Professional Documents
Culture Documents
To cite this article: Piseth Doung & Eiichi Sasaki (2020): Simplification of least-work
solution to an empirical method for prediction of shear lag stress in steel box moment
connections with internal diaphragms, Mechanics of Advanced Materials and Structures, DOI:
10.1080/15376494.2019.1710004
ORIGINAL ARTICLE
CONTACT Piseth Doung doung.p.aa@m.titech.ac.jp Department of Civil and Environmental Engineering, Tokyo Institute of Technology, Tokyo, Japan.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/umcm.
ß 2020 Taylor & Francis Group, LLC
2 P. DOUNG AND E. SASAKI
u00 1 k21 u1 p21 u2 ¼ u01 (7a) where rb is the bending stress calculated using the beam
theory (rb ¼ Mh/I).
u00 2 k22 u2 p22 u1 ¼ u02 (7b)
where k1, p1, u01, k2, p2, and u02 are the parameters represented
2.2. Application for cantilever beams
the geometry, material and load properties. To solve the sys-
tem of two second order differential equations above, we have In this study, a cantilever beam subjected to a point and
to rewrite this system as a system of four first order differential uniformly distributed loads was considered for the investiga-
equations by letting, u0 1 ¼ u3 and u0 2 ¼ u4 : Hence, we tion. The application aimed to predict the shear lag stress of
obtained the system of four first differential equations as: the rectangular box beam with the web depth-to-flange
u0 1 u3 ¼ 0 (8a) width ratio varied. Three series 200, 300, and 400 contained
the section properties as summarized in Table 1 below. All
0
u 2 u4 ¼ 0 (8b) series were subjected to a load which was equivalent to a
stress of 100 MPa at the support. By using the procedure in
u0 3 k21 u1 p21 u2 ¼ u01 (8c) Eqs. (9)–(11), the shear lag stress results are determined and
u0 4 k22 u2 p22 u1 ¼ u02 (8d) are shown in Table 2 below. As can be seen, additional
stress due to shear lag increased significantly in accordance
Eqs. (8a)–(8d) can also be written in forms of a matrix with the greater web depth for both the beam under a point
method as follows. and uniformly distributed loads. Moreover, it was observed
0 10 1 0 1 0 0 1
0 0 1 0 u1 0 u1 that with the same applied bending stress (100 MPa), the
B 0 0 0 1 CB u2 C B 0 C B u0 2 C cantilever beam under uniformly distributed load produced
B 2 CB C B C B C
@ k p2 0 0 A@ u3 A þ @ u01 A ¼ @ u0 3 A (9) twice shear lag stress greater than the cantilever beam under
1 1
p22 k22 0 0 u4 u02 u0 4 a point load.
The stress distribution across the beam flange and web cor-
responding to the additional bending moments due to shear 2.3. Simplification of stress distribution
lag can be summarized as:
" Calculations of shear lag stress following to Eqs. (10)–(11)
#
yi i z z3 2 requires solving a complicated mathematical problem with
0
rs ¼ Eh 1 i Is u 1 þ Iw u0 2 multiple steps before reaching the final solution. The
b iþ1 h h3 5h
procedure is also time consuming. A simplified method
(10) allows us to modify the stress given by assuming that shear
The total peak stress at the corner of the beam section lag occurs only in the flange alone to represent the shear lag
(included bending stress) can be calculated by giving y ¼ b in both web and flange of the beam. The modification factor
and z ¼ h, and expressed as: (bs) represents the effect of shear lag in the web obtained by
a compatible solution in a parametric study between the
i 2
rt ¼ rb 6Eh Is u0 1 þ Iw u0 2 (11) peak stresses in Eqs. (11) and (12) in relations with the ratio
iþ1 5h d/2b (b ¼ haft width of the beam flange).
4 P. DOUNG AND E. SASAKI
i Is P
rt ¼ rb þ bs Eh An (12)
iþ1I k
2iþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Figure 3. Column flange flexibility coefficient (af).
