the bending restraint effects of shorter spans on criti-
cal load values of beams with unequally spanned lateral bracings.
Typical design curves were also given from which critical load values
can be determined for different stiffness values and location of symmet-
rically placed lateral bracings. A curve was also given for the cases in
which providing lower stiffness than “full bracing” i.e. bracing stiffness
value that serves approximately equal to immovable support is thought
to be adequate in design. Finally, a comprehensive example is presented
to show the applicability of the proposed approach.
2. Results and discussions
2.1. Theory and model development
In this section of the study, Timoshenko's basic approach [10] for
critical moment of simply supported beams under uniform bending
moment was used to develop a model which can be used as a general
solution for beams that are laterally braced between the twisting sup-
ports at the level of their compression flanges. The model relates
Timoshenko's beam subjected to equal and opposite bending moments
at each end to an equivalent couple forces arising in the flanges of the
girders. The model and the derived analytical solutions were verified
by means of Finite Element (FE) analyses using ABAQUS commercial
software [1]. Quadrilateral, reduced integration, four-node shell
element S4R with sufficient meshing was used in the FE analyses to
model the flanges and the web of the studied three dimensional
237
beams, and the plate of the two dimensional compression flanges. S4R
is a common shell element type with four nodes, when modeling of
steel plates of columns, beams, and stiffeners is desired. Efforts were
undertaken to keep the meshing dimensional ratio equal to unity. At
least six elements were used across the width of the flanges. No local
buckling was observed in any of the FE analyses. In the linear buckling
analyses of the compression flanges, the members were fixed at one
end, free to slide in the longitudinal direction at the other end, while
free to warp at both ends. The deformations perpendicular to the
plane of compression flanges were not permitted in the two-
dimensional analyses. The three dimensional beams were fixed at one
end, free to slide in the longitudinal direction at the other end, while
free to warp and restrained against twist at both ends. Bracings were
modeled with linearly elastic transitional springs connecting one node
of the shell elements to the “ground”. The beam components in the per-
formed FE linear buckling analyses had elastic material properties. Gen-
erally speaking, the developed forces in the bracings are relatively little
while a significant increase in the resistance can be normally achieved
with the relatively soft bracings. As a rule of thumb the bracing forces
are normally within the range of 1–2% of the applied force in compres-
sion members. Based on the conditions mentioned above, any local
effects at the junction of the springs with the shell elements − that
might occur in nonlinear analyses with large deformations−were con-
sidered negligible.
The critical bending moment of a simply supported beam with
doubly symmetric cross-section subjected to uniform bending moment
about the strongest axis (y–y) can be calculated using the following
equation [10]:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Mcr ¼ ðπ=Lb Þ EIz GJ þ π 2 E2 Iz C w =L2b ; ð1Þ
where
Lb Beam span between the torsional restraints;
E and G Young's and shear moduli of beam material, respectively;
J, C w , and I z Cross-sectional properties including torsional constant,
warping constant, and moment of inertia about weakest
axis of beam, respectively, (see Fig. 2).
More accurate results can be obtained for I-beams with
monosymmetric cross-section considering Iz ¼ Izc þ ðt=cÞIzt [19];
where t, and c are distances from the neutral bending axis of the
monosymmetric cross-section to the centroids of tension and compres-
sion flanges, respectively; Iz , Izc , and Izt are lateral moments of inertia of
the cross-section, the compression flange, and the tension flange,
respectively, about the weak axis, i.e. “z” axis in Fig. 2. However, for
the purpose of this paper, Iz ≈2I zc can be also used to obtain a reasonably
conservative prediction of critical moment values for the common
I-shaped steel girder dimensions. Substituting C w with Iz h =4 for
2
doubly-symmetric I-beams (where h is the distance between the cen-
troids of the flanges), GJ ¼ C, and EC w ¼ C 1 , Eq. (1) can be rewritten as:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  
Mcr ¼ π2 EIz h=2L2b 1 þ CL2b =π2 C 1 : ð2Þ
Fig. 2. A model that relates critical moment of beams to critical load of their compression
flange.
238
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Setting γ LT ¼ 1 þ CL2b =π2 C 1 ¼ 1 þ 2C=Ncr h , Eq. (2) can be
written as:
Mcr ¼ ðNcr hÞ  γ LT :
Eq. (3) claims that critical moment of I-beams can be approximately
modeled by two opposing horizontal forces, γ LT Ncr , applied at the cen-
troids of the flanges, as shown in Fig. 2; where, the compression flange
of the beams is regarded as a column with critical load of Ncr ¼ π 2 EIzc =L2b .
Table 1 shows the variation of the γLT coefficient with different
values of torsional properties (i.e. CL2b =π2 C 1 ); where “γLT ¼ 1” repre-
sents beams with “zero” torsional stiffness, i.e. the web and the tension
flange give no contribution to controlling lateral torsional deformation
of the compression flange, and “γ LT ¼ ∞” represents beams with “infi-
nite” torsional stiffness.
Theoretically, knowing the buckling length of the compression
flanges, Eq. (3) can also be used to predict the critical moment value
for the cases of laterally restrained beams at the level of compression
flanges. To generalize Eq. (3), the buckling load of compression flange
and γ LT factor can be calculated from Eqs. (4)–(5), respectively.
Ncr ¼ π 2 EIzc =le ;
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
γ LT ¼ 1 þ Cle =π2 C 1 ¼ 1 þ 2C=Ncr h
wherele is the buckling length of the compression flange of the beams,
and in this paper is indicated as “le;∞”, “le;k”, or “Lb” which corresponded to
cases in which the compression flange of the beams are respectively lat-
erally restrained by means of immovable supports, compression flange of
beam is laterally restrained by means of elastic bracings, or the beam is
not laterally braced between the twisting supports.
Lateral torsional buckling of steel girders involves lateral movement
of their compression flange and the twist of the girder cross-section. In
the proposed approach presented in Eq. (3), the lateral translational
degree-of-freedom of the compression flanges is considered by study-
ing the buckling of the compression flanges (see Eq. (4)). While, the tor-
sional degree-of-freedom of the cross-section along the buckling length
of the compression flanges−i.e. “le ” obtained from Eq. (4)−is consid-
ered by the γLT coefficient (see Eq. (5)). It must be noted that in the pro-
posed approach the Vlasov condition [12] was assumed for the shape of
the cross-section of beams that remains unchanged during buckling
analyses. Therefore, no local buckling of cross-section was permitted.
2.1.1. Effects of different loading and boundary conditions
Eq. (3) was derived as a general solution, assuming the same bound-
ary and loading conditions as of Eq. (1), i.e. simply supported beam
subjected to uniform bending moment. There have been numerous
studies on critical moment of simple beams to statistically determine
conservative effects of different conditions such as bending moment
gradients, loading height and cross-sectional asymmetries on critical
moment values obtained by Timoshenko's approach (i.e. Eq. (1)). For
instance, for a simply supported doubly-symmetric unbraced beam
under a uniformly distributed load on the level of the top flange, the
modification coefficient to be applied to Eq. (1) was suggested [17] to
 
be equal to 0:8 ¼ 1:12  1:4−1 . Recommendations are limited to sim-
ple beams and can be often difficult to be justified for some specific
cases in practice when a combination of effects is being considered.
Table 1
Variation of the γ LT coefficient for beams with different torsional properties.
CL2b =C 1 0.0 0.1 10 100
γ LT 1.0 1.0 1.4 3.3
H. Mehri et al. / Structures 3 (2015) 236–243
2 For the purpose of further studies on the proposed approach, the au-
thors suggest an application of buckling length modification factor to ac-
count for the effects of distributed load at the level of top flanges, which
is a typical case in bridge application e.g. during the construction phase.
ð3Þ
2.1.2. Effects of imperfections and nonlinearities
Fig. 3 schematically shows a typical buckling curve recommended by
Eurocode 3 (2005), which is based on results from numerous laboratory
tests. Such curves relate critical compression load,Ncr, or critical bending
moment, M cr , values of columns or beams, respectively, to a reduction
factor, χ , which can be applied to e.g. plastic capacity of the cross-
section in order to calculate the corresponding design values. The effect
of different shapes and magnitudes of initial imperfections, residual
stresses, material nonlinearities, eccentricities, etc. are conservatively
considered in the buckling curves. Thus, using such buckling curves,
the corresponding design values can be calculated for given critical
load values of beams, or columns obtained from Eq. (3), or Eq. (4),
respectively.
Winter's rigid bar model [15] gives the minimum required stiffness
values for columns with equally spaced lateral bracings. For the cases
of unequally spaced bracings, it is proposed that the buckling length
can be conservatively considered equal to the length of the largest
ð4Þ span [18]. However for the unequally braced columns (or compression
flanges), the shorter portions of the beams create some bending re-
straint to the larger spans due to their different bending stiffness values.
ð5Þ In the present study, it is shown that this effect can be significant
depending on the bending stiffness ratio of the adjacent spans with
unequal lengths. Fig. 3 illustrates the possible overdesign problem
which can occur ignoring the mentioned effects; where point “A” repre-
sents the critical value obtained approximating buckling length equal to
the distance between the bracing points, and point “B” represents the
critical value obtained considering the bending restraint effects of the
shorter spans; χ A ; and χ B are design reduction factors corresponding
to the critical load values marked as “A” and “B”, respectively; N P , or
MP are plastic capacities of the column or the beam cross-section; and
λrel is lateral or lateral torsional slenderness ratios for the corresponding
column or beam, respectively.
The model explained in Section 2.1 was examined for beams with:
i) no lateral restraints (presented in Section 2.2), ii) lateral bracings that
served equal to immovable supports (presented in Section 2.3.), and
iii) lateral bracings that had elastic stiffness (presented in Section 2.4.),
all at the location of compression flanges:.
2.2. Laterally unrestrained beams
Empirical recommendations are commonly used by design engi-
neers to preliminary size the cross-section of bridge girders, which are
dependent on the length of girders, the width of their concrete decks,
and etc. Table 2 shows the values of γLT for typical I-shaped built-up
1000 ∞
Fig. 3. Schematic illustration of typical design curves for flexural or lateral torsional
10.1 ∞
buckling.
 H. Mehri et al. / Structures 3 (2015) 236–243
Table 2
Variations of γLT values for different IPE profiles, and common built-up steel girders. “Lb ”
and “d” are length and depth of beams, respectively.
Lb =d ¼ 16 Lb =d ¼ 20
IPE profiles 1.45–1.60 1.60–1.85
Built-up girders 1.40–2.30 1.50–2.60
bridge girders in steel–concrete composite bridges, and also for some
European Standard I-Beam profiles (IPE 100 to IPE 750). In the built-
up girders, the dimensions of cross-sections, i.e. the size of the flanges
and the webs, were calculated utilizing the commonly used empirical
methods for 108 beam cases (varying girder depth from 600 mm to
3000 mm, in 300 mm increments; deck width from 3000 mm to
12000 mm, in 3000 mm increments, and length from 9600 mm to
72000 mm). As can be seen in Table 2, typical γLT values vary within
the range of approximately 1.4 to 2.8.
In addition, Eq. (3) was used to calculate γ LT values for beam cases
(1)–(5) as shown in Fig. 4 varying span lengths between 6 and 36 m;
the critical bending moments were obtained using FE buckling analyses.
In Fig. 5 γ LT values obtained are plotted against CL2b =C 1 values. The
results were also compared to the curve obtained from Eq. (5). The
comparison verified that Eq. (5) gives a unique curve for both doubly-
and mono-symmetric beams with the assumptions made earlier.
2.3. Beams with immovable lateral restraints
In general, for many of the cases encountered in practice, much
higher stiffness values than the “ideal” stiffness are normally provided
to resist lateral deformations of compression members at the brace
points. The effect of unequal spanning of lateral bracings on critical
load values of compression flange of beams are investigated in this
section, assuming that lateral stiffness values of the bracings are large
enough to function fairly similar to immovable supports. Fig. 6 illus-
trates the central-span of the compression flange of a beam which is
laterally restrained with immovable supports at a αl distance from
either end, as an equivalent column with rotational spring stiffness of
βα EI=l at both ends; where EI is the bending stiffness of the cross-section
about the strong axis of compression flange, l is the length of the central-
span, and βα varies depending on α values.
For beam–column “BC” in Fig. 6, the general differential equation is:
EIw‴x þ Nw″x ¼ 0
where wx is the lateral deflection of the beam–columns “AB” and
“BC”.
Fig. 4. The studied beam cases: (1)–(5), each with varying lengths = 6, 12, 24 and 36 m.
239
Lb =d ¼ 24
1.80–2.15
1.60–2.80
Fig. 5. Values of γ LT for the beam cases: (1)–(5).
Applying boundary conditions ðwx2¼0 ¼ wx2¼l ¼ 0; EIw″x2¼0 ¼
EIw″x2¼l ¼ ðβα EI=lÞθÞ for beam–column “BC” gives:
cl ¼ −βα sinðclÞ=½1 þ cosðclÞ; ð7Þ
where c2 ¼ N=EI.
Eq. (7) has no close-form solution and should be solved numerically
for given values of βα . An analytical solution for the relationship
between α and βα was of interest.
For the beam–column “AB” in Fig. 6, applying boundary conditions
 
wx1¼0 ¼ wx1¼αl ¼ 0; EIw″x1¼0 ¼ 0; EIw″x1¼αl ¼ ðβα EI=lÞθ , and dividing
0
moment forceð βα EI=lÞθ by the slope at point B, wx1¼αl , gives the rota-
tional stiffness at point B:
2
βα ¼ ð3=α ÞðcαlÞ sinðcαlÞ=3½ð sinðcαlÞ−ðcαlÞ cosðcαlÞ; ð8Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
where cαl ¼ π N=Ncr , and N cr is the buckling load of the beam–
column “AB”.
Eq. (8) gives an exact solution for the magnitude of bending restraint
of the side-spans to the central-span. Substituting the resulted value of
βα from Eq. (8) into Eq. (7) gives the buckling length of the compres-
sion member, i.e. le;∞ , shown in Fig. 6. However, a simplified solution
for the relationship between α and βα , rather than Eq. (8) can be also
derived. The following relationship between α and βα can be written,
when second-order effects are neglected:
ð6Þ
βα ¼ 3=α: ð9Þ
Eq. (9) can be modified by applying the amplification factor η ¼
1=ð1−N=N cr Þ to consider the second-order effects of the compression
2
force on the deformations; where N and Ncr are approximately π 2 EI=l
Fig. 6. Bending restraint of side-spans with shorter length to the central-span of compres-
sion flange restrained by means of immovable lateral supports.
240 H. Mehri et al. / Structures 3 (2015) 236–243

2 Table 3
andπ2 EI=α 2 l , respectively. Thus, the relationship between βα andα can
Buckling length of compression flange with immovable restraints placed at partial length
be approximated by the following equation to consider the second- (αl).
order effects:
α ≈0.0a 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
  βα ∞ 28.2 13.6 8.22 5.37 3.58 2.36 1.48 0.85 0.37 0.0
βα ≈3=α 1−α 2 : ð10Þ
cl ≈2π 5.88 5.51 5.16 4.82 4.49 4.17 3.88 3.61 3.36 π
le;∞ =l ≈0.50 0.53 0.57 0.61 0.65 0.70 0.75 0.81 0.87 0.93 1.00
Comparisons between the results of Eqs. (8)–(10) showed that the
le;∞ =Lb ≈0.50 0.45 0.41 0.38 0.36 0.35 0.34 0.34 0.34 0.33 0.33
approximate solution obtained from Eq. (10) was in excellent agree-
a
ment with the exact solution obtained from Eq. (8) (see Fig. 7). There- This value represented lateral restraints at a distance which was “very” close to the
twisting supports but not equal to “zero”.
fore, by substituting the βα values obtained from Eq. (10) (for a given
value of α) into Eq. (7), the buckling load of the laterally restrained
compression flange can be calculated when lateral stiffness of the
braces is sufficiently large. Setting the determinant of the stiffness matrix to zero gives an exact
Table 3 shows the results of Eq. (7) for a number of βα values between solution for the normalized relationship between bracing stiffness and
“zero stiffness”, which represents nil rotational stiffness provided by the corresponding critical load values as in the following equation:
side-spans to the central-span, and “infinite stiffness” which means that
3 3
the side-spans act as a fully-clamped support to the central-span. kl =EI ¼ −ξðclÞ ½ cosðcαlÞ  cotð0:5clÞ− sinðcαlÞ=½ξcαl  sinðcαlÞ þ cotð0:5clÞ:
To summarize, for different positions of the lateral supports (αl), ð12Þ
critical moment values of beams can be calculated by substituting le;∞
obtained from Table 3 into Eqs. (3)–(5). Table 3 showed significant Similarly, for the asymmetric mode shape, applying boundary
 
decreases in the buckling length ratios, i.e. le;∞ =l, of the compression conditions wx1¼0 ¼ EIw″x1¼0 ¼ 0; wx1¼αl ¼ δ; w0x1¼αl ¼ θ for side-
flange in the central-span, especially for the small values of α. In bridge span beam, i.e. beam–column “AB”, and ðwx2¼0 ¼ δ; w0x2¼0 ¼
design practice, it is often assumed that the buckling length of the
θ; wx2¼0:5l ¼ EIw″x2¼0:5l ¼ 0Þ for the internal-span beam, i.e. beam–
braced girder is conservatively equal to the distance between the brac-
column “BC”, Eq. (13) gives a normalized solution between the bracing
ing points, if adequate stiffness larger than the recommended value is
stiffness and the corresponding critical load. For a given lateral stiffness
provided for the bracings [3]. The results in Table 3 show that this as-
value, critical load is the minimum value obtained from symmetric and
sumption can lead to a considerable over-design problem (it can be as
2 2 asymmetric calculations using Eqs. (12)–(13).
large as a factor of 4, see l =le;∞ for small values of α).
3 3
kl =EI ¼ −ξϑðclÞ ½ðα þ 0:5Þ  sinðcαl þ 0:5clÞ=½αξ  sinðcαlÞ þ 0:5ϑ  sinð0:5clÞ
2.4. Beams with elastic lateral restraints
ð13Þ

For practical uses, the method should be expanded for different

where ϑ ¼ 1=½ sinð0:5clÞ−0:5cl  cosð0:5clÞ.
stiffness values of lateral bracings. Bracings were represented as linearly
A typical design curve for the relationship between lateral bracings
elastic translational springs at symmetric locations with respect to the
stiffness and buckling length of the central-span for different locations
mid-span of the beams, which is mostly the cases in practice (see
of lateral bracings is shown in Fig. 9, as the results of linear FE buckling
Fig. 8). Depending on the magnitude of the lateral bracing stiffness,
analyses. Comparison between the FE results and the results of Eq. (12)
compression flanges of beams can deform in either symmetric or asym-
are made in Fig. 10; where the solid lines represented the FE results (for
metric shape with respect to the center-span (see Fig. 8).
kLb =N0 values between 0.0 and 200.0), and the dashed lines represent-
For the symmetric deformation shape, applying boundary conditions
  ed the results of Eq. (12), (for kLb =N0 values between 0.0 and 800.0);
wx1¼0 ¼ EIw″x1¼0 ¼ 0; wx1¼αl ¼ δ; w0x1¼αl ¼ θ for the side-span beam,
  and N0 ¼ π2 EI=L2b . The results obtained from Eq. (12) were expanded
i.e. beam “AB”, and wx2¼0 ¼ δ; w0x2¼0 ¼ θ; w0x2¼0:5l ¼ EIw‴x2¼0:5l ¼ 0 for
for more values of stiffness than the FE data to show the “excellent”
the internal-span beam, i.e. beam “BC”, Eq. (6) gives:
agreement between the two methods. The results showed that the rel-
 

atively small values of stiffness for symmetrically placed lateral brac-
ξc3  cosðcαlÞ þ k=EI −ξc3 αl  cosðcαlÞ−c2 δ 0
¼ ings, can significantly decrease the buckling length of the compression
ξc2  sinðcαlÞ −ξc2 lα  sinðcαlÞ−c  cotð0:5clÞ θ 0
ð11Þ flange, especially for the cases where α ≥0:25 (see Fig. 9). For the case
of α ¼ 1:0; the points indicated with (i) and (ii) in Fig. 9 represented
where ξ ¼ 1=½ sinðcαlÞ−cαl  cosðcαlÞ.
minimum bracing stiffness values (kLb =N0 being 13.2 and 81, respec-
tively) to change the buckling shapes from a single half-sine wave to

Fig. 7. Bending restraint of side-spans with shorter length to buckling length of central-
span, obtained from Eqs. (8)–(10). Fig. 8. Compression flange of beams with symmetrically placed lateral elastic bracings.
 H. Mehri et al. / Structures 3 (2015) 236–243
Fig. 9. Relationship between buckling length of the compression member shown in Fig. 8
and stiffness of lateral bracings.
Fig. 10. The curves showing ratios between critical loads of compression member shown
in Fig. 8 and critical load of the same member assuming the buckling length to be equal to
the distance between the bracing points.
two and three half-sine waves, respectively, where three half-sine wave
corresponds to the “ideal” bracing stiffness of the column [15]. These
thresholds values for columns with equally spaced lateral bracings i.e.,
Fig. 11. Investigations on shifting from a symmetric buckling mode shape, obtained from Eq. (12), to an asymmetric buckling mode shape, obtained from Eq. (13), for the compression
flanges shown in Fig. 8.
241
α ¼ 1:0, can be calculated by means of Winter's rigid bars model that
are hinged at the locations of braces.
Critical load values of the beam shown in Fig. 8 was normalized to
critical load values when − ignoring bending restraint of the side
spans−the buckling length is “conservatively” assumed to be equal to
2 2
the distance between the bracing points, i.e. l =le;k in Fig. 10. The results
shown in Fig. 10 verified that this assumption gives unsafe results for
relatively small values of brace stiffness for all locations of bracing. On
the other hand, providing relatively soft bracings, e.g. as little as
kLb =N0 ¼ 100 , significantly enhanced the critical load values. This
capacity is currently ignored in practice when the buckling length is
assumed to be equal to “l” [3].
Fig. 11 graphs the results obtained from Eqs. (12)–(13) for differ-
ent brace stiffness values and brace locations of the compression
flange shown in Fig. 8. Comparisons between the two curves showed
that shifting from a symmetric to asymmetric buckling mode shape
will not occur for small values of α (e.g. values between 0.0 and
0.6) for any stiffness values of the lateral bracings prior to buckling
of the internal-span. Fig. 12 shows the results of Eqs. (12)–(13) for
α ¼ 0:7 in which the FE results were also included. Obviously, the
2 2
minimum values for the critical load ratios (i.e. l =le;k ) of the internal-
span obtained from Eqs. (12)–(13) were in excellent agreement with
the FE results.
Fig. 13 also depicts required bracing stiffness values for the cases
if slightly larger buckling length than the corresponding to equiva-
lent to immovable supports, le;∞ , was desired in design. As it can be
concluded in Fig. 13, to approach fairly close to the largest possible
critical load of compression flange, a relatively large bracing stiffness
was required for the cases with small values of α which might not be
economical for particular cases in practice. For this reason, Fig. 13 en-
ables design engineers to examine the use of lower stiffness values to
achieve slightly lower load carrying capacity than corresponding to
“full bracing”.
3. Conclusions
Advanced commercial software packages are widely relied on to
model complex structures considering nonlinearities, imperfections,
etc. in bridge design practice. The results of such numerical analyses
should be checked by other approaches. Simplified models can greatly
242 H. Mehri et al. / Structures 3 (2015) 236–243

Acknowledgments

The financial support from “The Lars Erik Lundbergs

Stipendiestiftelse (Dnr 7/2013 and Dnr 2014/05)” to this study is
gratefully acknowledged.

