Computers and Structures 84 (2005) 1118 www.elsevier.

com/locate/compstruc

Deection of simply supported box girder including eect of shear lag

Chartree Lertsima a, Taweep Chaisomphob a,*, Eiki Yamaguchi b, Jaturong Sa-nguanmanasak a
a

Civil Engineering Program, Sirindhorn International Institute of Technology, Thammasat University, Pathumthani 12121, Thailand b Department of Civil Engineering, Kyushu Institute of Technology, Tobata, Kitakyushu 804-8550, Japan Received 8 October 2004; accepted 28 September 2005

Abstract The shear lag has been studied for many years. Nevertheless, most of the studies are related to the eect of the shear lag on stress distribution and very few have investigated the eect on deection, although some design codes have formulas for the eect of the shear lag on deection. In this conjunction, the present study carries out three-dimensional nite element analyses for various box girders to investigate the deection at the mid-span. The multimesh extrapolation is employed to ensure the accuracy. The present study thus reveals the inuence of the parameters that characterize the geometry of a box girder on the deection. It is also shown that the formulas adopted in the design codes underestimate the deection considerably. Based on the present numerical results, empirical formulas are proposed to compute the deection magnication factors that account for the dierence between the deections due to the nite element analysis and the beam theory. 2005 Elsevier Ltd. All rights reserved.
Keywords: Shear lag; Deection; Simply supported box girder; Three-dimensional nite element analysis; Parametric study; Multimesh extrapolation

1. Introduction The phenomenon of nonuniform normal stress distribution in the ange of a thin-walled member is called the shear lag. It has been known for many years and studied by many researchers. A concise but excellent literature review of research on the shear lag is provided by Tenchev [1]. However, most of the existing studies focus on stress distribution, and the eect of the shear lag on deection is rarely discussed. Nevertheless, British Standards Institute [2] and Japan Road Association [3] have formulas to evaluate such eect in the form of the eective width and they have been used for years in practice. However, even the denition of the eective width for evaluating deection
Corresponding author. Tel.: +66 2 986 9009x1907; fax: +66 2 986 9009x1900. E-mail address: taweep@siit.tu.ac.th (T. Chaisomphob). 0045-7949/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.compstruc.2005.09.008
*

does not seem to be very clear, while the eective width for the stress evaluation is dened explicitly in many papers [1,47]. In the past, analytical means was resorted to so as to investigate the shear lag phenomenon [713]. Some assumptions had to be employed to yield solutions inevitably in such an approach. The nite element method was also used, but the analysis model was often reduced to a two-dimensional problem due to the limitation of computer capacity [1,6]. With the advancement of computer technology, it is no longer an insurmountable task to deal with the shear lag phenomenon by the three-dimensional nite element method. In fact, recently the authors have conducted the nite element analysis by shell elements to study the eect of the shear lag on stress concentration in a simply supported box girder [14]. An eort was made to eliminate the discretization error by the multimesh extrapolation [15]. Loading conditions were also carefully treated.

