Engineering Structures 30 (2008) 30953104

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Experimental study on the ultimate bending resistance of steel tub girders with top lateral bracing
Byung H. Choi a, , Young-Suk Park b , Tae-yang Yoon a
a Civil Engineering Research Team, Research Institute of Industrial Science & Technology (RIST), 79-5 Yeongcheon, Dongtan, Hwaseong 445-813, South Korea b Department of Civil & Environmental Engineering, Myongji University, San 38-2 Namdong, Cheoin, Yongin 449-728, South Korea

article

info

a b s t r a c t
This paper presents a series of experimental test results on the flexural resistance of a steel tub girder with top lateral bracing. Three test girders were fabricated with high tensile strength SM570 TMC (nominal yield strength, Fy = 460 MPa) steel plates. The test girders had a half-size cross section of practical bridge girders whose depth and length were approximately 1000 mm and 12,000 mm, respectively. The pure bending test was conducted on the tub girder specimens using a universal test machine. Incremental nonlinear analyses were also performed on the tested girders and compared with the tested results. The incremental nonlinear analysis with finite element modeling was confirmed to be applicable to simulating inelastic buckling behavior and evaluating the ultimate bending strength. The effect of crossframes and top lateral bracings on flexural strength was then discussed. The flexural resistance capacity specified in the AASHTO LRFD design specifications was compared with the evaluated ultimate strengths, and the validity of the design specifications was verified. 2008 Elsevier Ltd. All rights reserved.

Article history: Received 10 January 2008 Received in revised form 13 February 2008 Accepted 7 April 2008 Available online 27 May 2008 Keywords: Buckling Beam Box girders Flexural strength Finite element method Nonlinear analysis

1. Introduction and background Composite box girders are frequently used for curved bridges for their superior torsional stiffness, economy, fabrication advantages, aesthetic appearance, etc. A box girder type called steel tub girder is generally known as an improved variant because it has a more efficient composite section with concrete slab as shown in Fig. 1. More detailed design guides have recently been published in AASHTO LRFD 3rd ed. [1] and the NSBA proposed design guideline [2]. Prior to the hardening of the concrete deck, the steel tub girder of an open section type has significantly lower torsional stiffness than that of the composite closed section. Moreover, the noncomposite tub girder must sustain the entire construction load including the wet concrete. Since the top flanges are compressed, and the depth of the web in compression is largest during this construction step in the AASHTO LRFD specifications [1], the non-composite section is susceptible to lateraltorsional buckling. Therefore, a critical construction stage of these girders is the duration of the casting of the concrete deck. Given the vulnerability to lateral torsional buckling in the non-composite state, the flexural resistance of the steel tub girder during this step could

Fig. 1. Typical cross section of composite steel tub girders.

Corresponding author. Tel.: +82 31 370 9588; fax: +82 31 370 9599.
E-mail address: bhchoi@rist.re.kr (B.H. Choi). 0141-0296/$ see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2008.04.014

be significantly affected by the spacing between brace points, i.e., unbraced length. To provide a nodal brace point against lateral torsional buckling for this critical stage, top lateral bracings and interior cross frames are inevitably installed. Top lateral bracings make the open-top trapezoidal section into a pseudo-box section, thereby increasing the torsional stiffness tremendously. Hence, top lateral bracings are mandatory in horizontally curved tub girders. They are also needed in tangent girders for the unintended and/or torsional loads that are unaccounted for as reported by Chen et al. [4]. In AISC LRFD, bracing members are usually classified as lateral bracing and torsional bracing [3]. Since the top lateral bracings of the tub girders are directly connected to the flange in compression to sustain the lateral displacement, they will most probably act as the brace point against the lateral torsional buckling of tub girders.