Appendix A. Solved example.

Determine critical moment, M cr , for the beams shown in Fig. 14. The
length of beams is 60.0 m between the twisting supports for both cross-
sections (S1: doubly-symmetric, and S2: mono-symmetric). The value
of α varies between 0.0 (unbraced beam), 0.1, 0.25, and 1.0 (equally
spaced lateral bracings); and stiffness of lateral bracings ðkLb =N0 Þ varies
between 25, 100, and ∞ (immovable lateral support).
Fig. 12. Comparisons between the results of Eqs. (12)–(13) and FE analyses for α ¼ 0:7.

help design engineers to either preliminary size the bridge girders and
their bracings, or check the FE results. A model was discussed in this
paper which can be used to calculate critical moment values of laterally
restrained beams at the level of their compression flanges for given
values of bracing stiffness. The model related buckling length of com-
pression flange of steel girders to their lateral torsional critical moment.
For this purpose, exact analytical solutions were derived to consider the
effects of bending restraints on critical load values, created due to un-
equal spanning of bracings. This effect has been neglected in practice
when the buckling length of compression members is assumed to be
equal to the largest distance between the bracing points. The presented Fig. 14. Bridge girders and their cross-sectional properties for the solved example.
paper showed that this assumption can give unsafe results for small
bracing stiffness values and can also lead to significant overdesign prob-
Solution:
lem even when relatively soft bracings are used.
Typical design curves were also given as the results of analytical so-
lutions for unequally spanned lateral bracings which give required lat-   
eral bracing stiffness to achieve the largest possible critical load N0 ¼ π2 2  105 50  6003 =12 =600002  10−6 ¼ 0:49 ½MN
values. One solved example examined the applicability of the approach
for 26 cases of mono-symmetric and doubly-symmetric beams, with i) For laterally braced cases with kLb =N0 ¼ 25:
different lateral bracing stiffness values and at different locations. The
results obtained using the model compared very favorably with those
from FE analyses. Besides, the time to perform the analyses using the ap-
proach was comparatively short. In addition, the model demonstrated
high potential to help in developing and understanding the theory of M cr , Eq. (3)
le;k =Lb N cr ½MN  γ LT , Eq. (5) FEA (%)
beam and column bracings. α ½MNm
Fig. 6 Eq. (4)
S1 S2 S1 S2 S1 S2

N. B.a 1.00 0.49 1.95 2.11 2.41 2.60 +0.8 −1.1

0.1 0.79 0.80 1.66 1.77 3.31 3.53 +4.2 +3.1
0.25 0.57 1.53 1.38 1.45 5.30 5.57 +4.0 +1.6
0.5 0.43 2.63 1.24 1.28 8.13 8.44 +1.7 −0.2
1.0 0.42 2.87 1.22 1.26 8.73 9.04 +2.5⁎ +1.4⁎
a
N.B.: No lateral bracing.
⁎ refers to a two-half-sine wave shape lateral torsional bucking mode of beam, as ob-
served in FEA.

ii) For laterally braced cases with kLb =N0 ¼ 100:

γ LT M cr , Eq. (3)
le;k =Lb N cr ½MN  FEA (%)
α Eq. (5) ½MNm
Fig. 6 Eq. (4)
S1 S2 S1 S2 S1 S2

0.1 0.59 1.41 1.41 1.48 4.98 5.24 +5.4 +0.6

0.25 0.43 2.67 1.23 1.28 8.23 8.53 +2.9 +0.2
0.5 0.36 3.85 1.17 1.20 11.23 11.56 +2.8 +0.6
1.0 0.33 4.45 1.15 1.18 12.75 13.07 +3.8⁎⁎ +1.4⁎⁎
Fig. 13. Lateral bracing stiffness requirements versus bracing location, when less than full ⁎⁎ refers to a three-half-sine wave shape of lateral torsional buckling mode of beam, as
 
bracing le ¼ le;∞ is desired in design. observed in FEA.
 H. Mehri et al. / Structures 3 (2015) 236–243 243

iii) For laterally braced beam cases with kLb =N0 ¼ ∞: [4] Egilmez O, Helwig T, et al. Buckling behavior of steel bridge I-girders braced by
permanent metal deck forms. J Bridge Eng 2012;17(4):624–33.
[5] Galambos TV. Guide to stability design criteria for metal structures. 5th ed. John
Wiley & Sons; 1998.
γ LT M cr , Eq. (3) [6] McCann F, Wadee MA, et al. Lateral stability of imperfect discretely braced steel
le;∞ =Lb Ncr ½MN FEA (%)
α Eq. (5) ½MNm beams. J Eng Mech 2013;139(10):1341–9.
Eqs. (7)&(12) Eq. (4) [7] Mehri H, Crocetti R. Bracing of steel-concrete composite bridge during casting of the
S1 S2 S1 S2 S1 S2
deck. Nordic Steel Construction Conf. 2012. Oslo, Norway; 2012.
0.1 0.44 2.55 1.24 1.29 7.92 8.23 +5.1 +2.6 [8] Plaut R. Requirements for lateral bracing of columns with two spans. J Struct Eng
0.25 0.39 3.24 1.19 1.23 9.69 10.01 +4.4 +1.3 1993;119(10):2913–31.
0.5 0.35 4.03 1.16 1.9 11.68 12.00 +0.6** −0.1** [9] Plaut RH, Yang Y-W. Behavior of three-span braced columns with equal or unequal
spans. J Struct Eng 1995;121(6):986–94.
1.0 0.33 4.53 1.14 1.17 12.95 13.28 +5.4** +2.9**
[10] Timoshenko S, Gere JM. Theory of elastic stability. New York: McGraw-Hill; 1961.
* and ** refer to 2nd and 3rd (i.e. two and three half-sine wave shapes, respectively) modes [11] Trahair N, Nethercot D. Bracing requirements in thin-walled structures. Developments
of lateral torsional buckling of beam, as observed in FEA. in Thin-Walled StructuresElsevier Applied Science Publishers; 1984. p. 93–130.
[12] Vlasov VZ. Thin-walled elastic beams. Jerusalem: Israel program for scientific
translations; 1961.
[13] Wang CM, Goh CJ, et al. An energy approach to elastic stability analysis of multiply
It is evident that there are only slight discrepancies between the
braced monosymmetric I-beams. Mech Struct Mach 1989;17(4):415–29.
results obtained from the proposed approach and the results from FE [14] Wang CM, Kitipornchai S, et al. Buckling of braced monosymmetric cantilevers. Int J
analyses, whereas FEA demands significantly more time than the pro- Mech Sci 1987;29(5):321–37.
posed approach (which can be done using e.g. an “Excel” sheet for all [15] Winter G. Lateral bracing of columns and beams. ASCE 1960;125(1):807–26.
[16] Yura J, Helwig T, et al. Global lateral buckling of I-shaped girder systems. J Struct Eng
the calculations). 2008;134(9):1487–94.
[17] Yura JA. Bracing for stability, state-of-the-art. Proceedings of the 13th Structures
References Congress. Part 1 (of 2), April 3, 1995–April 5, 1995, Boston, MA, USAASCE; 1995.
[18] Yura JA. Winter's bracing approach revisited. Eng Struct 1996;18(10):821–5.
[1] ABAQUS/CAE. V 6.13 [computer software]. RI: Simulia; 2013. [19] Yura JA. Fundamentals of beam bracing. Eng J 2001:11–26.
[2] Ålenius M. Finite element modelling of composite bridge stability; 2003.
[3] EC3. Design of steel structures — Part 1-1: general rules and rules for buildings. EN
1993–1–1:2005. Stockholm, Sweden: Swedish Standards Institute; 2005.
Paper III
 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 1 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

1 Scaffolding Bracing of Composite Bridges during

21 Construction
2 3 Hassan Mehri1 and Roberto Crocetti2

4 Abstract: Relatively little bracing of steel girders can significantly reduce the possibility of the risk of buckling failures,es, before
fore the composite
co
5 action between steel and concrete occurs. This paper intended, both experimentally and numerically, to study the torsional bracing bra perform-
perf
6 ance of a typical type of scaffoldings commonly used in bridge construction. Minor improvements in the structure ree of
ture o the scaffoldings
scaffold as well
w

7 as at their connections with steel girders were needed for this purpose, which are discussed in this paper. Inn the numerical
numemerical investigations,
mer vestiga the
8 effects of different initial imperfections on bracing properties of the scaffoldings were also investigated. The study was as extended
eex d to different
di
d
9 bridge lateral-torsional buckling slenderness ratios and geometries. Results showed that the proposed system greatly increased ncreased the load-
inc
incr
10 carrying capacity of the studied bridges with consideration of lateral-torsional buckling during the construction
onstru on phase. Brace forces
constru orc created in
fo
for
11 the scaffolding trusses were also measured in both the experimental and numerical investigations. ons. Finally,
ally, indications
in cations of bracebra moment ver-
12 sus in-plane moment values for different magnitudes of initial imperfections and lateral-torsional
-torsional buckling slenderness
nderne ratios are given.
DOI: 10.1061/(ASCE)BE.1943-5592.0000829. © 2015 American Society of Civil Engineers. ers.


13 Author keywords: Scaffold; Bracing; Composite bridge; Lateral-torsional buckling;
ing; Construction.
ling; Constr
Const ction.

14 Background and Problem Description  bridge

ridgee was lost by removing
remov only one secondary element (i.e., the 39
cross
ss beam).
beam 40
15 During the construction of composite bridges, the bare steel girders Mostst of the Swedish steel-concrete composite bridges are of 41
16 are prone to instability problems prior to composite action (i.e., e., the twin-girders
girders type. Discrete torsional bracings with a spacing of 42
17 when the concrete deck is not yet hardened) (Mehri and Crocetti ocetti 4–8
4 m in thee form of cross beam/diaphragm/frame are typically 43
18 2012). The girders may require different bracing systems, depend- depen adopted
ad
adop to brace the two main girders. Because of the relatively 44
19 ing on different boundary and loading conditions of the construction
con ruction l
large d
distance between the braces, the sizes of compression flanges 45
20 t
need at times to be increased to enhance the lateral-torsional buck- 46

processes. During launching for instance, the cantilever ever parts
ppart have
21 different boundary conditions than, for example, when the bridge bridg is re
ling resistance of bare steel girders. However, after the concrete 47
22 hoisted. Moreover, although the stresses are relatively atively small during
latively durin d
deck is hardened, the top flanges are needed primarily to provide 48
23 the lifting stage, they might be large enough gh to cause
c significant
significant
cant sufficient space for the shear studs used for composite action. The 49
24 insuffficient (Stithh et size of compression flanges could be reduced if beneficial bracing 50


deformations if the bracings and restraints ts are insufficient

insu
25 al. 2012). During concreting of the deck, ck, the girders
ders carry primarily
g rders effects from scaffolding could be taken into account. 51
26 the self-weight of concrete and steel. l. However, dynamic
dynamic effects andan More than 70 bridge accidents associated with timber scaffold- 52
27 possible load eccentricities of casting
ting concrete may also
sting a so occur.
occur ing failures, including two bridges in Sweden and Denmark in 2008 53
28 Many bridge failures have ve been reported (e.g., (e.g Scheer
Sche and and 2014, respectively, have been identified since 1970 in 19 coun- 54
29 55


Wilharm 2011), which occurredcurrredd in the construction phase ph during tries (Beale et al. 2012), mostly because of insufficient bracing.
30 lifting, launching, and, particularly,
particu rly,
particular ly, concreting of the deck.
d Among Thus, considering the mentioned problems and the fact that the in- 56
31 the causes for failure, a number involved
iinvoolved stability problems
pr in either stallation of the commonly used scaffoldings can cost up to 20% of 57
32 bridge or scaffolding
ing members. Providing
Providding
P ing more robust
r systems dur- the total construction budget of composite bridges, more efficient 58
33 ing the construction
ction of bridges might avoid avvoid possible failures that scaffoldings should be considered to also ensure the safety of the 59
34 60


can occur, for

or example,
e ample, because of human errors or lack of knowl- concreting process of composite bridges.
35 edge. In the case ofo the State
Sta Route 69 Bridge over the Tennessee
36 River, in 1996, two of thehe three
thre spans
span collapsed into the river when
37 one of the cross beams
beam was
w removed
remo
remov to allow the fixing of a connec- Introduction and Theory 61
38 tion (Zhao
(Zhao et al. 2009).
2009). This
20 Thi indicates that the stability of the entire
For the purpose of concreting a deck, two different types of scaf- 62
folding systems are currently used in Sweden: timber-frame and so- 63
called cup-lock scaffolds (Figs. 1 and 2). This paper proposes the 64
1
Ph.D. Candidate,
idate Division of Structural Engineering, Lund Univ.,
idate use of an improved generation of cup-lock scaffoldings (described 65
Box 118, Lund 22100, Sweden (corresponding author). E-mail: hassan in the Materials and Methods section) that can be installed on the 66
.mehri@kstr.lth.se bare steel girders prior to lifting or launching. The scaffoldings can 67
2
Professor, Division of Structural Engineering, Lund Univ., Box 118, 68
be used not only for a safe casting process, but also for stabilizing
Lund 22100, Sweden. E-mail: roberto.crocetti@kstr.lth.se 69
Note. This manuscript was submitted on August 1, 2014; approved on
the steel section of bridges by bracing them against out-of-plane
deformations. 70
June 29, 2015; published online on nnnnnnnn. Discussion period
open until nnnnnnnn; separate discussions must be submitted for Fig. 3 schematically shows the loading transfer mechanism 71
individual papers. This paper is part of the Journal of Bridge between the steel girders and the scaffolds for the two different 72
Engineering, © ASCE, ISSN 1084-0702. types of scaffolds. The efficiency of the typical cup-lock scaffolds 73

© ASCE 040nnnnn-1 J. Bridge Eng.

ID: hawalh Time: 07:13 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 2 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

Fig. 4. Stabilizing performance of the internal part of the scaffoldings F4 : 1

F1 : 1 Fig. 1. Cup-lock scaffoldings installed between steel girders in com-

F1 : 2 posite bridge (image courtesy of Britek AB)

Fig. 5. Cup-lock joint

jo details
detail
detail F5 : 1


also between
ween the girgirders near the top flanges (see Tcan and Tbar ,
ers nea 83
respectively,
ctively, in Fig. 33).
). The
he stee rods are normally fastened to the
steel ro 84
webeb of the girders and aaree pl placed approximately every 2.0 m 85
 between
etween
een the girders. It m must be mentioned that, in practice, some 86
bridges
ges
es have experienced lo local bending of the webs at the anchoring 87
points of the steel rods. Therefore, an alternative connection is also 88
of interest even for ththe preliminary function of the scaffolds, which 89
are carrying the
ar construction loads.
he co 90
The so-called cup-lock scaffolding system consists of a series of
Th 91
trusses iinterconnected by ledgers at cup-lock joints (Figs. 4 and 5). 92

Cup-lock joints provide fairly moment-stiff connections between 93
ledgers and bars. The cross sections of the bars and ledgers are
the le 94
currently identical in different projects. The scaffolding bars and
cu
cur 95
ledgers are f 48.3-mm steel pipes with 3-mm constant thickness in 96


different lengths. The trusses are normally placed every 1,200– 97

F2 : 1 Fig. 2. Cup-lock scaffoldings installedd at cantilevers
cantilevers
levers of composite
1,300 mm sitting on the bottom flanges of main girders. The cantile- 98
bridge (image courtesy of Britek AB)
F2 : 2 vers that also sit on the bottom flanges of the main girders are anch- 99
ored to the top flanges of the main girders by steel rods (see dashed 100
lines in Fig. 4). Fig. 4 illustrates the transferring mechanism of the 101


destabilizing forces of compressive flanges, F, between adjacent 102

bridge girders, where h ¼ depth of the girder cross section; D and 103
Dh are lateral movements of the top and bottom flanges, respec- 104
tively; u is the twist of the cross section; F1 –F4 are brace forces cre- 105
ated in the scaffolding bars; VA and VB are overturning and lifting 106


reactions generated from the destabilizing moment MT ; and S is the 107

transverse span. The cantilever parts carry only gravity loads (VC is 108
the vertical reaction of the cantilever parts under gravity loads) and 109
F3 : 1 Fig.. 3. Girders subj
subjected too construction
const
cons loads exerted by: (a) typical 110
do not contribute to the stabilizing action.
F3 : 2 cup-lock
lock scaffolds; (b) typical
typica timber-frame scaffolds 111
Using relatively little bracing can greatly increase the load-car-
rying capacity of compressive members (Winter 1960). Different 112
types of bracing are currently used in steel bridges (e.g., plan brac- 113
74 [Fig. 3(a)] is significantly
igni
igni reduced when the depth of the girders, ings or discrete torsional bracings). Generally speaking, different 114
75 and consequently the inclination of the bars, is lowered. In such a brace types give lateral and torsional restraints to the steel girders 115
76 case, timber-made rigid frames, instead of the cup-lock scaffolds, (Fig. 6). The structural system described in Fig. 4, with minor 116
77 are normally used with a spacing of 900–1,200 mm along the bridge improvements in its connections, is able to serve mainly as a contin- 117
78 span [Fig. 3(b)]. Moreover, for the relatively large depth of the steel uous torsional brace to the steel girders. The general approach for 118
79 girders, both the scaffolding spacing and the length of the cantile- the critical-moment prediction of torsionally braced beams (Taylor 119
80 vers can be extended. To transfer the tension forces generated by and Ojalvo 1966; Galambos 1998) can be used to obtain the critical- 120
81 the bending moments due to the presence of the cantilevers under moment value of the bridge girders, which are braced by the scaf- 121
82 construction loads, steel rods are used in the cantilever parts and foldings. Consequently, knowing the torsional stiffness of the 122

© ASCE 040nnnnn-2 J. Bridge Eng.

ID: hawalh Time: 07:13 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 3 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

F6 : 1 Fig. 6. Lateral-torsional deformation of steel bridge girders under con-

F6 : 2 struction load

123 scaffoldings (e.g., approximated as a k-frame) and the critical- Fig. 7. Boundary conditions
nditions off the
th test setup
tup F7 : 1
124 moment value, the design moment can be achieved using the code
125 specifications. More studies are required to give design recommen-
126 dations for the stiffness and the strength requirements of the scaf-
127 folding braces, which can be performed in future studies.
128 Using the torsional bracing potential of the scaffolding trusses
129 (Fig. 4), the twist of the girder cross section can be considerably


130 resisted. A proper lateral bracing system can serve as a complement
131 to the torsional bracing potential of the scaffolding trusses. In the
132 cup-lock scaffolding, a large number of ledgers (Fig. 5) are typically
133 placed between the trusses to brace the truss members in the out-of-
134 plane direction. The ledgers can contribute to an increase in the lat-
135 eral stiffness of the scaffolding system by creating a sort of
136 Vierendeel beam.
137 Providing adequate bracing is even more crucial for horizontally ally
138 curved bridges during construction, especially when the radius dius of
o
139 curvature tends to be small. In such a circumstance, the cross oss sec-
se
140 tion of the bridge may be subjected to large torsional nal moments
m ments

141 under construction loads (Fan and Helwig 2002;; Helwig eet al.
142 2007). In addition, if the scaffoldings are installed led girders
ed on main girde
143 before the erection of the bridge, they can brace ce the steel girders not
144 only during the concreting phase, but also o during,
uring, for example, Fig. 8. Strain gauges glued on the scaffolding bars


145 F8 : 1
launching or lifting processes. The other advantage
er adva ntage
age of the proposed
146 system is that both top and bottom flanges anges in popositive
sitive
itive and negative
negativ
147 bending moment regions, respectively, ctively, can be sst stabilized
abilized by the the girders. After the base plates and top chords of the scaffoldings 167
148 same system. This advantagee is not available for all brace types were fastened to the girders, the scaffolding bars were placed 168
149 (e.g., for lateral bracing in the form of corrugated
corrugate m metal decks between the hinged joints at both of their ends. By turning the 169


150 placed at the top flange level)

evel)) ((Egilmez
E ilmez et al. 2012
Egilmez 2012). must be noted
). It m threaded jacks (Figs. 7 and 8), the lengths of the bars could be 170
151 that the scaffoldings are already be being
eing used for the ppurpose of con- adjusted between the base plates and hinged joints attached to 171
152 creting the deck. Thus, possible rises in the bud budget of construction
budg the timber chord. Vertical and lateral deflections of both girders 172
153 due to use of thee scaffolding as a brac system are expected to be
bracingg syst were recorded with 14 potentiometers placed at different locations 173
154 negligible. The aimim of this paper was to pprovide indications of the along the beams (S1–S7 for the south girder and N1–N7 for the 174


155 bracing performan

performancee of the scaffoldings and the magnitudes of
th scaffoldin north girder). Fig. 8 shows the strain gauges glued on the outermost 175
156 brace forces created iin the sca
scaffolding members from the results of
scaffoldin top and bottom layers of the scaffolding bars to measure strains due 176
157 experimental
erimental and nunumerical
numeri investigations.
al inve
inves to both axial forces and possible bending in the plane of trusses. 177
In total, 32 strain gauges were glued on the bars of the four scaffold- 178
ing trusses: FR1, FR3, FR5, and FR7 (Fig. 9). The timber chord 179
158 als and
Materials an Method
Methods members of the scaffolding were double 45  120 mm2 (C24) 180
(Fig. 8), which is typical Swedish practice. 181
159 The results of two laboratory tests are presented in this paper: 182
Test Setup
TN1 (tested in the elastic range), where the two beams of the 183
160 Fig. 7 shows the boundary conditions at either end of the main specimens were braced by only one cross beam at midspan, and 184
161 beams of the test setup. The supports of the specimens were TN11 (tested until failure), where the main beams were braced by 185
162 designed to resist twisting of the cross section and to allow free the scaffoldings plus a cross beam at midspan. Fig. 9 schemati- 186
163 warping. Each of the girders was sitting on two bearings at the ends cally illustrates the setup for Test TN11, where the locations of 187
164 (free to slide at one end, but fixed at the other end), which were potentiometers are marked for both the south (S) and north (N) 188
165 welded to the rigid beams anchored to the reinforced floor. Efforts girders. The setup represented a portion of real bridges between 189
166 were performed to eliminate any possible tilt and misalignments of three adjacent cross beams. The two I-beams of the tested spec- 190

© ASCE 040nnnnn-3 J. Bridge Eng.

