See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/316897311

Distortional analysis of simply supported box girders with inner diaphragms

considering shear deformation of diaphragms using initial parameter method

Article  in  Engineering Structures · August 2017

DOI: 10.1016/j.engstruct.2017.05.004

CITATIONS READS

0 143

5 authors, including:

Yz Ren Qingrong Chen

Tsinghua University Southwest Jiaotong University
7 PUBLICATIONS   1 CITATION    2 PUBLICATIONS   0 CITATIONS   

SEE PROFILE SEE PROFILE

Bin Wang
Brunel University London
117 PUBLICATIONS   729 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Flexural, torsional, distortional behavior of girders with diaphragms View project

All content following this page was uploaded by Yz Ren on 08 November 2017.

The user has requested enhancement of the downloaded file.

 Engineering Structures 145 (2017) 44–59

Contents lists available at ScienceDirect

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

Distortional analysis of simply supported box girders with inner

diaphragms considering shear deformation of diaphragms using initial
parameter method
Yangzhi Ren a,b,⇑, Wenming Cheng b, Yuanqing Wang a, Qingrong Chen b, Bin Wang c
a
Key Lab of Civil Engineering Safety and Durability of China Education Ministry, Department of Civil Engineering, Tsinghua University, Beijing 100084, China
b
Department of Mechanical Engineering, Southwest Jiaotong University, No.111, North Section 1, Second Ring Road, Chengdu, Sichuan 610031, China
c
College of Engineering, Design and Physical Sciences, Brunel University, London, Uxbridge UB8 3PH, UK

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, the distortion of simply supported girders with inner diaphragms subjected to concentrated
Received 14 November 2016 eccentric loads is investigated using initial parameter method (IPM), in which the in-plane shear defor-
Revised 27 April 2017 mation of diaphragms is fully considered. A statically indeterminate structure was modeled with inner
Accepted 3 May 2017
redundant forces, where the interactions between the girder and diaphragms were indicated by a distor-
tional moment. Considering the compatibility condition between the girder and diaphragms, solutions
for the distortional angle, warping displacements and stresses were derived and further simplified by
Keywords:
establishing a matrix equation system. The validity of IPM was intensively verified by a finite element
Simply supported girder
Diaphragm
analysis and distortional experiments. Parametric studies were then performed to examine the effect
Eccentric load of the diaphragm number on the distortional angle, warping displacements and stresses under various
Shear deformation ratios of height to span of the girder and the diaphragm thicknesses. Besides, stabilities of the local
Distortion web plate and mid-span diaphragm were analyzed based on IPM for box girders with symmetrical three
Initial parameter method inner diaphragms. Results show that the local web plate will buckle before overall yielding with the
Finite element analysis increment of the eccentric loads Pj, and the mid-span diaphragm is constantly stable in the whole defor-
Experiment mation process. It shows that more attentions should be paid on the stability of the local web plate than
Local stability
overall yielding for girders subjected to eccentric loads.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Researches on the distortion of girders with inner diaphragms

During the past several decades, box girders have been widely
applied in buildings and bridges due to their large bending and tor-
sional stiffness. However, they are generally susceptible to the
cross-sectional distortion [1] under eccentric loads due to their
quadrilateral instability. Therefore, excessive distortional warping
and transversal bending stresses will be produced besides the
torsional and bending ones in box girders. In a special case, the dis-
tortional warping stresses may be significant to the torsional and
bending ones. In order to control the distortion, diaphragms are
installed at the interior of the girder, which can increase not only
the stability of the local plate, but also the resistance to the warp-
ing deformations and stresses [2–4].
⇑ Corresponding author at: Key Lab of Civil Engineering Safety and Durability of
China Education Ministry, Department of Civil Engineering, Tsinghua University,
Beijing 100084, China.
E-mail address: renyz66@mail.tsinghua.edu.cn (Y. Ren).
have been performed for many decades. The distortion of box gir-
der was initially studied by Dabrowski [5] who first formulated the
distortion of box girders with a symmetrical cross section. Li [6,7]
proposed that the shear strain of the cross section cannot be
ignored when the distortional shear rigidity is significant com-
pared to the distortional warping one for box girders. Wright [8]
proposed the Beam on Elastic Foundation (BEF) analogy for the dis-
tortion of girders with inner diaphragms, where the diaphragms
are analogous to inner supports. Based on BEF, Hsu [9,10] proposed
the Equivalent Beam on Elastic Foundation (EBEF) analogy consid-
ering the shear strain of the cross section, and found that the EBEF
analogy is more versatile than BEF due to its simplicity in analyzing
more complex problems such as non-uniform sections and multi-
span beams.
Interactions between the girder and diaphragms is the key issue
for the distortion of girders with inner diaphragms. A statically
indeterminate structure [11] was modeled with redundant forces
acting along the junctions between the girder and diaphragms.