ðiþ1Þð2iþ1Þ G
where, An ¼ 2i n ¼ n
EI ,
1
1ð2iþ2Þ I
2iþ1 Is , and k ¼ 1
b 2ð2i1Þ E n:
The extensive study was carried out for the cantilever beam 3.1. Shear lag stress due to column flange flexibility
under uniformly distributed load. The concept of simplifica-
tion is to modify the stress receiving from the assumption The non-uniform forces transferred from the beam flanges
apply to the column flange and cause an initial displacement
that shear lag occurs in the flange alone. For a cantilever
that can lead to an increase in peak stress and decrease in
beam under a uniformly distributed load, the simplified
mid-width stress in the beam flange. In order to calculate
peak stress can be expressed as:
the stress due to shear lag, the displacement of the column
flange must be received ahead based on the active compo-
i Is ql 1
rt ¼ rb þ bs Eh An 1 (13) nents, such as the column flange and diaphragm. The dis-
iþ1I k kl
placement of the column flange can be determined using the
where q is the uniformly distributed load. l is the beam flexibility coefficient (af) corresponding to the column
length. The calibration between Eq. (11) and (13) corre- width-to-thickness ratio, as shown in Figure 3. Then, the
displacement of the column flange is expressed as:
sponding to the ratio d/2b provided a construction of modi-
fication factor (bs), as shown in Figure 2 below. As seen q0 b0 3
wf ¼ af (15)
earlier, there were no difference between the modification D
factor bs for the cantilever beam under a point and uni- where b0 denotes the width of the column flange and D ¼
formly distributed load, which meant that it did not depend Etc3/12(1 2). For a box moment connection with internal
on load types, and the modification factor for the cantilever diaphragm, the total displacement of the beam flange at the
beam with a point load is usable. joint can be calculated using the axial stiffness superposition
between the two active components and expressed as:
d
bs ¼ 1:3463 þ 0:0816 (14) Fb Mb
2b d¼ ¼ (16)
K db ðKc þ Kd Þ
where bc expresses the column flange width. Mb represents
3. Box moment connections with the bending moment of the beam at the connection. db
internal diaphragms denotes the average depth of the beam. Kc and Kd denote
the axial stiffness of the column flange and diaphragm,
A box moment connection (T- or X-connection) with respectively. The stiffness of the diaphragm can also be
internal diaphragms deals with the shear lag problem that found in Doung and Sasaki [16, 22]. It could be noted that
enables cracking or failure at the joint when welding is uti- the system represents a cantilever beam when d ¼ 0.
lized. Prediction of shear lag stress of such a connection Using least-work solution with giving the system dis-
accounted for local effect of column flange flexibility includ- placement as an initial boundary condition, the simplified
ing the diaphragm, as firstly revealed by Doung and Sasaki stress distribution in the beam flange of the box moment
[16]. It was observed that when the flexibility was very low, connection with internal diaphragms can be summarized
the peak stress increased significantly. However, the flexible as follows.
column flange and diaphragm did not affect the stress For the connection with beam under a point load,
growth when the web depth of the beam was greater. This
yi i Is P d:k
investigation also included the relations between column rtp ¼ rb Ehbs 1 i An þ (17)
b iþ1 I k h
flange flexibility (included internal diaphragm) and web
depth of the beam on the peak stress due to shear lag and For the connection with beam under uniformly distributed
aimed to provide suggestions for the diaphragm selection. load,
MECHANICS OF ADVANCED MATERIALS AND STRUCTURES 5
Figure 4. von Mises stress developed in internal diaphragm box moment connection with the beam depth 300 mm.
yi i Is ql 1 d:k equation. We observed that for all the internal diaphragm
rtu ¼ rb Ehbs 1 i An 1 þ
b iþ1I k kl h box connections, the peak stresses given by FEM and man-
(18) ual equations provided well matching values. However,
stress distribution for the connection with the unequal web
depth-flange width was slightly different at the mid-width of
3.2. Finite element method the beam flange. This situation admitted that the simplified
manual approach for the stress distribution is suitable for
Finite element method (FEM) was carried out in order to the peak stress but has a slight difference at the mid-width
compare the stress distribution in the beam flange of the of the beam flange when the beam web depth is unequal to
box moment connection (T-connection) with internal dia- the flange width. However, since the stress at the mid-width
phragms with the theoretically simplified formulation using of the beam flange is not necessary for design and check,
Abaqus [23]. To ensure the reliability of FEM results, com- the manual approach is alternative to predict the peak stress
prehensive modeling must be carried out. The FEM consid- using either fourth or sixth order polynomial curve.