ID: hawalh Time: 07:14 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 4 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

F9 : 1 Fig. 9. Test setup and locations of potentiometers and strain gauges in Test TN11

191 imen were identical, with material properties of S355 (i.e., yielding
192 stress fy ¼ 355 MPa) and cross-sectional dimensions of 100
193 12 mm2 , 750  8 mm2 , and 150  15 mm2 for the top flanges,


194 webs, and bottom flanges, respectively. The bridge had longitudinal
195 and transversal spans of 15 and 2 m, respectively, building up a sys-
196 tem with a slenderness ratio for lateral-torsional buckling (i.e.,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
197 λLT ¼ MP =Mcr ) of 3.44, where MP is the plastic-moment 
198 capacity of the girder cross section; and Mcr is the critical moment
199 of the girders when they are braced only by the cross beam at mid-
200 span (i.e., Test TN1) obtained from linear eigenvalue analysis. Thee
201 cross beam at midspan consisted of a 2UPE (back-to-back) ack)
202 European standard U-profile, which was connected to thee mai main Fig 10. Locations
Fig.
F io of improvements on typical cup-lock scaffoldings F10 : 1
203 girders by four M12 high-strength bolts at each end. Loads ds were
we to func
function as a torsional bracing system F10 : 2
204 applied using a stiff H-shaped frame ð5:0  2:5 mÞ (Fig. (F g 7).7).

205 Between the loading frame and the girders, fourr shaft bearings bear gs
206 ( f 50 mm with 100-kN capacity) were installedd to reduce friction frictio
207 as much as possible. By this method, both girders rs were able to indi-
rders
208 vidually move in lateral directions with no bracing cing restraints from


209 the loading apparatus. Twelve discrete scaffolds FR1–FR122 in

scaffo dss (see FR1–FR12
210 Fig. 9) were attached to the main girders with with either 1,200- oro
211 1,300-mm spacing (as is commonn practice in Sweden). Swede
Swed n).
n)
212 In common practice, the cup-lock up-lock
lock scaffoldings arear not currently
cur
213 fastened to the steel girders. Moreover,reover, the scaffolding
scaffoldin bars
b are not


214 able to transfer tensile forces.rces. Minor

M nor changes were needed
need
ne to enable
215 the cup-lock scaffolding ng trusses (Fig.
(FFig. 44)) to also function
func
fun as a suita-
216 ble bracing system. m. The improvements
improvemennts involved mainly providing
improv
217 proper connections ons between the pr proposed
propos scaffoldings and main
sed sc
218 girders (i.e., Fi
Fig. 10, D1–D2) and introducing a mechanism
10, Details D1–D2)
D1–


219 that enableses the scaffolding

s affolding members to t resist tensile forces (i.e.,
220 Fig. 10,, Detail D3).D3) Simple
imple bolted connections
c between the scaf-
221 foldings
ings and girders were wer designed
design for to perform Test TN11 (Figs.
designe
222 10–12).
12).
12). The scaffolds
scaffol were bolted bol to the webs sitting on the bottom
223 flangess and were connected to the top flanges using T-shape pro-
224 files. In addition,
ddition, the improvement
ddition imp (illustrated as Detail D3 in Fig. Fig. 11. T-shaped connection between scaffolds and top flanges used F11 : 1
225 10) was made de simply
sim by using small spot welds between the jacks in Test TN11 (Detail D1 in Fig. 10) F11 : 2
226 and lower pipes. s. More
M proper improvements can be suggested for
227 the bridge-scaffolding market to suit different projects and to enable
3228 quick installations. However, this was outside the scope of this
229 paper. with semi-ideally hinged ends. Because of the mentioned reasons 235
230 At least three single-direction strain gauges are required for each and the fact that the scaffolding trusses were subjected mainly to 236
231 bar to separate the effect of axial force and bending moment in each torsion in the plane of trusses, only two strain gauges, rather than 237
232 member. However, bending in the bars was considered negligible three, were used to determine the axial force in the truss members. 238
233 owing to the use of relatively stocky bars (the length of the bars in The strain gauges were placed at the uppermost and bottommost 239
234 the test setup was shorter than approximately 1.2 m) in combination layers of each tubular cross section of the scaffolding bars at their 240

© ASCE 040nnnnn-4 J. Bridge Eng.

ID: hawalh Time: 07:14 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 5 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

241 midlength (Fig. 10). Mean values of the recorded strain pairs were Linear Eigenvalue Analyses of the Test-Setup Bridge 266
242 used to calculate the axial forces in the bars.
Table 1pshows the values of the lateral-torsional slenderness ratio, 267
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
λLT ¼ MP =Mcr , for different bracing configurations of the bridge, 268
243 Numerical Models where MP is the plastic-bending capacity of the girder cross section; 269
and Mcr is the critical moment of the bridge with different bracing 270
244 Abaqus/CAE 6.13 commercial software was used in the numerical
configuration obtained by linear eigenvalue analyses. 271
245 investigations. To introduce the initial geometric imperfections to
The results presented in Table 1 show that the scaffoldings 272
246 the girders, the buckling shapes obtained from linear eigenvalue
greatly increased the critical-moment value of the girders. More- 273
247 analyses were imported into the third-order incremental analyses.
over, the results show that relatively few stiffening effects were 274
248 Two different shapes and two different magnitudes of initial twists
achieved when a combination of the attached scaffolds ffo with a com-
ffold 275
249 were examined in this study, whereas the bottom flanges were 276
250 mon spacing of 1,250 mm and the commonly nly used
use cross
cr beams
always straight. The main beams and the cross beam were modeled 277
251 (i.e., a bracing span within the range of 4–8 8 m) was However,
as uused. Ho
using S4R shell elements with sufficient meshing, whereas the B31 278
252 an increase in the critical load of approximately ximately 27% was achi achieved
beam element was used to model both the steel and timber scaffold- 279
253 when a single cross beam was added att midspan m
mi to the scaffoldings
scaffoldin

ing members. Bending moments were released at both ends of the 280
254 spaced every 2,500 mm (i.e., representing ng the
resenting tth situation
ation in practice
prac
bars and timber beams. The effects of material nonlinearities were 281
255 where every second scaffold iss fastened fast to the
hee main
m girders).
rders Thus,
girders)
also taken into account in the third-order incremental analyses. The 282
256 utilizing the bracing potential of th scaffolds,
the scaffolds ds, the nnumber of
lds,
steel grade used was S355 with multilinear material properties 283
257 required permanent torsional onal bbracings
ional ings may be redureduced.
uce This may
duc
(stress values of 355, 360, 539, and 571 MPa corresponding to 284
258 lead to a reduction in steel consum consumption possibly to improved
ion and possib
strain values of 0.2, 1.8, 4.9, and 11.3%, respectively). The timber 285
fatigue performance ce of bridges.
259 beams of the top chord of the scaffolds were modeled with isotropic
260 material properties and a longitudinal elastic modulus of 11,000


261 Bracing Performance
Performance of the Temporary
Tempo Scaffoldings 286
MPa.
Fig. 13 shows that the scaffoldings
scaffol enhanced the lateral-torsional 287
buckling
ing resistance of the
kling th test-setup
est-se
est-setu bridge by a factor of 3.7 com- 288
262 Results and Discussions pared
aredd with the system w with only one cross beam at midspan (i.e., 289

263
264
265
The results of the experimental and numerical investigations
regarding the stiffness and strength of the proposed scaffolds are
presented in the following sections.
 comparison
mparison
arison between the re
result veri
tem, which
folding of composit
achi
b
ch can be achieved
results of Tests TN11 and TN1). This
fied the great bracing
verified
verifi performance of the proposed sys-
with minor changes in the current scaf-
composite bridges. In addition, the results showed that
290
291
292
293




F12 : 1 Fig.. 12. Connections

ction between
bet een
een scaffolds
sc
sca and bottom flanges of girders Fig. 13. Comparison between results of Tests TN1 and TN11 and FE F13 : 1

(Detail
ail D2 in Fig. 1
10) analyses of Tests TN11 and TN1 F13 : 2
F12 : 2

Table 1. Results
ts of Eigenvalue
Ei
E Analyses for Different Bracing Configurations of the Test Setup
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T1 : 1 Bracing configuration of test setup Mcr/MP λLT ¼ MP =Mcr
T1 : 2 Bare girders 0.03 5.77
T1 : 3 Only one cross beam at midspan 0.08 3.44
T1 : 4 Scaffolds applied every 2,500 mm without intermediate cross beam beambebeam 0.44 1.51
T1 : 5 Scaffolds applied every 2,500 mm with cross beam at midspan 0.56 1.34
T1 : 6 Scaffolds applied every 1,250 mm without intermediate cross beam 0.72 1.18
T1 : 7 Scaffolds applied every 1,250 mm with cross beam at midspan 0.72 1.18

© ASCE 040nnnnn-5 J. Bridge Eng.

ID: hawalh Time: 07:14 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 6 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

294 the bridge was able to tolerate relatively large out-of-plane defor- point loading as the loading condition in the test setup and uni- 357
295 mations. Ductility is normally considered to be an advantage that formly distributed loading applied on the timber chord of the scaf- 358
4296 provides a warning before a possible structural collapse. Therefore, foldings because it is the loading condition in practice, respec- 359
297 using the bracing potential of the scaffoldings, bridges may demon- tively). In Figs. 14 and 15, Scd and Crs stand for scaffolding and 360
298 strate relatively large out-of-plane deformations rather than a sud- cross beam. Comparisons showed that the load-twist curves were 361
299 den failure, which was often the case in the reported bridge collap- fairly identical for both loading conditions. Moreover, a combina- 362
300 ses that occurred during the concreting of decks. This can be tion of cross beams at midspan and the scaffolding (i.e., Setup 1 363
301 important to reduce fatalities and can lead to a safer failure in a case cases) could not significantly improve any stiffness or resistance 364
302 of possible collapse. values of the test-setup bridge, which was braced only by scaffold- 365
303 The shape of the initial imperfection can have significant influ- ing (i.e., Setup 2 cases). In addition, attaching every second scaf- 366
304 ence on the load-carrying capacity of compression members. Such folding also considerably increased the resistance cee of the bridge (i.e., 367
305 a capacity is generally the lowest when the same shape as the first Setup 3 cases). This can be a solution when a limitedmited enh
enhancement 368
306 buckling mode shape of the braced system is initially introduced to of resistance is desired in practice. 369
307 the girders. However, in practice, beams can have very different ffoldings was also
The bracing performance of the scaffoldings a examined
exami 370
308 on two new systems, which were modeled odeled
led
edd by
b duplicating
duplicatin the beams
beam 371

and complex initial imperfection in terms of both shapes and magni-

309 tudes than the critical ones (which are typically recommended by of the test-setup bridge with thee same cros ross
oss sections,
cross ns, as foll
follows: 372
310 codes). This was also true for the tested beams, which were initially Setup 4 was a single-span bridge,ge, an
and Setup 5 w was
wa a two-equ
o-equ
two-equal-span 373
pp
bridge with an internal support. Both Setup 4 and nd 5 brid
bridges had a 374
311 twisted with maximum values of u 0 ¼ Lb =3; 300h and u 0 ¼
total length of 30 m with h three intermediate
ermediate cross be ea
beams, resulting 375
312 Lb =1; 150h for the north and south girders, respectively, with asym-
313 metric distributions of geometric imperfections. The maximum val- ral-tors nal buckling, λLT , of 3.43 and
in slenderness ratios for lateral-tors
lateral-torsional 376
314 ues of the measured geometric imperfections were introduced to the
315 girders, and the results, which were in good agreement with the


316 results of tests considering the assumptions for material properties
317 of the timber beams of scaffoldings and ignoring the complex varia-
318 tion of geometric imperfections along the girders, are also shown in
319 Fig. 13. The data were also compared with the results of the numeri-
320 cal models for two initial twist values and two different initial
321 imperfection shapes, namely Case 1, where both top flanges had
322 sweeps in the same direction with maximum magnitude at midspan
323 and the bottom flanges were straight, and Case 2, where the similar ilar
324 initial imperfection as in Case 1 was introduced but with maximum ximum
325 magnitudes at different locations. Fig. 13 shows that Case 1, with wit a
326 maximum magnitude of u 0 =D0/h = Lb/500h [i.e., the recommended com ended

327 imperfection as determined by Eurocode 3 (SIS 2009)], 09)],
09 )], led tto the
328 more conservative values of resistance and out-of-plane of-plane deform
deforma-
329 perfections
tion compared with the three other initial imperfectionsrfections illustrated
330 in Fig. 13, where u 0 and D0 are the initial twistt of the girder cross


331 section and the initial lateral sweep of the top flanges,
he to anges, respectively;
332 Lb is the span length between the twisting wisting supports;
supports; and h iss the
su
333 depth of the girders. Although a study udy on the effects
effect of the shapes of Fig. 14. Three different scaffolding setups under four-point bending F14 : 1
334 initial imperfections (i.e., further herr than in Cases 1 an andd 2) on both load condition F14 : 2
335 brace forces and the resistance ce was of interest, this was
w outside the


336 scope of this paper. However, wever,, the worst imperfection

imperfect shape that
337 gives the most conservativevative resistance
resisttance is theoretically
res theoretical expected to
theoretica
338 be in the form of thee fi first
rst bucklin
buckling mo ode shape of the braced bridge,
mode
339 which was Casee 1 for the test-setup
test b
bridge. The recommended
340 shape for initial twist given by Eurocode 3 (i.e.,(i Case 1 with a maxi-


5341 mum magnitude o Lb =

nitude of =500d]
500d
500 d] was usedd iin the further finite-element
342 (FE) investigations.
vestigations.
343 Itt was of interes
interest to investigate
nvestig the bracing performance of the
nvestigate
344 scaffolding
ffolding when onlyo every
eve y second
se truss was fastened to the main
345 girders,s, as well aas when a rreduction in the number of intermediate
346 cross beamsams was
wa desired in practice. The following three different
347 confi
con figurat
g
scaffolding configurations were investigated numerically in the
348 men
mens
same bridge dimensions for these purposes: Setup 1, where the scaf-
349 foldings were installed every 1,200–1,300 mm plus a cross beam at
350 midspan; Setup 2, where the intermediate cross beam at midspan
351 was removed from Setup 1; and Setup 3, where the scaffoldings
352 were installed every 2,500 mm instead plus a cross beam at mid-
353 span. The investigations were carried out for two different loading
354 conditions, where the initial imperfection Case 1 with two different Fig. 15. Three different scaffolding setups under uniformly distributed F15 : 1
355 magnitudes was introduced to the girders. The results are shown in load applied on scaffolding (representing typical wet concrete loads)
356 Figs. 14 and 15 for the two different loading conditions (i.e., four- F15 : 2

© ASCE 040nnnnn-6 J. Bridge Eng.

ID: hawalh Time: 07:14 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 7 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

377 2.64, respectively. Although the cross-sectional dimensions of the bars were identical to the previous cases (i.e., Setups 1–5). The 394
378 bridge in Setup 4 were unrealistic, the bracing performance of the two main steel beams of the model were identical and had a 395
379 scaffolding on different span-to-depth ratios was of interest. A uni- depth of 2,500 mm ðLb =h ¼ 18Þ with top and bottom flanges of 396
380 formly distributed load was applied on the main girders, and the 500  35 and 800  45 mm2 , respectively, and a constant web 397
381 beams were given an initial twist of u 0 ¼ Lb =500h, as in Case 1. thickness of 20 mm. The bridge was torsionally braced by five in- 398
382 Fig. 16 shows that the scaffoldings enhanced the load-carrying termediate cross beams (each of them double UPE400 back-to- 399
383 capacity of the beams braced with cross beams, Mcrs, by approxi- back), which were placed every 7.5 m along the bridge span. 400
384 mately a factor of 4.0 for both cases; MScdþCrs is the load-carrying The scaffoldings enhanced the resistance of the bridge by factors 401
385 capacity of the bridges that were braced by both the scaffolding of 1.63 and 1.40, respectively, when material properties of S460 402
386 spanning 1,250 mm and the three intermediate cross beams. and S355 were considered for the girder cross sections. ct
cti It must be 403
387 Further investigation of the different slenderness ratios and noted that the curves in Fig. 17 were normalizedd to different
diffe
d magni- 404 6
388 transverse spans was required. The scaffoldings were installed ev- 405
tudes of plastic-moment capacities. This means ns that
t eeither the
389 ery 2,500 mm (i.e., every second scaffold is bolted to girders in 406
dashed lines (corresponding to the bridges es that werere braced
b on by
only
390 practice) on a single-span bridge model with a 45-m longitudinal 407
the intermediate cross beams) or the solid soolid lines (corresponding
sol (corr to
391 span and a 7-m transverse span (Setup 6), resulting in a slenderness 408

the bridges that were braced by both th thee ccross

cro s beams andan the scscaf-
392 ratio for lateral-torsional buckling of 1.16 (Fig. 17). The cross- 409
foldings) had the same initial stiffness
tiffness if the
tif hee curves
c ddepicted
were dep
393 sectional properties of both the timber chords and scaffolding 410
for M values instead.
The beams reached their ir ultimate
ul ate resistance because
becaause of
ecau
au o the yield- 411
ing of their compressionon fl flanges
anges caused warping and in-plane
aused by warpin 412
bending. Modern techniques
chniques enable the production of high-strength 413
steels with relatively
vely lower costs, and consequently,
nsequ more slender 414
beams can be dedesigned
gned for steel bridges. T Therefore, the scaffoldings 415


can be even ficient when the cros
n more eefficient cross section of the main gird- 416
ers can tolerate higher stress
ress va
values bebefore yielding. 417

 Brace Scaffolding Bars

ce Forces in the Scaf
S 418
FR1,, FR3, FR5, and FR7 ((Fig. 9) were chosen in the experimental 419
study to brace forces in the scaffolding bars at differ-
o measure the br 420
ent positions.
ns. Fig
ns Fig. 18 shows the test results for the measured axial 421
forces in the bars at maximum load level [u  u 0 ¼ 0:005 ðradÞ],
fo 422
co
corre
corresponding to an M=Mp value of approximately 0.45 (Fig. 13). 423
The axial
axi forces were calculated from the recorded strain values 424

by two strain
s gauges glued on each bar (32 gauges in total). The 425
results
result showed that relatively little axial force was generated in 426
the
th scaffolding bars, especially at the time of ultimate load, 427
whereas a very large improvement in the resistance value was 428


F16 : 1 Fig. 16. Resistance ratio versus twist ratio

io at midspan
m span for Setup 4 and achieved. The ultimate tensile axial capacity of the bars was 429
13
F16 : 2 5 bridges
approximately 100 kN. 430
According to the test data presented in Fig. 18, brace forces 431
were smaller near the twisting supports (e.g., in FR1) and were 432
larger close to the location of the maximum twist of the bridge 433


cross section as a single unit (i.e., in FR5 and FR7). The cross 434

F17 : 1 Fig. 17. Bracing performance of scaffoldings when applied every

F17 : 2 2,500 mm along the bridge span on the Setup 6 bridge with two differ- Fig. 18. Recorded axial forces in scaffolding bars at maximum load F18 : 1

F17 : 3 ent steel types point during Test TN11 F18 : 2

© ASCE 040nnnnn-7 J. Bridge Eng.

