See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245305189

Brace Forces in Steel Box Girders with Single Diagonal Lateral Bracing Systems

Article  in  Journal of Structural Engineering · August 2006

DOI: 10.1061/(ASCE)0733-9445(2006)132:8(1212)

CITATIONS READS
13 840

2 authors, including:

Kyungsik Kim
Cheongju University
19 PUBLICATIONS   177 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Kyungsik Kim on 31 May 2014.

The user has requested enhancement of the downloaded file.

 Brace Forces in Steel Box Girders with Single Diagonal
Lateral Bracing Systems
Kyungsik Kim1 and Chai H. Yoo, F.ASCE2

Abstract: Trapezoidal steel box girders 共also known as tub girders兲 may be at their critical stage during construction because the
noncomposite steel section must support both the wet concrete and the entire construction load. A lateral bracing system is usually
installed at the top flange level to form a quasi-closed box, thereby increasing the torsional stiffness during construction. Typical lateral
bracing includes a single-diagonal type 共SD type兲 and a crossed-diagonal type 共XD type兲. The brace forces in box girders with XD type
lateral bracing systems can now be estimated with good accuracy. In some horizontally curved box girders with SD-type systems,
substantial axial forces, possibly up to 35% of the total force, are developed in struts due to induced torsion and distortion, which are
neglected in previous studies. In this study, analytical equations were formulated to compute the brace forces developed in bracing
members by taking into account bending and torsional actions of tub girders with SD-type lateral bracing systems. Brace forces in both
diagonals and struts can be estimated by these new equations more accurately than with the previous procedures. Member forces due to
bending and torsion computed using the new equations were compared with those evaluated by three-dimensional finite element analyses,
and excellent correlation was found.
DOI: 10.1061/共ASCE兲0733-9445共2006兲132:8共1212兲
CE Database subject headings: Box girders; Steel; Bracing; Stiffness.

Introduction
Steel box girder systems have long been preferred superstructure
types for both horizontally curved and straight bridges because of
their structural and aesthetic advantages. A closed box girder in a
completed bridge has a superb torsional stiffness that may be 100
to more than 1,000 times that of a comparable I-girder 共Heins and
Hall 1981兲. This large torsional stiffness makes box girders par-
ticularly well suited for horizontally curved box girder bridges, in
which the bridge geometry may induce large torsional moments
in the girders. A typical box girder bridge system consists of one
or more steel tub girders that act compositely with a cast-in-place
concrete roadway deck. A schematic of a typical twin-girder
bridge cross-section is shown in Fig. 1.
Although the composite box section has a superb torsional
rigidity, the noncomposite steel section is critical when subjected
to large torsional loading during the early stages of bridge con-
struction prior to hardening of the concrete deck. The noncom-
posite dead load stress may account for up to 60–70% of the total
stress for a typical box girder bridge 共Topkaya and Williamson
2003兲. A lateral bracing system, sometimes referred to as a hori-
1
Research Engineer, GS Engineering & Construction, Seoul, Korea;
formerly, Postdoctoral Fellow, Dept. of Civil Engineering, Auburn Univ.,
Auburn, AL 36849-5337.
2
Professor, Dept. of Civil Engineering, Auburn Univ., Auburn, AL
36849-5337 共corresponding author兲. E-mail: chyoo@eng.auburn.edu
Note. Associate Editor: Donald W. White. Discussion open until
January 1, 2007. Separate discussions must be submitted for individual
papers. To extend the closing date by one month, a written request must
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on February 24, 2004;
approved on October 5, 2005. This paper is part of the Journal of Struc-
tural Engineering, Vol. 132, No. 8, August 1, 2006. ©ASCE, ISSN
0733-9445/2006/8-1212–1222/$25.00.
zontal top truss, is usually installed at the top flange level to form
a quasi-closed section, thereby increasing the torsional stiffness
during transport, erection, and construction. Single-diagonal
共SD-type兲 or crossed-diagonal 共XD-type兲 bracing systems are
typically used for a lateral bracing system and are shown in
Fig. 2. SD-type lateral bracing systems show a noteworthy char-
acteristic behavior that is not exhibited in XD-type lateral bracing
systems: Top flanges of box girders with SD types are subjected
to lateral bending, similar to that of continuous beams with inter-
mittent supports, due to the interaction between top flanges and
lateral bracing members. Since there are diagonals at both ends of
a strut in an XD-type bracing, the strut forces do not cause lateral
bending in the flanges. Although rudimentary research results
have been reported 共Heins and Hall 1981; AASHTO 1993兲, Fan
and Helwig 共1999, 2002兲 are believed to be the first to make a
significant contribution to our understanding of the mechanisms
involved in lateral bracing systems and internal cross frames.
They successfully presented analytical methods that can be used
to estimate the brace forces in both SD-type and XD-type lateral
bracing systems of box girders subjected to vertical loads and/or
applied torque. According to the equations given by Fan and
Helwig 共1999兲, strut forces are assumed to be induced only by the
bending of the box girder and its lateral force components due
to the inclined webs, regardless of whether it is an SD-type or
XD-type lateral bracing system. It is reasonable to assume that
strut forces are induced only by the bending of the box girder and
its lateral force components in all XD-type lateral bracing sys-
tems, but in the case of an SD-type lateral bracing system, a
considerable portion of the strut force developed is induced by the
torsional moment. In addition to the torsional effect, brace forces
in an SD-type system subjected to vertical bending can be more
exactly evaluated by considering a logical redistribution of lateral
force components 共the bending effect兲. A detailed analysis of this
redistribution of lateral forces will be given later. In this study,