ered a 3 D solid 8-node element (C3D8-R) incorporated
with the full model of the connections. The column was
restrained by pins at both ends. The steel material was char- 3.3. Relations between column flange flexibility and
acterized as a multi-linear isotropic behavior with strain web depth of beam on shear lag
hardening, as given by JSCE [24]. Young’s modulus of the The investigation observed that when the column flange
steel was assumed to be 205 GPa. A mesh size of one half flexibility was very low, there was no significant effect by
and one time of the element thickness was applied to the the depth of the beam web. This status was exhibited by the
thickness and the considerable portions of the connec- box connection without using diaphragms. The column
tion’s components. flange alone did not provide advantage in peak stress due to
A rectangular box 300x300x8 mm was used for the col- the deep web of the beam. Moreover, the box connection
umn with the length of 2 m. The beam size was matched to with very thin diaphragms may exhibit similar characteris-
the column but the depth of the beam was varied as tics. In this study, a numerical investigation has been con-
300 mm and 450 mm. The beam (Lb ¼ 1.6 m) was subjected ducted to observe the relations between the column flange
to a point and uniformly distributed loads equivalent to a flexibility (included internal diaphragm) and depth of the
bending stress of 100 MPa. The stress distribution was beam web on the peak stress due to shear lag. Two sample
recorded along the beam flange at the vicinity of the column beam series (series 200 and 400) of the box X-connections
flange (at welding region). are assessed, with beam subjected to a point load that is
The von Mise stresses in the internal diaphragm box con- equivalent to a bending stress of 100 MPa at the connection.
nections with 300 mm-beam depth are illustrated in The depth of the beam web varied from 0.75 b0 to 2 b0.
Figure 4. The stress distributions are also shown in Figures Similarly, the parametric study also considered the variation
5 and 6, respectively, and compared with the simplified of diaphragm flexibility by changing the thickness from 0.5tf
6 P. DOUNG AND E. SASAKI
Figure 5. Stress distribution in the beam flange for the internal diaphragm box moment connection with beam under a point load.
Figure 6. Stress distribution in the beam flange for the internal diaphragm box moment connection with beam under uniformly distributed load.
to 2.5tf. The thicknesses of the beam and column for series and beam web depth-to-flange width ratio for series 200 and
200 and 400 are 6 mm and 16 mm, respectively. 400, respectively. As seen, the peak stress ratio decreased
The results showed in Figures 7 and 8 present the nor- according to the thicker diaphragm. At the same time, it
malized peak stress ratio in terms of diaphragm thickness was also observed that the peak stress increased when the
MECHANICS OF ADVANCED MATERIALS AND STRUCTURES 7
web depth-to-flange width ratio of the beam was greater. 4.1. Connection with beam under a point load
However, when the diaphragm was thin, it was seemingly
The total peak stress due to shear lag is the combination
that the beam web depth did not affect much the peak
between the stresses derived by a cantilever beam and an
stress. From this study, it is suggested that when the dia-
initial displacement of the beam flange and can be seen in
phragm is thin (td 2/3tf), it is not required to modify the
Eq. (20).
peak stress by the modification factor bs.
