Elastic Web Buckling Stress and Ultimate Strength of H-Section Beams Dominated by Web Buckling

Numerical analyses and theoretic analyses are presented to study the elastic buckling of H-section beam web under combined bending and shear force. Results show that the buckling stress of a single web with clamped edges gives a good agreement with the buckling stress of an H-section beam web when the local buckling of the beam is dominated by the web buckling. Based on theoretic analyses, a parametric study is conducted to simplify the calculation of buckling coeﬃcients. The parameters involved are clariﬁed ﬁrst, and the improved equations for the buckling coeﬃcient and buckling stress are suggested. By applying the proposed method, the web buckling slenderness ratio is deﬁned. It is veriﬁed that the web buckling slenderness ratio has a strong correlation with the normalized ultimate strength of H-section beams when the buckling of the beams is dominated by web buckling. Finally, a design equation is proposed for the ultimate strength of H-section beams.


Introduction
e main girder in a steel frame bears large bending moment and shear force due to the horizontal load and the secondary beam load as shown in Figure 1; the global buckling and local buckling may occur at the end of the main girder. Current stability design methods of the H-section girder are mainly based on the calculation of elastic buckling stress. e elastic global buckling stress mainly depends on the interval of lateral supports and the width of flange, while the elastic local buckling stress mainly depends on the width-thickness ratio and depth-thickness ratio. However, the load condition and boundary condition may have a large influence on the buckling stress as well.
e theoretic studies on elastic local buckling have a long history: the traditional approach is to study the elastic buckling of a rectangular flat plate under assumed stress conditions and with various boundary conditions by using the energy method [1][2][3][4][5]. Yuan and Jin [6] proposed an extended Knatorovich method to solve the buckling problem of flat plates with various boundary conditions under compression and pure shear force and derived high-accuracy buckling coefficients. Kang and Leissa [7] formulated an exact solution procedure for the buckling analysis of flat plates with various boundary conditions under combined compression and bending force. Jana and Bhaskar [8] carried out buckling analyses of flat plates under nonuniform uniaxial compression by using Galerkin's method. To analyze the buckling stress of a web plate under complex tress conditions, Ritz's energy method using Fourier series functions is commonly adopted. Ikarashi and Suzuki [9] conducted Ritz's energy method to analyze the buckling stress of web plates with simply supported and clamped edges under combined bending and shear force, discovering that the results rapidly converged to the true solution as the number of Fourier series terms increased. Liu and Pavlovic [10,11] conducted Ritz's energy method to analyze the buckling stress of simply supported flat plates under patch compression and arbitrary loads, founding that the idealization of simple supports yielded sound agreements to many plates problems, when a plate was attached to other plates.
As above, efforts have been carried out on the elastic buckling of the rectangular flat plates. However, precise solutions for the buckling stress can only be obtained under simply assumed stress conditions, and it is still difficult to calculate the precise buckling stress of plates under complex stress condition, especially for the web of an H-section beam under combined bending and shear force as shown in Figure 1. To calculate the buckling stress of plates under complex stress condition, the approximate method is often used. Based on the parametric study, Suzuki and Ikarashi [12] proposed a series of approximate equations to calculate the buckling stress of web under combined compression, bending, and shear force. e equations were found to be too complex to be applied for practical uses, due to the number of parameters involved. It is highly necessary to develop a design formula with high accuracies and simple calculations.
It is assumed that the local stability and the load bearing capacity of an H-section beam are mainly dependent on the width-thickness ratio of the flange and depth-thickness ratio of the web. Kadono et al. [13] proposed the equivalent widththickness ratio which can be regarded as a major parameter to approximate the load bearing capacity of an H-section beam. Kimura [14] proposed design equations for the ultimate strength and plastic deformation capacity by using the equivalent width-thickness ratio, which were useful when the buckling of the beam was dominated by flange buckling. However, this method has not been proved applicable for the web buckling dominant H-section beams.
To evaluate the ultimate strength of web buckling dominant H-section beams, the high-accuracy equation for the elastic buckling stress of web is required. is study aims to solve the eigenvalue problem for an H-section beam web under combined bending and shear force by using Ritz's energy method and find out involved parameters. A parametric study is conducted to reveal the effect of parameters on the elastic buckling stress and to propose approximate equations for practical use. Based on the test results, the direct strength method is attempted to derive design equation for the ultimate strength of web buckling dominant H-section beams by using the proposed equations of elastic buckling stress.