ID: hawalh Time: 07:15 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 8 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

Fig. 20. Uplifting and overturning

ning reactions
r at maximum
m m load
lo level F20 : 1
for Setup 1 and 2 bridges F20 : 2
F19 : 1 Fig. 19. Failure of timber chord of Truss FR8 after load-carrying
F19 : 2 capacity was achieved


435 beam at midspan contributed to the resistance of the twist of the
436 bridge and consequently reduced the brace forces in the neighbor-
437 ing truss (i.e., FR7). As a result, some of the bars in FR5 showed
438 larger brace forces than those in FR7. Fig. 19 shows the failure of
439 the timber beam in Truss FR8, which occurred due to large out-
440 of-plane deformations after the maximum load-carrying capacity
441 was achieved. With respect to the center of the bridge, Truss FR8
442 had a symmetric location in relation to Truss FR5 (Fig. 9), which hich
443 had strain gauges glued to the bars and experienced the largest larges
7444 measured forces.
445 The distribution of brace moment values generatedd in the t e scaf-

446 folding trusses along the bridge span was of interest. st.
t. Destabilizing
Destabil
Destabi ing
447 moments at the location of each scaffold (see MT ¼ jVA orr VB j  S
448 in Fig. 4) can be obtained by knowing the uplifting ifting
ing and overturning
449 reaction values, VA and VB , which can be calculated
alculated
lated from the brace


450 forces in the scaffolding members. Fig. 20 shows hows

ws the distributions Fig. 21. Test and FE results for brace moment ratios in different F21 : 1
451 of calculated values for the uplifting ng and ovove rturning reactions,
overturning reactions trusses along test-setup bridge span F21 : 2
452 VA and VB , in Setups 1 and 2 under four-point loading condi- con
453 tions and at two different initial al imperfection magnitudes.
magnitu
magni des. Th The val-
454 ues were symmetric with respect spectct to the zero-load axis
a b
because it different trusses along the bridge spans directly depend on the corre- 476


455 was expected. The same trend was was observed when the th loading
l was sponding twist values of the bridge cross section at the location of 477
456 oldings instead,
applied on the scaffoldings insteead, with an exception
in exce
exc that the the trusses. The data depicted in Fig. 21 correspond to failure load 478
457 curves were shiftedd toward the compression
commpression axis.
axi This shift was level in the FE analyses, whereas the test data corresponded to the 479
458 obviously due too the contribution
contributi of the t scaffolding
sca bars to the maximum load point [u  u 0 ¼ 0:005 ðradÞ and maximum twist 480
459 applied gravity y loads.
oads. level [u  u 0 ¼ 0:0174 ðradÞ ¼ 1 . 481


460 rength
ength rrequirement
The strength quirement of any ki kind of bracing is normally The distributions of brace moment ratios along the spans for the 482
461 defined as a percent
percentil l
percentile of the applied load. The M=Mbr ratios for the Setup 4–6 bridges (with their different geometries and slenderness 483
462 two loading conditi
conditions (i.e.,
.e., four-point
four-
fou loading on the girders and ratios) are presented in Fig. 22. Similar to the Setup 1 bridge, the 484
463 tributed loading on the scaffoldings)
distributed s affo of the Setup 1 and 2 bridges brace moment distributions were locally disrupted at the locations 485
464 he test-setu
(i.e., the test-setup bridge w with or without a cross beam at midspan) of the cross beams in the Setup 4–6 bridges due to the contribution 486
465 nted in Fig. 21, where
are presented w M is the bending moment applied to of the cross beams to resisting against twist. It must be noted that 487
466 one girder; and Mbr b = MT is the corresponding brace moment in the although identical cross sections were used for the scaffolding bars 488
467 scaffolding used d to stabilize both girders. Approximately 2 and 4% and the timber chords in all Setup 1–6 bridges, the torsional stiffness 489
468 values for the brace moment ratios were obtained when the initial values of the trusses used in Setup 6 differed from the others due to 490
469 twist values of Lb =2000h and Lb =500h, respectively, were intro- different bridge depths and transverse spans. The results gave an in- 491
470 duced to the girders. Moreover, the presence of the cross beam at dication of maximum brace moment magnitudes, which were 492
471 midspan locally disturbed the distribution of the brace forces in the between approximately 2 and 4% of the maximum bending 493
472 scaffolding bars of the Setup 1 bridge for both loading conditions. moments in the main girders. 494
473 The brace moment ratios calculated from the test measurements (on Additional studies in the form of laboratory tests and numerical 495
474 Trusses FR1, FR3, FR5, and FR7) are also presented in Fig. 21. It investigations are needed to recommend stiffness and strength 496
475 must be noted that the magnitudes of the brace moments in the requirement criteria for a variety of bridge cases. Moreover, the 497

© ASCE 040nnnnn-8 J. Bridge Eng.

ID: hawalh Time: 07:15 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 9 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

Fig. 24. Scaffoldings applied in numerical modell of Y1504

Y Bridge F24 : 1

Table 2. Results of Eigenvalue Analysess off the

th Y1504 Bridge
Bridg (Maximum
(Maximu

Applied Bending Moment under Self-Weight

Self-Weig
Weighight of Wet Concrete
C =
28:4 MN  mÞ
F22 : 1 Fig. 22. FE results for brace moment ratios at different locations of
F22 : 2 trusses along spans for Setups 4–6 bridges corresponding to their maxi- 04 Bridge
Bracing configuration of Y1504 dge Mcr ðMN  mÞ T2 : 1

F22 : 3 mum load-carrying capacity Braced by only shear diaphragms

phragms
ms 11:4 T2 : 2
ear diaphragms
Scaffolds added to shear ms 11:4 T2 : 3


the conventional
ntional bracings
acings or the proposed
propo scaffolds, would not 523
improvee the resistance
resistanc of the bridge
b to system buckling. 524

 Conclusions
clu 525

onstruction
nstruction of com
The construction comp
composite bridges is a delicate operation, dur- 526
ing which h a number of accidents have occurred. This paper pre- 527
ssented the results
esults oof experimental and numerical investigations on 528
the ttorsional bracing performance of currently used scaffoldings in
th 529
F23 : 1 Fig. 23. System buckling of composite bridges during
ur g deck
dec the con
concreting of composite bridges. Minor structural improve- 530
14 ments in the scaffoldings, particularly at their connections with 531

F23 : 2 concreting
main ggirders, were briefly presented. The findings showed signifi- 532
c bracing performances of the proposed scaffoldings when they
cant 533
498 brace moment of the scaffoldings and consequently
uently the generated
equently wwere installed on different bridge dimensions with different slen- 534
499 forces in the scaffolding bars can be conservatively
tively approximated
rvatively 535


derness ratios for lateral-torsional buckling. This paper intended to

500 assuming a maximum expected twist wist value [e.g ðu  u 0 ÞÞ==
valuee [e.g., provide indications of load-carrying capacity and brace moment 536
501 u 0 ¼ 1]. values of the studied cases, which were representatives of real 537
bridges with different configurations. The scaffoldings enhanced 538
load-carrying capacity by factors of approximately 4.0 and 1.5 for 539
502 stem Buckling
Limitations against System Bu 540


the studied cases with slenderness ratios for lateral-torsional buck-

503 ling of 3.44 and 1.16, respectively. Moreover, the brace moments 541
torssional
to
Increasing the stiffness of the torsional ional bracings between
betw
bet the twist-
generated in the scaffolding trusses were in the range of 2–6% of 542
504 ing supports would not, in general,
gene ficantly enhance
significantly
signifi
signi e the criti-
505 the maximum in-plane bending moment in the main girders. The 543
cal-moment value bridges with system buckling problems
ue of the bridge
brace moments in the scaffolding trusses were highly dependent on 544
506 (Yura et al. 2008) 8) (Fig
(Fig. 23). confirm
fi
). This was confirmed
confi by two bridge fail-
545


507 ures, namely lyy the Y1504

1504 Bridge in Sweden
Swede and the Marcy Bridge in the magnitude of the introduced initial imperfections in the FE anal-
yses. Considering the fact that using the bracing potential of scaf- 546
508 the United
ted States. Fig.
F g 24 schematically
schematic demonstrates the cross-
foldings will not raise the construction costs, the bridge industry 547
509 sectional
onal dimensions
dimension of o the Y1504
Y1 Bridge, which collapsed due to
can greatly benefit from the bracing potential of the scaffoldings to 548
510 system
tem buckling, twisting
tw about
bout its longitudinal axis during the con-
enhance both the load-carrying capacity and the safety of bridges 549
511 cretingg of the deck.
dec Nine intermediate-shear
in diaphragms were des-
during construction. However, more studies in the form of labora- 550
512 ignated to serveser as discrete
dis torsional bracings in this bridge.
tory tests and numerical modeling are required to document the 551
513 Moreover, corrugated
corruga metalm decks were also designed to act as con-
stiffness and strength requirement criteria for a variety of bridge 552
514 tinuous lateral bracing
brac of the compression flanges in addition to 553
515 dimensions under different conditions.
their main function as stay-in-place formwork. However, both the
516 number of fasteners and the thickness of the metal sheets used in
517 construction were insufficient. The results of linear buckling analy- Acknowledgments 554
518 ses of the bridge are presented in Table 2. Ignoring the stabilizing
519 help of the corrugated metal sheets, the results confirmed that, with The financial support from the Lars Erik Lundbergs Stipen- 555
520 or without the scaffolds, the bridge would reach its system failure diestiftelse for this study, the research grant from Byggrådet to 556
521 under approximately 40% of the weight of wet concrete. Therefore, support the laboratory tests, and the support for required 557
522 increasing the total stiffness of the torsional bracings, using either scaffoldings from Britek AB are gratefully acknowledged. The 558

© ASCE 040nnnnn-9 J. Bridge Eng.

ID: hawalh Time: 07:15 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

 J_ID: JBEN ART NO: BEENG-1815 Date: 7-November-15 Page: 10 Total Pages: 11 4/Color Figure(s) ARTTYPE="TechnicalPaper"

559 assistance received from M. Molnar (Division of Structural Scheer, J., and Wilharm, L. (2011). Failed bridges: Case studies, causes 582
560 Engineering, Lund University), T. Lidin (Britek AB), and O. and consequences, John Wiley & Sons, New York. 583
561 Bengtsson (Centerlöf & Holmberg AB) is highly appreciated. SIS (Swedish Standards Institute). (2009). “Eurocode 3—Design of steel 584
structures—Part 2: Steel bridges.” SS-EN 1993-2:2006, Stockholm, 585
Sweden. 586
562 References Stith, J. C., Helwig, T. A., Williamson, E. B., Frank, K. H., Engelhardt, 587
M. D., Schuh, A. C., and Farris, J. F. (2012). “Comparisons of the 588
563 Abaqus/CAE 6.13 [Computer software]. SIMULIA, Providence, RI. computed and measured behavior of curved steel I-girders during 589
564 Beale, R., Andre, J., and Baptista, A. M. (2012) “A survey of failures of lifting.” J. Struct. Eng., 10.1061/(ASCE)ST.1943-541X.0000417, 590
565 bridge falsework systems since 1970.” Proc. ICE Forensic Eng., 1–10. 591
8566 165(4), 161–172. Taylor, A. C., Jr., and Ojalvo, M. (1966). “Torsional re rrestraint of lateral 592
567 Egilmez, O. O., Helwig, T. A., and Herman, R. (2012). “Buckling behavior buckling.” J. Struct. Div., 92(2), 115–130. 593
568 of steel bridge I-girders braced by permanent metal deck forms.” J. Timoshenko, S., and Gere, J. M. (1961). Theory eory of elastic stability, 594
569 Bridge Eng., 10.1061/(ASCE)BE.1943-5592.0000276, 624–633. McGraw-Hill, New York. 595 10
570 Fan, Z., and Helwig, T. A. (2002). “Distortional loads and brace Winter, G. (1960). Lateral bracing of columns umns and beams,
beams, Transactions
beam Transac 596
571 forces in steel box girders.” J. Struct. Eng., 10.1061/(ASCE)0733 Paper No. 3044, ASCE, Reston, VA. 597

572 -9445(2002)128:6(710), 710–718. 2008). “Global

(2
(200
Yura, J., Helwig, T., Herman, R., and Zhao, C. (2008). lobal lateral
la bbuck- 598
573 Galambos, T. V. (1998). Guide to stability design criteria for metal struc- ling of I-shaped girder systems.” JJ. Struct. Eng.
ng.,
g , 10.1061/(ASC
Eng., 1/(ASC
10.1061/(ASCE)0733 599
574 tures, 5th Ed., John Wiley & Sons, New York. -9445(2008)134:9(1487), 1487–1494.1494
1494. 600
575 Helwig, T., Yura, J., Herman, R., Williamson, E., and Li, D. (2007). Yura, J. A. (1995). “Bracing g for stability—State-of-the-art.”
ability—State-of-the
ability hee-a
he-a
e-art.” Proc., 13th 601
576 “Design guidelines for steel trapezoidal box girder systems.” Rep. CE, Reston, VA,, 88-10
Structures Cong., ASCE, 88-103. 602 11
577 FHWA/TX-07/0-4307-1, Center for Transportation Research at the Yura, J. A. (2001). “Fundamentals
ndamentalss of beam bracing.” Eng.
be bracing.” En J., 11–26. 603 12
578 Univ. of Texas at Austin, Austin, TX. Zhao, Q., Yu, B., ., Burdette, E. G., andd Hastings,
Has J. S. (2009). 604
579 Mehri, H., and Crocetti, R. (2012). “Bracing of steel-concrete composite “Monitoring g steel
teel girder stability for safer bridge erection.” J. 605


580 bridge during casting of the deck.” Proc., Nordic Steel Construction Cons Facil.,
Perform. Constr. Facil , 10.1061/(ASC
Facil. 0.1061/(ASC
10.1061/(ASCE)CF.1943-5509.0000048, 606
9581 Conf., Oslo, Norway. 98.
391–398. 607







© ASCE 040nnnnn-10 J. Bridge Eng.

ID: hawalh Time: 07:15 I Path: J:/ApplicationFiles/Journals/ASCE/JBEN/Vol00000/150139/Comp/APPFile/AC-JBEN150139

Paper IV
 Effects of geometric imperfections on bracing performance
of cross-beams during construction of composite bridges
Hassan Mehri1; Roberto Crocetti2; Joseph A. Yura3
1
Ph.D. candidate, Div. of Structural Eng., Lund Univ., Box 118, Lund 22100, Sweden. Email: Hassan.Mehri@kstr.lth.se
2
Professor, Div. of Structural Eng., Lund Univ., Box 118, Lund 22100, Sweden. Email: Roberto.Crocetti@kstr.lth.se

3Professor Emeritus in Civil Eng., Univ. of Texas at Austin, 10100 Burnet Rd., Bldg. 177, Austin, TX 78758. Email:

yura@mail.utexas.edu

Submitted on 19th July 2015, Journal of Engineering Structures.

Abstract
Steel girders normally require stability controls for different construction conditions before the lateral-
torsional support from the hardened concrete deck occurs. A number of failures have recently occurred
during the construction of composite bridges, which indicate the necessity for more investigations on
bracing requirements of steel bridges. Stability of slender girders and brace forces are greatly sensitive to
the girders' imperfections in terms of both magnitude and distribution along the bridges span. In this
paper, three large-scale experimental studies on a twin-I girder bridge varying the cross-beam location,
plus numerical investigations on three bridge structures were performed to determine the effects of
imperfection shapes on bracing performance of cross-beams in straight steel bridges. The results showed
that the current design recommendations can give unsafe and incorrect predictions for the brace forces
and should be improved in terms of both initial imperfection shapes and strength requirements of cross-
beams.

Keywordrd

Cross beam, torsional bracing, imperfection, composite bridge, construction

Introduction structural members, and only few experimental

Adequate braces are normally needed to avoid
lateral-torsional buckling of steel bridge girders
during construction. Improper bracing was one of
the main failure reasons of a number of recent
bridge collapses during construction [1].
Although, there have been a number of studies in
the field of bracing of compression members,
relatively little guidelines are available regarding
bracing requirements of steel bridge girders for
this critical design stage in the current code
specifications. Most of the previous studies
developed theoretical approaches for simple
studies on bracing of steel bridges have been done,
which are necessary to confirm findings from
numerical and theoretical investigations.
During construction, steel bridge girders are
typically braced by means of discrete torsional
bracings in the form of cross-beam/-frame/-
diaphragms. Although, several bridges with
different types of torsional braces were
numerically examined for the purpose of the
current study, this paper only focuses on the
experimental and the numerical results of the
cross-beam bracing of steel bridge girders.

1
Variation of initial imperfection values along the numerical model of the test bridges, the FE
bridges' spans can have significant influence on investigations are extended to two bridge cases: i)
the magnitude of brace forces. For stability a simply supported bridge which was designed by
analyses of simply supported steel bridges, a half- the authors utilizing the conventional empirical
sine-shape geometric imperfection is approaches for preliminary sizing the twin I-girder
recommended by Eurocode 3 [2, 3] to be bridges, and ii) an existing simply supported
introduced −in the out-of-plane direction− to the bridge in Sweden.
compression flanges of steel girders with a
maximum sweep value of 𝐿/500, where 𝐿 is the
bridge span between the end-supports. On the
other hand, AISC [4] recommends 𝐿𝑏 /500 as the
initial relative lateral displacement of the adjacent
brace points, where 𝐿𝑏 is the distance between the
brace points. Although the recommendations
might result in conservative load-carrying
capacities of the braced girder, they might not be a
representative for the “worst” imperfection shape
that maximizes the brace forces. The main Figure 1 Plan view of symmetric geometric imperfections
objective of the present paper is to study the studied in this paper.
geometric imperfection shape which maximizes
the brace forces in the cross-beams, and/or leads to
the lowest load-carrying capacity of the twin-I
girder bridges. The current design
recommendations [2, 4] for brace moment
requirements are also discussed in this paper for
three bridge cases with different configurations.
Figs. 1-2 illustrate different relevant imperfection
shapes which are investigated in this study, where
∆0 is maximum initial sweep of the compression
flange. In this definition of the geometric Figure 2 Plan view of asymmetric geometric imperfections
imperfection, the maximum sweep of the studied in this paper.
compression flanges was 𝑙/500 while the tension
flanges were always straight between the end-
Background
supports, where 𝑙 is the distance between the two
adjacent points with “zero” out-of-straightness The critical buckling moment of beams braced by
from the “perfect” axis. It must be noted that the means of discrete torsional braces with equivalent
number of braces, and the exaggerated curvatures continuous torsional stiffness value of 𝛽𝑇̅ =
for the original shape of the imperfect girders in 𝑛𝛽𝑇 /𝐿 can be obtained from Eq. (1), [5, 6]; where
Figs. 1-2 are used to demonstrate the concept for 𝑛 is the number of the intermediate torsional
the implemented shape of imperfections. braces; 𝐿 is the beam span, which should be
In the next section, current knowledge including replaced with 0.75𝐿 for the bridges that are braced
basic theoretical approaches and the present code with only one cross-beam at mid-span [6]; 𝛽𝑇 is
recommendations regarding stiffness and strength the effective torsional stiffness of the braced
requirements of the cross-beam braces are bridge cross-section; 𝑀𝑐𝑐,𝑏𝑏 , and 𝑀𝑐𝑐,𝑢𝑢𝑢𝑢 are the
presented. In Methods section, detailed critical moment values of the unbraced and the
information concerning both laboratory testing and braced beams, respectively; 𝐸 is the modulus of
numerical analyses is given. In the results sections, elasticity of the girder; 𝐶𝑢𝑢𝑢𝑢 , and 𝐶𝑏𝑏 are the
findings obtained from the experimental study and moment gradient factors for the unbraced and the
the numerical models are compared and discussed. braced beams, respectively; 𝐶𝑇 = 1.2 is a top
In the numerical investigations, in addition to the flange loading modification factor; 𝐼𝑧,𝑒 = 𝐼𝑧𝑧 +

2
(𝑡/𝑐)𝐼𝑧𝑧 ; 𝐼𝑧𝑧 and 𝐼𝑧𝑧 are the moments of inertia of
the compression and the tension flanges,
respectively, about the weak axis of the beam
cross-section; and 𝑡/𝑐 is the distance ratio of the
centroids of the tension and the compression
flanges from the neutral axis of the beam cross-
section.
2
𝑀𝑐𝑐,𝑏𝑏 = �𝐶𝑢𝑢𝑢𝑢 2
𝑀𝑐𝑐,𝑢𝑢𝑢𝑢 2
+ (𝐶𝑏𝑏 𝐸𝐼𝑧,𝑒 𝛽𝑇̅ )⁄𝐶𝑇
Before the concrete deck has hardened, it cannot
resist out-of-plane deformations of the top flanges
of the steel girders of composite bridges.
Therefore, the steel girders deform not only in-
plane of but also out-of-plane of the gravity loads
because of the presence of imperfections. Cross-
beams can discretely resist twisting of the bridge
cross-section by limiting the relative deformations
of the top and the bottom flanges of the adjacent
girders. This creates a brace moment, 𝑀𝑏𝑏 , at each
end of the cross-beam member; higher twists
generate larger brace moments; see Fig. 3.
main girders generated from the brace shear forces
(= 𝑀𝑇 /𝑆).
The brace moments at both ends of the cross-beam
member induce uplifting and overturning brace
shear forces, 𝑀𝑇 /𝑆, in the girders at the location of
the cross-beams. These brace shear forces reduce
one girder's vertical displacement and decrease the
other [7]. Consequently, the vertical deformations
(1)
∆1 and ∆2 caused originally by the brace moments
reduce the effective torsional stiffness of the
bridge cross-section, 𝛽𝑇 . Using Eq. (2), this girder
flexibility (= 1/𝛽𝑔 = 𝐿3 /12𝑆 2 𝐸𝐼𝑦 ) can be
considered in the calculation of the torsional
stiffness of the bridge cross-section [6, 8]; where
𝐼𝑦 is the strong axis moment of inertia of the main
girders, 𝛽𝑏 = the bending stiffness of the cross-
beam member, and 𝛽𝑠 = the bending stiffness of
the web stiffeners at the location of the cross-
beams.
1 1 1 1
= + + (2)
𝛽𝑇 𝛽𝑏 𝛽𝑠 𝛽𝑔

By neglecting the first term of Eq. (1), a

Figure 3 Brace forces generated in the cross-beam member
due to the twist of the braced cross-section.
In Fig. 3, 𝑀𝑇 is the destabilizing moment, 𝑀𝑏𝑏 is
the brace moment at the junction of the cross-
beam member and the main girder, 𝐹 is the
destabilizing force generated in each compression
flange, ℎ is the distance between the top and the
bottom flange centroids, 𝑆 is the transverse
distance between the two girders, 𝜃 is the twist of
the bridge cross-section at the location of the
cross-beams, ∆𝑐 and ∆𝑡 are the lateral movements
of the compression and the tension flanges,
respectively; and ∆ is the vertical deflection of the
conservative value can be calculated for the
“ideal” torsional stiffness required to achieve the
“full” load-carrying capacity, 𝑀𝑐𝑐,𝑏𝑏 ,
corresponding to buckling of individual steel
girder between the brace points. Providing at least
twice this ideal torsional stiffness is recommended
for design purposes [4, 9] to keep the brace forces
at reasonable magnitudes.
In reality however, all steel girders are imperfect
in terms of geometry, load eccentricities, support
conditions, and residual stresses. There have been
some deterministic and probabilistic studies on the
effects of imperfections on ultimate strength of
simple thin-walled beam members e.g. [10-12].
There have been also some studies in the field of
bracing requirements of imperfect compression
members, e.g. [7, 9, 13-15]. However, relatively
very few studies have been carried out on the
bracing performance of steel bridges considering
different imperfection shapes. Wang et al. [16]
found that the maximum torsional brace force
occurred in the brace which is closest to the
location of the maximum moment, when the shape
of initial imperfection provides the maximum twist
at the brace point and “zero” twist at its adjacent

3
braces. Such an imperfection arrangement may not 𝜃0
𝜃= − 𝜃0 (3)
reflect actual imperfection shapes in practice. In 𝛽𝑇,𝑖𝑖 𝑀 2
1−� �∙� �
addition, a consistency between the maximum 𝛽𝑇,𝑎𝑎𝑎 𝑀𝑐𝑐,𝑏𝑏
magnitudes of the studied imperfections cases may
be controversial. Therefore, despite the key role of 𝑀𝑏𝑏 = 𝛽𝑇,𝑎𝑎𝑎 ∙ 𝜃 (4)
imperfections on brace forces, there is relatively
limited information available on the effects of the This design approach will be also examined in this
imperfection shapes on the performance of article for the studied bridges.
torsional braces in common steel bridges during
construction.
A cross-sectional twist of 𝜃 = 𝜃0 = 1° = Methods
0.0175 [𝑅𝑅𝑅], was recommended [9] to estimate Test-program
strength requirements of torsional braces, i.e.
design brace forces; where 𝜃 and 𝜃0 are maximum Three full-scale tests were performed to study the
expected twist in addition to initial twist, and effects of initial imperfections on bracing
initial twist values of the brace bridge's cross- performance of the cross-beams in steel bridge
section, respectively. In [6] the initial twist was girders. In all test cases, the cross-beams were
changed to 𝜃0 = 𝐿𝑏 /500ℎ, to be consistent with placed at the mid-span. The only difference
the initial out-of-plane deflection 𝐿𝑏 /500 used for between the three tests was the locations of the
column and beam lateral bracing; where 𝐿𝑏 is the cross-beams across the depth of the main girders.
distance between the adjacent brace points, and ℎ More precisely, the cross-beams were placed:
is the distance between the centroids of the - At mid-depth of the main girders in
flanges. Based on the assumption of maximum test TN1, see Fig. 4,
expected twist being 𝜃 = 𝐿𝑏 /500ℎ, and 𝛽𝑇,𝑎𝑎𝑎 = - Near the bottom flanges of the main
2𝛽𝑖𝑖 being provided for the torsional braces, a girders in test TN2, see Fig. 5,
design brace moment for cross-beams is - Near the top flanges of the main
recommended [6] considering a linear relationship girders in test TN3, see Fig. 6.
between twists and brace moment values.
Moreover, AISC [4] also gives design
recommendations for the brace moment values of
the intermediate torsional braces based on the
assumptions of initial twists and provided brace
stiffness as described above. However, in practice
the initial imperfections of the twin girders or a
group of girders often are not identical and the
interaction between the girders can affect the
bracing requirements. Providing twice the ideal
stiffness for common steel bridge girders may not
always be practical due to relatively stocky cross-
section of the girders designed for traffic loads.
Two examples of such bridges (one is an existing
Figure 4 Test TN1, the cross-beam placed at mid-depth of the
bridge in Sweden) are investigated in this paper to main girders.
study brace moment magnitudes generated in
cross-beam members with smaller torsional
stiffness than twice the ideal value. Wang and
Helwig [17] developed the following modification
factor for the expected twist at any load level
when the brace stiffness provided is less than
2𝛽𝑖𝑖 :

4
Figure 5 Test TN2, the cross-beam was placed near the Figure 6 Test TN3, the cross-beam was placed near the
girders' bottom flanges. girders' top flanges.