1212 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006

 Fig. 1. Superstructure of tub girder bridge
presented are equations to predict brace forces in SD types for-
mulated analytically. Brace forces in an SD-type system com-
puted using the new analytical equations were compared with
those obtained from three-dimensional finite element analyses
共FEA兲, and an excellent correlation was found.
Background
For a quasi-closed box girder, a torsional analysis can be per-
formed using the equivalent plate method 共EPM兲 in which the
lateral bracing system is transformed into a fictitious plate with a
uniform thickness. Kollbrunner and Basler 共1969兲 developed
equations for the equivalent thickness of several types of bracing
systems by evaluating the strain energy stored in the system.
Dabrowski 共1968兲 presented similar equations to determine the
fictitious plate thickness based on the consistent deformation
theory. The quasi-closed box theory, or EPM, allows the torsional
properties of the tub girder to be approximated. The value of the
equivalent plate thickness is dependent on the bracing configura-
tion and cross-sectional areas of bracing members. The resulting
shear flow in a closed section, q, is equal to T / 共2A0兲 where T and
A0 are the torsional moment and the enclosed area of the box,
respectively. The shear flow acting on the fictitious plate is then
transformed to the axial forces acting on the diagonal members in
the lateral bracing system as
qb b D = Dtor + Dbend + Dlat
Dtor,SD = ± = ± T 共1兲
sin ␣ 2A0 sin ␣
Fig. 2. Lateral bracing system: 共a兲 SD type; 共b兲 XD type
JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006 / 1213
Fig. 3. Forces affecting bracing members: 共a兲 longitudinal
components; 共b兲 lateral components
qb b
Dtor,XD = ± = ± T 共2兲
2 sin ␣ 4A0 sin ␣
where Dtor,SD, Dtor,XD = diagonal forces due to applied torque in
the SD type and XD type, respectively.
It should be noted, however, that both torsion and vertical
bending induce forces in the lateral bracing members. Diagonals
in the lateral bracing system are subjected to the same total lon-
gitudinal deformation as the top flanges, as shown in Fig. 3共a兲. In
addition, the lateral force component resulting from the inclined
webs also affects the brace forces in the lateral bracing system.
The magnitude of the lateral force component can be evaluated
from the equivalent moment induced by the applied load on the
top flange, as shown in Fig. 3共b兲. Fan and Helwig 共1999兲 devel-
oped equations to predict brace forces in both SD- and XD-type
systems and proposed the following expressions for design
purposes
共3兲
S = Sbend + Slat 共4兲
where D , S = respective total forces in the strut and diagonal;
Dbend,Sbend = respective forces in the strut and diagonal due to
bending of a box girder; Dlat,Slat = respective forces in the strut and
diagonal due to lateral force components; and Dtor = diagonal
force from the torsional moment, determined using the EPM
关Eqs. 共1兲 and 共2兲兴. In addition, Fan and Helwig 共1999兲 recom-
mended designing the strut to carry the entire lateral load compo-
nent, i.e., Slat = swlat and Dlat = 0, where s = spacing between struts;
and wlat = lateral force component. This is essentially tantamount
to stating that the entire lateral force is carried by the struts alone,
and the torsional moment induces no forces in the struts at all. It
has been observed that there can be up to a 30% discrepancy
between the values predicted by Eq. 共4兲 and those from the FEA
in some box girder bridges 共Kim 2004兲. This lack of agreement
motivated the authors to undertake this research project.
Consider the three types of top flange lateral bracing systems
shown in Fig. 4. The box girder used in the example has a length
of 160 ft 共48.8 m兲 and is made up of 16 panels. Internal cross
frames are installed under every other strut location. Two differ-
ent arrangements 共A and B兲 are considered in the SD-type lateral

