BEHAVIOR OF STEEL BOX GIRDERS WITH TOP FLANGE BRACING

By Zhanfei Fan,1 Student Member, ASCE, and Todd A. Helwig,2 Associate Member, ASCE

ABSTRACT: A critical design stage for steel girders occurs during casting of the concrete bridge deck, when
the noncomposite steel section must support the wet concrete and the entire construction load. Although a
composite box girder has a high torsional stiffness in the completed bridge, during construction the open section
is relatively flexible in torsion. A horizontal truss system is usually installed at the top flange level to increase
the torsional stiffness and form a quasi-closed section. This paper presents results from a finite-element study
on the bending behavior of trapezoidal box girder systems during construction. The results show that large forces
can develop in the horizontal truss system due to vertical bending of the box girder. For truss systems with a
single diagonal, the forces induced from bending result in large lateral bending stresses in the top flanges of the
box. Many current design methods and computer programs do not consider the truss forces and top flange lateral
bending stresses induced from bending. Expressions to estimate the bending forces in the lateral truss system
as well as the lateral bending stresses in the top flange are developed. A numerical example illustrates the use
of the expressions.

INTRODUCTION system that is used, large lateral bending stresses may result
in the top flange. These large brace forces and girder stresses
Composite box girder bridges usually consist of steel girders are often not predicted by computer programs for box girder
of trapezoidal cross section with two top flanges and a con- analysis. An analytical method to evaluate truss forces and
crete slab. The closed cross section of the box in the completed girder stresses is therefore warranted.
bridge has a torsional stiffness that may be 100 to more than The paper will begin by presenting background information
1,000 times the stiffness of a comparable I-girder section. The and discussing recommendations by current design guides, fol-
large torsional stiffness makes box girders attractive for ap- lowed by an overview of the finite-element model used in the
plication in horizontally curved bridges in which the bridge analysis. Results from the finite-element analysis (FEA) will
geometry may result in large torques on the girders. In addi- then be presented; they show the effects of vertical bending
tion, there are a number of other structural, maintenance, and on horizontal truss forces and lateral bending of the top
aesthetic advantages that make box girders attractive for use flanges. Comparisons are made with the FEA results and pro-
in both curved and straight bridges. posed design equations that are developed in Appendix I. The
Although the torsional stiffness of a composite box girder last section of the paper summarizes the findings in the paper.
is large in the completed bridge, during transport, erection, and A numerical example is presented in Appendix II.
construction the girder consists of an open section with rela-
tively low torsional stiffness. This poses a major problem dur- BACKGROUND
ing the early stages of bridge construction when the steel sec-
tion may be subjected to large torques. Fig. 1 shows some The distribution of torsional moments in a curved box girder
typical bracing systems employed in box girders to increase may be obtained using a computer program for curved girders,
the torsional stiffness of the steel section. Internal cross frames or by approximate methods such as the M/R method discussed
[Fig. 1(a)] or diaphragms are used to control distortion of the by Tung and Fountain (1970). Although a three-dimensional
cross section from the applied torsion. Although external di- finite-element analysis may be performed on the box girder,
aphragms between adjacent girders can be used to increase the most current programs for curved girders are based on a grid
torsional stiffness of the bridge, they are mainly used only at analysis that makes use of line elements to model the girder.
supports due to aesthetic and fatigue concerns. A horizontal A program based upon a grid analysis typically requires the
truss fastened to the box near the top flanges is commonly cross-sectional properties such as the moment of inertia and
used to increase the torsional stiffness of the steel section. The
girder with the horizontal truss as shown in Fig. 1(b) or Fig.
1(c) is often referred to as a quasi-closed box girder.
The present paper focuses on the design and behavior of
the top flange horizontal truss that is used to increase the tor-
sional stiffness of the steel section during girder erection and
bridge construction. In particular, the effects of vertical bend-
ing of the box girder on the truss forces will be addressed.
Most current design methods neglect the effects of girder
bending stresses on the horizontal truss behavior; however, in
some cases, truss forces in the maximum bending region may
actually exceed the maximum member forces caused by the
torsional moments. In addition, depending on the type of truss
1
Res. Asst., Civ. Engrg. Dept., Univ. of Houston, Houston, TX 77204-
4791.
2
Asst. Prof., Civ. Engrg. Dept., Univ. of Houston, Houston, TX.
Note. Associate Editor: Ronald D. Ziemian. Discussion open until Jan-
uary 1, 2000. To extend the closing date one month, a written request
must be filed with the ASCE Manager of Journals. The manuscript for
this paper was submitted for review and possible publication on Decem-
ber 22, 1998. This paper is part of the Journal of Structural Engineer- FIG. 1. Bracing Systems for Box Girders: (a) Internal Cross-
ing, Vol. 125, No. 8, August, 1999. 䉷ASCE, ISSN 0733-9445/99/0008- Frame; (b) Single Diagonal-Type Horizontal Top Truss; (c) X-
0829–0837/$8.00 ⫹ $.50 per page. Paper No. 19903. Type Horizontal Top Truss