12

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

In the present study, the deection of a simply supported box girder at the mid-span, i.e. maximum deection, is investigated by the three-dimensional nite element method with shell elements. To this end, a parametric study is conducted for various values that characterize the geometry of a box girder. The multimesh extrapolation [15] is utilized to ensure the accuracy. Empirical formulas are then proposed to account for the dierence between the deections due to the nite element analysis and the beam theory. In all the analyses, a nite element program, MARC [16], is used. 2. Box girders to be analyzed Simply supported box girders (Fig. 1) under concentrated and uniformly distributed loads are analyzed. In the present study, an entire box girder is modeled as it is by shell elements: the shear lag problem is not reduced to a two-dimensional plane stress problem and no beam assumptions are implemented in the present analysis. In the three-dimensional nite element model, the load that is unique in the beam theory can be applied in various ways and the shear lag eect may vary by the way the load is applied. Therefore, following the previous study [14], two types of concentrated load shown in Fig. 2 and two types of uniformly distributed load shown in Fig. 3 are considered in the present study. Note that many researchers do not describe their loading conditions explicitly. Tenchev [1] is one of the very few researchers that have provided this kind of information: he has employed Loads C-2 and D-1 for his concentrated and distributed loads, respectively as shown in Figs. 2 and 3. 3. Deection evaluation The structural model described in the previous chapter is analyzed by the nite element method using shell elements. Although the nite element method is very versatile and powerful, the result may depend largely on nite element mesh employed in the analysis. To this end, the inuence of nite element mesh on the deection is rst studied: four nite element meshes of Meshes AD are used to evaluate the deection at the mid-span of a simply supported box girder (H/L = 0.2, B/H = 1.0, Tf/Tw = 1.0) under Load C-2. All the elements in each mesh are square and from Mesh A to Mesh D, the size of an element becomes ner in a consistent way, as may be realized in Fig. 4. Due to
2B Tw Y X Tf
(a) (b)
Fig. 1. Structural geometry of box girder: (a) cross section; (b) side view.

Mid-span section

Mid-span section

(a)

(b)

Fig. 2. Concentrated load: (a) Load C-1; (b) Load C-2.

Mid-span section

Mid-span section

(a)

(b)

Fig. 3. Distributed load: (a) Load D-1; (b) Load D-2.

Fig. 4. Finite element meshes.

2.4 2.3

Dm

w/w beam

2.2 2.1 2.0 0 500 1000 1500 2000 2500 3000 3500 4000 2 (mm2)

H
L

Fig. 5. Variation of deection with respect to representative element size.

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

5.0 4.0 3.0

13

B/H = 0.5: Load C-1 B/H = 1.0: Load C-1 B/H = 1.5: Load C-1 B/H = 2.0: Load C-1

B/H = 0.5: Load C-2 B/H = 1.0: Load C-2 B/H = 1.5: Load C-2 B/H = 2.0: Load C-2

2.0 1.0 0.0 0.025 0.050 0.075 0.100 0.125 H/L 0.150 0.175 0.200

(a)
5.0 4.0 3.0
B/H = 0.5: Load D-1 B/H = 1.0: Load D-1 B/H = 1.5: Load D-1 B/H = 2.0: Load D-1

The linear extrapolation can then be used to estimate the converged value of w/wbeam when the element size vanishes, as indicated by the dotted line in Fig. 5. This extrapolation method is what Cook et al. [15] call multimesh extrapolation. The converged value of w/wbeam, i.e. the point indicated by the arrow in Fig. 5, gives the deection magnication factor Dm that is sought in the present nite element analysis. 4. Parametric study

Dm

B/H = 0.5: Load D-2 B/H = 1.0: Load D-2 B/H = 1.5: Load D-2 B/H = 2.0: Load D-2

2.0 1.0 0.0 0.025 0.050 0.075 0.100 H/L 0.125 0.150 0.175 0.200

(b)

Fig. 6. Variation of Dm with respect to H/L(Tf/Tw = 1.0): (a) concentrated load; (b) distributed load.

symmetry, only a quarter of the box girder is analyzed. Yet the numbers of elements amount to 2080, 8320, 33,280 and 133,120 for Meshes AD, respectively. Fig. 5 shows the variation of the deection ratio w/wbeam, i.e. the ratio of the deection due to the nite element analysis w to the deection due to the beam theory wbeam, at the mid-span. This gure gives the variation with respect to the square of representative element size D. To be noteworthy, the line in Fig. 5 is almost straight in the vicinity of D2 = 0, which is in accordance with the description of Cook et al. [15]: the deection error is proportional to the square of element size. This observation in Fig. 5 may also serve as an evidence for the validity of the present numerical results.
2.5 2.0 1.5
Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5 Tf/Tw = 2.0 Tf/Tw = 2.0