Fig. 4. Top lateral bracing systems: 共a兲 SD-type A; 共b兲 SD-type B;
共c兲 XD type; 共d兲 cross-sectional properties; 共e兲 applied torque;
and 共f兲 applied vertical load 共1 in. = 25.4 mm; 1 ft= 0.305 m;
1 k / ft= 14.6 kN/ m兲
bracing system. Fig. 4 shows lateral bracing systems 关SD types A
in Fig. 4共a兲 and B in Fig. 4共b兲, along with an XD type in
Fig. 4共c兲兴, in addition to the box girder dimensions, and bracing
members shown in Fig. 4共d兲. Applied torque and vertical loads
are illustrated in Figs. 4共e and f兲, respectively.
Under torsional loading, the forces developed in the diagonals
of all three types were in good agreement with the results from
Eqs. 共1兲 and 共2兲, as expected. For strut forces, however, a discern-
able difference was detected. Contrary to the suggestion by Fan
and Helwig 共1999兲 in Eq. 共4兲, strut forces are developed under
torsional loads in SD-type bracing, as shown in Fig. 5. Strut
forces are induced in an SD-type bracing due to interactions be-
tween the top flanges and bracing members. A detailed investiga-
tion of a box girder braced by an SD-type system has been carried
Fig. 5. Strut forces due to torsion 共1 kip= 4.45 kN兲
1214 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006
out and the resulting data were incorporated into predictor equa-
tions for the determination of brace forces under a general
loading.
Finite Element Modeling
A general purpose finite element program, ABAQUS 共ABAQUS,
Inc. 2003兲, was used in the numerical analyses. The cross sections
of the box girders were built up with three-dimensional shell
elements 共S4R of ABAQUS兲. The S4R elements are four-node
doubly curved general-purpose shell elements characterized by
reduced integration with hourglass control. The solid diaphragms,
also modeled with S4R elements, were placed at both supports for
simply supported girders and also at the interior piers for three-
span continuous curved girders. The diagonals of the lateral brac-
ing system and the internal cross frames were modeled with
three-dimensional two-node truss elements 共T3D2 of ABAQUS兲.
A minimum of eight SR4 elements were used for each top flange
and ten SR4 elements were used to model the bottom flanges and
webs. Experience has shown that such grid refinement yields rea-
sonable accuracy for finite element numerical analyses of typical
tub girders. Both K- and X-shaped internal cross frames were
examined. Internal cross frames were placed at every other strut
location. In the case of K frames, the struts of the lateral bracing
system also acted as top transverse members of the internal K
frame. However, the strut force was computed from the lateral
bracing system at this stage of evaluation as the additional force
component induced as part of the internal cross frame would be
added at a later stage. Space beam elements 共B31 of ABAQUS兲
were used for the struts simply to avoid instability, as an unstable
situation occurs at the K joint where two legs meet if four truss
elements are used to model a K frame. The rotational degree of
freedom 共DOF兲 of the box was suppressed at the supports by
constraining vertical DOF of the nodes for the bottom corners of
the box cross section near the bearing device. Of these vertically
constrained nodes, at least one horizontal DOF was also con-
strained in order to prevent rigid body motion. As there is a solid
diaphragm installed at every support, this boundary condition is
believed to be equivalent to suppressing the rotational DOF of the
box within the extent of ignoring the elastic deformation of the
solid diaphragm itself. Linear elastic FEAs were carried out on
noncomposite steel structures using the mill specified modulus of
elasticity and Poisson’s ratio for the construction steel of
29,000 ksi 共200 GPa兲 and 0.3, respectively.
Fig. 6. Torsion and distortion components 共adopted from Fan and
Helwig 2002兲
 Fig. 9. Lateral force resultants from diagonal forces

ments occur at both ends of the box girder. This phenomenon is

attributable to the fact that there is no adjacent diagonal, as shown
Fig. 7. Axial deformation components: 共a兲 diagonals; 共b兲 struts in Fig. 9. For simplicity, consider top flanges and struts as a
separate two-dimensional structure in the horizontal plane, as
illustrated in Fig. 10共a兲. The two top parallel flanges may be
Bracing Member Forces due to Torsional Loads considered as beams connected to each other by the struts along
the span and the solid diaphragms at both ends. The net lateral
Fan and Helwig 共2002兲 studied the distortion of trapezoidal box loads induced from the diagonals acting on this simplified struc-
girders and separated the pure torsional components and distor- ture may be divided into two sets of force components, as shown
tional components from the applied torque as shown in Fig. 6. in Figs. 10共b and c兲. The forces shown in Fig. 10共b兲 are carried
The same model box girders as those in Fig. 4 were analyzed. The entirely by the two top flanges, while the forces shown in
diagonal member forces in lateral bracing systems are affected by Fig. 10共c兲 are carried by the two top flanges and the struts. The
longitudinal deformations of the top flanges and their lateral dis- net forces applied in the adjacent repeating panels and the corre-
placements as shown in Fig. 7共a兲. For the strut member, however, sponding deformation configurations are illustrated in Fig. 10共d兲.
lateral displacement components of the top flanges are a major Although DH is a function of torsional moment that varies along
parameter in the development of member forces as shown in
Fig. 7共b兲. Longitudinal deformations of the top flanges of a box
girder due to torsion alone are relatively small compared to those
due to vertical bending. Fig. 8 shows the relative lateral displace-
ments between two top flanges, defined as vA − vB as shown in
Fig. 7共b兲. The lateral force components, DH, shown in Fig. 9, the
vector sums of the two adjacent diagonal forces at each diagonal-
strut junction, are expressed as
DH = 共D− + D+兲sin ␣ 共5兲
− +
where D ,D = torsion-induced axial forces in two consecutive di-
agonals. It should be noted that D− and D+ are functions of the
torsional moment along the box girder and in opposite signs in
two adjacent bays in the case of an SD-type bracing system. It is
seen from Fig. 8 that the maximum differential lateral displace-