http://dx.doi.org/10.1016/j.engstruct.2017.05.004
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59
Moreover, the force method was applied to calculate redundant
forces, where elements in the stiffness matrix were obtained from
the finite strip method [12]. The numerical results were then
extended to multi-span bridges [13] and long-span curved bridges
[14]. An outstanding contribution was made by Suetake [15],
where the girder was regarded as an assembly of thin plates, and
the extended trigonometric series method (ETS) was applied to
analyze the stresses and deformations for box girders. Compar-
isons with FEM results show that ETS has a high accuracy. How-
ever, it is inconvenient to apply since there are many
simultaneous nonlinear equations to solve even for girders with
few diaphragms, e.g. there are up to 720 equations for a girder with
two diaphragms.
The wall thickness of diaphragms and the number of dia-
phragms in a girder will make a significant influence on the dis-
placements and stresses. Park [16,17] proposed a new beam
element with nine degrees of freedom per node for girders. Studies
showed that the distortional warping and transversal bending
stresses were reduced by increasing the diaphragm number. Simi-
lar conclusions can be drawn for straight multi-cell box girders
with diaphragms [18,19]. For horizontally curved bridges, the
rational spacing between adjacent diaphragms was provided [20]
according to the ratio between the distortional warping stress
and the bending stress. Using FEM, Zhang [21] found that the
rational number for diaphragms is 3 to 5 when the ratio of width
to height of the cross section is 1.5 and the rational number is 9
when the ratio is 4.5. Li [22] proposed a new finite element solu-
tion, and found that the distortional warping stress for cantilever
girders can be ignored when the spacing between adjacent
diaphragms is less than one fifth of the span; while for simply
supported and fixed girders, the spacing is less than one eighth
of the span.
Initial parameter method (IPM) was proposed first by Vlasov
[23] to analyze the non-uniform torsion of beams. Analogous to
IPM in non-uniform torsion, IPM can be extended to analyze dis-
tortions of girders. Considering the effect of shear strains of the
cross section, Xu [24,25] developed an equation with the variable
distortional angle, and established two categories of IPMs of the
fourth order, classified by the ratio of the distortional stiffness.
Harashima [26] proposed a distortional equation with a distor-
tional warping function, and established the fifth-order IPM. Both
IPMs in distortion have a high efficiency compared with FEM. How-
ever, IPMs has not been extended into the distortion for girders
with inner diaphragms.
l s
z j+1
z p(i+1) Pj
zj
z pi
t p(i+1)
z t pi
O
x K
y s
(a) simply supported box girder
with inner diaphragms
Fig. 1. Girder with inner diaphragms under eccentric loads Pj (j = 1,2,. . .,m).
45
For distortions of a girder with inner diaphragms, an assump-
tion of infinite-rigidity diaphragm was generally made in most
studies [16,20,27], where the in-plane deformation of diaphragms
was totally restrained and warping was free. Similar assumptions
can be found in distortion of curved box beams [28], where the dis-
tortional angle at the location of diaphragms is set as zero. How-
ever, the infinite-rigidity assumption is just an approximation,
which is not applicable to thin flexible diaphragms. The main
objective of this work is to investigate the distortion of simply sup-
ported girders with inner flexible diaphragms under concentrated
eccentric loads, where the in-plane shear deformation of dia-
phragms is fully considered. Interactions between the girder and
diaphragms are indicated by a distortional moment. Based on the
compatibility condition between the girder and diaphragms, solu-
tions for both the distortional angle and warping function are
obtained from the IPM. Taking a simply supported girder with 2,
5 and 9 diaphragms, respectively, as an example, the distortional
solutions from IPM were obtained, then verified by a FE analysis
and experiments. This was followed by a parametric study, in
which distortional deformations and stresses were investigated
in terms of the diaphragm number and thickness and the height
to span ratio of the girder. Based on the proposed IPM, stabilities
of both the local web plate and mid-span diaphragm were exam-
ined for girders with three symmetrical inner diaphragms. A series
of curves were obtained for the relations between the critical buck-
ling load and the position of diaphragms under various height to
width ratios of the cross section.
2. Structural model
Consider a simply supported box girder with inner diaphragms
under concentrated eccentric loads Pj (j = 1,2,. . .,m). The coordinate
system O-xyz is illustrated in Fig. 1a with the original O at the shear
center on one end of the girder. For analysis, the distances between O
and the mid-lines of webs B and D are marked by n1 and n2 in Fig. 1b,
respectively. The girder is made of a homogeneous, isotropic and lin-
early elastic material with its Young’s and shear moduli denoted by
E and G, respectively. The girder span is l. The thicknesses of web B
and D are t1 and t2 and their height is h, while the width of flanges
A and C is b and the thickness t3. The thickness for the ith diaphragm
is tpi (i = 1,2,. . .,n) and its mid line is marked by zpi, measured from O.
The load Pj is located on the top of web D at zj.
Fig. 2a shows that the eccentric load Pj can be decomposed into
three components [29,30] – the flexural load, the torsional and the
Pj
b
P j+1 A
t3
J N
t1 t2
x
s D B s
h
O
n1 n2
y
M
t3
C
Pin joint Flanges:A,C; Webs:B,D
Inner diaphragms Nodes:M,N,J,K
(b) cross section
46 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