The nite element analysis is conducted for various values that characterize the geometry of a box girder: H/L = 0.025, 0.05, 0.1, 0.15, 0.2; B/H = 0.5, 1.0, 1.5, 2.0; Tf/Tw = 0.5, 1.0, 1.5, 2.0. Fig. 6 shows the numerical results of Dm for Tf/Tw = 1.0. It can be seen that with the increase of each of H/L and B/H, Dm becomes larger in a monotonic fashion. Four loading conditions are applied in this numerical study. However, the two concentrated loads of Loads C-1 and C-2 give very little dierence as can be seen in Fig. 6(a). Fig. 6(b) indicates that the dierence due to the two distributed loads of Loads D-1 and D-2 is also subtle. Unlike stress [14], the way the load is applied is thus insignicant in the case of deection. Therefore, in the following numerical analyses, only Loads C-2 and D-2 are considered. In the relationships between Dm and H/L given in Figs. 7 and 8, the eect of Tf/Tw can be observed: as Tf/Tw increases, Dm becomes larger. The eect of Tf/Tw is therefore also signicant. Figs. 9 and 10 show the variation of Dm with respect to B/H. As stated earlier, Dm increases along with B/H. These gures further show that nonlinearity in the relationship between Dm and B/H is less than that in the relationship between Dm and H/L.
5.0 4.0
Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5 Tf/Tw = 2.0 Tf/Tw = 2.0

Dm

Dm

1.0 0.5 0.0 0.025

Dm

3.0 2.0 1.0

0.050

0.075

0.100 H/L

0.125

0.150

0.175

0.200

0.0 0.025

0.050

0.075

0.100 H/L

0.125

0.150

0.175

0.200

(a)
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.025 0.050 0.075
Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5 Tf/Tw = 2.0 Tf/Tw = 2.0

(c)
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.025

Dm

0.100 H/L

0.125

0.150

0.175

0.200

Dm

Tf/Tw 0.5 Tf/Tw = 0.5 Tf/Tw 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw 1.5 Tf/Tw = 2.0 Tf/Tw 2.0

0.050

0.075

0.100 H/L

0.125

0.150

0.175

0.200

(b)

(d)

Fig. 7. Variation of Dm with respect to H/L for various Tf/Tw (Load C-2): (a) B/H = 0.5; (b) B/H = 1.0; (c) B/H = 1.5; (d) (B/H) = 2.0.

14
2.5 2.0 1.5
Tf/Tw= 0.5 Tf/Tw = 0.5 Tf/Tw= 1.0 Tf/Tw = 1.0 Tf/Tw= 1.5 Tf/Tw = 1.5 Tf/Tw= 2.0 Tf/Tw = 2.0

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

5.0 4.0

Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5

Dm

1.0 0.5 0.0 0.025

Dm

3.0 2.0 1.0

Tf/Tw = 2.0 Tf/Tw = 2.0

0.050

0.075

0.100 H/L

0.125

0.150

0.175

0.200

0.0 0.025

0.050

0.075

(a)
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.025 0.050 0.075
Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5 Tf/Tw = 2.0 Tf/Tw = 2.0

(c)
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.025
Tf/Tw = 0.5 Tf/Tw = 0.5 Tf/Tw = 1.0 Tf/Tw = 1.0 Tf/Tw = 1.5 Tf/Tw = 1.5 Tf/Tw = 2.0 Tf/Tw = 2.0

0.100 0.125 H/L

0.150

0.175

0.200

Dm

Dm

0.100 H/L

0.125

0.150

0.175

0.200

0.050

0.075

0.100 H/L

0.125

0.150

0.175

0.200

(b)

(d)

Fig. 8. Variation of Dm with respect to H/L for various Tf/Tw (Load D-2): (a) B/H = 0.5; (b) B/H = 1.0; (c) B/H = 1.5; (d) B/H = 2.0.