Fig. 7 illustrates the setup for tests TN1, TN2, and TN3. Locations of the potentiometers are marked for
both the “South” (S) and the “North” (N) Girders. Twelve potentiometers, S/N1-6, recorded lateral
movements of the flanges, and two potentiometers, S/N7, recorded vertical deflections of the girders. The
two main girders were nominally identical with specified material properties of S355 (yield strength 𝑓𝑦 =
355𝑀𝑀𝑀), and cross-sectional dimensions of 100 × 12𝑚𝑚2, 750 × 8𝑚𝑚2 , and 150 × 15𝑚𝑚2 for the
top flanges, the webs, and the bottom flanges, respectively.

Figure7 Location of potentiometers, S1-7& N1-7, and strain gauges, NG1-4& SG1-4, for the tests TN1, TN2, and TN3.

The girders were spaced 2 m apart, and the simple
supported span was 15 m. Each girder was pinned
at one end, longitudinally free to slide at the other
end, while free to warp and restrained against twist
at their both ends, see Figs. 8-9. Loads were
applied using a fairly stiff H-shape loading
apparatus (5.0 × 2.5 m), resting on four shaft-
bearings in order to considerably reduce friction
between the girders and the loading apparatus −
thus minimizing possible lateral bracing effects,
see Fig. 10. Moreover, using the shaft bearings
overcame concerns regarding tipping effects of the
loading apparatus on the girders, and also
regarding possible eccentricity of the loading
points from the plane of the girders' webs. Bearing
were also used for the end-supports of the twin
girders, see Fig. 9. Efforts were undertaken to
eliminate possible tilts, eccentricities, and
misalignment of the main girders.

5

Figure 8 End-supports conditions of the top flanges.
Figure 9 End-supports conditions of the bottom flanges.
Figure 10 H-shaped rigid loading apparatus.
The cross-beam at mid-span consisted of two
European standard channels 𝑈𝑈𝑈140
(𝑏𝑏𝑏𝑏 𝑡𝑡 𝑏𝑏𝑏𝑏)and was fastened to the main
girders at different depths during tests TN1, TN2,
and TN3, see Fig. 4-6, using four 𝑀12 high
strength bolts at each end. Locking washers were
used at both ends of the bolts, and efforts were
6
undertaken to adequately tighten the bolts in order
to prevent looseness and slip at the bolted
connections. Bolts and washers were replaced with
the new ones prior to each test. To measure brace
forces generated in the cross-beam member, eight
strain gauges were glued on the outermost flanges'
surfaces of the cross-beam at 200mm distance
from the web of the main girders; see NG1-4 and
SG1-4 in Fig. 7.
Finite element model
ABAQUS commercial software [18] was used in
the numerical investigations. In addition to the FE
models of the test bridges, two single span bridges
with different lateral-torsional slenderness ratios,
𝜆𝐿𝐿 = �𝑀𝑃 /𝑀𝑐𝑐,𝑏𝑏 , were also numerically
studied; where 𝑀𝑃 is the theoretical plastic
moment capacity of the girders cross-section, and
𝑀𝑐𝑐,𝑏𝑏 is the first lateral-torsional critical moment
value of the braced bridge obtained from a
buckling analysis. One of the bridges is in service
in Sweden. Details of these two bridges will be
discussed later. The cross-beams were attached to
the girders and a third-order incremental analysis
was performed, after a deformed shape of the
“unbraced” girders obtained from the linear
eigenvalue analyses was assigned to the main
girders as initial geometric imperfections. The
girders were stress-free at the initial state of the
analyses. The cross-beams of the test bridges TN1,
TN2, and TN3 were attached to the vertical
stiffeners of the main girders at four-nodes at mid-
span in order to conform to the experimental work.
Several relevant shapes of the imperfections
shown in Figs. 1-2 were examined in this study,
while bottom flanges were always straight
between the end-supports. The main girders and
the cross-beams were modeled using S4R shell
elements with sufficient meshing at three
directions with the dimension ratios close to unity.
The effects of material nonlinearities and large
deformations were also taken into account in the
third-order incremental analyses. The steel grades
used were S355 (for the test-models, and the
cross-beams) and S460 (for Bridge setup-2, and
Bridge 1530, see Section 4.6) with multi-linear
material properties (stress values of 355, 360, 539,
and 571 MPa corresponding to strain values of 0.2,
1.3, 5.5, and 11.3% for S355 steel type; and stress
values of 460, 466, 561, and 605 MPa critical moment value of the braced girders under
corresponding to strain values of 0.2, 1.2, 4.0, and four point bending moment obtained from the
11.2% for S460 steel type). The loading condition linear buckling analyses.
applied on the top flanges was four-point bending
moment for the test bridges TN1, TN2, and TN3
in order to conform to the experimental study. The measured geometric imperfections of the
test girders
Before loading the test bridges, out-of-straightness
Results and discussions
of both flanges of the girders were measured at
Torsional brace stiffness of the test bridges
one meter increments along the span using four
The contribution to the cross-section torsional tightened strings. Figs. 11-12 show the normalized
stiffness given by a cross-beam member in a magnitudes of the lateral sweep of the flanges, and
reverse-curvature-bending is three times larger the calculated twist of the cross-sections of both
than the contribution given by a cross-beam girders, respectively.
member in a single-curvature-bending. However,
the contribution of girders' flexibility should be
only considered in a reverse-curvature-bending of
the cross-beam member. Table 1 shows torsional
stiffness values that were calculated for the cross-
beam of the test bridges TN1, TN2, and TN3,
using Eq. (2) and information from literature [6].
Table 1 Brace stiffness values of the studied bridges
109 [𝑁. 𝑚𝑚/𝑅𝑅𝑅].

Set-up 𝛽𝑔 𝛽𝑏 𝛽𝑠 𝛽𝑇
TN1 2.0 7.2 10.0 1.4
TN2* - 2.4* 5.8 1.7*
TN2 2.0 7.2 5.8 1.2
TN3 2.0 7.2 5.8 1.2
* Figure 11 Measured lateral sweep of the top flanges, 𝑢𝑡.𝑓. ,
The values corresponded to single-curvature-bending of the
cross-beam member for test TN2. and the bottom flanges, 𝑢𝑏.𝑓. .

Substituting the critical moment value

corresponding to buckling occurring between the
brace point and the end supports into Eq. (1), a
minimum required stiffness value of 𝛽𝑇 = 0.04 ×
109 𝑁. 𝑚𝑚/𝑅𝑅𝑅 can be obtained for the test
girders to achieve a “full” bracing condition. The
provided values in Table 1 are 30-43 times larger
than the minimum required stiffness for full
bracing, so the girders were fully braced at the
location of the cross-beam (i.e. at mid-span). In
addition, the first three eigenvalues obtained from
linear buckling analyses of the three tested
configurations, namely TN1, TN2, and TN3, were
identical. The 𝑀𝑐𝑐,𝑏𝑏 /𝑀𝑃 ratios were 0.08, 0.12, Figure 12 Calculated twist values, 𝜃0 , of the test girders TN1,
TN2, and TN3.
and 0.16 corresponding to the 1st, 2nd, and the 3rd
buckling modes, respectively; where 𝑀𝑃 is the The twist was calculated as the difference between
theoretical plastic bending capacity of the the lateral sweep of the top and the bottom flanges
individual girder, and 𝑀𝑐𝑐,𝑏𝑏 is the corresponding divided by the girder depth. Although minor

7
tolerances (e.g. ±2mm) for the measured of the flanges. The south girder had a considerably
geometric imperfections may be expected from the larger initial twist at mid-span in test TN2,
utilized method, the accuracy of the data was 𝜃0 = 𝐿/1111𝑑 ≈ 1°, compared to initial twists in
adequate for the purpose of this study. The girders tests TN1 and TN3, 𝜃0 ≈ 0.35°. Consequently, as
had eccentric sweeps with respect to the center of shown in Fig. 13, the south girder experienced
bridge, and they were twisted with different relatively large twist values during the test TN2
magnitudes in different directions. Moreover, the compared to the tests TN1 and TN3, see Fig. 13.
south girder had smaller lateral sweep but
significantly larger twist compared to the north
girder. Measurements for the test bridges TN1,
TN2, and TN3 showed maximum twist
imperfection values of (+𝐿/7500ℎ, +𝐿/75000ℎ,
and + 𝐿/2143ℎ), respectively, for the north
girder, and (−𝐿/2941ℎ, −𝐿/1111ℎ, and −
𝐿/3261ℎ), respectively, for the south girder;
where 𝐿 is the bridge span; ℎ is the distance
between the centroids of the girders' flanges.
Furthermore, positive values show the lateral
movements of the flanges towards the “north”.
The influences of both the magnitudes and the Figure 13 Strength versus twist of the south girder during
distributions of the geometric imperfections of an tests TN1, TN2, and TN3, plus FE results.
individual girder on the load-carrying capacity of
twin-I shaped bridges, as well as on brace
moments generated in the cross-beam members
are discussed in the following sections. It must be
noted that although possible eccentricities and
residual stresses can have some minor effects on
the test results, studying their effects were out of
the scope of this paper. Moreover, the test girders
were loaded well below in their elastic
limits (𝑀𝑚𝑚𝑚 /𝑀𝑃 = 0.12; 𝑀𝑚𝑚𝑚 /𝑀𝑦 = 0.16);
where 𝑀𝑚𝑚𝑚 is the maximum applied bending
moment, and 𝑀𝑦 is the elastic bending moment
capacity of the girders' cross-section. Therefore,
any influence of possible permanent stresses Figure 14 Strength versus twist of the north girder during
persisting after each test on the result of tests TN1, TN2, and TN3, plus FE results.
subsequent test was considered negligible. However, the ultimate resistances of the three test
bridges, TN1, TN2, and TN3, were approximately
Effect of geometric imperfections on the load- the same. The three test results showed that the
carrying capacity of the test bridges test girders exhibited higher load carrying
capacities than the expected buckling load
Figs. 13-14 show the strength ratio versus twist of corresponding to buckling of the girders between
the south and the north girder, respectively, the brace point and the end
obtained from performing tests TN1, TN2, and supports (𝑀𝑐𝑐,𝑏𝑏 ⁄𝑀𝑃 = 0.081). The 𝑀𝑚𝑚𝑚 /𝑀𝑃
TN3. Twists at mid-span were calculated from the values were 0.124, 0.108, and 0.111 for the test
relative lateral movements of the top and the bridges TN1, TN2, and TN3, respectively). This
bottom flanges of the girders. The maximum can be because of a combination of few reasons
magnitudes of the initial twists along the bridge e.g. interaction of two girders with different
span were calculated for the test bridges TN1, imperfections in opposite directions, slight friction
TN2, and TN3 based on the lateral displacements at supports and at shaft bearings, etc.

8
FE results were in good agreement with the tests
TN1, TN2, and TN3 results for the south girder
both in terms of initial stiffness values and load-
carrying capacities, when the maximum values of
the measured imperfections were introduced to the
girders at mid-span. However, the results for the
north girder are more erratic as shown in Fig. 14.
The test results are in good agreement with the FE
analyses in terms of load-carrying capacities and
initial stiffness values, but the girder locally
demonstated significantly lower stiffness values
for a period of loading during the tests TN1, TN2,
and TN3.
The out-of-plane deformations of twin-I shape
bridges are greatly sensitive to initial geometric
imperfections of their individual girders along the
bridges spans. Generally speaking, an accurate
prediction of lateral-torsional deformations of a
bridge is difficult. Fig. 15 schematically shows the
measured deformed shapes of the flanges of the
north and the south girders (the top flanges
graphed with solid lines, and the bottom flanges
depicted with dashed lines), plus the distortions of
the bridges' cross-section at the location of the
cross-beams. It must be noted that the deformation
values are corresponding to the maximum
recorded twist for each test at mid-span, i.e. after
the load carrying capacities were achieved. Fig. 15
shows that only the north girder buckled between
the cross-beam and the twisting supports during
the three tests, while the south girder had
significantly lower twists and lateral deformations
at the quarter spans. As shown in Fig. 3 and
explained earlier, in Twin-I girders, one girder is
locally more loaded than the other due to the brace
shear forces in the cross-beams. Fig. 11 shows that
both girders had initial sweeps toward the “north”
direction. Therefore, following the initial sweeps,
the north girder was subjected to larger bending
moment than the south girder, and reached its
ultimate resistance earlier than the south girder.
Comparisons between Figs. 11-12 and Fig. 15, see
the plan views, show that the south girder
deformed following the shape of its initial
imperfections, while, the north girder changed its
deformed shape from a half sine wave to a sine
wave shape. This may be a reason for the complex
load-twist curves of the north girder shown in Fig.
14 for the three tests.
Figure 15 Lateral deformations of the top (solid lines) and the
bottom (dashed lines) flanges of the north (N. G.) and the
south (S. G.) girders at “maximum recorded twists”.
The cross-sections of the test bridges in Fig. 15
show that both girder cross-sections twisted in the
same direction during the three tests, creating
reverse-curvature-bending in the cross-beams,
regardless of the location the cross-beam member
across the depth of the main girders. The observed
behaviors were also numerically investigated by
studying the effects of different imperfection
shapes on the lateral-torsional deformations of
steel bridge girders. The results are presented in
the following sections.
Linear buckling analyses were performed on the
test bridges TN1, TN2, and TN3. The
corresponding critical moment ratios, 𝑀𝑐𝑐,𝑏𝑏 /𝑀𝑝 ,
computed for the 1st, 2nd, and 3rd eigenvalues are
marked in Fig. 16 along with a schematic
illustration of the deformed shape of the top
flanges (with solid lines) and the bottom flanges
(with dashed lines). The results of the third-order
nonlinear analyses are also shown in Fig. 16 when
three different symmetric initial imperfection
shapes, i.e. imperfection cases (1)-(3) in Fig. 1,
were introduced to the girders. Results showed
that the shape of initial imperfection had
significant effects on both the initial stiffness
values and the load-carrying capacities of the
girders. The imperfection shape case (3) has a
shape similar to the first buckling mode of the
braced bridge and led to the lowest load-carrying
capacity and to the lowest initial stiffness. The
girders reached the higher load-carrying
9
capacities, ultimately towards the critical moments the girders had an asymmetric imperfection shape
corresponding to the second and the third with respect to the center of the bridges.
eigenvalues when imperfection shapes cases (1) Comparison between the curves in Figs. 16 and 17
and (2), respectively, were initially introduced to showed that the imperfection cases (4)-(7) with a
the girders. In both imperfection cases (1) and (2), maximum initial sweep value of 𝑙/500,
at the ultimate load levels, the girders had shapes approximately resulted in the load-carrying
similar to the corresponding initial imperfection capacities within the same range as case (3), i.e.
shape. Investigations were extended to TN1 corresponding to the “lowest” eigenvalue.
bridges with a maximum imperfection value of
𝑙/2000 for cases (1)-(3). The effects of the
magnitudes of initial imperfections on both initial
stiffness and load-carrying capacity were
significant for cases (1) and (3), as seen in Fig. 16,
and were negligible for imperfection case (2). The
girders reached their ultimate load carrying
capacities when local yielding occurred at the
location of maximum out-of-plane deformations,
i.e. at fourth-span distances from the end supports.
To summarize, introducing the recommended
imperfection shape according to the EC3, i.e. case
(1), the test girders deformed in the form of 2nd
Figure 16 The effect of initial imperfection cases (1)-(3), as
eigenmode shape of the braced bridge and resulted illustrated in Fig. 1, on the test bridges TN1, TN2, and TN3.
in approximately 50% higher load-carrying
capacity than the expected value corresponding to
the buckling of the girders between the brace
points. Whereas, the provided stiffness values of
the cross-beams were considerably larger than the
minimum required stiffness for the full bracing
conditions. For imperfection cases (1)-(2) see
Fig.1, the girders deformed in the out-of-plane
direction according to a similar shape as that of the
initial imperfection. It should be pointed out that
chosen magnitude of the initial imperfection for
these girders were relatively large.
In practice however, the magnitudes of Figure 17 The effect of imperfection cases (4)-(7) on the test
imperfections often vary in a complex pattern bridges TN1, TN2, and TN3.
along the girders' span, and the distribution of the
imperfections (including geometric imperfections, It must be noted that, the deformed shapes of the
residual stresses, and eccentricities) are less likely test bridges were in the form of deformations
to be perfectly symmetric with respect to the illustrated in Fig. 15, i.e. only one girder buckled
center of bridges, such as cases (1)-(2). Therefore, between the brace points and the end supports,
the asymmetry in the initial imperfection profiles when initial imperfection shape cases (4)-(6) with
may create a situation in which the girders can a maximum magnitude of ∆0 = 𝑙/500 were
snap-back to a deformed shape corresponding to introduced to the girders. However, due to the
their lowest lateral-torsional buckling capacity. interaction between the two girders, the imperfect
Studying all possible imperfection shapes is test bridges carried slightly higher loads than the
obviously not possible. However, some relevant critical moment corresponding to buckling
imperfections shapes were also investigated which between the cross-beam and end-supports when
might occur in practice, see cases (4)-(7) in Figs. the above mentioned imperfections were assigned
1-2. In imperfection cases (4)-(7), at least one of to the girders. For case (7) with a maximum sweep
magnitude of ∆0 = 𝑙/500 and for cases (4)-(7)

10
with smaller imperfection magnitude of ∆0 = bending and the warping of cross-beams are also
𝑙/2000, both girders deformations were in the investigated in this paper, see sections 4.5 and 4.7.
form of the first buckling mode of the braced Fig. 19 shows normalized values of the maximum
bridge, i.e. both girders buckled between the cross- recorded stress values in the cross-beam member
beam and the end-supports. Therefore for the latter
during each test TN1, TN2, and TN3; where 𝑆𝑦 is
imperfection cases, the south girder with smaller
the elastic section modulus of the cross-beam
bending moments than the north girder−due to
section; 𝑀 is the bending moment applied to the
brace shear forces illustrated in Fig. 3−carried
girders under the four-point loading condition, and
higher load to reach its critical moment level after
𝑀𝑃 is the plastic bending capacity of the girders'
buckling of the north girder occurred. Fig. 15
cross-section. Fig. 19 also shows FE results for the
shows that the test bridge experienced stiffening
maximum stress values, i.e. maximum value of
effects due the deformation of the south girder
SG1-SG4 and NG1-NG4, computed at the same
altering from a half-sine-wave to a sine-wave
locations of the strain gauges, in order to
shape.
demonstrate the effects of the magnitude of
imperfections on the brace stresses. Results
Effect of initial imperfections on in-plane brace showed identical stress curves for TN1, TN2, and
moments of the test bridges TN3 configurations for each of the imperfection
magnitude. In addition, the FE results also showed
The effects of different shapes of geometric
that stress values were directly proportional to the
imperfections on magnitudes of bracing forces
magnitude of the imperfections, which has also
developed in the cross-beam members were also
been observed by Wang and Helwig [17]. Further
investigated in this article, and the results are
numerical investigations on the brace forces were
presented in the following sections.
carried out when different imperfection shapes
Fig. 18 shows stress values in the longitudinal axis illustrated in Figs. 1-2 with maximum sweep value
direction of the cross-beams measured by strain of 𝑙/500 were assigned to the girders of the test
gauges NG1-NG4 and SG1-SG4 during the tests bridges.
TN1, TN2, and TN3. Test results showed that
stress values in the cross-beam member were very
small during the tests TN1, TN2, and TN3.
Significantly large stiffness values were provided
for the cross-beams of the test bridges (𝛽𝑇,𝑖𝑖 /
𝛽𝑇,𝑎𝑎𝑎 ≈ 30 gives a brace force modification
factor of 0.51, see Eq. (3)). Moreover, the
measured initial imperfections of the main girders
were also small. The two mentioned reasons above
reduced the developed forces in the cross-beams
of the test bridges.
Test results also showed that, the cross-beam was
subjected to out-of-plane bending moments. The
out-of-plane bending moments of the cross-beam
were considerably larger near its connection to the
north girder (i.e. stress values at NG1-4) compared
to the out-of-plane bending moments obtained
from measurements at SG1-4. In practice, the Figure 18 Stress values recorded in the cross-beam member
cross-beams are designed to satisfy their in-plane during the tests TN1, TN2, and TN3.
strength and stiffness requirements during
construction, and they may not be sufficiently stiff
in their out-of-plane direction. The Out-of-plane

