Beruflich Dokumente
Kultur Dokumente
Abstract: Trapezoidal steel box girders are often used in curved bridges due to the large torsional stiffness that results from the closed
cross section. However, the torsional loading on the box girders can cause the cross section to distort from its original shape, which results
in distortional stresses in the plates of the box section. The distortional stresses are controlled by internal cross frames that restrain the
Downloaded from ascelibrary.org by Syracuse University Library on 07/27/13. Copyright ASCE. For personal use only; all rights reserved.
cross section from distortion. Current design specifications do not account for forces in the bracing members due to box girder distortion.
This paper presents an analytical study on the distortional behavior of box girders with a trapezoidal cross-sectional shape. Typical
torsional loads on curved box girders are discussed, and the distortional components of these applied torsional loads are studied. The
distortional components from different torsional loads on trapezoidal box girders are identified and used to derive expressions for the
brace forces in the internal cross frames for quasi-closed box girders. The results from the approximate equations for brace forces due to
cross-sectional distortion are verified by three-dimensional finite-element analyses.
DOI: 10.1061/共ASCE兲0733-9445共2002兲128:6共710兲
CE Database keywords: Box girders; Steel; Stiffness; Torsion.
Fig. 3. The effect of the bending stresses on the top flange can be
modeled as an axial force that is approximately equal to M /h,
where M is the bending moment and h is the girder depth. The
force M /h acts on the curved segment in the longitudinal direc-
tion, as shown in Fig. 3共a兲. Due to the curvature of the segment,
the force resultant of (M /h)d is developed on the flange in the
lateral direction. Dividing the lateral force by the length of the
segment, ds, a distributed lateral load equal to M /(Rh) is de-
rived, as shown in Fig. 3共b兲. Due to the nature of bending
stresses, the bottom flange is subjected to the same magnitude of
the lateral load in the opposite direction. As a result, these two
horizontal components form a distributed torsional load equal to
M /R acting on the girder 共Fig. 4兲. Therefore, the effect of the
curvature can be approximately analyzed by a straight model in Fig. 4. Equivalent torsional loads on curved box girder
which the flange segments are subjected to longitudinal forces of
M /h as well as external distributed loads of M /(Rh) in the lateral
direction, as demonstrated in Fig. 3共b兲. Tung and Fountain 共1970兲
used this approximation in the development of the M /R method
for the torsional analysis of curved box girders.
The distortional behavior of a box girder is dependent on the
manner in which the external torque is applied to the girder. For
simplicity, rectangular sections are considered in this section. The
distortional behavior of trapezoidal sections will be covered later.
A torsional load, either comprised of a vertical couple or a hori-
zontal couple, can be modeled as a uniform torsional component
superimposed on a distortional component, as demonstrated in Fig. 5. Torsion and distortion of rectangular box girder due to
Figs. 5 and 6. The rectangular thin-walled box has a respective vertical forces
depth and width of h and b. Each couple can be modeled by the
uniform torsional components given in Fig. 5共b兲 or 6共b兲 superim-
posed on the distortional components given in Fig. 5共c兲 or 6共c兲.
Although the boxes in Figs. 5共a兲 and 6共a兲 are subjected to the
same magnitude of torsion (m T ), the resulting distortional
stresses are opposite in direction since the distortional loads
shown in Figs. 5共c兲 and 6共c兲 are opposite. The pure torsional
components shown in Figs. 5共b兲 and 6共b兲 generate a uniform
Saint-Venant shear flow along the circumference of the box girder
cross section, and warping stresses due to this torsional compo-
nent are usually negligible. However, significant distortional
Fig. 6. Torsion and distortion of rectangular box girder due to
horizontal forces
by examining how the transverse force components in the distor- Fig. 9. Brace forces in cross frame under distortional loads
Downloaded from ascelibrary.org by Syracuse University Library on 07/27/13. Copyright ASCE. For personal use only; all rights reserved.