Fig. 10. Lateral forces from diagonals: 共a兲 lateral forces; 共b兲 lateral
Fig. 8. Differential lateral displacements of top flanges due to torsion force affecting lateral bending; 共c兲 lateral force affecting struts; and
and distortion 共1 in. = 25.4 mm兲 共d兲 deformation of two repeating adjacent panels

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006 / 1215


Fig. 11. Lateral forces from cross-frames: 共a兲 lateral forces; 共b兲
lateral force affecting lateral bending; 共c兲 lateral force affecting
bracing members; 共d兲 deformation of two repeating adjacent panels;
and 共e兲 interface forces
the girder length, strut forces are approximated from lateral forces
that are assumed to be the same in magnitude in two adjacent
panels, a reasonable approximation as any difference will be
small. The force developed in the strut member can be expressed
共Kim 2004兲 by examining the equilibrium of forces illustrated in
Fig. 10共d兲.
共2s兲3
192I f
冉 冊
Stor = DH 共6兲
b 共2s兲3
+ EAD
2AS 192I f
where Stor = strut force due to pure torsional component;
I f = second moment of inertia of one top flange about z axis;
AS = cross-sectional area of a strut; and s = spacing between struts.
Axial forces are also developed in struts due to distortional
components, as illustrated in Fig. 8. Diagonal members of internal
cross-frames resist distortional deformations of the box cross sec-
tion and consequently the resulting member forces are transferred
to lateral bracing, as illustrated in Fig. 11共a兲. The magnitude of
the horizontal force components, KH, can be evaluated approxi-
mately by multiplying horizontal forces from distortional compo-
nents, denoted as qH,dist in Fig. 6, by the internal cross-frame
spacing, 2s, which yields 共Fan and Helwig 2002兲
KH = 2sqH,dist = 冉
s M a
− ew
2h R b
冊 共7兲
Duplicating the procedure used in the development of forces in
struts due to pure torsional components illustrated in Fig. 10共a兲,
1216 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006
the net forces applied in the adjacent repeating panels and corre-
sponding deformation configurations are illustrated in Fig. 11共d兲.
Although KH is a function of bending and/or torsional moment
that varies along the girder length, bracing member forces can
also be approximated from the lateral forces that are assumed to
be the same in magnitude within two adjacent panels. Since the
lateral stiffness of the web is negligible 共Kim 2004兲, individual
top flanges should be in equilibrium in the horizontal plane by
three brace force components only. Thus, summing the three
brace force resultants acting on one flange in Fig. 11共e兲 yields
Sdist = − Ddist sin ␣ 共8兲
where Sdist, Ddist = forces in struts and diagonals due to distortion.
Forces in diagonal members can be determined by their elonga-
tions and force-deformation relationships represented by Hooke’s
law. The elongation of the diagonal member, ␦D, shown in Fig. 7,
is given by
␦D = 共v1 + v2兲sin ␣ + u cos ␣ 共9兲
where v1 = relative displacement of the top flange in the lateral
direction as shown in Fig. 11共d兲; and v2 = axial elongation of the
strut; u = relative displacement of panel point in the longitudinal
direction. It is noted that the relative displacement, u, in the lon-
gitudinal direction of each panel point is not significant due to
torsion, thus neglected. Each top flange is assumed to bend in a
manner similar to a continuous beam between panel points, as
shown in Fig. 11共d兲. The interface forces between the top flange
and bracing assemblies are shown in Fig. 11共e兲. The relative
lateral deflection between two consecutive panel points, v1, is
computed to be
v1 =
共2s兲3 KH
192EI f 2
冉+ Sdist 冊 共10兲
where E = modulus of elasticity. Similarly, elongations of the strut,
v2, and the diagonal, ␦D, are given by
bSdist
v2 = 共11兲
EAS
LDDdist
␦D = 共12兲
where AS,AD = cross-sectional area of the strut and diagonal,
respectively; and LD = length of the diagonal. Substituting
Eqs. 共10兲–共12兲 into Eq. 共9兲 and solving Eqs. 共8兲 and 共9兲 simulta-
neously for Ddist yields
ADASs3 sin ␣
Ddist = ± KH
48ASLDI f + 2ASADs3 sin2 ␣ + 48ADbI f sin2 ␣
共13兲
where the positive sign denotes SD-type A and the negative sign
SD-type B. Equation 共13兲 was derived to compute the lateral di-
agonal member force in a simple-span box girder, but it can also
be extended to continuous girders. However, care must be exer-
cised near interior supports where solid diaphragms are installed
and it is recommended that forces in the lateral diagonal due to
distortion be neglected at the panels immediately adjacent to in-
terior solid diaphragms.