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999 / 829

torsional constants as input for the program. A difficult stage the FEA are nearly coincident. Negative values of the truss
to model in these programs is the behavior of the steel section forces indicate compression. The torsional loading causes the
during construction when the top flange truss forms a quasi- diagonals within a panel of an X-type truss to have equal mag-
closed box girder. The torsional analysis of a quasi-closed box nitudes of force; however, one diagonal is in compression
girder is usually performed using the equivalent plate method while the other is in tension. The diagonals of adjacent panels
(EPM) developed by Kollbrunner and Basler (1969). Two in an SD-type truss typically alternate between compression
types of truss systems that are usually considered consist of a and tension.
single diagonal (SD-type) system as shown in Fig. 1(b) or an The straight girder in Fig. 4 was only subjected to the tor-
X-system (X-type) as shown in Fig. 1(c). For the purpose of sional moment caused by the uniformly distributed torque. The
discussion in the present paper, the diagonals within a panel torsional loading for most applications, however, is caused by
of the X-type truss will be referred to as diagonal X1 or di- gravity loading on curved girders. The EPM was also tested
agonal X2, as shown in Fig. 1(c). According to the EPM, the
top lateral truss system is treated as a fictitious plate so that
the torsional properties of the box can be approximated during
the structural analysis. Once the distribution of torsion is
known, the shear flow, q, can be determined using traditional
torsional theory for box girders. Fig. 2(a) shows that q is a
function of the applied torsion, T, and the enclosed area of the
box, A0. The shear flow acting on the fictitious plate is then
transformed to diagonal member forces in the lateral truss, as
demonstrated in Figs. 2(b) and 2(c). The diagonals are de-
signed to carry the resulting axial member forces from the
torsional analysis. Many design guides (Heins 1975; Highway
1982; Guide 1993) recommend the EPM for the design of the
diagonals of the horizontal truss.
The struts for the top flange truss are typically designed to
carry the horizontal component of the applied load that is a
result of the sloping web of a trapezoidal box girder. Fig. 3
demonstrates the transformation of the vertical load into a web
shear and a horizontal component, p. Design guides (Highway
1982; ‘‘Chapter’’ 1997) provide recommendations for the de- FIG. 4. Pure Torsion on Simply Supported Straight Box Girder
sign of the struts for the lateral load component and account
for lateral bending of the flanges between the struts. One of
the assumptions in these recommendations for evaluating the
required forces in the struts and the lateral bending stresses in
the top flange is that the top and bottom flanges each resist
half of the horizontal web components of the applied load
( p/2). The half acting on the bottom flange does not generate
any top flange lateral bending stress or forces in the struts. For
a truss panel length of s, the recommended design tensile force
for the struts is therefore equal to ps/2.
To demonstrate the accuracy of the EPM, the straight girder
shown in Fig. 4 was analyzed with a three-dimensional finite-
element model that will be discussed in the next section of the
paper. Twist of the girder was prevented at the ends and a
uniformly distributed torque of 22.2 kN ⭈ m/m (5 kip ⭈ ft/ft)
was applied. Forces in the truss members as determined using FIG. 5. Brace Forces in Diagonals due to Torsion: (a) X-Type
Truss; (b) SD-Type Truss
the EPM were based on the value of the torque at the middle
of the corresponding truss panel. An SD-type and an X-type
truss were considered. Fig. 5 shows that the EPM has excellent
agreement with the FEA. The data points from the EPM and

FIG. 2. Diagonal Forces from Equivalent Plate Method (EPM)

FIG. 3. Horizontal Component of Applied Loads on Top

Flanges FIG. 6. Curved Continuous Box Girder

830 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999

on the curved girder shown in Fig. 6. The radius of curvature buckling of the brace. The graph in Fig. 8(c) shows that the
of the girder is 291 m (955 ft), which results in a 19.7⬚ sub- forces in the struts from the FEA results are significantly larger
tended angle for every 100 m of girder length (6⬚/100 ft). The than values predicted using the current design method. The
girder is nonprismatic, with two cross sections—sections P main source for the difference between the FEA results and
and N—used in the respective positive and negative moment current design methods for the diagonals and the struts is due
regions as shown in Fig. 6. The distributed load of 24.1 kN/m to vertical bending stresses in the box girder. In addition, part
(1.65 kips/ft) on each of the top flanges simulates the gravity of the error in the strut forces is due to the erroneous assump-
load from wet concrete as well as other construction loads. tion that the bottom flange resists half of the horizontal load
The horizontal top truss is an X-type system with 64 panels component, p. With the exception of the girder self-weight,
along the length of the bridge and a panel size of approxi- the entire lateral component from the applied load should be
mately 3.05 m (10 ft). Torsional moments on the girder are applied to the top flange. Additional results demonstrating the
caused by the horizontal curvature. The moment diagrams effect of the girder bending as well as a method to determine
from bending and torsion are shown in Fig. 7. Since the girder the bending component will be presented later in the paper.
is symmetrical about point C, the diagrams for only half of
the girder length are shown. The top flange truss consists of FINITE-ELEMENT MODEL
WT6 ⫻ 13 sections for the diagonals and L4 ⫻ 4 ⫻ 5/16
members for the struts. These members were selected using The three-dimensional finite-element program ANSYS (AN-
the EPM based on the maximum torque of 836 kN ⭈ m (616 SYS 1996) was used to study the behavior of box girders with
kip ⭈ ft) that occurs to the left of the support at point B. a horizontal truss located at the top flange. The cross sections
The quasi-closed box girder was analyzed with a three- of the box girders were built up with eight-node shell ele-
dimensional finite-element model that will be discussed in the ments. Shell elements were also used to model transverse stiff-
next section. Fig. 8 shows graphs of the member forces along eners and solid diaphragms. The solid diaphragms were used
the girder length predicted by the EPM and the resulting val- at the supports. Internal K-frames were spaced at 6.1 m (every
ues from the FEA results. Negative values of the forces indi- two panels of the truss—20 ft) for most of the analyses; how-
cate compression. The graphs of the diagonal forces show that ever, isolated cases were also run with K-frames spaced at 3.05
the results predicted using the EPM have poor agreement with m (10 ft) to determine the effect on lateral bending stresses in
the FEA results. The EPM significantly underestimates the the flanges. The diagonals of the K-frame were modeled with
member forces at several locations along the length. There are beam elements to ensure geometric stability, while the top lat-
a number of locations in which the EPM predicts tension in erals consisted of two-node truss elements. Two-node truss
members that are actually in compression from the FEA re- elements were also used to model the diagonals and struts of
sults, which therefore may lead to potential problems with the top lateral truss system. The top flange truss was fastened
to the girder at the junction between the top flange and the
web. Twist of the girders was prevented at the supports.
At internal K-frame locations, the struts of the top flange
truss also act as the top laterals of the internal K-frames. The
purpose of the internal K-frames is to control distortion of the
box section. The distortional forces in the diagonals of a K-
frame cause the two top laterals [L1 and L2 in Fig. 1(a)] to
have equal magnitudes of compression and tension. If the strut
from the top flange lateral truss was also part of the K-frame,
the axial strut forces from the FEA results were the average
FIG. 7. Bending and Torsional Moment Diagrams: (a) Bending of the forces in the two halves [L1 and L2 in Fig. 1(a)],
Moment; (b) Torsional Moment thereby eliminating the K-frame component. The distortional
analysis for K-frames of quasi-closed box girders will be dis-
cussed in a later paper.
The paper will focus on the design of the top flange lateral
truss in box girders. The box girders that were considered were
symmetric about the vertical axis of the cross section and had
a uniform depth along the length.