Fig. 2. Loading decomposition, deformations and stresses of girders.

distortional load. In Fig. 2b, the cross-section rigidly rotates around where Bi (i = 1, 2, 3, 4) are coefficients determined by the
O with a torsional angle h under torsional loads. Fig. 2c illustrates boundary conditions, and ui are defined as
the transversal deformations in the web and the flange under dis- u1 = cosh(k1z)sin(k2z), u2 = cosh(k1z)cos(k2z), u3 = sinh(k1z)cos
tortional loads, where uM and vM are horizontal and vertical dis- (k2z), u4 = sinh(k1z)sin(k2z)
placements at node M, respectively. The variation of the right where ki (i = 1,2) are distortional coefficients.
angle at node N is defined as the distortional angle v, given by vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vs
usffiffiffiffiffiffiffiffiffi u ffiffiffiffiffiffiffiffiffi
v = v1 + v2. Moreover, the warping displacement wd and the stress 1ut 4EIc EIc 1ut 4EIc EIc
rd, produced by the distortional moment Bd, are shown in Fig. 2d. k1 ¼ þ ; k2 ¼ 
2 EIt GIk 2 EIt GIk
There also exists the shear stress sd along the cross-section profile,
developed by the distortional moment Md, as shown in Fig. 2e. Relations between the warping function and the distortional
This article will focus on the distortional deformation and stres- angle are [7]
ses of a simply supported box girder with inner diaphragms sub-
EIt 000
jected to concentrated eccentric loads. vðzÞ ¼  W ðzÞ; Bd ðzÞ ¼ EIt W 0 ðzÞMd ðzÞ ¼ EIt W 00 ðzÞ ð3Þ
EIc