Finite Element Analysis.
e loading condition of the part of an H-section beam (main girder) between the column and the secondary beam as shown in Figure 1 can be approximately regarded as Figure 2. e internal moment on the beam is assumed to be linearly varied along the length direction axis (x-axis), which can be expressed as where Mb is the left end bending moment, β is moment gradient (0 ≤ β ≤ 2), and the right end bending moment is (1 − β)Mb. e relationship between Mb and shear load Q s is expressed as In this study, numerical simulations are conducted using the finite element program ABAQUS to analyze the H-section beam. e boundary condition and load condition of an H-section beam between the column and the secondary beam ( Figure 1) are assumed as shown in Figure 3. e left beam end is connected with the column, and the right end with a stiffener is connected with the secondary beam. For simplification, a cantilever beam model using shell elements is used in FEA ( Figure 3). e left end is completely fixed, the edges of web and flange on the right end are set as rigid edges to form a rigid plane, and the displacement along z-direction of the rigid plane is also constrained. On such boundary condition, the shear load Q s and moment (1 − β)Mb are acting on the center point of the rigid plane in which Q s and M b satisfy equation (2). e FEA includes linear elastic buckling analysis and large deformation analysis. e elastic buckling analysis is performed first, and then the distribution of the initial geometric imperfection is assumed as the first elastic buckling mode obtained from the elastic buckling analysis. e load condition of a beam end ( Figure 1) is close to a cantilever beam with β � 1. For this reason, numerous experimental studies on the H-section cantilever beams have been reported [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30], which have important reference significance for the study of local buckling. e shear span L may be close to the interval of the secondary beam (Figure 1), and the normal value of L/D should be approximately 4-6. According to the Code for Design of Steel Structure [31], for normal strength steel (Q235), L/B should be no more than 16 to ensure the elastic global stability. In this paper, 158 sets of reported experimental data [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] of H-section cantilever beams (with β � 1) are collected as shown in Table 1. According to the data, it can be confirmed that, in most cases, 4L/D ≤ 6 and L/B ≤ 16. For comparison, two test results [30] and FEA results with normal dimensions are shown in Figure 4 (L/D � 1400/350 � 4, L/B � 1400/ 175 � 8, t w � 4.5, and t f � 16) and Figure 5 (L/D � 1400/ 350 � 4, L/B � 1400/175 � 8, t w � 6, and t f � 9). e load versus deflection response of the beam is generally affected by the initial geometric imperfection (IMP) and residual stress. As shown in Figures 4(a)-4(c) and Figures 5(a)-5(c), for both web buckling dominant and flange buckling dominant beams, the FEA results agree well with the test results, by employing proper initial geometric imperfections (with IMP ≈ D/400). However, the actual distribution of the initial geometric imperfections and the residual stresses involved in the beams is unclear in the reported paper  [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30], and it is difficult to exactly analyze the load versus deflection response of the beams. In addition, the load versus deflection response is slightly affected by the loading program (including monotonic loading and various kinds of cyclic loading) as indicated by Kimura [14].
us, it is difficult to include these effects in the evaluation method to estimate the ultimate strength or the plastic deformation capacity of a beam. To estimate the ultimate strength of a beam, a simple calculation method is preferable for practical use.
Although the initial geometric imperfection, residual stress, loading program, and so on may cause a large deviation in the test, the relatively conservative evaluation method should be produced. In the following research, the direct strength method is conducted to investigate the correlation between the buckling slenderness ratio and the ultimate strength based on the test results [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] subjected to the web buckling dominant beam. e primary study is to propose high-accuracy formulas to calculate the elastic buckling stress.

eoretic Analysis.
To analyze the elastic buckling stress of a web, a simplified model conducted by the theoretical energy method is used here, in which the H-section beam web is regarded as a single web with the boundary condition shown in Figure 6. e end edges AB and CD are set as rigid edges and constrained by a pin and a roller, respectively. e out-of-plane displacement of the other two longer edges AC and BD is constrained to keep straight, and the rotation around the x-axis is constrained as well. As shown in Figure 7, the bending stress σ(x,y) in the web can be expressed as where σb is the maximum value of bending normal stress. M b can be expressed as where A w � dt w is the section area of web and A f � 2bt f is the section area of flange. e shear stress is assumed to be uniformly distributed, the shear force Q s is expressed as where τs is the web shear stress. According to the above expression, the ratio of τs to σb can be approximately expressed as where λw � L/d is the aspect ratio of web. e total potential energy of the web under combined bending and shear force is where U is the strain energy and V b and V s are the external work due to bending and shear force, respectively. U, V b , and V s can be expressed as follows: where W is the out-of-plane displacement function and D w is the flexural rigidity:     Mathematical Problems in Engineering 5