11

Figure 19 Normalized maximum stress values in the cross-
beam of the test bridges TN1, TN2, and TN3.
Figs. 20-21 show the results of the numerical
investigations on the in-plane brace moments
developed in the cross-beam of bridge TN1 at the
location of NG1-NG4, and SG1-SG4 strain
gauges, respectively.
Figure 20 Normalized values of the in-plane brace
moments,𝑀𝑏𝑏,𝑦 , in the cross-beam of the test bridge TN1 near
the north girder.
Figure 21Normalized values of the in-plane brace
moments,𝑀𝑏𝑏,𝑦 , in the cross-beam of the test bridge TN1 near
the south girder.
12
Results showed that imperfection case (1), as
illustrated both in Fig. 1 and Fig. 19, gave the
largest brace moments with a maximum
magnitude of approximately 6 percent of the
applied maximum bending moment in the main
girders under four-point bending condition.
Similar to the test data, results also showed that
the cross-beam was subjected to relatively smaller
bending moments at its south girder junction than
at the other end. Moreover, the cross-beam
experienced single-curvature bending when
imperfection cases (2) and (7) were introduced to
the girders and was subjected to reverse-curvature
bending when imperfection cases (1) and (3)-(6)
were assigned to the main girders. Results showed
that a single-curvature bending, as occurred for
imperfection case (2)& (7), is a possibility when
the cross-beam is placed at mid-depth of girders.
Brace moment values obtained from Eq. (4) are
also plotted in Fig. 20 which show unsafe
predictions for the strength requirements of the
test bridge especially for the loads near the first
critical moment value. Although the FE results and
Eq. (4), are based on the different definitions for
the magnitudes of imperfections, the results
obtained from the two approaches considering
identical imperfection magnitudes were also in a
poor agreement with predictions of Eq. (4).
Effects of imperfections on out-of-plane brace
moments of the cross-beam in the test bridges
Recorded strains during tests TN1, TN2, and TN3
showed that the cross-beam at mid-span was under
biaxial bending moment, see Fig. 18. The
secondary stresses created due to the out-of-plane
bending of the cross-beam were considerably
larger than the stresses created from the in-plane
bracing performance of the cross-beam. It must be
emphasized that efforts were undertaken to
correctly connect the data channel ports to the
right strain gauges.
Fig. 22 shows a plan view of the deformed shapes
of both flanges of the test girders TN1, TN2, and
TN3. The sketches refer to deformations observed
during FE analyses for different imperfection
cases, in which the top and the bottom flanges are
graphed by the solid and the dashed lines,
respectively. Geometric imperfection cases (1)-(3)
and (4)-(7) with maximum magnitude of 𝑙/500
led to the lateral-torsional deformation shapes of
(a)-(c), and (d), respectively. Clearly, the cross-
beam had to follow the curvature of the girders at
its both junctions. Thus, the secondary moments
occur from the girders' rotation about its weak axis
at the junction with the cross-beam. Fig. 22
indicates that the largest secondary moments (i.e.
out-of-plane bending moment, 𝑀𝑏𝑏,𝑧 ) can be
expected for the asymmetric curvature of the
girders, which had the largest rotations at the
junctions with the cross-beam. For the test bridges,
the secondary stresses at either end of the cross-
beam member are expected to be significant for
geometric imperfection cases (3)-(7) and be
negligible for imperfection cases (1)-(2).
Moreover, the out-of-plane moments are also
expected to be larger at the north girder junction
than the other end for the imperfection cases (4)-
(7), i.e. the deformed shape case (d) in Fig. 22, as
was also observed during the tests.
Figure 22 Plan view of the deformed shapes of the top (solid
lines) and the bottom (dashed lines) flanges of the test bridges
TN1, TN2, and TN3 with different initial imperfection cases,
𝜃0 = 𝑙/500ℎ.
Fig. 23 shows the out-of-plane bending moments
computed from the stress values at the outermost
flanges surfaces of the cross-beam at 200mm
distance from the web-center of the north girders
of the test bridge TN1. Introducing imperfection
cases (3)-(7) to the main girders of the test bridges,
at least one of the girders deformed in an
asymmetric shape, generating considerable
secondary brace moments. The out-of-plane brace
moments were approximately as large as 3 percent
of the applied bending moment of the main
girders. Although, the test bridges had only single
cross-beam between the twisting supports, the out-
of-plane brace moments should be taken into
account in sizing of cross-beams when buckling of
girders between the braces is a possibility. It must
be noted that the magnitude of the out-of-plane
bending moments solely depends on the out-of-
plane rotation and the lateral bending stiffness of
the connection between the cross-beam and the
main girder, and it is independent of in-plane
bracing performance of cross-beams.
Figure 23 Normalized values of the out-of-plane brace
moments,𝑀𝑏𝑏,𝑧 , in the cross-beam of the test Bridge TN1.
Effect of geometric imperfections on bracing
performance of Bridge setup-2 and Bridge 1530
Numerical investigations were extended to the two
other single span bridges with different lateral-
torsional slenderness ratios illustrated in Fig. 24,
in which one of the bridges, Bridge 1530, is an
existing bridge in Sweden.
Uniformly distributed loads were applied to the
top flanges of the girders representing construction
loads in practice. Initial imperfection cases (1)-(7)
were introduced to Bridge setup-2 in which the
dimensions were obtained utilizing the empirical
approaches commonly used to preliminary size the
twin-I shape bridge girders. Bridge setup-2 had
longitudinal and transversal spans of 45.0m and
7.0m, respectively, with length over depth ratio of
18 (i.e. the depth of the girders was 2500𝑚𝑚).
There were five intermediate cross-beams.
Although using cross-frames instead of cross-
beams is more suitable for such bridges with
13
relatively deep girders, two European standard
channel profile UPE400 with 7.5m spacing were
used for the purpose of this study, creating a
system with lateral-torsional slenderness ratio (i.e.
𝜆𝐿𝐿 = �𝑀𝑃 /𝑀𝑐𝑐,𝑏𝑏 ) of 1.15; where 𝑀𝑐𝑐,𝑏𝑏 is the
first lateral-torsional critical moment value of the
entire bridge obtained from a buckling analysis
under uniformly distributed loading conditions
applied on the top flanges.
Figure 24 Dimensions of the two numerically studied bridge
cases, Bridge setup-2 and Bridge 1530, in addition to the test
bridges TN1, TN2, and TN3.
Bridge 1530 had longitudinal and transversal span
of 30.0m and 4.0m, respectively with length to
depth ratio of 27.5 creating a system with
slenderness ratio for lateral-torsional buckling,
𝜆𝐿𝐿 =0.85 (thus, stockier than the previous bridge).
A sine-wave shape with inflections at the brace
points here named case (8), see Fig. 2, was also
considered in the FE analyses in addition to the
geometric imperfections cases of (1)-(7). Although
imperfection case (8) is unlikely to occur in
common bridge applications, one might want to
check the imperfection shape in the form of
buckling of girders between the brace points.
Torsional bracing stiffness of Bridge setup-2
and Bridge 1530
In Bridge setup-2, an eigenvalue analysis showed
that the ratio between the critical moment,
corresponding to buckling between the brace
points, and the theoretical plastic capacity of
girders should be 𝑀𝑐𝑐,𝑏𝑏 /𝑀𝑃 = 0.81. In order to
14
achieve such a value, the torsional stiffness
required, according to Eq. (1), would be 𝛽𝑇,𝑖𝑖 =
2.7 × 1010 𝑁. 𝑚𝑚/𝑅𝑅𝑅. Such a stiffness is
approximately 5% larger than the brace stiffness,
𝛽𝑇,𝑎𝑎𝑎 , used for the cross-beams of Bridge setup-2;
where 𝛽𝑇,𝑎𝑎𝑎 and 𝛽𝑇,𝑖𝑖 are the provided (or actual),
and the ideal torsional stiffness of the cross-beams.
In Bridge 1530, an eigenvalue analysis indicated
that the local yielding of the bridge girders would
occur before buckling between the cross-beams;
i.e. the 1st eigenmode was greater than the plastic
bending capacity of the girders' cross-section.
Using 𝑀𝑐𝑐,𝑏𝑏 = 𝑀𝑃 , the required stiffness from
Eq. (1) is 4.1 𝐸9 𝑁. 𝑚𝑚/𝑅𝑅𝑅, which is 51% less
than the brace stiffness (8.4𝐸9 𝑁. 𝑚𝑚/𝑅𝑅𝑅)
provided for the cross-beams.
Consequently, in both Bridge setup-2 and Bridge
1530, the provided stiffness for the bracings were
less than twice the ideal stiffness values. Bridge
1530 is as an example of a real bridge which
shows that code recommendation of using at least
twice of the ideal torsional stiffness may not
always occur in practice. The effects of different
shapes of geometric imperfections on the load-
carrying capacity and on the developed brace
moments in the cross-beam members of the
described bridges are presented in the following
sections:
Effect of geometric imperfections on load-
carrying capacity of Bridge setup-2 and Bridge
1530
The shape of the first lateral-torsional buckling
mode of Bridge setup-2 was between the end-
supports−i.e. not between the brace points−due to
inadequate torsional stiffness of the bracings as
described earlier.
Maximum geometric imperfections of ∆0 = 𝑙/500
were introduced to the girders for all imperfection
cases (1)-(7) as illustrated in Figs. 1-2. Results are
shown in Fig. 25. Imperfection case (1) resulted in
the smallest value of both resistance and stiffness.
However, geometric imperfection shape cases of
(4)-(6) gave only slightly larger resistance and
stiffness values than the values for the case (1).
The bridge demonstrated significantly lower
ultimate resistance compared to the 1st eigenmode
when an initial imperfection in the form of the 1st
buckling mode, i.e. case (1), was initially given to
the girders. The girders reached their ultimate
resistance due to continuous loss of lateral-
torsional stiffness initiated from yielding in some
parts of the cross-sections under in-plane bending
stresses and stresses generated from the out-of-
plane deformation.
In addition to the imperfection magnitude of
∆0 = 𝑙/500 for cases (1)-(7) as described above,
two other analyses were also performed for Bridge
setup-2 varying the magnitudes of imperfection
and the stiffness of the cross-beams:
i) smaller imperfection magnitude of ∆0 =
𝑙/2000 for case (1),
ii) imperfection magnitude of ∆0 = 𝑙/500 for
case (1), the cross-beams were replaced with
IPE600 (European I-shape profile), and the
width and the thickness of web stiffeners
were increased to 325mm and 25mm,
respectively. The replacements provided
actual torsional stiffness of approximately
twice the ideal brace stiffness, i.e. 𝛽𝑇,𝑎𝑎𝑎 ≈
2.0𝛽𝑇,𝑖𝑖 .
Both latter cases demonstrated relatively larger
load-carrying capacity and initial stiffness values
compared to the case (1) with ∆0 = 𝑙/500 and
𝛽𝑇,𝑎𝑎𝑎 = 0.95𝛽𝑇,𝑖𝑖 .
Figure 25 Effect of initial imperfection shapes (1)-(7) on the
load-carrying capacity of Bridge setup-2.
Fig. 26 shows moment-twist curves for the
different imperfection cases (1)-(8) obtained from
FE investigations of Bridge 1530 taken at the
location of the cross-beams closest to the midspan.
Similar to Bridge setup-2, the first lateral-torsional
buckling mode of Bridge 1530 was lateral-
torsional buckling of the entire bridge between the
twist supports due to relatively small torsional
stiffness assigned for the braces. As it can be
observed from Fig. 24, the cross-section of the
bridge is designed with relatively wide flanges at
mid-span compared to the depth of the girders.
This has led to a relatively stocky bridge cross-
section resulting in smaller out-of-plane
deformations of the bridge, compare twist values
in Fig. 26 and Figs. 16-17&25. The load carrying
capacities of the bridge were approximately
identical when initial imperfection cases (1)-(8)
were introduced to the bridge. No buckling
occurred for any of the imperfection cases and the
bridge reached its ultimate capacity because of
loss of lateral-torsional stiffness due to partial
yielding of the girders. Similar to Bridge setup-2,
the load-carrying capacities were significantly
lower than the critical elastic moment value; the
smaller lateral torsional slenderness ratio the
greater detrimental effects of inelasticity on load
carrying capacity of bridges.
Figure 26 Effect of initial imperfection shapes (1)-(8) on
load-carrying capacity of Bridge 1530.
Effects of imperfections on brace moments of
the cross-beams in Bridge setup-2 and Bridge
1530
Brace moments created in the cross-beam member
at mid-span, i.e. at the location of maximum twist,
were also computed for Bridge setup-2 and Bridge
1530. Similar to the test bridge TN1, the results
for Bridge setup-2 and Bridge 1530 also showed
that the in-plane brace moments of the cross-
beams were considerably larger at one end
compared to the other. Therefore, only the larger
brace moments for Bridge setup-2 and Bridge
15
1530 are reported in Figs. 27-28. Comparisons
between Figs. 20-21 and 27-28 showed that for the
studied bridge cases, the brace moments were
smaller for the stockier bridges with smaller
lateral-torsional slenderness ratios.
Figure 27 Normalized values of the in-plane brace
moments,𝑀𝑏𝑏,𝑦 , in the cross-beam of Bridge setup-2.
Figure 28 Normalized values of the in-plane brace
moments,𝑀𝑏𝑏,𝑦 , generated in the cross-beam of Bridge 1530.
Figs. 20-21 and 27-28 also suggested that both
geometric imperfection shape cases (1) and (5)
should be separately assigned to the girders of
twin-I girders in FE analyses as the “worst”
geometric imperfection shapes which maximize
brace forces created in the cross-beams of twin-I
shape bridges. However, the case (1) gave
reasonably conservative results for the brace
moments of the all studied bridge cases. Similar to
the test bridge, Figs. 27-28 also show that Eq. (4)
gave significantly unsafe results for the brace
moment requirements for both Bridge setup-2 and
Bridge 1530.
16
Applying imperfection cases illustrated in Figs. 1-
2, both Bridge setup-2 and Bridge 1530 were
deformed in the form of their first lateral-torsional
buckling mode shapes, which are perfectly
symmetric with respect to center of the bridges,
see Figs. 25-26. Consequently, as the symmetric
deformations create nil out-of-plane rotations in
the connection of cross-beams with main girders,
the out-of-plane bending of the cross-beam of
Bridge setup-2 and Bridge 1530 are expected to be
negligible for any of the imperfection shapes. The
FE analyses confirmed that no out-of-plane
bending moments were developed in the cross-
beams of Bridge setup-2 and Bridge 1530 for any
of the imperfection cases illustrated in Figs. 1-2.
Asymmetric cross-beams
Bridge 1530 had a single channel at each cross-
beam location. The attachment of the channel web
to the girder's stiffener creates an eccentric
connection that develops warping torsional
stresses in the channel profile as illustrated in
Fig. 29. The warping stresses should be added to
the bending stresses generated from the bracing
performance of the cross-beams. Eccentricities of
overturning and uplifting forces, see 𝑀𝑇 /𝑆 in Fig.
3, from the shear center of such a cross-beam
generate torsional forces at both ends of brace
member. For the cross-beams with channel profile
for instance, dividing the cross-section into two
angles, the warping stresses in the flanges can be
approximately calculated by studying the bending
stresses created from an equivalent couple
moments, 𝑀𝑇𝑇 , placed at the level of the flanges
centroids, about the neutral axis of the angles, see
Fig. 29; where 𝑑 is the depth of cross-section; 𝑡𝑓 is
the thickness of the flanges; and 𝑇𝑤 is the warping
component of the torsion generated due to
eccentricity of 𝑀𝑇 /𝑆 forces from the shear center
(S.C.) of the channel profile.
The eccentricity of the torsional warping forces,
i.e. “e”, depends on the thickness of stiffeners and
the attachment configuration. The distribution of
torsional warping forces along the cross-beam
span also depends on the warping boundary
conditions at both sides of the cross-beam.
However for the sake of simplicity in order to
demonstrate the concept, the webs of the cross-
beams of Bridge 1530 were attached to the vertical
stiffeners with no eccentricity in the FE analyses.
Figure 29 An approximate method to calculate torsional
warping stresses in the flanges of the cross-beams with
channel profiles.
Figures 30-31 show significant warping moments,
𝑀𝑇𝑇 , calculated directly from the stress values at
both ends of the cross-beam of Bridge 1530 at
mid-span for imperfection cases (1)-(7).
Imperfection case (8) created no warping at the
cross-beam.
Figure 30 Normalized values of warping moment,𝑀𝑇𝑇 , at the
junction of the cross-beam of Bridge 1530 with the north
girder.
Imperfection cases (2)-(3) and (7) resulted in
torsions at both ends of the cross-beam with equal
magnitudes and acting at the same direction.
While, the cross-beam of Bridge 1530 at mid-span
was subjected to torsions at it's both ends acting at
opposite directions−due to the reverse-curvature
bending of the cross-beam member−when
imperfection cases (1), and (4)-(6) were
introduced to the main girders. Utilizing channel
profile significantly affected the stress values in
the longitudinal axis direction of the cross-beam of
Bridge 1530. For this reason, using single cross-
beams with eccentric cross-sections is not
recommended by the authors for practice.
Figure 31 Normalized values of warping moments,𝑀𝑇𝑇 , at
the junction of the cross-beam of Bridge 1530 with the south
girder.
Conclusions
A number of collapses have been recently reported
during construction of composite bridges, in which
improper bracing was one of the main reasons of
the failures. Initial imperfections have great
influence on the brace forces developed in the
cross-beams, and often on the load-carrying
capacity of slender girders. Strength requirements
of bracings are defined based on the assigned
imperfection shapes and magnitudes in either
analytical or numerical studies. Current code
specifications such as Eurocode 3 and AISC have
different recommendations for the shape and the
magnitudes of initial imperfections for stability
analyses of steel bridge girders. The shape of
initial imperfection in reality can be significantly
different from the code recommendations. The
effects of geometric imperfection variation along
the span of bridges on their load-carrying capacity
and on their brace forces are even more
complicated when a number of girders
interconnected with torsional braces are being
studied. This study showed that the interaction
between the two girders greatly influences the
response of the entire bridge. Three full-scale
laboratory tests on a twin-I girder bridge varying
the location of the cross-beam across the depth of
girders, plus third-order analyses on three different
bridges (with different lateral-torsional slenderness
ratios) were carried out and discussed in this
paper. Several relevant imperfection shapes were
17
introduced to the girders of studied bridge in the
numerical studies. Due to presence of brace shear
forces with opposite directions generated in the
main girders at the location of each cross-beam,
one of the girders will be subjected to larger
vertical loads than the other. Consequently, it will
reach its ultimate resistance earlier. The results
also showed that the slender steel bridge girders
can reach a load-carrying capacity which is
significantly larger than the theoretical critical
load for the braced girders. This occurs when
following the initial deformed shape of the
imperfect girders requires lower energy than
snapping to the shape of first buckling mode.
Moreover, both the tests and the FE investigations
showed that the shape of geometric imperfections
can have a dominant effect on distortion of the
braced bridges' cross-section resulting in either a
single-curvature or a reverse-curvature bending of
the cross-beam members regardless of the location
of the cross-beam member across the depth of
main girders.
The investigations suggested that the code
recommendations regarding the shape of
imperfections for stability analyses of bridges
should be improved. In addition to the current
half-sine shape between the end-supports
recommended by Eurocode 3, this article
recommends that a similar imperfection shape but
with a maximum magnitude at quarter span of
bridges, i.e. with an eccentric distribution be also
considered. In addition, out-of-plane bending and
torsional forces could be created in the cross-
beams. The current AISC design recommendation
also showed significantly unsafe results for the
brace moment requirements for the three studied
bridges. The study also showed that significant
warping stresses can be developed in the cross-
beams with asymmetric cross-sections, and
avoiding such profiles as a cross-beam member in
bridge applications is suggested.
Acknowledgements
The financial supports from The Lars Erik
Lundbergs Stipendiestiftelse (Dnr 7/2013 and Dnr
2014/05) to this study, the research grant from
Byggrådet (Dnr 12/2012) to support the laboratory
tests are gratefully acknowledged. Moreover, the
assistance received from M. Molnar (Div. of
18
Struct. Eng., Lund Univ.) and T. Helwig (Div. of
Struct. Eng., Univ. of Texas in Austin) is highly
appreciated.
References
[1] H. Mehri and R. Crocetti, "Bracing of steel-
concrete composite bridge during casting of
the deck," presented at the Nordic Steel
Construction Conf., Oslo, Norway 2012.
[2] EC3, "Design of steel structures –Part 2: Steel
bridges," in SS-EN 1993-2:2006 ed.
Stockholm, Sweden: Swedish Standards
Institute, 2009.
[3] EC3, "Design of steel structures –Part 1-1:
General rules and rules for buildings," in EN
1993-1-1:2005, ed. Stockholm, Sweden,
Swedish Standards Institute, 2005.
[4] AISC, "Specification for structural steel
buildings," vol. ANSI/ AISC 360-05, ed.
Chicago, Illinois, 2005.
[5] T. V. Galambos, Guide to stability design
criteria for metal structures, 5th ed., John Wiley
& Sons, 1998.
[6] J. A. Yura, "Fundamentals of beam bracing,"
Engineering J., pp. 11-26, 2001.
[7] T. A. Helwig, J. A. Yura, and K. H. Frank,
"Bracing forces in diaphragms and cross
frames," in Structural Stability Research
Council Conf., April, 1993, pp. 6-7.
[8] H. Milner, "The design of simple supported
beams braced against twisting on the tension
flange," Institution of Engineers (Australia)
Civ. Eng. Trans., 1977.
[9] J. Yura, B. Phillips, S. Raju, and S. Webb,
Bracing of steel beams in bridges: Center for
Transportation Research, Bureau of Eng.
Research, University of Texas at Austin, 1992.
[10] B. Schafer and T. Peköz, "Computational
modeling of cold-formed steel: characterizing
geometric imperfections and residual stresses,"
J. of Const. Steel Research, vol. 47, pp. 193-
210, 1998.
[11] Y. L. Pi and N. S. Trahair, "Nonlinear inelastic
analysis of steel beam-columns, II:
Applications," J. of Struct. Eng.-ASCE, vol.
120, pp. 2062-2085, Jul 1994.
[12] E. Lindgaard, E. Lund, and K. Rasmussen,
"Nonlinear buckling optimization of
composite structures considering “worst”
shape imperfections," International J. of Solids
and Structures, vol. 47, pp. 3186-3202, 2010.
[13] G. Winter, "Lateral bracing of columns and
beams." vol. Vol. 125, No. 1, ed: ASCE, 1960,
pp. 807-826
[14] Z. P. Bazant and Y. Xiang, "Postcritical
Imperfection-Sensitive Buckling and Optimal
Bracing of Large Regular Frames," J. of
Struct. Eng., vol. 123, pp. 513-522, 1997.
[15] J. A. Yura and G. Li, "Bracing requirements
for inelastic beams," in 2002 SSRC Annual
Stability Conf., 2002.
[16] L. Wang and T. A. Helwig, "Critical
imperfections for beam bracing systems," J. of
Struct. Eng., vol. 131, pp. 933-940, 2005.
[17] L. Q. Wang and T. Helwig, "Stability bracing
requirements for steel bridge girders with
skewed supports," J.of Bridge Eng., vol. 13,
pp. 149-157, Mar-Apr 2008.
[18] ABAQUS/CAE, V 6.13 [computer software]:
RI, Simulia., 2013.