3096

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

Note, however, that the top lateral bracing is designed mainly to increase only the torsional stiffness of tub girders [2] by making the U-shaped open tub girders as a quasi-closed box section. According to the NSBA design guideline [2], the internal cross-frames, not the top lateral bracings, should be utilized as brace points to prevent girder top flanges from buckling. Moreover, Article 6.11.3.2 of the AASHTO LRFD specifications [1] stipulates that the provisions of the I-section girder shall be applied only to the top flanges of tub sections for critical stages of construction, and that the unbraced length should be taken as the distance between interior crossframes or diaphragms. It is also remarked in the commentary of Article 6.11.3.2 that top lateral bracing attached to the flanges at points where only struts exist between the flanges may be considered as brace points at the discretion of the engineer. In addition, the test results obtained by Chen et al. [4] suggested that the bending capacity of the tub girder could be less than the nominal flexural resistance specified by the AASHTO LRFD design specifications [5] when test specimens were braced only by the top lateral bracing. The experimental studies by Chen et al. [4] were conducted on a series of scaled test models of a U-shaped girder with the X-brace truss system or metal deck panels fixed to the top flanges. The detrimental effects of shortening in the top lateral bracing were also found to decrease the U-shaped girder bending capacity. Current design provisions for U-shaped girders consider the brace points as unyielding support. For the test case, the experimental load capacity was 25% less than the nominal flexural resistance calculated using the design specifications [5]. The cross-frames installed in between the I-shaped girders act as the brace points to prevent the girder flanges from buckling in compression. It is unlikely that the I-shaped girders, and the internal cross-frames of box girders installed in an inner section, will maintain the original cross-sectional shape against sectional distortion and effectively control distortional stresses. According to the AASHTO Guide specifications for curved girder bridges [6], the interior cross-frames should be installed with proper space and sufficient rigidity to limit the distortional-to-bending normal stress ratio to less than 10% and it is also stipulated that the maximum cross-frame spacing should not exceed 10 m (30 ft) in all horizontally curved tub girders. Therefore, the required spacing for interior cross-frames may differ from what is required as brace points which should be more compact to establish an unbraced design length such as the strut spacing of the top lateral bracing system. Note that bracing members are no longer utilized once the concrete deck has cured as they could represent a significant portion of a girder steels weight and cost; hence, the need to minimize them. Still, there is no existing codified design method to consider the top lateral bracings as the brace points of the noncomposite tub girder design. Moreover, the structural stability of the non-composite tub girder under positive bending as usually encountered during the concrete deck construction stage has not been properly addressed in previous research. As such, verifying the flexural strength of the non-composite steel tub girders and comparing them with those specified in the design specifications such as the AASHTO LRFD [1] served as a strong motivation for the authors to conduct this study. At the same time, investigating whether the top lateral bracings could perform as brace points for the non-composite tub girder sections in positive flexure was required to illustrate that the unbraced length of the steel tub girder could be determined based on the strut spacing of the top lateral bracings. The objective of this experimental study was to investigate the following: (1) Inelastic buckling of non-composite tub girders at failure (2) Experimental and analytical evaluation of ultimate flexural resistance in positive flexure (3) Validity of top lateral bracings as brace points for the determination of unbraced length.

Fig. 2. Fabrication of test specimens.

An approach to estimating the forces in the lateral bracing members due to flexure and top-flange bending stresses was recently proposed by Fan and Helwig [7]. They found that the current design method based on the EPM (equivalent plate method) developed by Kollbrunner and Basler [8] underestimated the member forces of the top lateral bracing. They attributed the difference to the neglected effects of girder vertical bending and erroneous assumptions in terms of the lateral force components of applied loading due to sloping webs. Based on the equations proposed by Fan and Helwig [7], a realistic section of the top lateral bracing could be selected for the steel tub girder models in this study. Previously recommended design guides [9] (Highway, 1975) were also used as reference. Test specimens of the non-composite tub girders and experimental set-up are presented followed by the results of the bending test. The ultimate flexural behaviors of non-composite tub girders are represented through the test results and finite element analyses. The ultimate flexural strengths from the bending test and finite element analysis of the tub girder models are then compared with the currently available design specifications. The last section of this paper summarizes the findings of this study. 2. Test specimens and set-up 2.1. Steel tub girder specimens Three tub girder models were prepared with various strut spacings as shown in Fig. 2. The total length of the test model was 12,000 mm. In particular, the middle part of 7200 mm served as the specimen region subjected to uniform bend-loading. The depth of the girder was 1017 mm. The general dimensions and properties of

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104 Table 1 General dimensions and properties of test girders (Unit: mm)
H D tw buf tuf blf tlf rt Lr Lp

3097

Table 3 Top lateral bracings for the test girders Test girders Bracing strength Bracing type A B C L75 75 6 L90 90 10 L100 100 10
Pn