冑 冉冊
spaced along the girder length, and must possess sufficient stiff- 2
ness. The current design recommendations 共Highway 1982; Guide 2V a 2L DK
D⫽ 冑D V2 ⫹D H
2
⫽ h 2⫹ ⫽ V (2)
1993兲 provide some guidelines on the minimum stiffness and h 2 h
spacing requirements for the cross frames to control the distor- in which L DK ⫽length of the K-frame diagonals. Fig. 9共b兲 dem-
tional stresses, however, no strength requirements are provided. onstrates that the distortional loads cause the two diagonals of a K
Although brace forces are developed in the internal cross frame to develop equal magnitudes of axial forces, D, however,
frames as a result of distortion, these forces are not considered in one is in tension and the other is in compression. Similarly, equal
current design methods 共Highway 1982; Guide 1993兲. The most amounts of tension and compression, S, are developed in the two
common type of internal cross frame in modern box girder sides of the struts from the distortional loads. Eqs. 共1兲 and 共2兲
bridges is the K frame. Fig. 9 shows an internal K frame sub- imply that if the components of the distortional loads 共H and V兲
jected to a concentrated distortional load represented by the forces that act on the K frame are known, the axial forces in the K frame
H and V. Since no torsional moment results from the components can be determined. The next section will focus on identifying the
of the distortional load, V and H must satisfy the equilibrium components of the distortional loads. Although identifying these
distortional components can at first appear somewhat compli-
cated, once these components are identified, relatively simple de-
sign expressions can be developed for the forces induced in the
K-frame members.
components. Therefore, if the equivalent pure torsional compo- Fig. 11. Pure torsional and distortional components in a vertical
nents are known 关Fig. 5共b兲 or 6共b兲兴, they can be subtracted from couple
the applied torques 关Fig. 5共a兲 or 6共a兲兴 to determine the distortional
components 关Fig. 5共c兲 or 6共c兲兴. A pure torsional load consists of a
shear component proportional to the Saint-Venant torsional shear Therefore, utilizing the expression q i ⫽b i T/(2A 0 ) for the shear
flow, which possess a constant intensity along the perimeter of the flow in the ith plate, the component of the pure torsional load on
cross section of the box 共Wright et al. 1968; Fan 1999; Helwig the top plate is 关Fig. 11共c兲兴
and Fan 2000兲. Equilibrium and mechanics of thin-walled mem-
bers can be used to determine the intensity of the uniform shear pb b2
q t,pure T ⫽ ⫻b⫽ p (4)
flow, f, as shown in Fig. 10. A torsional moment of T⫽ f (2A 0 ) 2A 0 2A 0
results from this uniform shear flow, where A 0 is the enclosed The distortional components on the girder plates can be de-
area of the box. Therefore, for an applied torsional moment of T, rived by subtracting the pure torsional components from the ap-
the equivalent pure torsional components of this applied torsional plied torsional load. For example, the distortional component on
load are represented by the uniform shear flow shown in Fig. 10 the top plate of the box is
in which f ⫽T/(2A 0 ). If this distributed load along the perimeter
of the cross section is replaced with the shear loads on the indi- pb 2 a2
q t,dist⫽p t ⫺q t,pure T ⫽2p tan ⫺ ⫽⫺ p (5)
vidual plates, the shear load on the ith plate of the box is equal to 2A 0 2A 0
b i T/(2A 0 ) where b i is the width of the plate. The determination
in which the relations of 2A 0 ⫽h(a⫹b) and tan ⫽(b⫺a)/(2h)
of b i should exclude any plate overhangs. For a rectangular box,
have been applied. Other torsional and distortional components
the distributions shown in Figs. 5共b兲 and 6共b兲 are the pure tor-
on the web and bottom plates of the box can be similarly derived
sional loads that can be verified using these criteria.
and the results are given in Table 1. The expressions given in
The pure torsional load can then be subtracted from the ap-
Table 1 represent the magnitude of the force components and any
plied torque to obtain the distortional components. For example,
negative signs 关such as the one given on q t,dist in Eq. 共5兲兴 have
if the applied torque consists of two opposing vertical forces,
been omitted since the directions of the pure torsional and distor-
p (kN/m), as shown in Fig. 11共a兲, the resulting distributed tor-
tional components are shown in Figs. 11共c and d兲.
sional load is pb (kN m/m). The vertical forces may be resolved
into components in the webs and the top plate as demonstrated in
Fig. 11共b兲 for a trapezoidal cross section. These forces can be
resolved using static equilibrium that results in the following Table 1. Pure Torsional and Distortional Components
forces:
Torsional load by Torsional load by
p t ⫽2p tan , (3a) Component vertical forces horizontal forces
p w ⫽p/cos , (3b) b2 b
q t,pure T p p
2A0 a⫹b
p b ⫽0 (3c)
in which the subscripts t, w, and b refer to the top plate, web bc c
q w,pure T p p
plates, and the bottom plate, respectively. The intensity of the 2A0 a⫹b
‘‘shear flow’’ of the equivalent pure torsional load is pb/(2A 0 ).
ab a
q b,pure T p p
2A0 a⫹b
a2 a
q t,dist p p
2A0 a⫹b
ac c
q w,dist p p
2A0 a⫹b
b
ab p
q b,dist p a⫹b
2A0
Fig. 10. Saint-Venant shear flow in box girder
冉 冊
Downloaded from ascelibrary.org by Syracuse University Library on 07/27/13. Copyright ASCE. For personal use only; all rights reserved.