EFFECT OF BOX GIRDER BENDING ON TRUSS

BEHAVIOR
FEA results were presented earlier in the paper for girders
subjected to pure torsion as well as combined bending and
torsion. Although the EPM did a good job of predicting forces
in the top horizontal truss for the case of pure torsion, there
were large differences between the EPM and the FEA results
for the case of combined bending and torsion.
The straight girder shown previously in Fig. 4 was further
studied to focus on the effect of box girder bending on the
truss behavior. The 48.8 m long (160 ft) girder was simply
supported with a uniformly distributed load of 24.1 kN/m
(1.65 kips/ft) on each top flange, as shown in Fig. 9. An SD-
type and an X-type truss were considered in the analysis. The
sizes of the diagonals were WT6 ⫻ 13 for the X-type system
FIG. 8. Brace Forces for Three-Span Curved Girder: (a) Axial and WT6 ⫻ 20 for the SD-type system, while the struts were
Forces in Diagonal X1; (b) Axial Forces in Diagonal X2; (c) Axial L4 ⫻ 4 ⫻ 5/16 for both systems.
Forces in Struts Fig. 10 shows a graph of the FEA results for the forces in
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999 / 831
 to the top flange. The lateral load component for the beam in
Fig. 9 is p = 6.0 kN/m (0.41 kips/ft). The current method
would therefore use wLT = 3.0 kN/m, which results in a max-
imum lateral bending moment of 2.3 kN ⭈ m (1.7 kip ⭈ ft) for a
panel length of 3.0 m (10 ft). Converting the lateral bending
moment to a stress gives a maximum value of 3.48 MPa (0.50
FIG. 9. Simply Supported Girder Subjected to Transverse ksi). However, the FEA result for the maximum lateral bend-
Load ing stress is approximately 6.89 MPa (1.0 ksi), which is ap-
proximately twice the predicted value by the current method.
This indicates that the assumption in the current method that
the top and bottom flanges each carry half of the horizontal
component of the load is incorrect. The lateral stresses on the
top flange are induced by the full component of the vertical
load applied on the top of the web ( p, as shown in Fig. 3).
Therefore, the estimation of the lateral bending stress during
the construction of the concrete deck must use the full hori-
zontal components of the gravity loads applied at the top
flange. Although the girder self-weight may be divided be-
tween the top and bottom flange, this component is relatively
small compared to the weight of the wet concrete and other
construction loads, and may be conservatively added to the
FIG. 10. Brace Forces in Simply Supported Girder due to total weight applied to the top flange.
Bending
A more serious issue regarding lateral bending stresses oc-
curs in girders with the SD-type truss system. As demonstrated
the diagonals and struts for both truss systems that were stud- by Fig. 13, the top flange lateral bending stresses are signifi-
ied. Negative values indicate compression. The two diagonals cantly larger than those in flanges with the X-type truss sys-
in a given panel of the X-type truss have identical forces and tem. For the SD-type truss, the largest lateral bending stresses
are represented by a single data point in Fig. 10. The graphs often range from 20% to 30% of the maximum vertical bend-
show that large axial forces develop in the truss despite the ing stress. The maximum lateral bending stress for the results
absence of torsional moments. The distribution of the brace shown in Fig. 13 is approximately 25% of the maximum axial
forces along the length has the same shape as the bending bending stress. In addition to having much larger magnitudes,
moment diagram, implying that vertical bending of the box the distribution of the lateral bending stress for the SD-type
girder causes these forces. The forces in the X-type truss sys- truss is quite different from that of the X-type truss. The wave-
tem are larger than the corresponding forces in the SD-type lengths of the lateral bending stresses are twice as large for
truss system. All of the forces are larger than values predicted
using the current design method (Highway 1982) that was out-
lined earlier in the paper. Since there is no applied torsion, the
current design method would predict zero diagonal forces and
a tensile force of 9.2 kN (2.1 kips) in all struts caused by the
horizontal component of the applied load.
The reason why forces develop in the truss with no torsional FIG. 11. Brace Forces in Horizontal Truss due to Vertical
moment applied can be understood by considering the distri- Bending Moment
bution of vertical bending stresses in a trapezoidal box girder.
For quasi-closed box girders, the neutral axis of the steel sec-
tion is usually near middepth of the girder, as shown in Fig.
11. The horizontal truss system therefore connects to the girder
in a region of high bending stress. Due to compatibility, the
truss must experience the same strains as the box girder in the
axial direction. This is why the diagonal forces shown in Fig.
10 are all in compression.
Before the effects of vertical bending of the box girder are
addressed, it is important to discuss the bending behavior of
the top flanges of the box girders. In addition to stresses
caused by vertical bending of the box girder, the top flanges FIG. 12. Longitudinal Stresses in Top Flange of Box Girder
may also experience significant stresses due to lateral bending. with X-Type Truss
Fig. 12 shows the longitudinal stresses at the middle (point
M) and the edges (points L and R) of one of the top flanges
for the straight box girder shown in Fig. 9 with the X-type
truss system. The differences between the stresses at the edge
and those in the middle of the top flange represent the lateral
bending stresses developed in the top flanges. The lateral load
component acting between panel points of the top flange truss
causes these lateral-bending stresses. As discussed earlier, the
current method assumes that the top and bottom flanges each
carry half of the lateral load component, p. The top flange is
typically modeled as a continuous beam (between panel
points) for which the maximum moment due to the lateral load
component would be (wLT s2/12), where s is the panel length FIG. 13. Longitudinal Stresses in Top Flange of Box Girder
and wLT is the portion of the lateral load component applied with SD-Type Truss