Comparison w/design force

Dbend Pn /Dbend

984.8

1014.0

10.5

200.0

10.5

1100.0

20.0

47.5

3578.3

953.0

(kN) 103.7 134.9 104.0

(kN) 56.3 52.7 34.6 1.8 2.6 3.0

Sbend (kN) 43.7 27.7 13.5

Pn /Sbend

2.4 4.9 7.7

and usually used nowadays. The test girder specimens denoted as A, B, and C and shown in Fig. 4 had strut spacings of 1200 mm, 2400 mm, and 3500 mm, respectively. The unbraced length of the test specimens was in between Lr and Lp . As such, they had a noncompact section in the inelastic buckling region. Angle (L) shapes made of SS 400 steel were used for the top lateral bracing members for test specimens whose nominal yield strength was 240 MPa. The design force equations for the top lateral bracings of steel tub girders were recently established by Fan and Helwig [7]. Their design equations for the forces induced in the diagonals (Dbend ) in the struts (Sbend ) and resulting lateral bending stress fL bend due to vertical bending are expressed as (especially for single-diagonal (SD) type trusses):
Dbend fL bend fx Top s cos

K1 1.5s = 2 Sbend bf tf

Sbend = Dbend sin ;

(1)

where K1 and K2 are parameters defined as:

K1 = d Ad

b As

sin +
2

s3

2b3 f tf

sin ;
2

K2 =

d Ad

2b
As

sin
2

(2)

Fig. 3. Stressstrain curves of SM 570 TMC steel plate coupons.

Table 2 Material properties of steel plates used Member Top flange Bottom flange Web Width (mm) 200.0 1100.0 1014.0 Thickness (mm) 20.0 10.5 10.5 Yield strength (MPa) 533 510 510

where s = spacing of struts (panel length), = acute angle between the top flange and the diagonal, bf and tf = respective values of the width and thickness of the top flange, d = length of a diagonal, and Ad and As = respective cross-sectional areas of the diagonals and struts. The cross-sectional dimensions of the top lateral bracings for the test girders are shown in Table 3. Their compressive strength (Pn ) was then compared with the applied forces as calculated from the proposed design (Eqs. (1)(2)). As seen from Table 3, the top lateral bracings were designed to resist the design forces sufficiently as proposed by Fan and Helwig [7]. 2.3. Initial geometric imperfections Prior to testing, the initial geometric imperfections of each girder model were measured using a laser displacement sensor. Fig. 5 shows the many types of initial imperfections found in the test girders. Lateral displacements in the top flanges between the strut spacing were measured at 200-mm intervals. Out-ofplane deflections of the web panels were also obtained in a region between the transverse stiffeners, possibly affecting the ultimate buckling strength of the model girders considerably due to the lateral torsional buckling mode and web bend-buckling mode. The maximum values of the geometric imperfections between nodal lines were measured for all test girders (Table 4). The maximum tolerance value for the web flatness deviation (f ) due to welding was taken as w/120 based on Article 3.5 of the Bridge Welding Code [10]. The maximum tolerance limit (c ) for the crookedness (straightness) of the top flanges in the lateral direction was based on Article 11.4.13.4 of the AASHTO Specifications Division II [11], i.e., should not exceed Lb /1000 where Lb denotes the unbraced length of the top flange. The maximum values of the initial geometric imperfections as

the girder section are presented in Table 1. The limiting unbraced lengths denoted as Lr and Lp were for the slender section and noncompact section, respectively, as calculated by the equations in Article 6.10.8.2.3 of the AASHTO LRFD [1]. The main girder members of the test specimens consisted of SM 570 TMCP (thermo-mechanical control process) steel plates whose nominal yield strength was estimated to be 460 MPa. As the representative property of high-strength steel, the stressstrain curve of the steel plate coupon did not have an obvious yield point as presented in Fig. 3 unlike the other steels stressstrain curve. The yield strength values were obtained by the ASTM tension test on coupons cut from steel plates to fabricate the test girder models and presented in Table 2. 2.2. Arrangement of the top lateral bracing The top lateral bracing for the experimental study consisted of a single diagonal (SD) type truss, which is more economical

3098

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

Fig. 4. Top lateral bracing of test specimens: (a) Specimen A; (b) Specimen B; (c) Specimen C.