冉 冊
same procedure outlined above and the results are also presented 1 M
in Table 1. The directions of the force components are indicated q H,pure T ⫽ ⫹ew (7b)
in Figs. 12共b and c兲. 4h R
The torsional and distortional components derived above act in
the planes of the girder plates. For the purpose of comparison
with finite-element results, static equilibrium can be used to trans-
q V,dist⫽
1
冉
M a
⫺ ew
2 共 a⫹b 兲 R b 冊 (8a)
distortional components in the vertical direction at each joint can Eqs. 共7兲 and 共8兲 are based on the assumption that the torsional
be derived from the results in Table 1 as loads induced by curvature are in the same direction as the tor-
sional loads induced by eccentric loads. In the positive bending
1 h moment region 共compression in the top flanges兲, this implies that
q V ⫽ q w,dist cos ⫽ p (6a)
2 2 共 a⫹b 兲 the eccentric load is located outside the centerline of the girder as
demonstrated in Fig. 16共a兲. Therefore, a positive e should be used
The horizontal components at each of the top joints can be de-
to represent the eccentricity of a vertical load that is located out-
rived as
side the girder centerline. If the eccentric load is on the interior
1 1 1 side of the centerline, as shown by Fig. 16共b兲, a negative eccen-
q H ⫽ q t,dist⫹ q w,dist sin ⫽ p (6b) tricity 共e兲 should be used.
2 2 4
The results of these components for both the pure torsional and
distortional loads are presented in Figs. 13 and 14 for the respec-
Brace Forces due to Distortion of Quasi-Closed Box
tive cases of torques from opposing vertical and horizontal forces. Girders
The results presented in Figs. 13 and 14 can be used to derive
the pure torsional and the distortional components under the com- Although the expressions for the distortional components derived
bined action of eccentric vertical loads 共Fig. 2兲 and the girder in the last section are useful for a theoretical analysis on box
curvature 共Fig. 4兲. Fig. 15共a兲 shows a box subjected to the tor- girder distortion, design engineers will generally not carry out
sional loads induced by both the girder curvature and an eccentric such an analysis. Most engineers utilize commercial software that
load. The pure torsional and distortional components in these total employs a grid analysis, which generally does not reflect the ef-
applied torsional loads can be derived by replacing p in Figs. 13 fects of distortion nor the internal brace forces. However, if the
and 14 with ew/b and M /(Rh), respectively. The resulting total distortional loads are known, design expressions can be derived
pure torsional components and distortional components, as de- for the forces in the internal cross frames. Once the distortional
noted by q V,pure T , q H,pure T and q V,dist , q H,dist in Figs. 15共b and c兲, loads are determined, the distortional forces on the internal K
can be obtained by the principle of superposition, which yields frames, represented by H and V in Fig. 9共a兲, can be obtained
the following solutions: using refined methods for box girder distortion, such as the beam-
on-elastic-foundation analogy 共Wright et al. 1968兲. However,
many of these refined methods are not convenient for use in de-
V⫽
sK M a
冉
⫺ ew
2 共 a⫹b 兲 R b 冊 (9a)
sign due to the mathematical complexity. Therefore, a simple Substituting the above expressions into Eqs. 共1兲 and 共2兲 results in
model that is easy to apply is desirable for determining these the following equations for the K-frame forces:
Downloaded from ascelibrary.org by Syracuse University Library on 07/27/13. Copyright ASCE. For personal use only; all rights reserved.
冉 冊
brace forces.