832 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999

 lateral load component of the applied load. Using the principle
of superposition, the force components from each of the above
sources can be evaluated and the resultant forces summed.

Truss Forces due to Vertical Bending of Box Girder

The force that is induced in the truss due to vertical bending
of the box girder is a function of the stiffness of the truss and
the bending strain in the girder at the level of the connection
to the horizontal truss. For simplicity, the level of the truss
FIG. 14. Lateral Bending Moments in Top Flanges with SD- connection was assumed to be at the junction between the web
Type Truss and the top flange of the box girder. This is a conservative
assumption (due to the larger bending stress at the top flange
level) and results in reasonable estimates of the truss forces
for most practical truss positions. Fan (1999) presents a more
general solution for cases in which the truss connection may
be placed significantly below the top flange (>10% of the
girder depth).
A derivation of the forces induced in the truss due to vertical
bending is presented in Appendix I. The derivation of the truss
forces is based on a longitudinal stress of fx Top at the middle
of the top flange, given by the expression fx Top = M/Sx Top. The
girder moment, M, is taken at the middle of the panel under
consideration, while Sx Top is the section modulus based on the
extreme fiber distance to the top flange. The following equa-
FIG. 15. Top Flange Stresses with SD-Type Truss; K-Frames tions estimate the force in the diagonals, Dbend, the force in the
Spaced at 3.05 m struts, Sbend, and the resulting lateral bending stress fL bend for
the two truss types:
the SD-type as for the X-type truss. In an SD-type truss, the
large strut force developed from vertical bending results in a SD-Type Trusses
large lateral force applied directly to the flange, therefore caus-
ing lateral moments in the top flange, as shown in Fig. 14. fx Top s cos ␣ 1.5s
Dbend = ; Sbend = ⫺Dbend sin ␣; fL bend = 2 Sbend
Although the horizontal component from the applied load does K1 b f tf
cause lateral bending, most of the lateral bending in a girder (1a–c)
with an SD-type truss results from the large strut forces caused
from vertical bending of the box girder. Although the internal X-Type Trusses
K-frames were spaced at 6.1 m (20 ft) in the above analysis,
Fig. 15 shows that providing additional K-frames has very fx Top s cos ␣
Dbend = ; Sbend = ⫺2Dbend sin ␣; fL bend = 0
little effect on the lateral bending stresses. The graph in Fig. K2
15 shows the lateral bending stresses when K-frames were (2a–c)
spaced at 3.05 m (10 ft). The distributions of lateral bending
stresses from Figs. 13 and 15 for the two K-frame spacings where K1 and K2 = parameters defined by
are nearly identical. Using a closer spacing of internal K- d b s3
frames provides better control over distortion of the box sec- K1 = ⫹ sin2 ␣ ⫹ 3 sin2 ␣ (3)
Ad As 2b f tf
tion; however, the entire cross section simply rotates and the
top flanges still bend laterally with an effective span of two d 2b sin2 ␣
truss panels. K2 = ⫹ (4)
Ad As
Since there are diagonals at both ends of a strut in the X-
type truss, the strut forces do not cause lateral bending in the where s = spacing of struts (panel length); ␣ = acute angle
flanges. Top flange lateral bending in a girder with an X-type between the top flange and the diagonal; bf and tf = respective
truss is caused by the horizontal component from the applied values of the width and thickness of the top flange; d = length
load, as outlined above. of a diagonal; b = distance between the middle of the top
flanges; and Ad and As = respective cross-sectional areas of the
PROPOSED EQUATIONS diagonals and the struts. The diagonal forces, Dbend, will have
the same state of stress (compression or tension) as the top
The forces that develop in the top flange horizontal truss flange at the point under consideration. Unlike the axial forces
can result from the following three components: due to torsion, for X-type trusses both diagonals have the iden-
tical brace force (with the same sign). It is therefore important
1. Torsional moment on the girder to maintain the sign convention for the forces when adding
2. Vertical bending of the box girder the bending component to the torsional component. For both
3. Lateral component of the applied load truss types, the state of stress in the struts is opposite that of
the diagonals, as given by the negative sign in (1b) and (2b).
A complete design of the truss would therefore take all three The diagonal and strut forces given in the above expressions
of these potential components into consideration. Results have are a function of the member size (larger/stiffer members de-
already been presented that demonstrate that the EPM provides velop larger forces). It is therefore necessary to pick a diagonal
good estimates of the truss forces that result from the applied and strut size to compute these components. Depending upon
torsion. Expressions will be presented in this section that can the location along the girder length, trial sizes may be selected
estimate the truss forces and top flange lateral bending stresses as a function of the torsional requirements given by the EPM.
that result from vertical bending of the box girder and the For regions of high bending moments and high torsional mo-
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999 / 833
ments, it is recommended to start the design process by dou- where STot and DTot = respective total forces in the strut and
bling the force requirement for torsion (based on the EPM) to diagonals; and fL Tot = total lateral bending stress. DEPM in (6b)