Table 4 Initial geometric imperfections Test Top-flange crookedness (mm) specimens Web flatness (mm)

Measured Unbraced AASHTO Diff. Measured Width AASHTO Diff. length A B C Fig. 5. Initial geometric imperfection types: (a) Top-flange lateral crookedness; (b) Web out-of-flatness. 1.6 2.9 3.9 1200.0 2400.0 3500.0 1.2 2.4 3.5 0.4 0.5 0.4 3.9 4.6 6.8 1014. 8.5 1014. 8.5 1014. 8.5

4.5 3.8 1.7

measured from the test girders were compared with the specified tolerance limits (Table 4). Fig. 6 shows measurement of geometric

imperfections in the web panel of a test girder utilizing the laser displacement sensor.

Fig. 6. Measurement of initial geometric imperfections in the test girders: (a) laser sensor (b) data logger.

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104 Table 5 Ultimate flexural strengths from the pure bending test Test specimens
Lb (mm) Mcr (108 N mm)

3099

Diff. (%)

Test result A B C 1200 2400 3500 61.9 57.0 44.6

AASHTO LRFD 58.4 49.9 42.0 6.0 14.3 6.1

3. Test results and comparative studies 3.1. Ultimate bending behavior of non-composite tub girders The failure modes of test girder models under pure bending are illustrated in Fig. 9. As the strut spacing decreased, and unbraced length Lb approximated Lp , local buckling in the top flanges and webs was clearly observed at failure. As shown in Fig. 9, the failure mode of test model A appears to be a complex mode with obvious local deformations in the top flanges, webs, and top lateral bracings. On the other hand, in the failure mode of test model B as shown in Fig. 10, the lateral torsional buckling mode in between the struts was more clearly formed; some local deformations in the webs and flanges might be affected by the lateral torsional buckling behavior. Fig. 11 shows the failure mode of test model C, where the lateral torsional buckling mode is more dominant and almost no local buckling modes were observed in the webs or flanges. As the unbraced length increased in the steel tub girders, lateral torsional buckling between the top lateral struts was found to tend toward the more dominant failure mode. The test failure modes revealed some inflection points formed at the strut locations in the top lateral displacements due to the bending-compressive force. The struts of the top lateral bracing would likely be able to establish a nodal point against the lateral torsional behavior of the top flanges and webs if they have sufficient stiffness against local buckling failure. 3.2. Assessment of design equations on ultimate flexural strength In Table 5 and Fig. 12, the ultimate flexural strengths of tub girder specimens are presented and compared with the bending resistances from AASHTO LRFD [1]. The strength curve represented in Fig. 12 was suggested in the AASHTO specifications (Article 6.10.8.2.3 of AASHTO LRFD [1]) to provide a valid bending resistance of the compression flange due to lateral torsional buckling. Nominal flexural resistance varied along unbraced length Lb as determined based on the doubly symmetric I-sections (Timoshenko and Gere 1961 [12]; AASHTO LRFD, 2004). If Lb Lp , then Fnc = Rb Rh Fyc where Lp = Limiting unbraced length to achieve the nominal flexural resistance of Rb Rh Fyc under uniform bending (mm) = 1.0rt 2.4. Test set-up and apparatus Fig. 7 shows the set-up of a universal testing machine with capacity of 1000 metric tons for the ultimate introduction of uniform bending load into the test specimens. The specimen region was located in the middle part up to the L/3 size of the model girders as shown in Fig. 7 and was expected to be subjected to pure moment by two-point loading. As shown in Fig. 8, several LVDT sensors were installed at the centerline points beneath the bottom flange of each girder specimen to measure the vertical bending and distortional displacements. Electronic strain gauges were also mounted on some top-lateral bracing members as well as the girder main members.
E Fyc

Fig. 7. Test set-up for uniform bending (1000-ton UTM): (a) facial view; (b) side view.

Fig. 8. Measurement set-up for the loading test.

Rb = Web load-shedding factor determined as specified in Article 6.10.1.10.2 Rh = Hybrid factor determined as specified in Article 6.10.1.10.1 The nominal flexural resistance of a member braced at or below the non-compact limit is expressed as a linear function of the unbraced length as illustrated in Fig. 12. In other words, for the case of Lp Lb Lr , Fnc = Cb 1 1 Fyr Rh Fyc Lb Lp Lr Lp Rb Rh Fyc Rb Rh Fyc

(3)

for the slender region where the unbraced length is larger than Lr (Lb Lr ),

3100

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

Fig. 9. Ultimate failure modes from test model A: (a) entire failure mode; (b) web bend-buckling; (c) flange local bucking; (d) top diagonal bracing deformation.

Fig. 10. Ultimate failure modes from test model B: (a) bend-loading; (b) ultimate bending; (c) flange lateral displacement; (d) local deformations.