Mathematical Problems in Engineering
where E is Young's modulus and υ is the Poisson ratio (υ � 0.3). σ crw and τ crw in equations (9) and (10) are the critical value of σb and τs, which are defined as where k bw and k sw denote the buckling coefficients due to σ crw , τ crw , and k bw /k sw � α (equation (6)). e out-of-plane displacement W is expressed by a double Fourier series function as follows: where e mn is the series coefficient and f m (x) and g n (y) are the Fourier series functions. Assuming the edges of web are clamped, the functions f m (x) and g n (y) can be expressed as Here, W, f m (x), and g n (y) in above equations are replaced with ω, µ m (x), and v n (y), respectively, as follows:  Equations (8)-(10) can be written as the following equivalent equations: According to the stationary value theory of total potential energy, the conditional expression for the critical state of local stability is To solve the buckling coefficients k bw and k sw in equation (20), a generalized eigenvalue analysis is required. e involved parameters to be input are the aspect ratio λ w , moment gradient β, and stress ratio α � k sw /k bw (equation (6)). e theoretic analysis using Ritz's energy method is presented above. e energy method using Fourier series functions showed sound convergences, small amount of computations, and high accuracies in the previous studies [9]. Moreover, the energy method has been validated using the finite element analysis (FEA), which showed that the buckling coefficients of a single web with clamped edges gave good agreements with those of an H-section beam web when the rotations of flanges were clamped [12]. However, relevant studies are rather limited towards the condition when flanges are not clamped.

Elastic Buckling of an H-Section Beam and a Single Web.
When the out-of-plane displacement of flanges is constrained, the flange buckling will be prevented to allow web buckling to occur only. For a normal H-section beam, the lower flange (in compression) is not constrained, and the web buckling and flange buckling may occur simultaneously.  (1 -β)σ b Figure 7: Stress distribution of web. condition (using equation (6)) are analyzed as well by the theoretical energy method presented above. It is shown that the shear buckling coefficient k sw of H-section beams with clamped flanges (Case 1) obtained by FEA is close to the coefficient k sw of single webs obtained by theoretical energy method, which corresponds to the previous research [12]. When the flanges are not clamped (Case 2: free flanges), the coefficient k sw of H-section beams with t f > 10 mm obtained by FEA is close to the coefficient k sw of single webs as well. However, when t f < 10 mm, the k sw of H-section beams are smaller than k sw of single webs due to the effect of flange buckling.
For comparison, three typical buckling modes of H-section beams obtained by FEA with free flanges are given in Figure 9, and three typical buckling modes of single webs obtained by theoretical analysis are given in Figure 10. e theoretical buckling mode is expressed by equation (14), where e m,n (m � 1, 2, . . ., 20; n � 1, 2, . . ., 10) is the first-order eigenvector. By applying the Wolfram Mathematica program, the web buckling modes expressed by equation (14) can be drawn as shown in Figure 10.
As shown in Figure 9(a), the buckling mode of an H-section beam (with t f � 12 mm and free flanges) is controlled by web buckling, in which the web buckling mode is basically the same as the buckling mode (Figure 10(a)) of a single web with identical shape and stress condition (λ w �2400/400 � 6, β � 1, and α � 0.4028), and the shear buckling coefficients are close to each other (8.64 ≈ 8.84).
As shown in Figure 9(b), even with a slightly buckled flange, the local buckling of an H-section beam (with t f � 10 mm and free flanges) is dominated by web buckling, and the buckling mode of the beam web is similar to the buckling mode (Figure 10(b)) of a single web with identical shape and stress condition (λ w �6, β � 1, and α � 0.3403), and the shear buckling coefficients are close to each other (8.22 ≈ 8.67), which corroborate that the edge condition of a web can be regarded as clamped condition when the local buckling of an H-section beam is dominated by web buckling.
However, when the local buckling of an H-section beam (with t f � 6 mm and free flanges) is dominated by flange buckling, the buckling mode (Figure 9(c)) of the beam web is different from a single web buckling mode (Figure 10(c)) with identical shape and stress condition (λ w � 6, β � 1, and α � 0.2153).
rough comparisons, it is concluded that the buckling of an H-section beam with free flanges is dominated by web buckling when the flange buckling does not occur or slightly occur, and the boundary condition of the longer web edges is close to clamped condition. erefore, to calculate the elastic local buckling stress of an H-section beam dominated by web buckling, the analytic model can be simplified as a single web with the boundary condition as shown in Figure 6 and with the stress condition as shown in Figure 7. e presented theoretical analytic method with high accuracy and small amount of computations is valuable for studying the buckling stress of an H-section beam web.