select trial member sizes, and then checking the bending be- is the diagonal force from the torsional moment determined
havior using the above equations. using the EPM. The stress given in (6c) includes the top flange
The above expressions were derived assuming uniform mo- stress that results from both the lateral and vertical bending of
ment and therefore a uniform stress within the panel length. the box girder. Additional lateral bending stresses result due
For cases with variable moment, it is recommended to begin to the curvature of the girder; however, design guides (High-
the procedure by calculating the values of Dbend using the stress way 1982) provide recommendations for these components.
at the middle of the panel, followed by calculating Sbend based Some of the struts of the top flange lateral truss are also used
on the average value of Dbend for the two adjacent panels. For as a top lateral for the internal K-frames that control distortion
nonprismatic girders with a flange transition within the panel, of the box. The distortional forces that result in the K-frames
( fx Top s) in (1a) and (2a) can be replaced with 兺ni=1 fi si, where will be presented in a later paper.
fi is the average top flange stress in the ith segment (within It is important to maintain the proper sign convention while
the panel) that has the length of si. It is also conservative to superimposing the forces. For example, torsional loading
simply use the largest longitudinal girder stress within the causes one of the diagonals of an X-type system to be in com-
panel. The bending forces in SD-type trusses are also a func- pression, while the other one is in tension; however, bending
tion of the size of the top flange. If a flange transition occurs causes the same state of stress in both diagonals.
within a panel, it is conservative to use the width and thickness
of the larger flange in the above expressions. The width and FEA RESULTS AND PROPOSED EQUATIONS
thickness of the larger flange are used, since this produces
larger forces in the diagonal of an SD-type truss. The bending The proposed expressions from the previous section were
forces in an X-type truss are not a function of the flange sizes. compared with the FEA results on the straight girder previ-
The numerical example in Appendix II demonstrates the use ously shown in Fig. 9. Fig. 16 shows a comparison between
of the proposed equations. the proposed solutions and the FEA results for the truss forces
Since the top flange truss picks up force due to the vertical along the girder length. The total forces in the struts and the
bending of the box girder, it also contributes to the box girder diagonals were calculated using superposition, as given in (6a)
bending stiffness. As a result, the actual fx Top is less than that and (6b). The proposed equations have good agreement with
predicted using M/Sx Top. Although Fan (1999) discusses the the FEA results. Fig. 17 shows a comparison of the bending
effects of the top flange truss on the box girder bending stiff- stresses in the flange from the FEA results and the value ob-
ness, neglecting the effect results in conservative estimates of tained by adding the lateral bending stresses from (1c) and
the truss forces and lateral bending stresses in the top flange. (5c) to the vertical bending stress. The curves show that the
proposed solutions provide a good envelope of the maximum
and minimum stresses from the FEA results. In design, the
Truss Forces due to Lateral Components of Applied value of interest would be the maximum lateral bending stress,
Loading which would be a function of the bending moment and the
Based upon the results shown previously, the entire lateral flange size. In a nonprismatic section, the engineer should be
load component, p, should be applied to the top flange (instead able to locate a few critical sections along the girder length
of assuming that the bottom flange carries half). In an SD-type that need to be checked.
truss, the struts carry the majority of the lateral load compo- The proposed expressions were also checked using the
nent. In an X-type truss, however, the struts and the diagonals three-span curved girder previously shown in Fig. 6. Fig. 8
work together to carry the lateral load component. Expressions
for the strut and diagonal forces due to the lateral load com-
ponent in an X-type truss are presented in Appendix I. How-
ever, the diagonal force due to the lateral load component in
an X-type truss is relatively small compared to the bending
and torsional components. For simplicity, it is therefore rec-
ommended to design the strut to carry the entire lateral load
component. This results in the following design expressions
for both SD-type and X-type truss systems:
ps2
Dlat = 0; Slat = ps; fL lat = (5a–c)
2b f2 tf
Dlat, Slat, and fL lat in the above expressions = diagonal force,
strut force, and lateral bending stress due to the lateral load FIG. 16. Proposed Equations and FEA for Truss Forces in
component. All other terms in the above expressions have been Straight Girder
previously defined. The expression for fL lat comes from mod-
eling the top flange as a continuous beam supported at the
truss panel points, for which the maximum moment is given
by ps2/12. Applying the flange section modulus of b 2f tf /6 re-
sults in the expression given in (5c).
For design, it is necessary to superimpose the forces and
stresses due to vertical bending, the lateral load components,
and the torsional moments. This therefore leads to the follow-
ing expressions:
STot = Sbend ⫹ Slat; DTot = DEPM ⫹ Dbend ⫹ Dlat (6a,b)
FIG. 17. Proposed Equations and FEA for Top Flange
fL Tot = fL bend ⫹ fL lat ⫹ fx Top (6c) Stresses (SD-Type Truss)