Fnc = Fcr Rb Rh Fyc

Fyr = Compression-flange stress at the onset of nominal yielding

Cb Rb 2 E 2 L
b rt

where Fcr = Elastic lateral torsional buckling stress (MPa) =

Lr = Limiting unbraced length to achieve the onset of nominal

yielding in either flange under uniform bending considering the compression-flange residual stress effects (mm) = rt FE yr

within the cross-section including residual stress effects but excluding compression-flange lateral bending taken as the smaller value between 0.7Fyc and Fyw , but not less than 0.5Fyc Cb = Moment gradient modifier. With the exception of hybrid sections with Fyw < 0.7Fyc , the term Fyr is always equal to 0.7Fyc . For the case of uniform major-

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

3101

Fig. 11. Ultimate failure modes from test model C: (a) ultimate loading state; (b) flange lateral displacement; (c) inflection point at strut.

axis bending when Cb is equal to 1.0 and when rt is defined as:

rt = bf c
1 Dc tw D2 3 bf c tf c hd

12

h d

where Dc = Depth of the web in compression in the elastic range (mm) The AASHTO inelastic transition curve Eq. (3) appears to match the test results at a maximum of 6% difference. Thus, Eq. (3) can be said to provide a rational guideline for the determination of the bending resistance of non-composite steel tub girders. The test result comparison also revealed that a parabolic curve similar to that proposed by Choi et al. [15] and Choi [16] could better simulate the ultimate bending resistance in the inelastic buckling range. 3.3. Numerical analysis
Fig. 12. Comparison of the ultimate failure strength.

3.3.1. Finite element modeling Steel tub girder models were simulated using a commercial finite element code, ABAQUS [13]. The model in the middle specimen region was assumed to be subjected to a uniform bending state. Thus, two-point loading was simulated by applying a concentrated load at a total of four nodes on the edges of the specimen region with the end boundary conditions as shown in Fig. 13. Although there were various finite elements available in ABAQUS, only the four-node plate/shell element S4F was used to represent the flanges and webs of the tub girder models for its simple yet quantitatively stable performance. A series of analytical investigations conducted previously on steel plate buckling such as that by Lee and Yoo [14] showed that a minimum of four square four-node shell elements was sufficient to model a flange or a web panel. In this study, the top flanges were divided into four square elements for an appropriate aspect ratio of the elements; the bottom flanges and webs were divided into eight square elements for sufficient accuracy based on the studies of Choi et al. [15] and Choi [16]. For simplicity, however, the top lateral bracing and cross-frames were modeled by CBAR, a frame element. 3.3.2. Incremental nonlinear analysis A series of incremental nonlinear analyses was performed on the steel tub girder models. To model the initial geometric imperfections, the out-of-flatness of web panels was simulated using the web-bend local buckling mode factored by the AASHTOspecified maximum tolerance (f ) value; the crookedness of the top flanges was modeled by the lateraltorsional buckling mode factored by the AASHTO-specified maximum tolerance value (c ). The maximum tolerance (f ) of web flatness was determined based on the D/120 ratio as suggested by the Bridge Welding Code [10]. On the other hand, the maximum tolerance (c ) of flange crookedness was determined from the L/1000

Fig. 13. Finite element modeling.

proportion. L denotes the longitudinal compression member length; in this study, however, it could be taken in two ways: the lateral bracing spacing (Lb ) and the total specimen length (Ls ). In the numerical analysis in this study, the geometric imperfections were considered for the web-bend buckling and the lateraltorsional local buckling. The eigenvector corresponding to the first eigenvalue of the elastic buckling analysis was used for the imperfection shape and the amplitude of the imperfection was determined as stated above. Thus, elastic buckling analyses were performed prior to the incremental nonlinear analysis. A bilinear stressstrain curve that was linearly elastic and perfectly plastic was assumed for the material constitutive relationship of structural steel. The von Mises yield criterion was adopted for the material nonlinear analysis.

3102

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

Fig. 14. Analysis result comparison for the ultimate failure mode: (a) Specimen A; (b) Specimen B; (c) Specimen C.