Fig. 12. Interface forces due to lateral force components 共wlat兲
Bracing Member Forces due to Vertical Bending
Equations for forces in SD-type lateral bracing members due to
vertical bending in a tub girder may be derived based on the
following assumptions:
1. Internal cross frames are assumed to be installed at every
other strut location.
2. Longitudinal and lateral force components acting on top
flange and bracing system of a box girder subjected to verti-
cal bending are considered separately as shown in Figs. 3共a
and b兲.
Fig. 13. Lateral displacements of top flanges due to lateral force
components 共wlat兲
JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006 / 1217
3. The webs have negligible lateral resistance against lateral
bending of the top flanges. This assumption was verified by
analyzing a number of examples of hypothetical box girders,
where the web contribution to resisting top flange lateral
bending was less than 3% 共Kim 2004兲.
The first assumption has an important implication for the behavior
of box girders with SD-type lateral bracing systems. If internal
cross frames are installed at every strut location, the lateral brac-
ing system and internal cross-frames interact substantially differ-
ently from those described in this study, and it is expected that
this complex behavior will be the subject of further study. Fan
and Helwig 共1999兲 successfully derived equations to predict the
brace forces induced by longitudinal stresses on the top flanges as
shown in Fig. 3共a兲. Although Fan and Helwig 共1999兲 recom-
mended that, for simplicity, the strut be assumed to carry the
entire lateral force components, a logical redistribution of lateral
force components to both struts and diagonals leads to a more
accurate estimation of the brace forces.
Consider an SD-type lateral bracing system subjected to lateral
force components, as shown in Fig. 3共b兲. The interactive forces
between the top flanges and lateral bracing members are shown in
Fig. 12. Interactive forces QA and QB can be determined from
simple equilibrium considerations
QA + QB = 2swlat 共14兲
QA = 2Dlat sin ␣ + Slat 共15兲
QB = Slat 共16兲
where s = spacing between struts; and Dlat,Slat = forces in the diag-
onal and strut due to lateral force components, respectively. Sub-
stituting Eqs. 共15兲 and 共16兲 into Eq. 共14兲 and rearranging yields
Slat = swlat − Dlat sin ␣ 共17兲
The induced lateral force component due to inclined webs causes
the top flange to bend in a manner similar to that seen in a con-
tinuous beam between panel points. As shown in Fig. 13, the
relative lateral deflection between two adjacent panel points, v1, is
determined by superimposing the deflections due to the distrib-
uted lateral force components, wlat, and the concentrated load, QA
wlat共2s兲4 QA共2s兲3
v1 = − 共18兲
384EI f 192EI f
The elongations of the strut, v2, and the diagonal, ␦D, are
Slatb
v2 = 共19兲
EAS
DlatLD
␦D = 共20兲
EAD
Substituting Eqs. 共18兲–共20兲 into Eq. 共9兲 and solving for Dlat yields
the following expression for the diagonal member force induced
by lateral loading
24ADbsI f sin ␣
Dlat = wlat 共21兲
24ASLDI f + AD共ASs3 + 24bI f 兲sin2 ␣
Table 1. Diagonal Forces Three-Span Continuous Horizontally Curved Girder 共kips兲
Proposed equations
Panel Difference
number Dbend Dlat Dtor Ddist Sum FEA 共%兲
1 −1.78 0.83 −33.82 −0.32 −35.09 −33.36 5.2
2 −4.90 0.83 30.91 −0.88 25.97 24.93 4.2
3 −7.29 0.83 −25.66 −1.31 −33.42 −31.93 4.7
4 −8.96 0.83 18.68 −1.60 8.96 8.29 8.1
5 −9.90 0.83 −10.60 −1.77 −21.44 −19.62 9.3
6 −10.13 0.83 2.01 −1.81 −9.09 −10.11 −10.1
7 −9.63 0.83 6.45 −1.72 −4.07 −2.17 87.3
8 −8.40 0.83 −14.18 −1.51 −23.26 −24.37 −4.6
9 −6.46 0.83 20.58 −1.16 13.80 15.47 −10.8
10 −3.79 0.83 −25.01 −0.68 −28.64 −29.78 −3.8
11 −0.40 0.83 26.86 −0.07 27.22 28.69 −5.1
12 3.51 1.07 −25.54 0.57 −20.39 −22.38 −8.9
13 8.08 1.07 20.41 1.31 30.86 33.20 −7.0
14 13.33 1.07 −10.87 2.16 5.69 3.77 50.9
15 16.97 1.32 −3.68 2.57 17.17 18.26 −6.0
16 22.79 1.32 23.87 47.98 45.25 6.0
Diaphragm
17 22.70 1.32 5.97 29.99 28.69 4.5
18 16.69 1.32 14.02 2.52 34.55 34.15 1.2
19 12.80 1.07 −28.19 2.07 −12.25 −13.94 −12.1
20 7.33 1.07 37.15 1.19 46.74 47.59 −1.8
21 2.55 1.07 −41.50 0.41 −37.47 −37.99 −1.4
22 −1.64 0.83 41.88 −0.29 40.78 40.74 0.10
23 −5.26 0.83 −38.87 −0.94 −44.24 −43.87 0.8
24 −8.15 0.83 33.10 −1.46 24.32 24.57 −1.0
25 −10.32 0.83 −25.18 −1.85 −36.52 −36.11 1.1
26 −11.77 0.83 15.72 −2.11 2.67 3.15 −15.0
27 −12.49 0.83 −5.34 −2.24 −19.24 −18.77 2.5
Symm.
Note: 1 kip= 4.45 kN.