834 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999

 induced by vertical bending produce large lateral bending
stresses in the top flange. Top flange lateral bending stresses
also result from the horizontal component of the applied load
due to the sloping webs of trapezoidal girders; however, the
current design method incorrectly assumes that the top and
bottom flange equally resist the horizontal load component.
Equations were developed and presented to predict the top
flange truss forces and the lateral bending stresses due to box
girder bending and the lateral component of the applied load.
The equations have good agreement with finite-element results
for straight and curved box girders.

APPENDIX I. DERIVATION OF TRUSS FORCES DUE

TO VERTICAL BENDING
The following assumptions apply to this derivation:

1. The box girder is a prismatic section subjected to a state

of uniform moment.
2. The web of the girder provides negligible lateral resis-
tance to the top flange.
FIG. 18. Proposed Equations and FEA for Truss Forces in
Curved Girder 3. Effects of horizontal curvature on the bending behavior
are neglected.
had shown that only applying the EPM had poor agreement
with the FEA results. Fig. 18 shows a graph of the FEA results Fig. 19 shows the interactive forces between the bracing
and the proposed equations. The contribution of the diagonals members and the top flanges for (1) the X-type; and (2) the
in carrying the lateral load component ( p), as outlined in the SD-type truss systems. The force F is the resultant from the
appendix, has been considered; however, the difference is not uniform vertical bending moment loading in the girder. The
too significant if only the struts are designed to carry p. There longitudinal strains in the box girder and the top flange hori-
are a few isolated points (around flange transitions and solid zontal truss are identical due to compatibility. Due to sym-
diaphragms) where there is a discernable difference between metry, for a given truss, all diagonals will therefore have the
the proposed solutions and the FEA results; however, even in same force, Dbend, and the same sign (compression or tension).
these regions, the proposed expressions provide reasonable Similarly, for a given truss, all struts will have the same force,
agreement with the FEA results. The proposed expressions Sbend, and the same sign.
generally have good agreement with the FEA results along the
entire length of the girder. X-Type Truss
In design, the engineer would typically not have to apply The lateral interactive forces, Q, between the truss and the
the proposed equations along the entire girder length, but in- top flange must be identical at all joints due to symmetry of
stead only in critical regions to size the top flange lateral truss the structure, as shown in Fig. 19(a). Equilibrium of the top
members. Critical regions will generally occur around supports flange yields Q = 0, since the lateral stiffness of the web is
or other regions of high bending moment and torsion. It is assumed negligible, as outlined above; however, the flange
important to note, however, that in some instances the maxi- may still experience some rigid body movement. Since Q = 0,
mum compression in the diagonal members may occur in a the lateral bending stress due to the vertical bending will be
region of moderate moment and torsion. The maximum com- equal to zero, as given in (2c). Equilibrium of a joint gives
pression force in the diagonal of the top flange truss of the the force in the strut as a function of the diagonal force
three-span box girder previously shown in Fig. 6 occurs 29.0
m (95 ft) to the right of support B. The numerical example in Sbend = ⫺2Dbend sin ␣ (7)
Appendix II shows the calculation of the diagonal and strut
force in this region. The maximum compression in the struts
usually occurs in the region with the highest top flange tensile
stress.
Although the results presented in the current paper have
focused on the behavior of the top flange truss in box girders,
the bending problems discussed may also arise in lateral brac-
ing systems for I-shaped girders. A lateral truss is often pro-
vided near the bottom flange of I-girder bridges for bracing
against wind loads. Since this truss connects to the girders near
the bottom flange, the same bending induced forces outlined
earlier in the paper will develop in these truss members. If the
truss system used is either an SD-type or an X-type truss in-
stalled between two girders, the equations presented in the
current paper are directly applicable.

SUMMARY AND CONCLUSIONS

Current design methods for the top flange lateral truss in
box girders underestimate the magnitude of the member forces
because they do not account for the effects of vertical bending FIG. 19. Interactive Forces between Top Flanges and Horizon-
of the box girder. For trusses with a single diagonal, the forces tal Truss: (a) X-Type Truss; (b) SD-Type Truss

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999 / 835

 FIG. 21. Lateral Displacement of Top Flanges due to Strut
Force

which is the expression given in (1b).