4.0 3.0

6.0
H/L = 0.025 H/L = 0.10 H/L = 0.20 H/L = 0.05 H/L = 0.15

5.0 4.0

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

Dm

Dm

2.0 1.0 0.0 0.5 1.0 B/H 5.0 4.0 3.0 1.5 2.0

3.0 2.0 1.0 0.0 0.5 1.0 B/H 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 1.5 2.0

(a)

(c)
H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

H/L = 0.05 H/L = 0.15

Dm

2.0 1.0 0.0 0.5 1.0 B/H 1.5 2.0

Dm

(b)

0.5

1.0 B/H

1.5

2.0

(d)

Fig. 9. Variation of Dm with respect to B/H (Load C-2): (a) Tf/Tw = 0.5; (b) Tf/Tw = 1.0; (c) Tf/Tw = 1.5; (d) Tf/Tw = 2.0.

Figs. 11 and 12 give the variation of Dm with respect to Tf/Tw. Dm grows with the increase of Tf/Tw. Nonlinearity in the relationship between Dm and Tf/Tw is very small: the curves in Figs. 11 and 12 are almost straight. Some design codes [2,3] provide formulas for the shear lag eect on deection. Figs. 13 and 14 present Dm calculated through those formulas together with the present numerical results. While Dm due to the design formulas exhibits the same tendencies as those observed in the present nite element analysis, discrepancy is evident: Dm due to the design formulas considerably underestimates Dm by the nite element analysis except where H/L is small.

The dierence between the Dm values due to the formulas in the two design codes is also evident: Dm due to Japan [3] is consistently larger than that due to British [2]. The existing design formulas for the evaluation of deection are not satisfactory.

5. Empirical formulas Based on the present numerical results, a regression analysis is conducted. And the following empirical formulas for directly evaluating Dm are obtained:

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

4.0 3.0

15

6.0
H/L = 0.025 H/L = 0.10 H/L = 0.20 H/L = 0.05 H/L = 0.15

5.0 4.0

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

Dm

Dm

2.0 1.0 0.0

3.0 2.0 1.0 0.0

(a)
5.0 4.0 3.0

0.5

1.0 B/H

1.5

2.0

0.5

1.0 B/H

1.5

2.0

(c)
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0
H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

H/L = 0.05 H/L = 0.15

Dm

2.0 1.0 0.0 0.5 1.0 B/H 1.5 2.0

Dm

0.5

1.0 B/H

1.5

2.0

(b)

(d)

Fig. 10. Variation of Dm with respect to B/H (Load D-2): (a) Tf/Tw = 0.5; (b) Tf/Tw = 1.0; (c) Tf/Tw = 1.5; (d) Tf/Tw = 2.0.

3.0 2.5 2.0

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

6.0 5.0 4.0

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

Dm

Dm

1.5 1.0 0.5 0.0 0.5 1.0 T f /T w 4.0 3.0 1.5 2.0

3.0 2.0 1.0 0.0 0.5 1.0 T f /T w 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.5 1.5 2.0

(a)

(c)
H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

H/L = 0.05 H/L = 0.15

Dm

2.0 1.0 0.0 0.5 1.0 T f /T w 1.5 2.0

Dm

1.0 T f /T w

1.5

2.0

(b)

(d)

Fig. 11. Variation of Dm with respect to Tf/Tw (Load C-2): (a) B/H = 0.5; (b) B/H = 1.0; (c) B/H = 1.5; (d) B/H = 2.0.

Concentrated load: Dm a1 B=H where a1 = 4.12(H/L)1.77, a2 = 83.53(H/L)4.5, a3 = 41.43(H/L)2.33, a4 = 76.53(H/L)3 18.09(H/L)2 + 1.54(H/L) + 0.97. Distributed load: Dm a1 B=H 1:6 a2 T f =T w a3 B=H T f =T w a4 ; 2
2:3

where 1 a1 = 7.78(H/L)1.8, a2 = 579.02(H/L)4 + 204.73(H/L)3 21.89(H/L)2 + 0.97(H/L) 0.01, a3 = 50.46(H/L)2.48, a4 = 39.6(H/L)3 8.77(H/L)2 + 0.25(H/L) + 1. It is noted that the above formulas are applicable for 0.025 6 H/L 6 0.2, 0.5 6 B/H 6 2.0 and 0.5 6 Tf/Tw 6 2.0. Fig. 15 illustrates some comparison between Dm due to the empirical formula and the present nite element analysis (FEA). Good agreement can be recognized in these

a2 T f =T w a3 B=H T f =T w a4 ;