i Is P d:k
rs ¼ rs0 þ rsd ¼ Ehbs An þ (20)
iþ1I k h
4. Empirical stress evaluation
By substituting parameters Is, I, An, and k into Eq. (20), the
The empirically simplified method in this study focused on peak stress due to shear lag derived in by a cantilever beam
evaluating the simplified peak stresses given by Eqs. (17) can be simplified to
and (18) with an empirical chart associated with shear lag
b0 P
parameter g. Okumura and Ishizawa initially introduced the rs0 ¼ bs g (21a)
d Aw 0
empirical method concerned with shear lag parameter, deal-
ing with the shear lag stress in rectangular box knee connec- rffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
E 4i2 1 S
tions. The method relates with the section properties of the g0 ¼ 9 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
G iþ1
beam and the axial force in the column flange. The peak ð3 þ SÞ ð3 þ SÞ 3 þ 2ði þ 1ÞS
stress due to shear lag given by Okumura and Ishizawa is
expressed as: (21b)
Similarly, the peak stress due to shear lag in accordance
b 0 Fi with the initial displacement of the beam flange once, is
rs ¼ g (19) deduced as:
d Aw
E
rsd ¼ bs dg (22a)
where b0, d, and Aw are the width, depth, and total web area b0 d
of the beam, respectively. Fi is the total forces in the column rffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi
G 2i þ 1 1
flange. g denotes the shear lag parameter, which propor- nd ¼ 6 :i rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(22b)
E 2i 1
tioned with the ratio of flange-to-web area of the beam. The ð3 þ SÞ 3 þ 2ði þ 1ÞS
shear lag parameter g was further simplified to an empirical
chart. Further studies by Nakia et al. [25] and Hwang et al. where b0 represents the flange width of the beam. d and Aw
[20] were conducted in order to provide more precise pre- denote the depth and area of the beam web. S is the ratio of
diction of the peak stress due to shear lag in the box knee web-to-flange area of the beam (Aw/Af). i denotes the order
connections. They proposed the new empirical chart for the of exponent using in the presumed longitudinal displace-
shear lag parameter with redefining the connection model. ment of the beam flange. The shear lag parameters g0 and
For a box moment connection with internal diaphragms, gd for i ¼ 4 can be simplified to an empirical chart as shown
additional shear lag stress was also observed in relation to in Figures 9 and 10, respectively. As seen in Figure 9, the
the ratio of web-to-flange area of the beam. Herein, an shear lag parameter g0 increased to the peak at S ¼ 1.75, and
empirical method for shear lag parameter is proposed for then gradually decreased according to the great value of
both the connections with point and uniform loads. ratio of web-to-flange area of the beam. However, the shear
8 P. DOUNG AND E. SASAKI
Figure 11. Shear lag parameter g0 for internal diaphragm box moment connec-
tion with beam under uniformly distributed load in terms of ratio S.
Figure 9. Shear lag parameter g0 for internal diaphragm box moment connec-
tion with beam under a point load.
Figure 12. Shear lag parameter g0 for internal diaphragm box moment connec-
tion with beam under uniformly distributed load in terms of ratio L/b.
Figure 10. Shear lag parameter gd.
E 2i 1 S 1
g0 ¼ 9
G ði þ 1Þ2 ð3 þ SÞ2 ðL=bÞ
0 1
lag parameter gd in Figure 10 started with the highest value rffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
@ði þ 1Þ: L : G: 2i þ 1 3þS
and rapidly decreased when S 5. Nonetheless, the param- 1A
b E 2i 1 3 þ 2ði þ 1ÞS
eter gd steadily decreased when S > 5, as meant when the
area of the beam web is five times greater than the area of (24)
the beam flange, the peak stress due to shear lag in accord-
The plot of g0 in Figure 11 showed that the parameter
ance with the initial displacement of the beam flange did
depends on two geometric parameters such as the ratios of
not change much. It could be noted that for a cantilever sys-
web-to-flange area (S ¼ Aw/Af) and length-to-haft flange
tem, the load P represents the shear force V, so that the
width (L/b) of the beam. It was also observed that the shear
total peak stress can be calculated as:
" lag parameter g0 for the connection with beam under uni-
#
b0 V E formly distributed load is identical to that under a point
rs ¼ bs g þ d gd (23) load when L/b ¼ 100. In accordance with the ratio L/b, the
d Aw 0 b0
parameter g0 in Figure 12 increased dramatically up to
L/b ¼ 10, and became stable when L/b > 10, for all values of S.
Using the manner from Figures 11 and 12, the simplified
4.2. Connection with beam under uniformly
chart was invented for the parameter g0 which involves with
distributed load
the optimal parameters g1, g2, and g100 and can be seen in
The peak stress due to shear lag for the internal diaphragm Figures 13 and 14. The procedure requires, firstly, calculat-
box moment connection with beam under uniformly distrib- ing the optimal parameters g1, g2, and g100 in Figure 13
uted load can be further simplified using the same proced- with respect to the ratio S. After that, regarding to the ratio
ure for a point load case. The total peak stress can be L/b, the shear lag parameter g0 can be determined using the
calculated using Eq. (23). The shear lag parameter gd for a empirical chart in Figure 14.
point load can also be used for uniformly distributed load
case. However, it is required to calculate shear lag parameter
4.3. Results comparisons
g0 as it depends on the length-to-haft flange width ratio of
the beam. The shear lag parameter g0 for the connection The peak stress calculation was compared with the FEM and
with beam under uniformly distributed load is expressed as: the existing test results given by Tanabe [26] for the internal
MECHANICS OF ADVANCED MATERIALS AND STRUCTURES 9
Table 3. Stress comparison for connections with beam under a point load.