Previous Study. Based on theoretic analyses, Suzuki and
Ikarashi [12] proposed approximate equations for the buckling coefficients of web with clamped edges as follows: where As indicated by Suzuki and Ikarashi [12], the proposed equations are only applicable when 1 < ß ≤ 2, and the equations are too complex to be applied for practical uses. ese defects should be annihilated. In this study, a new method based on parametric studies is proposed to simplify the calculation method and to expand the application range. e first study is to formulate the equation under simple stress condition such as pure shear, uniform bending, and unequal bending. e interaction between the buckling coefficients of shear and bending is then studied to suggest an approximate formula for the calculation of buckling stress.

Buckling Coefficient of Web under Pure Shear Force.
Let k bw � 0 (without considering the effect of bending stress); the critical conditional expression equation (20) can be written as follows: According to equations (17), (19), and (24), the shear buckling coefficient k sw0 is related to aspect ratio λ w only. Result ( Figure 11) shows that k sw0 converges to 8.98 in the case of the infinitely long web, which agrees well with previous research [2]. For the finite length web, when λ w is larger than 1, the analyzed result corresponds to the approximate equation suggested by Moheit [3] as follows:

Buckling Coefficient of Web under Unequal Bending
Moment. Let k sw � 0 (without considering the effect of shear stress); the critical conditional expression equation (20) can be written as follows:  N).

(26)
According to equations (17), (18), and (26), the bending buckling coefficient k bw0 is related to λ w and β. According to Bijlaard' research [4], for an infinitely long plate under uniform bending (β � 0), the buckling coefficient k bw0 was 39.6. In this study, the analyzed results ( Figure 12) show that k bw0 converges to 39.6, which shows a good agreement with Bijlaard' research [4]. k bw0 is only slightly larger than the lower limit 39.6 when λ w ≥1. As shown in Figure 13, when β > 0, the larger value of β is, the higher value of k bw0 is. By changing the abscissa of Figure 13 into β/λ w , Figure 14 can be obtained. It is shown that the curves of k bw0 versus β/λ w with various β almost overlap with each other, and they can be approximated by the following equation:

Interaction Curve of the Buckling Coefficients under Combined Bending and Shear Force.
For an H-section beam web, the combined bending and shear stress must be considered. To investigate the interaction between bending and shear stress, k sw versus λ w curves and k bw versus λ w curves with β � 2 and various A f /A w are analyzed based on the critical conditional expression equation (20). As shown in Figure 15, k sw versus λ w curves obtained by considering the combined bending and shear stress are lower than k sw0 versus λ w curve obtained by considering the shear stress only. When λ w is small enough, all k sw versus λ w curves converge to k sw0 versus λ w curve. As shown in Figure 16, k bw versus λ w curves obtained by considering the combined bending and shear stress are lower than k bw0 versus λ w curve obtained by considering the bending stress only. When λ w is large enough, all k bw versus λ w curves converge to k bw0 versus λ w curve. e beams may present various buckling modes due to the different configurations [32,33]. According to the theoretical analyzed results in this study, the elastic local buckling modes of a single web can be roughly divided into three types (shear type, bending type, and intermediate type). ree web buckling modes with A f /A w � 1 and β � 2 and different aspect ratio (λ w � 6, 8, and 12) are shown in Figure 17; the symbols of k sw versus λ w and k bw versus λ w are shown in Figures 15 and 16, respectively. e web with λ w � 6 presents shear type buckling mode (Figure 17(a)) due to the relatively large shear stress, in which the similar shapes of buckling waves are observed along the length direction. e web with λ w � 12 presents bending type buckling mode ( Figure 17(c)) due to the relatively large bending stress, in which the buckling waves concentrate close to the web ends. e web with λ w � 8 presents intermediate bucking mode when the effects of shear and bending stress are comparable (Figure 17(b)). For all webs, the buckling coefficients k sw are always smaller than k sw0 and the buckling coefficients k bw are always smaller than k bw0 .
As above, k sw0 and k bw0 can be regarded as the upper limits of k sw and k bw , respectively. By taking the ordinate as k sw /k sw0 and taking the abscissa as k bw /k bw0 , Figures 15 and  16 can be expressed by Figure 18. It is shown that all the interaction curves (with β � 2 and various A f /A w ) overlap with each other, and they can be evaluated by an approximate equation as follows: Figure 19 shows a large number of analytical data with various λ w , β, and A f /A w in the range of 1 ≤ λ w ≤ 40, 0 ≤ ß ≤ 2, and 0.3 ≤ A f /A w ≤ 2.5; generally, results of k bw /k bw0 versus k sw /k sw0 distribute around the curve of equation (28). erefore, equation (28) can be regarded as an interaction formula to calculate k bw and k sw .