Fig. 15. Local buckling mode from the finite element analysis results: (a) Specimen A; (b) Specimen B; (c) Specimen C. Table 6 Results of finite element analysis of ultimate bending behavior Specimen type
Lb

Test result
Mcr

Analysis result Failure mode Web bending Lateral torsional Lateral torsional
Mcr

Diff. (%)

(mm)

(108 N mm)
A B C 1200 2400 3500 61.9 57.0 44.6

(108 N mm) 58.9 54.2 42.0 5.1 5.2 6.2

3.3.3. Results of the finite element analysis The lateraltorsional global buckling, lateraltorsional local buckling, and web-bend buckling (flange local buckling) modes were applied to each finite element analysis model. The ultimate bending strength values were quantitatively evaluated and presented together with the ultimate failure mode (Table 6). The bend-buckling behavior of non-composite tub girders with slender section was more sensitive to the lateraltorsional buckling mode. On the other hand, as the unbraced length decreased to approximate the compact unbraced length Lp value, webbend local buckling should have more influence on the ultimate bending strength. The failure modes resulting from the nonlinear incremental analysis for each tested model are represented in Fig. 14 and compared with the test results in more detail in Fig. 15. Note that the figures show the perfect match between the local buckling modes in the top flanges and web of model A and lateral torsional buckling modes of models B and C. Therefore, based on the comparison with the ultimate failure modes from the test,
Table 7 Effect of intermediate cross-frame on the ultimate bending resistance Intermediate cross-frame
Lb (mm)

the ultimate bend-failure behavior of steel tub girders could be adequately simulated through incremental nonlinear analysis. The applied bending moment and vertical displacement M curves are presented in Fig. 16 together with the two different geometric imperfection amplitudes based on the AASHTO and AWS specifications and measured value. The difference in the measured and AASHTO and AWS tolerances for the geometric imperfections had little effect on the ultimate strengths. This phenomenon is consistent with observations made in other research [15,16]. As the unbraced length of the compact section is reduced to that for the compact section, flexural ductility behind the ultimate bending strength becomes more pronounced. Based on the result of the finite element analysis on flexural ductility in post-buckling, however, a considerably lower evaluation compared to the test result was possible. The difference in the finite element analysis result related to post-buckling behavior may be attributed to the adoption of the linearly elastic and perfectly plastic (460 MPa) simplified model for the material properties of high-strength steel. As shown in Fig. 3, the steel for test specimen SM570 TMC (Fy = 460 MPa) usually has a strain hardening pattern unlike that of conventional steels, i.e., begins immediately after passing the yielding point. Table 6 shows that the top lateral bracings of the test models should have sufficient stiffness to attain the design bending strength specified by the AASHTO specifications. The results of the incremental nonlinear analysis showed maximum values that were about 6% less than the test results and approximately 3% less than the estimations of the AASHTO design specifications. 3.3.4. Alternative bracing scheme comparison As explained earlier, the top flanges of the tub girder specimens were laterally supported by only the strut members in the test

Bracing type

Buckling mode

Mcr (108 N mm)

Diff. (%) AASHTO LRFD 49.9 49.9 14.3 8.6

FE Analysis Yes No 2400 2400 L 90 90 10 L 90 90 10 Local, LTB Local, LTB 57.0 54.2

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104

3103

Fig. 18. Comparison of M curves along the cross-frame installation scheme.

connected vertically to the cross-frame members was analyzed for test specimen B. The analysis results are compared in Table 7 and Fig. 17. Based on the comparison, the failure modes from the analysis models were similar regardless of the vertical crossframes used. A 4.9% difference in the ultimate bending strength was noted between the two models. The lower strength value is for the case wherein the vertical cross-frame was removed, thereby exceeding the bending resistance suggested by the AASHTO design specifications. The comparison of the M curves is plotted in Fig. 18. The comparative study revealed that the two bracing scheme models have a minimal effect on flexural ductile capacity in addition to the bending resistance. Therefore, only the strut members of the top lateral bracing system should be able to provide sufficient stiffness to withstand lateraltorsional buckling behavior. Hence, vertical cross-frame spacing should be designed only considering the distortional behavior in the box section especially for the composite section after the concrete deck is hardened. 4. Summary and conclusion This study examined the ultimate flexural resistance of noncomposite steel tub girders in positive flexure by presenting a series of experimental test results. Three test girders were fabricated with high tensile strength SM570 TMC (nominal yield strength, Fy = 460 MPa) steel plates. The test girders had a half-size cross section of practical bridge girders. A series of pure bending tests was conducted on the non-composite steel tub girders using a universal test machine. The flexural resistance capacity specified in the AASHTO LRFD design specifications was compared with the tested ultimate strengths. The validity in the broad application of the corresponding AASHTO design strength estimations to steel tub girders was also demonstrated. Fan and Helwigs design equations for the top lateral bracings were found to be effective in the SD type bracing design for the test girders. Still, the applicability to more general cases should be studied further especially for the strut members as a brace point. The incremental nonlinear analysis of the finite element model effectively simulated the ultimate bending behavior of non-composite tub girders for the ultimate strength, mode, and brace forces. The consistency with the test results verified the applicability of incremental nonlinear analysis with finite element modeling to simulating inelastic buckling behavior and evaluating the ultimate bending strength. Nonetheless, additional quantitative studies may be required to investigate the minimum required bracing stiffness for the ultimate flexural resistance. The effect of the intermediate cross-frames on flexural strength was quantitatively examined using the verified analytical models. The