Peak stress ratio, rma/rb
Series Model î0 îd rs Current FEM Test Diff. (%) Okumura Diff. (%)
200 B200 150 6 0.73 2.70 46.30 1.45 1.48 – 2.33 1.13 23.54
B200 200 6 0.80 2.34 53.83 1.55 1.64 – 5.52 1.12 31.56
B200 300 6 0.85 1.86 73.34 1.72 1.87 – 7.90 1.10 41.08
300 B300 200 8 0.70 2.85 44.93 1.45 1.55 – 6.45 1.14 26.38
B300 300 8 0.80 2.34 58.43 1.59 1.69 – 6.05 1.12 33.71
B300 450 8 0.85 1.87 76.81 1.77 1.87 – 5.23 1.10 41.00
400 B400 300 16 0.73 2.71 49.44 1.49 1.50 – 0.63 1.13 24.78
B400 400 16 0.80 2.34 60.45 1.60 1.62 – 1.16 1.12 31.09
B400 600 16 0.85 1.86 79.86 1.80 1.86 – 3.23 1.10 40.75
600 B600 400 20 0.70 2.86 48.46 1.48 1.64 – 9.56 1.14 30.68
B600 600 20 0.80 2.34 63.66 1.64 1.86 – 11.91 1.12 39.77
B600 800 20 0.84 2.00 77.11 1.77 2.00 – 11.60 1.11 44.77
Tanabe Test 2 0.80 2.34 131.70 2.07 – 2.10 1.37 1.45 31.01
Table 4. Stress comparison for connections with beam under uniformly distributed load.
Peak stress ratio, rmax/rb
Series Model î0 îd rs Current FEM Diff. (%) Okumura Diff. (%)
200 B200 150 6 0.73 2.70 52.34 1.52 1.57 3.13 1.18 25.05
B200 200 6 0.79 2.34 63.44 1.63 1.71 4.62 1.12 34.64
B200 300 6 0.85 1.86 82.88 1.83 1.92 4.95 1.07 44.52
300 B300 200 8 0.70 2.85 55.56 1.56 1.63 4.36 1.21 25.56
B300 300 8 0.79 2.34 72.14 1.72 1.76 2.32 1.12 36.43
B300 450 8 0.85 1.87 93.04 1.93 1.95 1.10 1.07 45.30
400 B400 300 16 0.73 2.71 61.36 1.61 1.59 1.29 1.18 26.11
B400 400 16 0.79 2.34 74.32 1.74 1.71 2.09 1.12 34.49
B400 600 16 0.85 1.86 96.55 1.97 1.96 0.37 1.07 45.53
600 B600 400 20 0.70 2.86 62.61 1.63 1.77 8.02 1.21 31.59
B600 600 20 0.79 2.34 81.30 1.81 1.99 8.72 1.12 43.64
B600 800 20 0.84 2.00 97.19 1.97 2.15 8.33 1.08 49.82
10 P. DOUNG AND E. SASAKI
[21] T. Okumura, and N. Ishizawa, The design of knee joints for [24] JSCE. Standard specifications for steel and composite structures,
rigid steel frames with thin walled section, Trans. Japan Soc. Japan Society of Civil Engineers, Japan, 2007.
Civil Eng., vol. 1968, no. 153, pp. 1–18, 1968. DOI: 10.2208/ [25] H. Nakai, T. Miki, and Y. Hashimoto, On limit state design
jscej1949.1968.153_1. method considering shear lag phenomenon of corner parts of
[22] P. Doung, and E. Sasaki, Load-deformation characteristics and steel rigid frames, Doboku Gakkai Ronbunshu (JSCE),
performance of internal diaphragm connections to box col- vol. 1992, no. 455, pp. 95–104, 1992. DOI: 10.2208/jscej.1992.
umns, J. Thin-Walled Struct., vol. 143, pp. 106221, 2019. DOI: 455_95.
10.1016/j.tws.2019.106221. [26] A. Tanabe, Fatigue retrofitting of steel bridge frame piers with
[23] Dassault-Systemes. Abaqus/CAE User’s Manual, Simullia, high seismic performance, Dissertation, Tokyo Institute of
Velizy-Villacoublay, France, 2017. Technology, Tokyo, 2005.