16
3.0 2.5 2.0

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

6.0 5.0 4.0

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

Dm

Dm

1.5 1.0 0.5 0.0 0.5 1.0 T f /T w 4.0 3.0 1.5 2.0

3.0 2.0 1.0 0.0 0.5 1.0

T f /T w

1.5

2.0

(a)
H/L = 0.025 H/L = 0.10 H/L = 0.20

(c)
8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.5

H/L = 0.05 H/L = 0.15

H/L = 0.025 H/L = 0.10 H/L = 0.20

H/L = 0.05 H/L = 0.15

Dm

1.0 0.0 0.5 1.0 T f /T w 1.5 2.0

Dm

2.0

1.0 T f /T w

1.5

2.0

(b)

(d)

Fig. 12. Variation of Dm with respect to Tf/Tw (Load D-2): (a) B/H = 0.5; (b) B/H = 1.0; (c) B/H = 1.5; (d) B/H = 2.0.

2.0 1.5

Present British (1982) Japan (2002)

2.5 2.0

Dm

Dm

1.0 0.5 0.0 0.025

1.5 1.0 0.5

Present British (1982) Japan (2002)

0.050

0.075

0.100

0.125

0.150

0.175

0.200

0.0 0.025

0.050

0.075

0.100

0.125

0.150

0.175

0.200

(a)
5.0 4.0

H/L

(c)
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.025

H/L

Dm

3.0 2.0 1.0 0.0 0.025

Present British (1982) Japan (2002)

Present British (1982) Japan (2002)

Dm

0.050

0.075

0.100

0.125

0.150

0.175

0.200

0.050

0.075

0.100

0.125

0.150

0.175

0.200

(b)

H/L

(d)

H/L

Fig. 13. Comparison of Dm (Load C-2): (a) B/H = 0.5, Tf/Tw = 1.0; (b) B/H = 2.0, Tf/Tw = 1.0; (c) B/H = 0.5, Tf/Tw = 2.0; (d) B/H = 2.0, Tf/Tw = 2.0.

gures. To quantify the accuracy, the percentage error of Dm calculated from the proposed formula is dened as Dm Emp Dm FEA e 100 %; 3 Dm FEA where Dm Emp and Dm FEA are the Dm values obtained from the proposed empirical formula and the present nite element analysis, respectively. The overall accuracy of the proposed formula for each loading condition is then calculated as the mean square error by the following equation [1]: v u N u1 X 2 Dm t 4 ei ; e N i1

where ei is the error computed by Eq. (3) for a present nite element result and N is the number of the present nite element results for a loading condition. Since in the present study, the combination of the parameters has required 80 box girders to be analyzed for each loading, N in Eq. (4) is equal to 80. Using Eq. (4), the mean square error is found to be 1.06% and 1.58% for concentrated and distributed load, respectively. 6. Concluding remarks Finite element analyses were conducted extensively for various simply supported box girders to investigate the mid-span deection. The whole girder is modeled by shell

C. Lertsima et al. / Computers and Structures 84 (2005) 1118

2.0 1.5 2.5 2.0

17

Present FEA British (1982) Japan (2002)

Dm

1.0 0.5 0.0 0.025

Dm

1.5 1.0 0.5

Present British (1982) Japan (2002)

0.050

0.075

0.100

0.125

0.150

0.175

0.200

0.0 0.025

0.050

0.075

0.100

0.125

0.150

0.175

0.200

(a)
5.0 4.0
Present British (1982) Japan (2002)

H/L

(c)
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.025
Present British (1982) Japan (2002)

H/L

Dm

3.0 2.0 1.0 0.0 0.025

0.050

0.075

0.100

0.125

0.150

0.175

0.200

Dm

0.050

0.075

0.100

0.125

0.150

0.175

0.200

(b)

H/L

(d)

H/L

Fig. 14. Comparison of Dm (Load D-2): (a) B/H = 0.5, Tf/Tw = 1.0; (b) B/H = 2.0, Tf/Tw = 1.0; (c) B/H = 0.5, Tf/Tw = 2.0; (d) B/H = 2.0, Tf/Tw = 2.0.