Improved Equations.
Taking k sw /k bw � a and substituting equations (25) and (27) into the interaction formula equation (28), the approximate equations for k bw and k sw are obtained as follows: where α is the stress ratio as shown in equation (6). e proposed equations (equations (29) and (30)) are far simpler than the previous ones (equation (21) and (22)). For verification and comparison, the proposed curves and FEA results with various cases of A f /A w and β are shown in Figures 20-23. It is found that equation (21) gives good agreements with the FEA results when β � 2 and β � 1, as shown in Figures 20 and 21. However, when β � 0.5 and β � 0.1, equation (21) does not agree with FEA as shown in Figures 22 and 23. As indicated previously, the equations (equations (21) and (22)) can only apply to the range of 1 ≤ β ≤ 2. is defect is overridden in this study. As shown in equation (30) gives good agreements with the FEA results for all cases. e proposed equations with high accuracies are applicable for the full range (0 ≤ ß ≤ 2), which are valuable for further studies.

(36)
By taking the ordinate as the normalized ultimate strength τ max (obtained from test data) and taking the abscissa as the equivalent width-thickness ratio (b/t f ) eq , Figure 24 can be obtained. ere are 94 test data in the range of σ crw >1.5σ crf and 64 test data in the range of σ crw ≤ 1.5σ crf , in which σ crw is calculated by using equations (12) and (29) and σ crf is calculated by equation (35). It is shown that τ max has a strong correlation with (b/t f ) eq , and it is reasonable to suppose that the local buckling is dominated by flange buckling when σ crw > 1.5σ crf . e calculation of τ max has been suggested by Kimura [14] for the flange buckling dominant H-section beams, and τ max of the beams under monotonic loads can be expressed as follows: By taking the ordinate as τ max + 0.01 L/D and the abscissa as (b/t f ) eq in the range of σ crw > 1.5σ crf , Figure 25 can be obtained, in which equation (37) gives a good agreement with test results. Moreover, the distribution of τ max obtained by cyclic tests is slightly higher than that obtained by monotonic tests as indicated by Kimura [14]. However, the correlation between τ max and (b/t f ) eq is not strong when σ crw ≤ 1.5σ crf , due to the effect of web buckling as shown in Figure 24. As demonstrated in equation (36), (b/t f ) eq contains the width-thickness ratio of flange and the depththickness ratio of web only. us, it is insufficient to regard (b/t f ) eq as the major parameter to evaluate the ultimate strength when the buckling is dominated by web buckling.