The forces that develop in the diagonals are a function of
FIG. 20. Elongation of Diagonal of Horizontal Truss the axial stiffness of the diagonals and struts, as well as the
lateral stiffness of the top flange of the box girders. When the
To develop an expression for the force in the diagonal, con- truss is subjected to a strain in the axial direction of the box
sider a panel of the truss system between two struts, as shown girder, the strut ties the two flanges of the box girder together.
in Fig. 20. The axial elongation of all diagonals is the same The top flanges of the box girder will bend due to the strut
due to symmetry and is defined as ⌬diag. The elongation, ⌬diag, force, Sbend. The top flange will bend like a continuous beam
can be expressed as (Fig. 20) between panel points, as shown in Fig. 21, with a relative
lateral deflection between two consecutive panel points, v1,
⌬diag = u cos ␣ ⫹ v sin ␣ (8) given by
where u and v = relative displacements of a diagonal in the Sbend(2s)3 ⫺Dbend s3sin ␣ ⫺Dbend s3sin ␣
longitudinal and lateral directions (of the top flange), respec- v1 = = = (15)
192EIf 24EIf 2Eb f3 tf
tively. The relative displacement, v, is the same as the elon-
gation of the struts, for which basic engineering mechanics where If = moment of inertia of the top flange in the lateral
and (7) yield the following expression: direction; and bf and tf = respective width and thickness of the
top flange. The elongation of the diagonal members can be
Sbend ⫺2Dbend sin ␣ expressed similar to (8) by
v= b= b (9)
EAs EAs
⌬diag = u cos ␣ ⫹ (v1 ⫹ v2)sin ␣ (16)
where E = elastic modulus; b = panel width (strut length); and where u = elongation of the top flanges in a panel, defined by
As = strut area. The relative displacement, u, is the same as (10); and v 2 = elongation of the struts defined by
the elongation of the girder at the truss. If the longitudinal
stress in the girder at the truss connection is fx Top, Hooke’s law Sbend ⫺Dbend sin ␣
yields the following expression: v2 = b= b (17)
EAs EAs
fx Top s Eq. (11) still applies to diagonals. Substituting (10), (11),
u= (10) (15), and (17) into (16) and rearranging yields
E
where s = panel length. fx Top s cos ␣
Dbend = (18)
Similarly, the elongation of the diagonal is given by K1
Dbend d b s3
⌬diag = d (11) K1 = ⫹ sin2␣ ⫹ 3 sin2␣ (19)
EAd Ad As 2b f tf
where d = length of the diagonal; and Ad = diagonal area. which are the expressions given in (1a) and (3).
Substituting (9)–(11) into (8) and rearranging yields The lateral bending stress in the flange due to the force Sbend
is given by
fx Top s cos ␣
Dbend = (12) 1.5s
K2 fL bend = Sbend (20)
b f2 tf
d 2b sin ␣ 2

K2 = ⫹ (13) which was found by dividing the maximum flange lateral mo-
Ad As
ment of [Sbend(2s)/8] by the section modulus (b 2f tf )/6.
which are the same expressions given in (2a) and (4).
X-Type Truss Forces due to Lateral Load
SD-Type Truss Components

The joints between the truss and the flanges can be divided A force of ps results at each truss panel point due to the
into two types, A and B, as shown in Fig. 19(b). At joint A lateral load component, p. The diagonals and the struts in an
only a strut is connected to the flange, while a strut and two X-type truss resist this force. The longitudinal deformation of
diagonals frame into joint B. Since the lateral stiffness of the the top flanges is assumed negligible. Therefore, if the elon-
web is negligible, lateral equilibrium of a top flange shows gation of the struts is v, the elongation of the diagonals, ⌬diag
that the interactive forces Q and Sbend are equal in magnitude is
and opposite in direction. Horizontal equilibrium of joint B in ⌬diag = v sin ␣ (21)
Fig. 19(b) gives
The axial forces in struts and diagonals are given by the fol-
Sbend = ⫺Dbend sin ␣ (14) lowing expressions:
836 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999
 v fx Top s cos ␣
Slat = EAs (22) Dbend =
b K2

Dlat = EAd
⌬diag
d
v
= EAd sin ␣
d
(23) 冉 ⫺122.8 MPa ⫻
1,000 kPa
MPa 冊(3.05 m)cos(45⬚)
= = ⫺71.2 kN
The equilibrium at the joints requires 3,720 m⫺1
(29)
Slat ⫹ 2Dlat sin ␣ = ps (24)
The bending forces in both diagonals X1 and X2 is ⫺71.2 kN
Substitution of (22) and (23) into (24) results in the following (⫺160 kips) compression.
expression for v:
ps Total