Large brace forces in box girders are often developed during s KL DK M a
D⫽⫾ ⫺ ew (10a)
erection and construction when no composite action is available 2A 0 R b
冉 冊
from the concrete deck. The top plate of the box girder during this
s Ka a M
stage consists of two top flanges on each of the webs and a hori- S⫽⫾ ew⫺ (10b)
zontal truss, as shown in Fig. 17. The top lateral bracing system is 4A 0 b R
provided to resist the torsional moments induced by the dead As shown earlier in Fig. 9共b兲, the two diagonals and struts in a K
loads of the girder and the concrete. The U-shaped steel section frame are subjected to equal magnitudes of tension and compres-
with the top lateral truss system is commonly referred to as a sion due to the distortional loads, which is represented by the ⫾
quasi-closed box girder. The two most common types of horizon- signs in Eqs. 共10a兲 and 共10b兲.
tal trusses are demonstrated in Fig. 17共b兲. Each bracing panel Eqs. 共10a兲 and 共10b兲 will be further studied by comparison
consists of either an X or a single diagonal member 共SD-type兲. K with the FEA results in the next section. Eqs. 共10a兲 and 共10b兲
frames are provided under the lateral struts of the top flange only address the K-frame forces that are generated by the cross-
trusses and are usually spaced every one or two bracing panels sectional distortion of the quasi-closed box girders. Other box
along the length of the truss. Internal and external solid dia- girder deformations may also cause forces in the diagonals and
phragms are normally used at the girder supports. No intermedi- struts of the internal K frames. For example, a uniform force in
ate external cross frames or diaphragms between the adjacent both sides of the lateral struts develops due to bending of box
girders are assumed in this derivation. girders as discussed in a previous paper by Fan and Helwig
As discussed earlier, the distortional loads on box girders are 共1999兲. Live loads also add additional brace forces to the internal
resisted by both the out-of-plane shears in the through-thickness cross frames as a result of the cross-sectional distortion of the
direction of the girder plates and the in-plane shears on the cross fully closed box girders. However, as mentioned above, the dis-
sections of the girder. The out-of-plane distortional stiffness de- tortion in the closed box will be significantly less than in the
pends mainly on the thickness of the plates, along with the shape quasi-closed box. The total forces in the box girder cross frames
and dimensions of the cross section. For a completed box girder, will be studied in a later paper.
this stiffness will usually be relatively large due to the closed
profile and the large thickness of the top plate 共the concrete slab兲.
For a quasi-closed section, however, the top ‘‘plate’’ is only a Comparison with Finite Element Analysis Results
horizontal truss. Between panel points of the top lateral truss, the
individual plate elements of the box section are relatively flexible Three-dimensional FEA was performed to verify the expressions
in out-of-plane bending. Therefore, the out-of-plane bending stiff- developed in the last section for the brace forces in internal cross
ness of quasi-closed box girders can be neglected, which leads to frames. The FEA program ANSYS 共ANSYS 1996兲 was used in
a conservative estimate of the distortional warping stresses and the investigation. The girder plates in the top and bottom flanges
the brace forces in the K frames. A further simplification can be and the webs of the quasi-closed box girders were modeled using
made by considering the fact that most torsional loads are distrib- shell elements, while the bracing members in the top lateral truss
uted along the girder length. A simple solution of the brace forces system and the internal K frames were modeled with either beam
can be obtained from the product of the K-frame spacing and the or truss elements. Shell elements were also used to model the
intensity of the distortional loads formulated in the last section. solid diaphragms at the supports and the stiffeners in the box
girders. Analyses were conducted on both straight and curved box
girders.
The straight girder studied was a simple span quasi-closed box
girder subjected to distributed top flange loads, as shown in Fig.
18共a兲. The load was equivalent to an eccentric distributed vertical
load w⫽48.2 kN/m 共3.3 kips/ft兲 with an offset of 0.53 m 共20.83
in.兲 to the centerline of the girder. This loading condition arises in
girders with an unsymmetrical bridge deck, which frequently oc-
curs in the fascia girder of bridges. The girder dimensions as well
as the bracing member sizes are shown in Fig. 18共b兲. Both X- and
SD-type horizontal trusses were considered. The girder length is
Fig. 17. Quasi-closed box girders 共a兲 cross section and 共b兲 plan view
divided into 16 bracing panels, each with a length of 48.8/16⫽3.1
of truss types
m 共10 ft兲. K frames were provided at every other lateral strut and
Acknowledgments
The writers would like to thank Dr. Karl Frank and Dr. Joseph
Yura of the University of Texas at Austin, Dr. Reagan Herman of
Fig. 21. K-frame forces in curved continuous box girder the University of Houston, and J. C. Liu, P.E. and John Vogel,
P.E. of TxDOT for their assistance and support in this study.