2.6 Present FEA: Tf/Tw=0.5 Present FEA:Tf/Tw=0.5 Eq. (1): Tf/Tw=0.5 Eq. (1):Tf/Tw=0.5 2.4 Present FEA: Tf/Tw=1.0 Present FEA: Tf/Tw=1.0 2.2 Eq. (1): Tf/Tw=1.0 Eq. (1): Tf/Tw=1.0 Present FEA: Tf/Tw=1.5 Present FEA: Tf/Tw=1.5 2.0 Eq. (1): Tf/Tw=1.5 Eq. (1): Tf/Tw=1.5 1.8 Present FEA: Tf/Tw=2.0 Present FEA: Tf/Tw=2.0 Eq. (1): Tf/Tw=2.0 Eq. (1): Tf/Tw=2.0 1.6 1.4 1.2 1.0 0.8 0.025 0.050 0.075 0.100 0.125 0.150 0.175 0.200 (a) H/L

2.6 Present FEA: Tf/Tw=0.5 Present FEA: Tf/Tw=0.5 Eq. (2): Tf/Tw=0.5 Eq. (2): Tf/Tw=0.5 2.4 Present FEA: Tf/Tw=1.0 Present FEA: Tf/Tw=1.0 2.2 Eq. (2): Tf/Tw=1.0 Eq. (2): Tf/Tw=1.0 Present FEA: Tf/Tw=1.5 Present FEA: Tf/Tw=1.5 2.0 Eq. (2): Tf/Tw=1.5 Eq. (2): Tf/Tw=1.5 1.8 Present FEA: Tf/Tw=2.0 Present FEA: Tf/Tw=2.0 Eq. (2): Tf/Tw=2.0 1.6 Eq. (2): Tf/Tw=2.0 1.4 1.2 1.0 0.8 0.025 0.050 0.075 0.100 0.125 0.150 0.175 0.200
H/L

Dm

(c)

Present FEA: Tf/Tw=0.5 8.0 Present FEA: Tf/Tw=0.5 Eq. (1): Tf/Tw=0.5 Eq. (1): Tf/Tw=0.5 7.0 Present FEA: Tf/Tw=1.0 Present FEA: Tf/Tw=1.0 Eq. (1): Tf/Tw=1.0 Eq. (1): Tf/Tw=1.0 6.0 Present FEA: Tf/Tw=1.5 Present FEA: Tf/Tw=1.5 5.0 Eq. (1): Tf/Tw=1.5 Eq. (1): Tf/Tw=1.5 Present FEA: Tf/Tw=2.0 Present FEA: Tf/Tw=2.0 4.0 Eq. (1): Tf/Tw=2.0 Eq. (1): Tf/Tw=2.0 3.0 2.0 1.0 0.0 0.025 0.050 0.075 0.100 0.125 0.150 0.175 0.200 (b) H/L

Fig. 15. Dm due to proposed formulas and nite analysis: (a) Load C-2 with B/H = 0.5; (b) Load C-2 with B/H = 2.0; (c) Load D-2 with B/H = 0.5; (d) Load D-2 with B/H = 2.0.

elements and the accuracy is ensured by the multimesh extrapolation. The present study has then revealed that the deection magnication factor Dm increases with the increase of each of H/L, B/H and Tf/Tw. To be interesting, while the dependence of Dm on H/L has been found nonlinear, the relationship between Dm and Tf/Tw is almost linear. It has been also shown that the existing formulas