Equations for Brace Forces ADASs3 sin ␣

Ddist = ± KH
48ASLDI f + 2ASADs3 sin2 ␣ + 48ADbI f sin2 ␣
Forces in diagonals, D, and struts, S, in the SD-type lateral
bracing system due to vertical bending and torsional loading are 共27兲
summarized as follows:

D = Dbend + Dlat + Dtor + Ddist 共22兲 Sbend = − Dbend sin ␣ 共28兲

S = Sbend + Slat + Stor + Sdist 共23兲

where subscripts bend, lat, tor, and dist denote the brace forces
induced by vertical bending, lateral load, pure torsion, and distor-
tion, respectively. They are

f xTops cos ␣
Dbend = 共24兲
LD b s3
+ sin2 ␣ +
AD AS 24I f
where f xTop = longitudinal stress in the middle of the top flange

24ADbsI f sin ␣
Dlat = wlat 共25兲
24ASLDI f + AD共ASs3 + 24bI f 兲sin2 ␣

b Fig. 14. Comparison of diagonal forces in straight girder due to

Dtor = ± T 共26兲 torsion 共1 kip= 4.45 kN兲
2A0 sin ␣

1218 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006

Fig. 15. Comparison of strut forces in straight girder due to torsion
共1 kip= 4.45 kN兲

Slat = swlat − Dlat sin ␣ 共29兲

Fig. 17. Three-span continuous horizontally curved tub girder
共2s兲3 共1 ft= 0.305 m; 1 in. = 25.4 mm兲
192I f

冉 冊
Stor = DH 共30兲
b 共2s兲3
+ tems as shown in Figs. 4共a and b兲. The box girder properties and
2AS 192I f
and
Sdist = − Ddist sin ␣ 共31兲
Equations 共24兲, 共26兲, and 共28兲 were adopted from Fan and Helwig
共1999兲. It should be noted that all force components are functions
of bending and/or torsional moments. Although Dlat and Ddist are
fairly small except near interior supports, it is the authors’ opinion
that force components that approach 20% of the other major
forces must not be ignored. A detailed quantitative comparison of
each contribution is given in Table 1.
In the case of nonprismatic flanges, the flange second moment
of inertia with respect to a vertical axis passing the centroid of the
flange may be determined by taking a weighted average moment
of inertia reflecting the transition length ratio within the panel.
vertical and torsional loading are given in Fig. 4. Fig. 14 com-
pares forces in the diagonals determined by Eq. 共22兲 with results
from the FEA. The strut forces computed from Eq. 共23兲 are com-
pared with those obtained from the finite element analysis in
Fig. 15. Excellent agreement is evident between proposed equa-
tions and FEA for strut forces. As indicated in Eq. 共4兲, Fan and
Helwig 共1999兲 assumed torsional loads not cause any strut forces.
Axial forces in the SD-type lateral bracing members of the
straight box girders shown in Fig. 4 subjected to vertical bending
are comparatively shown in Fig. 16. Note that the forces deter-
mined from Eqs. 共22兲 and 共23兲 are only 6% greater than those
obtained from FEA, while the values computed from Fan and
Helwig 共1999兲 estimate brace forces 14% greater than those from
FEA at the mid-span. This improvement is believed to be due to a
logical redistribution of the lateral force components due to ver-
tical bending to both the struts and diagonals.

Comparison of Lateral Bracing Forces

The lateral bracing forces computed from Eqs. 共22兲 and 共23兲 were
compared with the results from FEAs of a simply supported
straight box girder with two different types of lateral bracing sys-

Fig. 16. Forces in bracing members in straight girder due to vertical Fig. 18. Bending and torsional moment diagrams of three-span
bending 共1 kip= 4.45 kN兲 continuous horizontally curved girder 共1 k ft= 1.356 kN m兲