冉 冊
v= (25)
EAs 2EAd Using (6b), the total axial forces are ⫺97.9–71.2 = ⫺169.1
⫹ sin2␣ kN (38.0 kips) for diagonal X1 (compression) and 97.9–71.2
b d
= 26.7 kN (6.0 kips) for diagonal X2 (tension). The FEA re-
Substituting (25) back into (22) and (23) and rearranging re- sults are ⫺159.2 kN for X1 and 31.5 kN for X2. Note: If the
sults in the following expressions: diagonal force due to the lateral load component is considered
using (27), the results are ⫺162.2 kN for X1 and 33.6 kN for
d
Slat = ps (26) X2.
Ad K2
b sin ␣ Strut Force
Dlat = ps (27)
As K2 Using a bending moment of 5,670 kN ⭈ m, the axial diagonal
forces due to bending in the panel to the left of the above
where K2 is given in (13). Since the forces in the diagonal are
panel [80.8 m (265 ft) to support A] can be calculated as above
generally small compared to the forces due to bending and
to give Dbend = ⫺51.0 kN (⫺11.5 kips). The axial force in the
torsion, the method presented in the main text designs only
strut between these two panels due to bending is calculated by
the struts for the force ps.
(2b) using the average Dbend from the two adjacent panels
APPENDIX II. NUMERICAL EXAMPLE
Consider the three-span box girder shown in Fig. 6. Deter-
Sbend = ⫺2Dbend sin ␣ = ⫺2 冉⫺51.0 kN ⫺ 71.2 kN
2
冊
mine the axial forces in the bracing panel in the interior span ⭈ sin(45⬚) = 86.4 kN (19.4 kips) (30)
located 29.0 m (95.0 ft) to the right of support B [or 83.8 m
(275 ft) from support A]. The bending and torsional moments The lateral component of the applied loading is p = 24.1 ⫻
in the middle of the panel are 7,907 kN ⭈ m (69,985 k ⭈ in.) and 0.25 = 6.0 kN/m (0.4 kips/ft), which results in ps = 6.0 ⫻
⫺514 kN ⭈ m (⫺4,554 k ⭈ in.), respectively. Assume WT6 ⫻ 13 3.05 = 18.3 kN (4.1 kips). The total strut force is therefore
diagonals [Ad = 24.6 cm2 (3.8 sq. in.)] and an L4 ⫻ 4 ⫻ STot = 86.4 ⫹ 18.3 = 104.7 kN (23.5 kips). The FEA results
5/16 strut [As = 15.5 cm2 (2.4 sq in.)] for the X-type truss. yielded 94.4 kN. If (26) is used, which accounts for the di-
agonals carrying a portion of the lateral load component, Slat
Torsion = 8.6 kN, which results in STot = 86.4 ⫹ 8.6 = 95.0 kN.

Using the equations from the EPM as outlined in Fig. 2, the ACKNOWLEDGMENTS
shear flow is given by q = T/(2A0). A0 is the area enclosed by
The writers would like to thank the Texas Department of Transporta-
the box. The average width of the box is (304.8 ⫹ 190.5)/2 tion (TxDOT) for providing funding for this research investigation. In
= 247.7 cm (97.5 in.) and the depth is 228.6 cm (90 in.), from addition, the writers would like to extend special thanks to Karl Frank,
which A0 = 247.7 ⫻ 228.6 = 5.66 m2 (8,775 sq in.). The shear Joseph Yura, Reagan Herman, Chris Gilchrist, Bryan Chen, and Michael
flow is therefore q = 514/[2(5.66)] = 45.4 kN ⭈ m (0.26 kips/ Lopez, as well as TxDOT engineers Arnie Cohen and John Vogel.
in.). The width of the box at the top is 3.05 m (120 in.), while
␣ = 45⬚, which therefore gives DEPM = qb/(2 sin ␣) = 45.4 ⫻ APPENDIX III. REFERENCES
3.05/[2 sin(45⬚)] = ⫾97.9 kN (⫾22.0 kips). Diagonal X2 will
ANSYS users manual: Version 5.3. (1996). ANSYS, Houston, Pa.
be in tension, while X1 will be in compression. ‘‘Chapter 13: Four LRFD design examples of steel highway bridges.’’
(1997). Highway structures design handbook: Volume II. American Iron
Bending and Steel Institute and National Steel Bridge Alliance, Chicago.
Fan, Z. (1999). ‘‘Field and computational studies of steel trapezoidal box
The bending moment of 7,907 kN ⭈ m (69,985 k ⭈ in.) causes girder bridges,’’ PhD dissertation, College of Engrg., University of
a bending stress at the top flange equal to ⫺122.8 MPa Houston, Tex.
(⫺17.81 ksi) in compression. The length of the diagonal is Guide specifications of horizontally curved highway bridges. (1993).
431 cm (170 in.), while the panel width is 305 cm (120 in.), American Association of State Highway and Transportation Officials,
Washington, D.C.
from which Heins, C. P. Jr. (1975). Bending and torsional design in structural mem-
bers. Lexington Books, Lexington, Mass.
d 2b sin2␣ 431 cm 2 ⫻ (305 cm) ⫻ sin2(45⬚)
K2 = ⫹ = 2 ⫹
Highway structural design handbook: Volume 2. (1982). AISC Marketing,
Ad As 24.6 cm 15.5 cm2 Pittsburgh.
Kollbrunner, C. F., and Basler, K. (1969). Torsion in structures—An en-
= 37.2 cm⫺1 = 3,720 m⫺1 (28) gineering approach. Springer, New York.
⫺1 ⫺1
Tung, D. H. H., and Fountain, R. S. (1970). ‘‘Approximate torsional
With K2 = 3,720 m (94.5 in. ), the resulting diagonal force analysis of curved box girders by the M/R method.’’ AISC Engrg. J.,
due to bending is given by 7(3), 65–74.

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1999 / 837