Web Buckling Slenderness Ratio and New Design Equation.
To evaluate the ultimate strength of a web buckling dominant H-section beam, not only the depththickness ratio but also other parameters such as aspect ratio, bending gradient, and section areas should be considered. To avoid complex calculations, a direct strength method based on the calculation of elastic buckling stress is employed to investigate the relationship between the normalized ultimate strength τ max and web buckling slenderness ratio S w , in which S w is defined as follows: where τ crw is the shear buckling stress which is calculated by using equations (13) and (30); M crw is the bending buckling moment which is calculated by using equations (12), (29), and (39): By changing the abscissa of Figure 24 into S w , Figure 26 can be obtained. By comparison, the data dispersion in Figure 26 is smaller than that in Figure 24 when σ crw ≤ 1.5σ crf , and it is reasonable to suppose that the local buckling is dominated by web buckling when σ crw ≤ 1.5σ crf . erefore, the new defined web buckling slenderness ratio (S w ) can be regarded as a major parameter to evaluate the normalized ultimate strength (τ max ) of a web buckling dominant H-section beam. However, the test results are lower than the Euler curve equation (40), due to the inelastic buckling: τ e is the normalized elastic buckling strength. e following asymptotic equation (41) is attempted to evaluate τ max , in which τ max converges to equation (40) with a large value of S w , and τ max converges to 1 with a small value of S w : (41) By substituting equation (40) into equation (41), (41) can be written as (42) As shown in Figure 26, equation (42) takes the lower limit in the range of σ crw ≤ 1.5σ crf . However, it has been indicated [14] that τ max may be larger than 1 for an inelastic buckling H-section beam. To avoid underestimations, equation (43) is attempted: (43) Figure 27 shows the relationship between τ max and S w in the range of σ crw ≤ 1.5σ crf only. By taking the maximum value of equations (42) and (43), (44) is obtained: It is shown that proposed equation (44) takes the lower limit of test results, and the upper limit is about its 125%.
is means that the deviation caused by initial geometric imperfection and residual stress is lower than 25% in the tests. e proposed design equation (44) produces good predictions for the test results of the ultimate strengths of the web buckling dominant H-section beams, and the application range is σ crw ≤ 1.5σ crf .
Moreover, according to the test data (Figure 27), the normalized ultimate strength τ max is not affected by the loading program (including monotonic loading and various kinds of cyclic loading) when the buckling is dominated by web buckling. For both monotonic tests and cyclic tests, the dispersion is small.

Equation 37
Monotonic test (58 data) Cyclic test (36 data) e deviation in Figure 27 is about 25%, meaning that the beams with the same buckling slenderness ratio may have different normalized ultimate strengths with 25% deviation. In Section 2.1 (Figures 4 and 5), the FEA results have shown that the influence of the geometric imperfections with D/800 ∼ D/200 only causes approximately 5% deviation, which is far smaller than 25%. In addition, other influences such as residual stresses, material characteristics, and the test methods may also cause deviations to a certain degree. However, these could not be the primary reason for the large deviation. As mentioned previously, the direct strength method is used in this study, which calculates the elastic buckling stress and buckling slenderness ratio to predict the normalized ultimate strength.
e beams with the same buckling slenderness ratio do not necessarily mean they have the same normalized ultimate strength.
e beams with different configurations may have different buckling behaviors and different normalized ultimate strengths, even though they have the same value of buckling slenderness ratio. erefore, to improve the evaluation method of the normalized ultimate strength, further parameters and their influences should be studied to reduce the deviation.

Summary and Conclusions
eoretic analysis by Ritz's energy method for the H-section beam under combined bending and shear force is presented. e theoretic analysis was verified against the FEA when the buckling of the beam is dominated by web buckling. A parametric study based on the stress separation concept is conducted to simplify the calculation method for buckling coefficients. e design equation based on direct strength method for the normalized ultimate strength of a web buckling dominant H-section beam is proposed. e conclusions are drawn as follows: (1) Even when the flange is slightly buckled, the buckling mode and buckling stress of an H-section beam web show no obvious difference with a single web with clamped edges, when the local buckling of the beam is dominated by web buckling. e analytic model of a web buckling dominant H-section beam can be simplified by a single web with clamped edges. (2) Without considering the effect of shear stress, the bending buckling coefficient k bw0 of web under unequal bending stress is related to the aspect ratio λ w and moment gradient β, and when λ w > 1, k bw0 can be approximately expressed by equation (27). e shear buckling coefficients k sw0 (equation (25)) and bending buckling coefficient k bw0 (equation (27)) can be regarded as the upper limits of the buckling coefficients k sw and k bw of web under combined bending and shear stress, respectively. e interaction curve of the buckling coefficients can be expressed by equation (28), and the approximate equations for calculating the buckling coefficients (equations (29) and (30)) are proposed.
(3) According to a number of tests, it is shown that the normalized ultimate strength τ max has a strong correlation with the equivalent width-thickness ratio (b/t f ) eq in the range of σ crw > 1.5σ crf , whereas τ max has a strong correlation with the web buckling slenderness ratio S w in the range of σ crw ≤ 1.5σ crf . It is reasonable to assume that the local buckling is dominated by the flange buckling when σ crw > 1.5σ crf , whereas it is dominated by the web buckling when σ crw ≤ 1.5σ crf . (4) e distribution of the normalized ultimate strength τ max obtained by cyclic tests is slightly higher than that obtained by monotonic tests, when the local buckling is dominated by flange buckling. However, τ max is not affected by the loading program when the buckling is dominated by web buckling.