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006 / 1219

Table 2. Strut Forces Three-Span Continuous Horizontally Curved Girder 共kips兲
Proposed equations
Strut Difference
number Sbend Slat Stor Sdist Sum FEA 共%兲
1 2.36 3.54 0.93 0.35 7.17 7.72 −7.1
2 4.31 3.54 −1.68 0.63 6.80 6.81 −0.1
3 5.74 3.54 2.22 0.85 12.35 12.06 2.4
4 6.67 3.54 −2.58 0.98 8.61 9.01 −4.5
5 7.08 3.54 2.74 1.04 14.40 14.02 2.7
6 6.98 3.54 −2.70 1.03 8.85 9.42 −6.0
7 6.37 3.54 2.47 0.94 13.32 12.96 2.7
8 5.25 3.54 −2.04 0.77 7.52 7.89 −4.6
9 3.62 3.54 1.41 0.53 9.11 8.87 2.6
10 1.48 3.54 −0.59 0.22 4.65 4.46 4.3
11 −1.10 3.45 −0.42 −0.14 1.79 2.01 −11.1
12 −4.10 3.37 1.56 −0.55 0.29 −0.56 −48.3
13 −7.57 3.37 −2.90 −1.01 −8.11 −7.47 8.6
14 −10.71 3.28 4.43 −1.37 −4.38 −5.78 −24.3
15 −14.06 3.19 −5.74 −0.75 −17.35 −16.06 8.1
Diaphragm
16 −13.93 3.19 −5.69 −0.73 −17.15 −15.85 8.2
17 −10.43 3.28 4.03 −1.34 −4.45 −5.55 −19.8
18 −7.12 3.37 −2.73 −0.95 −7.42 −6.80 9.2
19 −3.49 3.37 1.32 −0.47 0.73 −0.39 88.6
20 −0.32 3.45 −0.12 −0.03 2.98 3.19 −6.5
21 2.44 3.54 −0.96 0.36 5.38 5.36 0.3
22 4.74 3.54 1.84 0.70 10.82 10.54 2.6
23 6.53 3.54 −2.53 0.96 8.50 9.05 −6.1
24 7.81 3.54 3.02 1.15 15.51 15.10 2.7
25 8.58 3.54 −3.31 1.26 10.07 10.88 −7.5
26共Symm.兲 8.83 3.54 3.41 1.30 17.08 16.62 2.8
共Note: 1 kip= 4.45 kN兲.

The validity of the proposed formulation was also checked for
the case of a three-span curved girder. Pertinent information on
the dimensions and loading are given in Fig. 17. The box girder
consists of three different types of cross sections, with different
plate thicknesses but the same heights and widths. X-shaped in-
ternal cross frames were installed at every other strut location.
The lateral bracing system has a total of 54 panels 共16+ 22+ 16兲.
Fig. 18 shows bending and torsional moment diagrams for the
three-span continuous horizontally curved girder subjected to a
vertical load of 3.3 k / ft 共48.18 kN/ m兲. Tables 1 and 2 compara-
tively show forces in diagonals and struts in the lateral bracing
system of the three-span continuous model. As can be seen from
Table 1, the diagonal forces computed from Eq. 共22兲 are in fairly
good agreement with those from the FEA. The greatest discrep-
ancy occurs in the diagonals subjected to the lowest forces. As it
is likely that the same member will be used for all the diagonals
in a bracing system, this discrepancy in the diagonals subjected to
the lowest forces appears to be nonconsequential. Strut forces

Fig. 19. Diagonal forces in lateral bracing system of three-span Fig. 20. Strut forces in lateral bracing system of three-span
continuous horizontally curved girder 共1 kip= 4.45 kN兲 continuous horizontally curved girder 共1 kip= 4.45 kN兲