19
Paper V
 End-warping bracings during the construction of steel
bridges
Hassan Mehri1; Roberto Crocetti2
1
Ph.D. candidate, Div. of Structural Eng., Lund Univ., Box 118, Lund 22100, Sweden. Email: Hassan.Mehri@kstr.lth.se
2
Professor, Div. of Structural Eng., Lund Univ., Box 118, Lund 22100, Sweden. Email: Roberto.Crocetti@kstr.lth.se

Submitted on 2nd December 2015, Journal of Bridge Engineering, (ASCE).

Abstract

Slight bracing of steel bridges so as to control end-warping of their flanges near the end-supports can be
effective in preventing a system buckling failure of them, which if it occurs often has drastic
consequences in terms both of fatalities and costs. Relatively little information is available, however,
regarding the stiffness and strength requirements of such bracings, which can be of great benefit in
connection with steel bridges that are prone to system buckling. In the present paper, the causes of the
failure of Bridge Y1504 in Sweden, in which corrugated metal sheets were employed to stabilize the
trapezoidal girder during the concreting of the deck, are discussed. The adequacy of the metal sheets and
of their attachments to the steel girder was very much questioned in failure investigations carried out
following the bridge's collapse. The results of seven full-scale laboratory tests of different configurations
of bracings used in a twin I-girder bridge to resist end-warping at support points are presented here.
Bracing forces in the plan bracings of the test bridge were compared with those obtained through an
approximate analysis suggested by Eurocode 3 for the preliminary analysis of such bracings. The bracing
forces present in the cross-beam of the test bridge were obtained for different brace configurations. The
bracings, both of the plan bracing type and of corrugated metal sheets that were employed, were found to
effectively enhance the load-carrying capacity of the test bridge. Only relatively small forces were
generated in the bracings involved to achieve such a substantial increase (by a factor of greater than 2.0)
in the load carrying capacity of the test bridge in question.

Keyword
Plan bracing, corrugated metal sheet, bridge, construction, warping, system buckling

Introduction improved substantially by increasing either the

stiffness or the number of intermediate cross-
Global lateral-torsional (or system) buckling (see bracings employed [1-3].
Fig. 1) is a failure mode that has occurred during
A modification of Timoshenko's basic approach
the construction of several steel bridges of the
[4] has been made [5], one representing an
single-span type having a relatively small span-to-
extension of the effective length method, its
width ratio. Examples of failures of this type are
providing an approximate solution, as presented in
the collapse of the Bridge Y1504 in Sweden [1]
Eq. (1), for predicting the critical moment capacity
and of the Marcy Bridge in New York City [2],
of a single beam having a particular set of
both of which occurred in 2002. The load-carrying
boundary conditions, where 𝑀𝑐𝑐,𝑦,0 is the critical
capacity of bridges in which the dominant failure
mode is that of system buckling cannot be moment value of the beam involved; 𝐶𝑏 is the

1
moment gradient factor which takes accounts of
loading effects [6]; 𝐸 and 𝐺 are the elasticity and
the shear moduli, respectively; 𝐿 is the beam span;
𝑘𝑥 and 𝑘𝑧 are the effective buckling length factors
with respect to the longitudinal and vertical axes;
and 𝐼𝑧 , 𝐽, and 𝐶𝑤 are the in-plane moment of
inertia, the St. Venant torsional constant, and the
warping constant of the cross-section of the beam,
respectively.
π π2 ECw
Mcr,y,0 = 𝐶𝑏 � � ∙ �EIz GJ�1 +
kzL GJ(k x L)2
Figure 1 Global lateral-torsional (or system) failure mode, 𝜃
being twist of bridge cross-section at mid span.
Substituting (2𝐸𝐼𝑧 ), (2𝐺𝐺), and (2𝐸𝐼𝑧 ℎ02 /4 +
2𝐸𝐼𝑦 𝑆 2 /4) for the 𝐸𝐼𝑧 , 𝐺𝐺, and 𝐸𝐶𝑤 terms in Eq.
(1), Yura et al. [3] derived an approximate solution
(presented in Eq. 2) from which the system
buckling capacity of a twin I-girder bridge can be
estimated, where 𝑆 is the transversal span between
the centroids of adjacent girders; 𝐼𝑦 is the in-plane
moment of inertia of a single girder; 𝐼𝑧,𝑒𝑒𝑒 =
𝐼𝑧,𝑐 + (𝑡/𝑐)𝐼𝑧,𝑡 ; 𝐼𝑧,𝑐 and 𝐼𝑧,𝑡 are the lateral
moments of inertia of the compression and tension
flanges of the cross section about their centroids;
and 𝑡 and 𝑐 are the distance of the centroids of the
tension and compression flanges from the centroid
of the cross-section.
𝜋 2 𝑆𝑆
𝑀𝑐𝑐,𝑦,𝑔 = 𝐶𝑏 ∙ �𝐼𝑦 𝐼𝑧,𝑒𝑒𝑓
(𝑘𝑧 𝐿)2
Yura et al. [3] indicated the solution to be also
applicable to bridges having a trapezoidal cross-
section. As based on Eq. (2), when system
2
buckling is the dominant failure mode of a bridge,
the bridge's load carrying capacity can be
improved theoretically by providing the regions
near the end-supports (and perhaps the mid-span
too, so as to prevent, if applicable, further modes
of system buckling from occurring) with warping
restraints, these reducing the effective buckling
length factor, 𝑘𝑧 , in Eq. (2). The improved critical
moment can be estimated by taking account of the
effective buckling length factor to be 𝑘𝑧 =
0.5~0.6 for the both-end-fixed condition that can
(1)
be achieved by use of adequately stiff bracings so
as to effectively prevent lateral rotation of the
compression flanges involved. Bracings of this
type can be either in the form of typical plan
bracings or of corrugated metal sheets that are
appropriately attached to the main girders.
Fig. 2 shows schematically an analogy between
the shear diaphragm action of the corrugated metal
sheets and of the plan bracings that takes place in
composite bridges. Plan bracings and corrugated
metal sheets of typical types, if they possess
sufficient shear stiffness and strength, can
presumably provide the compression flanges with
reliable warping restraints [2, 7, 8]. Although the
plan bracings that are attached to the top flanges of
steel bridges are not particularly attractive to
constructors, due to their possible conflicting with
the reinforcing bars, those bracings located near
the supports on a relatively short portion of the
bridge span there may be less problematical for
constructors. Such bracings can be installed
temporarily below the top flanges prior to
concreting of the deck in question and can be
removed once the concrete deck has hardened.
The corrugated metal sheets have also been used
frequently as a stay-in-place formwork during
concreting of the deck in the case of composite
bridges. In the United States, it is common
practice that the corrugated metal sheets of the
end-closed type are spanned between the cold-
formed angle profiles that are already welded to
the top flanges of the main girders, this enabling a
leveling of the sheets in the case of variations in
the flange thicknesses and also for the relative
(2) camber of the adjacent girders. Eccentricities
imposed by the presence of such angle profiles
increase to a large extent the overall shear
flexibility of the panels involved, this reducing the
effectiveness of the metal sheets in providing a
stabilizing system for the compression flanges in
question [8]. As a result of this, the current
AASHTO [9] forbids employing metal sheets as a
stabilizing system in steel bridge applications.
Figure 2 Analogy between stabilizing performance of
corrugated metal sheets and plan bracings, when they are
attached to top flanges of steel bridge girders
In Norway, corrugated metal sheets that are
relatively thick (𝑡 = 3~5 𝑚𝑚) are used
frequently to provide a stay-in-place formwork of
trapezoidal girders for steel bridges, their being
welded directly to the top flanges involved. Under
such circumstances, the eccentricity of the sheets
from the centroid of top flanges is relatively small.
In this way, the overall shear flexibility of the
sheets reduces substantially compared with the
situation in which the installation of sheets on the
top flanges involves the use of nail and screw
fasteners having the possibility of slips at the
attachments.
Bridges of twin I-girder type having a relatively
large transversal span between the main girders
are the most common practice in Sweden. Such
bridges, compared with the multi I-girder or
trapezoidal types require deeper metal sheets with
sufficient stiffness in bending so as to transfer
appropriately the construction loads between the
main girders. The deeper sheets increase the
thickness of the concrete deck involved, this
making the use of metal sheets less attractive in
the construction of Twin-I girders with a wide
transversal span. Corrugated metal sheets with
reasonable sizes, however, have been used also in
Sweden with a dual function of stay-in-place
formwork and stabilizing system during the
construction of bridges of trapezoidal cross-
section. The Bridge Y1504 was designed with a
longitudinal span of 65 m in length and having
nine intermediate cross-diaphragms between the
end-supports of it so as to control a possible
distortion of the bridge cross-section. The
corrugated metal sheets of the Bridge Y1504 were
attached to its top flanges by the use of nail
fasteners. The cross-section of the bridge twisted
suddenly with respect to its longitudinal axis when
only a quarter volume of the first concreting
sequence was poured. The end-bearings
experienced severe damage because of excessive
warping of the cross-section there. Apparently, by
assuming the function of the stay-in-place sheets
as a formwork only, the constructor involved
replaced the prescribed type of the sheets with an
equivalent or better alternative, the one which was
probably available or cheaper in the market; see
Fig. 3. The diameter of the seam fasteners was
changed also from 6.3 mm (with 150 mm
spacing) to 4.8 mm (with 300 mm spacing). The
edge nails employed had a diameter of 4.5 mm
and were placed at each valley of the cross section
of the sheets. Calculations regarding the risk of
system buckling for the steel girder and of global
buckling for the metal sheets were missing in the
design documents. A report issued after the failure
evaluations of the Bridge Y1504 revealed that the
actual shear stresses of the critical nail fasteners at
end-support points of it to be greater than their
nominal shear resistance [10].
Figure 3 Prescribed and used corrugated metal sheets in
Bridge Y1504
The accuracy of the current guidelines, such as
[11, 12] for example, for determining the nominal
shear stiffness of a certain corrugated sheet with
typical arrangement of fasteners are at the times
difficult to be truly justified [10]. Although the
folded stiffeners of the employed sheets enhance
the in-plane bending stiffness of the panels, they
increase to a considerable extent the warping
flexibility of them compared with a similar profile
without such stiffeners. This flexibility leads to the
development of large forces in the edge fasteners
3
since the panel lacks sufficient shear rigidity to
properly distribute the shear stresses involved
between its edge fasteners. The use of two nails
instead of one at each valley − one at each side of
such stiffeners − could increase significantly the
shear stiffness of the panels and could reduce
presumably the risk of failure. The eccentricity of
gravity loads from the shear center of the cross-
section of the Bridge Y1504 created torques along
the span which can be the greatest at the end-
bearing points if a half-sine lateral imperfection
shape be assumed for the steel girder. The end
panels there that were installed near the end-
bearings (see Fig. 1), thus, need to be attached all
the way around to the steel girder and the cross-
diaphragms involved, this enabling them to
function more appropriately as a shear diaphragm.
The end panels, however, were attached neither to
the support cross-diaphragms nor to the
intermediate cross-diaphragms in the Bridge
Y1504, this causing the corner fasteners of the end
panels became overloaded during the concreting
process. The progressive shear failure of the edge
fasteners near the end-supports converted the
closed section to an open cross-section, this
reducing dramatically the torsional stiffness of the
cross-section. The torsional stiffness of the open
section was apparently inadequate to resist the
torque developed at the support points.
The present article reports the results of seven full-
scale tests that were performed to investigate
warping restrain performance of typical corrugated
metal sheets and of two configurations of plan
bracings, when they were installed on top flanges
of a twin I-girder bridge near its end-supports. The
brace forces in the bars of the plan bracings
employed were also calculated on the basis of the
recorded axial strains in them. The brace moments
in the intermediate cross-beam of the test bridge
were calculated also on the basis of the measured
strains at its both ends. The test results regarding
the performance of the bracings employed so as to
resist end-warping of the test bridge were
compared with each other and discussed. These
brace forces were compared also with the ones
obtained through an approximate method that is
used commonly by the design engineers for the
analysis of such systems.
4
Test program
Prior to the present test program, three tests
namely TN1-TN3 were carried out in the elastic
range of the test bridge, in which the location of a
single cross-beam at mid span varied across the
depth of girders. The study investigated the effects
of the shape of initial imperfections on the brace
forces develop in the cross-beam involved. The
present paper reports the results of seven full-scale
tests in different brace configurations as following:
i) Test TN4, in which a single-panel Z-type
bracing was installed near the end-supports of
the test bridge, this providing a braced span
of 1200 mm at its each end (see Fig. 4). One
cross-beam was installed also at mid-height
of the girders so as to control the twist of
them at mid-span (see Fig. 5);
ii) Test TN5, in which a double-panel Warren-
type bracing was installed near the end-
supports of the test bridge, this providing a
braced span of 2400 mm at its each end (see
Fig. 6). One cross-beam was installed also at
mid-span of the bridge;
iii) Test TN6, in which the cross-beam of the test
setup TN4 was removed at the mid span;
iv) Test TN7, in which the cross-beam of the test
setup TN5 was removed at the mid span;
v) Test TN8, in which the corrugated metal
sheets of an end-closed type were installed
near each end-support (see Fig. 7), their
providing a braced panel of 2400 mm in
length at each end of the test bridge for which
the two edges of each panel were attached to
the main girders. The nail fasteners employed
fastened every second valley of the
corrugated sheets involved to the top flanges
of the main girders. One cross-beam was
installed also at the mid-span;
vi) Test TN9, in which similar brace
configuration to the test setup TN8 was
employed, however, the nail fasteners
attached the metal sheets to the main steel
girders at every valley of them;
vii) Test TN10, in which similar brace
configuration to the test setup TN9 was
employed, however, the two free edges of
 each brace panel were fastened to a
transversal angle profile that was spanned
between the top flanges of the main girders
(see Fig. 8).
The cross-beam at mid-span consisted of two
European standard channel profiles UPE 140
(back to back), its being attached to the web-
stiffeners at the mid span of the test bridge by the
use of four 𝑀12 bolts of high-strength type. The
bolts were assured to be tightened sufficiently for
which locking washers were used so as to prevent
Figure 4 Z-type plan bracing with a 1200 𝑚𝑚 brace span near
each end support of test bridge during tests TN4 and TN6.
Figure 6 Warren-type plan bracing with a 2400 𝑚𝑚 braced
span near each end support during tests TN5 and TN7.
possible looseness and slip at the connections
between the main girders and the cross beam in
question. The corrugated metal sheets, nail
fasteners, and screws employed were replaced
with new ones prior to the tests TN9-TN10. Seam
fasteners were self-drilling screws of 4.2 ×
13 mm in size attaching the adjacent sheets at
400 mm spacing. The cartridge used in the nail
gun was of Extra Heavy Powder Actuated type
and the nails employed were of Kit X-ENP MXR
D4.5 mm type; see Fig. 9.
Figure 5 Cross-beam at mid-span during tests TN4-TN5 and
TN8-TN10.
Figure 7 Corrugated metal sheets with a 2400 𝑚𝑚 brace span
near each end support employed during tests TN8-TN10; photo
corresponds to maximum load level during test TN10.
5
Figure 8 Cold-formed angle profiles employed at free-edges of
each brace panel near end supports during test TN10.
The details of test setups TN4-TN10 − this
including the location of potentiometers, LVDTs,
and Strain gauges employed − are presented in
Fig. 10, where (S) stands for the “South” girder
and (N) for the “North” one. The lateral
movements of both flanges of each girder were
measures at one-third, mid, and two-thirds of span
points by potentiometers, whereas the lateral
movements of the top flanges at the brace points
were measured by LVDTs. The brace moments at
each end of the cross-beam involved were
calculated on the basis of the data that was
recorded by eight strain gauges SG1-SG4 and
NG1-NG4; see Fig. 10. These gauges were
mounted at approximately 200 mm distance from
the web of the main girders on the top most
surface of the top flange and on the bottom most
surface of the bottom flange of the cross-beam in
question. One pair of the strain gauges of a single-
direction type were also glued at a same cross-
section of each bar of the plan bracings employed
during the tests TN4-TN7 so as to calculate the
axial forces in the bars (marked as 𝐹1 − 𝐹10 in
Fig. 10) on the basis of a mean value of the
recorded pair strains. These gauges were attached
at the mid-length of the tubular bars in question at
the uppermost and the bottommost layers of their
cross-section.
The main girders of the test bridge were identical
in terms of size and material properties, having a
6
Figure 9 Nail fasteners of X-ENP type and cartridge employed
during tests TN8-TN10.
top flange of 100 × 12 mm2, a bottom flange of
150 × 15 mm2 and a web of 750 × 8 mm2 in
size. The bridge span in longitudinal direction was
15000 mm in length and the girders were placed
2000 mm apart. Each girder of the bridge was
pinned at one end, free to slide at the other end,
and restrained against twist at its both ends. The
girders were of S355 type having a nominal
yielding strength of 355 MPa. To simulate a four-
points-flexure loading condition, the point-loads
were transferred via an H-shape rigid beam that
was hoisted on four high quality shaft-bearings so
as to reduce the amount of friction between the
loading apparatus and the main girders. The test
program was designed carefully to eliminate any
misalignment and tilt of the main girders, and the
eccentricity of the brace bars involved at the
intersection points of them. The use of locking
washers reduced substantially any slip at the
connection of the brace bars with the main girders.
The brace bars employed were of tubular section
type having an external diameter of 33.7 mm and
a constant thickness of 2 mm. These brace bars
were attached at each end to the top flanges of the
main girders by the use of two 𝑀10 bolts of a
high-strength type. The corrugated metal sheets
employed were 1.0 𝑚𝑚 in thickness and had a
cross-sectional geometry as shown in Fig. 9.
 Figure 10 Location of potentiometers, LVDTs, and strain gauges employed during tests TN4-TN10.
The geometric imperfections of the test girders
that were measured prior to each test
The initial out-of-straightness of the top, 𝑢𝑡.𝑓. , and
bottom, 𝑢𝑏.𝑓. , flanges of the test girders were
measured prior to each test, their normalized value
being presented in Figs. 11-12.
Figure 11 Initial out-of-straightness of “North” and “South”
girders for both flanges of them prior to tests TN4-TN7.
These values correspond to a nominal distance of
the flanges involved from a string that was
stretched tightly between the end supports.
Although a tolerance of approximately ±2 mm in
these measurements can be expected to occur, the
accuracy of them was adequate for the purpose of
the present study which was to assure the
permanent deflections of the flanges in the lateral
direction remain negligible after each test.
Figure 12 Initial out-of-straightness of “North” and “South”
girders for both flanges of them prior to tests TN8-TN10.
It was found that the north girder had a maximum
initial out-of-straightness of 𝐿/685 for its top
7
flange and of 𝐿/780 for its bottom flange. The
south girder, however, had relatively smaller
geometric imperfections than the north girder, its
having maximum initial out-of-straightness of
𝐿/1060 for its top flange and of 𝐿/1160 for its
bottom flange. The measurements confirmed that
the permanent deflections of the test girders after
each test in lateral direction of them were
negligible. This was expected to be true since the
girders were being loaded within their elastic
range (𝑀𝑚𝑚𝑚 /𝑀𝑃 ≈ 0.2) during the tests TN4-
TN10.
Estimations of the buckling capacity of the test
bridges TN4-TN10 on the basis of the effective
buckling length method
The effects of warping restraints provided by the
bracings employed in the present study on
improving the critical moment value of the test
bridge can be well estimated taking account of an
appropriate value for k z in Eqs. (1) or (2). In other
words, the critical moment of a bridge restrained
by a certain warping bracings increases
approximately by a factor of 𝐿𝑏 /(𝑘𝑧 𝑙) compared
with a situation of it without such a restraints,
where 𝐿𝑏 and 𝑙 are the unbraced length of the
compression flange without and with the warping
restraints. The 𝑘𝑧 values of 0.5~0.6 and 0.7~0.8
seem to be reasonable for both-end fixed and for
one-end fixed and the-other-end free to warp
boundary conditions − its value within the range
depending upon the stiffness of the provided
warping restraints. The lower values within the
Table 1 The critical moment values of the test bridges TN4-TN10 estimated on the basis of effective buckling length method.
The critical
Test 𝑳𝒃 [𝒎𝒎]
moment ratio
TN4 𝑀𝑐𝑐1,𝛽,𝑘1 /𝑀𝑐𝑐,𝛽 7500
TN5, TN8-TN10 𝑀𝑐𝑐1,𝛽,𝑘2 /𝑀𝑐𝑐,𝛽 7500
TN6 𝑀𝑐𝑐1,0,𝑘1 /𝑀𝑐𝑐,0 15000
TN7 𝑀𝑐𝑐1,0,𝑘2 /𝑀𝑐𝑐,0 15000
The effective buckling length factor of the test
bridges TN4 and TN6 having a single panel
8
suggested ranges there correspond to a situation in
which adequately stiff warping restraints are
provided for the bridge in question. The critical
moment value of the test bridges TN4-TN10 were
estimated in the basis of the effective buckling
length method explained above and being
compared in Table 1 with the ones that are
obtained through a numerical analysis. Where
𝑀𝑐𝑐,0 is the critical moment of the test bridge
without any bracing system along its span; 𝑀𝑐𝑐,𝛽
is the critical moment of the test bridge having
only one cross-beam at its mid-span; 𝑀𝑐𝑐1,𝛽,𝑘1 and
𝑀𝑐𝑐1,𝛽,𝑘2 are the critical moment values that
correspond to the first eigenmodes of the test
bridge if the bracings as described for the test
TN4, [𝐿𝑏 = (15000/2) mm, 𝑙 = 𝐿𝑏 −
1200 mm], and TN5, [𝐿𝑏 = (15000/2) mm, 𝑙 =
𝐿𝑏 − 2400 mm], are employed; and 𝑀𝑐𝑐1,0,𝑘1 and
𝑀𝑐𝑐1,0,𝑘2 are the critical moment values that
correspond to the first eigenmode of the test bridge
if bracings as described in the Method section for
TN6, [𝐿𝑏 = 15000 mm, 𝑙 = 𝐿𝑏 − 2400 mm], or
TN7, [𝐿𝑏 = 15000 mm, 𝑙 = 𝐿𝑏 − 4800 mm], are
employed. In Table 1, 𝑀𝑐𝑐,0 = 0.029 and 𝑀𝑐𝑐,𝛽 =
0.081 , respectively, are obtained by taking
account of effective buckling length factor to be
𝑘𝑧 = 1.0 in Eq. (1) and the unbraced span of
girders being 15000 mm and 7500 mm in length.
The critical moments calculated by the effective
buckling length method agreed well with the ones
obtained through a three dimensional buckling
analysis of the tests bridges TN4-TN10 performed
in Abaqus [13].
𝑳𝒃 /(𝒌𝒛 𝒍)
𝒍 [𝒎𝒎] 𝒌𝒛 𝒌𝒛 Numerical
(= 𝟎. 𝟓~𝟎. 𝟔) (= 𝟎. 𝟕~𝟎. 𝟖) analyses
6300 − 1.7~1.5 1.5
5100 − 2.1~1.8 2.0
12600 2.4~2.0 − 2.0
10200 3.0~2.5 − 2.9
bracing of Z-type were more closer to the upper
bound of the suggested range for 𝑘𝑧 , whereas they
were more closer to the lower bound of the
suggested range when the test bridge had a double
panel bracing of Warren type near its each support
(i.e. tests TN5 and TN7). These indicate the
double panel bracing employed were more
effective than the single panel brace type in
creating a fixed-end boundary condition for the
lateral rotation of the compression flanges of the
test bridge.
value to force the girders buckle between the brace
point and the end-supports, both girders deformed
in a fashion much similar to the shape of the
second eigenmode of the test bridge TN4 (with a
critical moment value of 0.21𝑀𝑃 ); see Figs. 13-14.
Apparently, the beams followed their initial shape
of imperfection and reached a load level which is
much greater than the critical moment value of
0.12𝑀𝑃 that corresponds to the first eigenmode of
the test bridge TN4.