Dm

Dm

Dm

Present FEA: Tf/Tw=0.5 Present FEA: Tf/Tw=0.5 8.0 Eq. (2): Tf/Tw=0.5 Eq. (2): Tf/Tw=0.5 7.0 Present FEA: Tf/Tw=1.0 Present FEA: Tf/Tw=1.0 Eq. (2): Tf/Tw=1.0 Eq. (2): Tf/Tw=1.0 6.0 Present FEA: Tf/Tw=1.5 Present FEA: Tf/Tw=1.5 Eq. (2): Tf/Tw=1.5 Eq. (2): Tf/Tw=1.5 5.0 Present FEA: Tf/Tw=2.0 Present FEA: Tf/Tw=2.0 Eq. (2): Tf/Tw=2.0 4.0 Eq. (2): Tf/Tw=2.0 3.0 2.0 1.0 0.0 0.025 0.050 0.075 0.100 0.125 0.150 0.175 0.200 (d) H/L

adopted in the design codes tend to underestimate Dm considerably. Based on the present numerical study by the nite element method, empirical formulas have been proposed to compute Dm. These formulas yield Dm in very good agreement with the values obtained by the nite element analysis.

18

C. Lertsima et al. / Computers and Structures 84 (2005) 1118 [5] Sedlacek G, Bild S. A simplied method for the determination of the eective width due to shear lag eects. J Constr Steel Res 1993;24:15582. [6] Lee CK, Wu GJ. Shear lag analysis by the adaptive nite element method. 1: Analysis of simple plated structures. Thin Wall Struct 2000;38:285309. [7] Tahan N, Pavlovic MN, Kotsovos MD. Shear-lag revisited: The use of single Fourier series for determining the eective breadth in plated structures. Comput Struct 1997;63(4):75967. [8] Reissner E. Least work solutions of shear lag problems. J Aeronaut Sci 1941;8:28491. [9] Winter G. Stress distribution in and equivalent width of anges of wide, thin-wall steel beams. NACA Technical Note No. 784, 1940. [10] Reissner E. Analysis of shear lag in box beam by the principle of minimum potential energy. Q Appl Math 1946;6(3):26878. [11] Malcolm DJ, Redwood RG. Shear lag in stiened box girders. J Struct DivASCE 1970;96:140319. [12] Kristek V, Evans HR. A hand calculation of the shear lag eect in unstiened anges and in anges with closely spaced stieners. Civil Eng Pract Des Eng 1985;4:16390. [13] Song QG, Scordelis AC. Formulas for shear-lag eect of T-, I-, and box beams. J Struct EngASCE 1990;116(5):130618. [14] Lertsima C, Chaisomphob T, Yamaguchi E. Stress concentration due to shear lag in simply supported box girders. Eng Struct 2004;26(8): 1093101. [15] Cook RD, Malkus DS, Plesha ME. Concepts and applications of nite element analysis. 3rd ed. New York: John Wiley and Sons; 1989. [16] MARC Analysis Research Corporation. MARC manuals Vol A-D, Rev K.6., 1994.

Acknowledgements The present research was partially supported by the Thailand Research Fund through the Royal Golden Jubilee Ph.D. Program (Grant no. PHD/0055/2543) and by the Ministry of Education, Science, Sports and Culture of Japan through Grant-in-Aid for Scientic Research (B) (no. 15360244). It is also the outgrowth of the academic agreement between Sirindhorn International Institute of Technology, Thammasat University and Faculty of Engineering, Kyushu Institute of Technology, and has been partially supported by the two academic bodies. These supports are gratefully acknowledged. References
[1] Tenchev RT. Shear lag in orthotropic beam anges and plates with stieners. Int J Solids Struct 1996;33(9):131734. [2] British Standards Institution. Steel, concrete and composite bridges, BS 5400, Part 3 Code of practice for design of composite bridges. London: England, 1982. [3] Japan Road Association. Design specications for highway bridges: Part II. Steel bridges. Tokyo: Japan, 2002. [4] Moatt KR, Dowling PJ. Shear lag in steel box girder bridges. Struct Eng 1975;53(10):43948.