1220 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006

determined from Eq. 共23兲, shown in Table 2, compare very well
with those from the FEA. Again, the greatest discrepancy occurs
in the struts subjected to the lowest forces, and thus once more
appears to be nonconsequential. The forces in the diagonals and
struts of the lateral bracing system of the model given in Fig. 17
computed analytically from Eqs. 共22兲 and 共23兲, along with the
values obtained using the equations presented by Fan and Helwig
共1999兲, are comparatively shown in Figs. 19 and 20 and com-
pared with those from the FEA. As the diagonal forces due to
lateral load components and distortion that were neglected in Fan
and Helwig 共1999兲 are relatively small, significant differences
between the two analytically determined values were not antici-
pated and Fig. 19 confirms this. In the case of strut forces, how-
ever, significant discrepancies are evident in Fig. 20 between the
two analytically determined values. Fan and Helwig 共1999兲 un-
derestimate the maximum strut forces by 38% at the interior pier
and 22% at the maximum positive bending moment location. Un-
like the case of the diagonals, this discrepancy deserves judicious
attention as the strut forces predicted by Fan and Helwig 共1999兲
are on the unconservative side. The reason for this lack of agree-
ment may be attributable to the fact that Fan and Helwig 共1999兲
neglected to account for the force components due to torsion
and/or distortion in the determination of strut forces.
Concluding Remarks
New equations to estimate the forces in diagonals and struts in a
lateral bracing system of tub girders have been suggested in re-
cent years. It has been observed, however, that there are signifi-
cant discrepancies between the member forces computed from
these equations and those obtained from FEA in box girders with
an SD-type lateral bracing system. Forces in both the struts and
diagonals in SD-type bracing are affected by the longitudinal de-
formations and lateral displacements of the top flanges. Longitu-
dinal deformations are primarily due to vertical bending, while
lateral displacements are caused by vertical bending, lateral force
components due to inclined webs, torsion, and distortion. The
strut forces in an XD-type bracing induced by torsion are very
small compared with those developed in an SD-type bracing be-
cause of the self-equilibrating nature of the crossed diagonals
within a panel. Although the major portion of the strut forces is
caused by vertical bending and lateral force components, torsion
and distortion may contribute up to 35% of the strut force in a
typical horizontally curved continuous tub girder. Note that the
lateral bracing system in composite tub girders is only required to
accommodate the load during the construction. Once the concrete
roadway deck is completely hardened, the composite concrete
deck takes over the structural function provided by the lateral
bracing system. Although the bracing system is of a temporary
nature, there has been a reported case of failure of the lateral
bracing during construction 共Chen 2002兲. These temporary, yet
important, primary load carrying bracing members can now be
designed based on more accurate predictions.
Acknowledgments
The work presented in this paper was supported in part by the
Huff Eminent Scholar fund and Alabama Department of Trans-
portation Project No. 930-563. These financial supports are grate-
fully acknowledged. The opinions and conclusions expressed or
JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006 / 1221
implied in this paper are those of the authors. They are not nec-
essarily those of the funding agencies.
Notation
The following symbols are used in this paper:
A0 ⫽ enclosed area of the box cross section;
AD ⫽ cross-sectional area of diagonal in lateral
bracing;
AS ⫽ cross-sectional area of strut in lateral bracing;
a ⫽ web spacing of trapezoidal box girder at bottom
flange;
b ⫽ web spacing of trapezoidal box girder at top
flange;
D ⫽ total diagonal force in lateral bracing;
D− , D+ ⫽ Dtor in two repeating adjacent panels of lateral
bracing;
Dbend ⫽ diagonal force due to vertical bending;
Ddist ⫽ diagonal force due to distortion;
DH ⫽ horizontal force component transferred from Dtor;
Dlat ⫽ diagonal force due to lateral force component;
Dtor ⫽ diagonal force due to torsion;
E ⫽ modulus of elasticity;
e ⫽ eccentricity of vertical load relative to box girder
centroid;
f xTop ⫽ longitudinal stress in the middle of top flange;
h ⫽ box girder depth;
I f ⫽ second moment of inertia of top flange with
respect to the vertical centroidal axis;
KH ⫽ horizontal force component transferred from
internal cross-frames;
M ⫽ bending moment;
LD ⫽ length of diagonal in lateral bracing;
LS ⫽ length of strut in lateral bracing;
QA, QB ⫽ interactive forces between the top flange and
bracing members;
q ⫽ shear flow in a closed section;
R ⫽ radius of curvature of the girder;
S ⫽ total strut member force;
Sbend ⫽ strut force due to vertical bending;
Sdist ⫽ strut force due to distortion;
Slat ⫽ strut force due to lateral force components;
Stor ⫽ strut force due to pure torsion;
s ⫽ spacing between struts;
T ⫽ torsional moment;
u ⫽ displacement in longitudinal direction;
v ⫽ displacement in lateral direction;
v1 ⫽ relative displacement of top flange in lateral
direction;
v2 ⫽ axial elongation of strut;
w ⫽ uniformly distributed vertical load;
wlat ⫽ lateral force component;
␣ ⫽ horizontal angle between diagonal and top
flange;
␦D ⫽ axial elongation of lateral diagonal; and
␸ ⫽ vertical angle of inclination web of trapezoidal
box girder.
References
ABAQUS, Inc. 共2003兲. ABAQUS analysis user’s manual: Version 6.4,
Pawtucket, R.I.
American Association of State Highway and Transportation Officials
共AASHTO兲. 共1993兲. Guide specifications for horizontally curved
highway bridges, Washington, D.C.
Chen, B. S. 共2002兲. “Top-lateral bracing systems for trapezoidal steel
box-girder bridges.” Ph.D. dissertation, Univ. of Texas at Austin.
Dabrowski, R. 共1968兲. Curved thin-walled girders: Theory and analysis,
Cement and Concrete Association, London, 共translated from the Ger-
man original兲.
Fan, Z., and Helwig, T. A. 共2002兲. “Distortional loads and brace forces in
steel box girders.” J. Struct. Eng., 128共6兲, 710–718.
Fan, Z., and Helwig, T. A. 共1999兲. “Behavior of steel box girders with top
1222 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2006
View publication stats
flange bracing.” J. Struct. Eng., 125共8兲, 829–837.
Heins, C. P., and Hall, D. H. 共1981兲. Designer’s guide to steel box girder
bridges, Bethlehem Steel Corporation, Bethlehem, Pa.
Kim, K. 共2004兲. “Research on horizontally curved steel box girders.”
Ph.D. dissertation, Auburn Univ., Alabama.
Kollbrunner, C. F., and Basler, K. 共1969兲. Torsion in structures: An en-
gineering approach, Springer, New York.
Topkaya, C., and Williamson, E. B. 共2003兲. “Development of computa-
tional software for analysis of curved girders under construction
loads.” Comput. Struct., 81, 2087–2098.