The load-carrying capacity of the test bridges

TN4-TN10

This section deals with the load-carrying capacity

of the test bridge measured during the tests TN4-
TN10 in which different configuration of bracings
against warping at its support points were being
investigated. In a previous work carried out by the
authors [14] it was found that the shape of initial
imperfections to have substantial effects both on
the load carrying capacity of the studied bridges
Figure 13 Measured lateral deflections of both flanges of
and on the brace forces generated in the cross- north girder at one-third and two-thirds of bridge span during
beams involved. In some of the bridge cases test TN4.
studied there, the girders followed the shape of the
initial imperfection they had and reached a load-
carrying capacity which was much greater than the
critical moment value corresponding to their
lowest eigenmode. Similar trends to this were
observed also during the tests TN4-TN10, which
are being presented here. The solid lines in Figs.
13-21 represent the lateral deformation, ∆, of the
top flanges and the dashed ones graph the lateral
deformation of the bottom flanges of the main
girders at their one-third and two-thirds of span,
where, 𝑀𝑃 is the theoretical plastic bending
capacity of the cross-section of the girders; and Figure 14 Measured lateral deflections of both flanges of
𝑀𝑐𝑐2,𝛽,𝑘1 is the critical moment values that south girder at one-third and two-third of bridge span during
correspond to the second eigenmodes of the test test TN4.
bridge if the bracings as described for the test TN4 The first eigenvalue of the test bridge TN5 is 34%
are employed. The normalized critical moments greater than for the TN4 configuration of it,
that were calculated in the previous section however, the load carrying capacity of the test
together with their corresponding buckling modes bridge TN5 was slightly smaller than the one that
are shown also in the figures. Having smaller was measured for the test bridge TN4 (i.e.
initial imperfections, the south girder 𝑀/𝑀𝑃 ≈ 0.21 and 𝑀/𝑀𝑃 ≈ 0.19, respectively,
demonstrated, overall, smaller out-of-plane were measured for the test bridges TN4 and TN5).
deformations compared with the ones measured This occurred because the north girder of the test
for the north girder. bridge TN5, in contrast to of the test bridge TN4,
Although the stiffness of the cross-beam involved did change substantially its initial imperfection
was 30 times greater than the minimum required from a half-sine to an S shape, this being similar to

9
its first eigenmode that corresponds to the lowest
critical moment value of it.
The south girder of the test bridge TN5, however,
still demonstrated a trend similar to one occurred
for this girder during the test TN4, its being
deformed laterally in a half-sine shape with
comparatively smaller magnitudes of deformation
than the ones were measured for the north girder.
Figure 15 Measured lateral deflections of both flanges of
north girder at one-third and two-thirds of bridge span during
test TN5.
The normalized load carrying capacity versus the
lateral deflection of the north girder at one-third
and two-thirds of its span during the tests TN6-
TN7 are presented in Figs. 16-17.
Figure 16 Measured lateral deflections of both flanges of
north girder at one-third and two-third of its span during test
TN6.
Similar to the tests TN4-TN5, such relatively
slight bracings increased the load-carrying
capacity of the unbraced main girders by a factor
of greater than 2.0. Both girders buckled similar to
their first eigenmodes during the tests TN6-TN7,
the south girder being deformed still with a
smaller magnitude of out-of-plane deformations
than the north girder. Although during the tests
10
TN6-TN7 the unbraced length of the individual
girders were rather unrealistic, the test data for the
load carrying capacity and the brace forces in the
truss bars can be relevant for investigations of the
end-warping bracing of the bridge against system
buckling.
Figure 17 Measured lateral deflections of both flanges of
north girder at one-third and two-third of its span during test
TN7.
Corrugated metal sheets with proper attachments
can reduce also substantially the warping
deformations of the flanges to which they are
attached. The stabilizing performance of them,
however, is greatly affected by the shear flexibility
of the sheets and the occurrence of slip at their
attachments with the steel girders involved. The
normalized load carrying capacity versus the
lateral deformation of the north girder at one-third
and two-thirds of its span during the tests TN8-
TN10 are shown in Figs. 18-20. Similar to the
tests TN4-TN5, the lateral deformations of the
south girder were still much smaller than the ones
measured for the north girder. It was found that the
sheets employed enhanced the load carrying-
capacity of the bridge by a factor of 2.2~2.6 with
respect to the critical moment value of it with only
one cross-beam installed at its mid-span. Having
the same brace span as the test bridge TN5, the
test bridges TN8-TN10 achieved a similar load
carrying capacity to the one that was measured
during the test bridge TN5. The metal sheets
employed during the tests TN8-TN10, however,
demonstrated shear flexibilities in a greater extent
than the Warren type of bracings did during the
test TN5. The improvements carried out during the
tests TN9-TN10 in the number of attachments for
the sheets employed enhanced 9-16% the load-
carrying capacity of the test bridge TN8.

Figure 18 Measured lateral deflections of both flanges of
north girder at one-third and two-third of its span during test
TN8.
Figure 19 Measured lateral deflections of both flanges of
north girder at one-third and two-third of its span during test
TN9.
Fig. 20 Measured lateral deflections of both flanges of north
girder at one-third and two-third of its span during test TN10.
These improvements, which are described in the
Method section, involved attachment of the metal
sheets to the steel girders at their every valley
during the tests TN9-TN10 instead of every
second valley during the test TN8. For each brace
panel involved near the supports, a four-edges-
attached instead of a two-edge-free boundary
condition was another improvement that took
place prior to the test TN10 in the attachments of
the sheets to the steel girders involved.
Figure 21 Measured lateral deflections of top flange at brace
points (Station N1) during tests TN4-TN5 and TN8-TN10.
The shear stiffness of the different bracings
employed in the present study can be better
compared with each other in Fig. 21 in which the
lateral deformations of the north girder at the brace
points (see Station N1 in Fig. 10, its having a
1200 mm distance from the end supports during
the test TN4, and a 2400 mm distance from the
end-support during the tests TN5 and TN8-TN10)
are depicted in the same figure. Attaching the free
edges of the brace panels to the stiffening angle
profiles employed during the test TN10 had little
impact on improving the shear stiffness of them,
however, this resulted in an approximately 10%
greater load carrying capacity compared with the
ones obtained during the tests TN5 and TN9.
Brace forces in the truss bars
Axial brace forces in the truss-bars (see F1-F10 in
Fig. 10) during the tests TN4-TN7 were calculated
on the basis of the mean value of the recorded pair
strains at each bar involved. The brace forces
obtained for the tests TN4-TN5 are presented here
in Figs. 22-23, their being normalized by an axial
force in the compression flange of each girder − a
virtual value that was obtained from dividing the
maximum in-plane bending moment in each girder
by the distance between the centroids of its
flanges. Overall, forces with relatively slight
magnitudes were generated in the truss bars
involved during the tests TN4-TN7. The struts that
11
are marked as 𝐹1 and 𝐹10 in Fig. 10 had a distance
of 187 mm in length from the support points, as
the results, they contributed also to stabilizing the
main girders.
𝑞 plus any external loads in the lateral direction
(wind load for example); the force 𝑁 can be
obtained from 𝑁 = 𝑀𝑚𝑚𝑚 /ℎ if the bracings
restrain the compression flange of a girder with a
constant height of ℎ; and 𝑀𝑚𝑚𝑚 is the maximum
in-plane bending moment in the girder being
restrained.

Figure 22 Normalized values of brace forces generated in

truss bars involved based on recorded strains during test TN4.

These brace forces measured in the truss bars were

compared here with the results of an approximate
method which is used commonly by the design
engineers for analysis of such bracings. In the
analysis of bracing systems that provide the
compression flanges of steel girders with a lateral
stability, Eurocode 3 [15] suggests taking account
of a half-sine shape of geometric imperfection
having a maximum sway of ∆0 = 𝛼𝑚 𝐿/500 for the
flanges to be restrained, where 𝛼𝑚 =
�0.5(1 + 1/𝑚), 𝑚 being the number of girders
being restrained, and 𝐿 is the length of bridge Figure 23 Normalized values of brace forces generated in
truss bars involved based on recorded strains during test TN5.
span. The recommendation described above − that
is also being used as a common practice among The larger the lateral displacements of the
the design engineers to estimate the brace forces in bracings, ∆𝑞 , result in the greater brace forces in
such bracings − yields, by chance, a similar
the bars. The lateral displacements at the brace
imperfection in terms of the shape and magnitude
points should be kept as small as possible in
to the ones which were measured for the
practice so as to rationalize the magnitude of brace
compression flange of the north girder prior to the
forces generated in the bars. In the previous
tests TN4-TN5; see Figs. 11-12. For convenience,
studies [6, 16] it has been found that providing
Eurocode 3 allows also the effects of geometric
bracings with adequate stiffness beyond a certain
imperfection to be replaced with an equivalent
value (two times of the ideal stiffness value
destabilizing force of 𝑞 = 8𝑁 × (∆0 + ∆𝑞 )/𝐿2 obtained by the Winter’s model [16], for example)
being uniformly applied in the lateral direction to results in small deformations, typically within the
the compression flange in question, where ∆𝑞 is magnitude of the geometric imperfections
the lateral deflection of the bracing system due to involved at the brace points. Four relevant

12
buckling modes of (i)-(iv) as shown in Fig. 24, for
each of the test bridges TN4 and TN5 were studied
here, where ∆𝑞 = ∆0 = 𝛼𝑚 𝑙/500; 𝑞1 and 𝑞2 are
calculated based on the Eurocode recommendation
as described above; and 𝑙 is the length of each
half-sine wave in the presumed buckling mode
shape. These result in 𝑞1 = 1.9 × 10−6 (𝑀𝑚𝑚𝑚 /ℎ)
and 𝑞2 = 3.7 × 10−6 (𝑀𝑚𝑚𝑚 /ℎ) [𝑁/𝑚𝑚].
(i)-(iv) are presented in Tables 2-3. The results
obtained by use of equivalent lateral load model
agreed fairly well with the test results presented
earlier in Figs. 22-23.
The approximate method had a conservative
prediction of the brace forces for almost half of the
truss bars in question, however, resulted in unsafe
values with relatively large discrepancies for some
of the truss bars. Although the bracing forces
developed in the truss bars of such an imperfect
structural system are far complex to be accurately
determined, such a method can be very valuable
for design engineers in the preliminary design
stage. For instance, a design on the basis of the
largest force obtained by the approximate method,
together with taking account of an appropriate
factor of safety would provide a reasonable size
for the bracings employed in the test bridges TN4-
TN5.
Table 2 Absolute values of normalized brace forces in truss
bars of test bridge TN4.

Case 𝑭𝟏,𝟏𝟏 𝒉 𝑭𝟐,𝟗 𝒉 𝑭𝟑,𝟖 𝒉

𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎
(𝒊) 0.010 0.035 0.015
Figure 24 Equivalent lateral load models for geometric (𝒊𝒊) 0.002 0.001 0.008
imperfections so as to estimate brace forces in truss bars (𝒊𝒊𝒊) 0.012 0.035 0.015
involved during test TN4-TN5. (𝒊𝒊) 0.004 0.002 0.016
The brace forces in the truss bars of the tests Test TN4 0.025 0.031 0.010
bridges TN4-TN5 obtained for the load cases of

Table 3 Absolute values of normalized brace forces in truss bars of test bridge TN5.

Case 𝑭𝟏,𝟏𝟏 𝒉 𝑭𝟐,𝟗 𝒉 𝑭𝟑,𝟖 𝒉 𝑭𝟒,𝟕 𝒉 𝑭𝟓,𝟔 𝒉

𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎 𝑴𝒎𝒎𝒎
(𝒊) 0.012 0.030 0.002 0.028 0.011
(𝒊𝒊) 0.002 0.000 0.002 0.001 0.006
(𝒊𝒊𝒊) 0.015 0.037 0.004 0.029 0.012
(𝒊𝒊) 0.004 0.000 0.004 0.002 0.013
Test TN5 0.017 0.008 0.037 0.055 0.025

Effects of warping restraints employed on the test bridge had fairly similar imperfections during
the tests TN4-TN10 in terms of the shape and
brace forces generated in the cross-beam
magnitude (see Figs. 11-12), this enables the brace
forces generated in the cross-beam of the test
The magnitude of brace forces generated in the
bridges TN4-TN10 to be comparable with each
cross-beams can be affected greatly by the shape
other ignoring the effects of initial imperfections
and magnitude of geometric imperfections that the
into account. In all, eight strain gauges were glued
main girders have initially [14]. Each girder of the

13
on the farthest layers of the flanges for the cross-
beam of the test bridge, these recording the strain
values at approximately 200 mm distance from the
web of the main girders; see SG1-SG4 and NG1-
NG4 in Fig. 10. The largest values of the normal
stress, 𝜎𝑚𝑚𝑚 , at the location of the stain gauges
SG1-SG4 and NG1-NG4 were normalized by the
in-plane bending moment of the main girders, 𝑀;
divided by the in-plane elastic section modulus of
the cross-beam’s section, 𝑆𝑏𝑏,𝑦 , see Fig. 25. It was
shown earlier in Fig. 21 that the bracings
employed during the tests TN8-TN10
demonstrated greater shear flexibilities compared
with the plan bracing of the same braced-span did
during the test TN5. The brace forces in the cross-
beam of the test bridges TN8-TN10 were also
greater than the ones obtained for the test bridge
restrained by the plan bracings employed during
the test TN4-TN5; see Fig. 25. The stiffer were the
warping bracings, the smaller brace forces were
generated in the cross-beam.
Figure 25 Normalized value of maximum stress calculated in
cross-beam based on recorded strains during tests TN4-TN5
and TN8-TN10.
In contrast to the test TN4 in which both girders of
the test bridge deformed in a half-sine-wave shape,
the compression flange of the north girder buckled
in an S shape fashion during the tests TN5, and
TN8-TN10. The asymmetric out-of-plane
deformation (with respect to the mid span of the
test bridge) of the north girder during the tests
TN5 and TN8-TN10 resulted in greater lateral
bending in the cross beam in question compared
with its situation during the test TN4 where the
north girder deformed laterally in a fairly
symmetric fashion. The contribution of the lateral
bending of the cross beam to its longitudinal
stresses at the location of strain gauges SG1-SG4
14
and NG1-NG4, therefore, was lesser during the
test TN4 than the others.
Conclusions
Global lateral torsional buckling is a failure mode
that has occurred in construction of the Marcy
Bridge in New York and of the Bridge Y1504 in
Sweden, both in 2002. The corrugated metal
sheets were designed so as to stabilize the steel
girder during the concreting of it deck. The
adequacy of the employed sheets and their
attachments were questioned in the failure cause
evaluations.
The use of slight bracings against warping of a
bridge near its end-supports can enhance greatly
the load-carrying capacity of such bridges. These
can be achieved by either the typical plan bracings
or the corrugated metal sheets of end-closed type
having appropriate attachments with the main
girders. The present paper reports the results of
seven full-scale laboratory tests in which the plan
bracings and metal sheets employed were installed
near the supports of it and enhanced the load-
carrying capacity of the test bridge by a factor of
greater than 2.0. The metal sheets employed here,
however, showed significantly larger lateral
deflections than the utilized plan bracings did.
Attaching the free edges of the brace panels to the
stiffening angle profiles spanning between the top
flanges of the adjacent girders had little impact on
improving the shear stiffness of them, however,
this resulted in an approximately 10% greater load
carrying capacity as compared both with the
similar sheets without use of such stiffening angles
and with the truss bracings employed having
similar length of brace span. The effective
buckling length method was examined also to
predict the critical moment value of the test bridge
that was restrained by the warping restraints
employed. It turned out that this method can be
very valuable in preliminary design of steel
bridges that are restrained by such partial span
bracings.
Relatively small forces were generated in the truss
bars in order to achieve these marked
improvements in the load-carrying capacity. The
brace forces in the employed truss bars were
compared with the ones obtained through an
approximate method being used frequently by the
design engineers for the analysis of such bracings.
The results obtained by the approximate method −
which is suggested also by Eurocode 3 − agreed
fairly well with almost half of the brace forces
measured during the tests whereas, the predication
gave unsafe results with very large discrepancy for
few of the bars. Although the bracing forces in the
truss bars of a real bridge in practice having
complex imperfection shapes are very difficult to
be properly determined, such a method can be of
great benefit for design engineers in the
preliminary design stage.
Finally, the brace forces at both ends of the cross
beam of the test bridge were calculated in the basis
of the recorded strains, this occurred varying the
configurations of the warping restraints of the test
bridge. Relatively small forces were generated in
the cross beam during the tests. The brace forces
in the cross beam in question were greater when
the metal sheets rather than the plan bracings
employed were installed on the test bridge. It was
found that the shape of imperfections affected in a
great extent the axial stress values generated in the
flanges of the cross-beam involved due to its
lateral bending.
Acknowledgements
The financial support from “The Lars Erik
Lundbergs Stipendiestiftelse” to this study, the
research grant from Byggrådet to support the
laboratory tests, and the material support by
Structural Metal Decks Ltd for the required
corrugated sheets is gratefully acknowledged.

References
[1] H. Mehri and R. Crocetti, "Scaffolding bracing of composite bridges during construction " Journal of
bridge Engineering (ASCE), vol. DOI:10.1061/(ASCE)BE.1943-5592.0000829, 2015.
[2] H. Mehri and R. Crocetti, "Bracing of steel-concrete composite bridge during casting of the deck,"
presented at the Nordic Steel Construction Conference 2012, Oslo, Norway 2012.
[3] J. A. Yura, T. Helwig, R. Herman, and C. Zhou, "Global lateral buckling of I-shaped girder systems,"
Journal of Structural Engineering, vol. 134, pp. 1487-1494, 2008.
[4] S. Timoshenko and J. M. Gere, Theory of elastic stability. New York: McGraw-Hill, 1961.
[5] R. D. Ziemian, Guide to stability design criteria for metal structures 6th Ed., 6 ed. Hoboken, New Jersey:
John Wiley & Sons, 2010.
[6] J. A. Yura, "Fundamentals of beam bracing," Engineering Journal, pp. 11-26, 2001.
[7] J. A. Yura and J. A. Widianto, "Lateral buckling and bracing of beams, a re-evaluation after the Marcy
Bridge collapse," in Structural Stability Research Council - 2005 Annual Stability Conference, Apr 6 - 9
2005, Montreal, QB, Canada, 2005, pp. 277-294.
[8] O. Egilmez, T. Helwig, and R. Herman, "Buckling behavior of steel bridge I-girders braced by permanent
metal deck forms," J. of Bridge Eng., vol. 17, pp. 624-633, 2012.
[9] AASHTO, "AASHTO LRFD bridge specificatios," in Third Edition, ed: American Assosiation of State
Highway and Transportation Officials, 2005.
[10] H. Mehri, "Bracing of steel bridges during construction; theory, full-scale tests and simulations," Ph. D.
Thesis (TVBK-1049), Division of Structural Engineering, Lund University, Sweden, Jan. 2016.
[11] L. D. Luttrell, "diaphragm design manual," ed: Steel Deck Institute 1981.
[12] ECCS, "European recommendations for the application of metal sheeting acting as a diaphragm," in ECCS
publication, ed, 1995.
[13] ABAQUS/CAE, V 6.13 [computer software]: RI, Simulia., 2013.
[14] H. Mehri, R. Crocetti, and J. A. Yura, "Effects of geometric imperfections on bracing performance of cross-
beams during construction of composite bridges," Engineering Structures, Submitted on 19 July 2015.
[15] EC3, "Design of steel structures –Part 2: Steel bridges," in SS-EN 1993-2:2006 ed. Stockholm, Sweden:
Swedish Standards Institute, 2009.
[16] G. Winter, "Lateral bracing of columns and beams." vol. Vol. 125, No. 1, , ed: ASCE, 1960, pp. 807-826

15