Fig. 16. M curves under bend-loading: (a) Specimen A; (b) Specimen B; (c) Specimen C.

region. In other words, the struts of the top lateral bracing system in the specimen region were not shored up by a vertical crossframe. To investigate the effect of the vertical cross-frame on the establishment of a braced point and the ultimate bending strength of tub girders, an additional analysis model wherein the struts were

Fig. 17. Ultimate failure mode comparison along the alternative cross-frame installation scheme: (a) w/ Cross-frame; (b) Strut only.

3104

B.H. Choi et al. / Engineering Structures 30 (2008) 30953104 Steel Construction; 2006. [4] Chen BS, Yura JA, Frank KH. Top lateral bracing of steel U-shaped girders. Rep no FHWA/TX-02/1395-2F. Austin (TX): Center for Transportation Research, University of Texas; 2002. [5] AASHTO. LRFD bridge design specifications. 2nd ed. Washington (DC): American Association of State Highway and Transportation Officials; 1998. [6] AASHTO. Guide specifications for horizontally curved steel girder highway bridges with design examples for I-girder and box-girder bridges. Washington (DC): American Association of State Highway and Transportation Officials; 2003. [7] Fan Z, Helwig TA. Behavior of steel box girders with top flange bracing. J Struct Eng ASCE 1999;125(8):82937. [8] Kollbrunner CF, Basler K. Torsion in structures An engineering approach. New York: Springer; 1969. [9] Highway structural design handbook. Pittsburgh (PA): AISC Marketing; 1982. [10] ANSI/AASHTO/AWS D1.5-96. Bridge welding code. Washington (DC): A joint publication of the American Association of State Highway and Transportation Officials and the American Welding Society; 2002. [11] AASHTO. Standard specifications for highway bridges. 17th ed. Washington (DC): American Association of State Highway and Transportation Officials; 2002. [12] Timoshenko SP, Gere JM. Theory of elastic stability. 2nd ed. New York: McGraw-Hill; 1961. [13] ABAQUS. Analysis users manual version 6.5. Pawtucket (RI): ABAQUS, Inc.; 2004. [14] Lee SC, Yoo CH. Strength of curved I-girder web panels under pure shear. J Struct Eng, ASCE 1999;125(8):84753. [15] Choi BH, Kang YJ, Yoo CH. Stiffness requirements for transverse stiffeners of compression panels. Eng Struct 2007;29(9):208796. [16] Choi BH. Design requirements for longitudinal stiffeners for horizontally curved box girders. Ph.D. thesis. Auburn (AL): Auburn University; 2002.

strut members of the top lateral bracing system were found to be able to establish a brace point which made the non-composite tub girder attain the flexural resistance capacity specified in the AASHTO LRFD design specifications. The cross-frames should be designed as intermediate diaphragms for limiting the distortional stress. Acknowledgments This research was supported in part by a grant from the POSCO research program and the Construction Core Technology Program (05CCTRD09-High Performance Construction Material Research Center) funded by the Ministry of Construction and Transportation of the Korean government. The authors deeply appreciate their financial support. The opinions and conclusions expressed or implied in this paper are those of the authors and are not necessarily those of the funding agencies. References
[1] AASHTO. LRFD bridge design specifications. 3rd ed. Washington (DC): American Association of State Highway and Transportation Officials; 2004. [2] NSBA. Practical steel tub girder design. Chicago (IL): National Steel Bridge Alliance; 2005. [3] AISC. Steel construction manual. 13th ed. Chicago (IL): American Institute of