3. Distortion of girders without diaphragm Substituting Eq. (2) into Eq. (3), the matrix equation is given by
ZðzÞ ¼ UðzÞB ð4Þ
In distortion, when the shear stiffness of the cross section has a
significant value in comparison with the warping one, the influ- where
ence of shear strain of the cross section on deformation and stres- 2 3
 EI
EIt
c
u0001 ðzÞ  EIEIct u0002 ðzÞ  EIEIct u0003 ðzÞ  EIEIct u0004 ðzÞ
ses cannot be ignored [6,7]. The distortional differentiate equation 6 u1 ðzÞ u2 ðzÞ u3 ðzÞ u4 ðzÞ 7
6 7 T
can be expressed as [7] UðzÞ ¼ 6 7; B ¼ B1 ; B2 ; B3 ; B4 ;
4 0
u1 ðzÞ u2 ðzÞ 0
u3 ðzÞ 0
u04 ðzÞ 5
0000 EIt u001 ðzÞ u002 ðzÞ u003 ðzÞ u004 ðzÞ
EIt WðzÞ  EIc WðzÞ00 þ EIc WðzÞ ¼ m0d ð1Þ
GIk ð5Þ
where It, Ic and Ik are distortional warping, frame and shear con- Z(z) is the state vector in IPM,
stants, respectively; W(z) is the distortional warping function; md
is the distributed distortional moment. The number of the apostro- Bd ðzÞ Md ðzÞ T
ZðzÞ ¼ fvðzÞ; WðzÞ;  ; g ð6Þ
phes of W(z) indicates the first, second and forth differentiations EIt EIt
with respect to the z-coordinate.
The boundary conditions for a simply supported girder are
Under concentrated distortional loads, md = 0, and the solution
for Eq. (1) is given as [26] vð0Þ ¼ 0; Bd ð0Þ ¼ 0; for the initial end z ¼ 0;
X
4
WðzÞ ¼ Bi ui ðzÞ ð2Þ vðlÞ ¼ 0; Bd ðlÞ ¼ 0; for the ultimate end z ¼ l:
i¼1
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 47

Correspondingly, the state vectors on both ends are small circles indicate the joints between the girder and diaphragms
  where redundant forces are located.
M d ð0Þ
Zð0Þ ¼ 0; Wð0Þ; 0; T; ZðlÞ In order to analyze the interactions between the girder and dia-
EIt
 T phragms, two assumptions are made:
Md ðlÞ
¼ 0; WðlÞ; 0;
EIt (1) Self balance of the in-plane forces of diaphragms

For z = 0, Z(0) = U(0)B, we have

Under distortional loads, the summation of all vertical and hor-
B ¼ ½Uð0Þinv  Zð0Þ ð7Þ izontal redundant forces and moments are zero for diaphragms.
That is
where [U(0)]inv is the inverse matrix of U(0).
Substituting Eq. (7) into Eq. (4) yields in the state vector Z(z) X X
r X X
r X X
r X X
r
Hej ¼ 0; V ej ¼ 0 Hej yej þ V ej xej ¼ 0
ZðzÞ ¼ PðzÞ  Zð0Þ ð8Þ e¼A;B;C;D j¼2 e¼A;B;C;D j¼2 e¼A;B;C;D j¼2 e¼A;B;C;D j¼2

ð11Þ
where P(z) is the transfer matrix, and P(z) = U(z)[U(0)]inv.
Eq. (8) is the standard form of IPM for the distortion of girders where xej and yej are the distances between the redundant forces Vej,
without diaphragms. However, the transfer matrix P(z) is compli- Hej and the original O.
cated. Based on the relations between ui(z) (i = 1,2,3,4) and their Based on the self balance assumption, only the distortional
differentiations (see Eq. (A1)–(A3) in Appendix A), the matrix P component for redundant forces is reserved. Referred to the forma-
(z) is simplified as tion of external moment Mj [26,27], the distortional components of
2 3 the redundant forces are gathered and indicated by a concentrated
SC 000
1 ðzÞ KC 000
2 ðzÞ SKC 000
3 ðzÞ KC 000
4 ðzÞ
distortional moment Mpi for the ith diaphragm. Therefore, the
6 S
C 1 ðzÞ C 2 ðzÞ SC 3 ðzÞ C 4 ðzÞ 7
6 7 interactions between the girder and diaphragms can be indicated
PðzÞ ¼ 6 K
7 ð9Þ
4 S
K
C 01 ðzÞ C 02 ðzÞ SC 03 ðzÞ C 04 ðzÞ 5 by a moment Mpi (i = 1,2,. . .,n). Only Mpi,, opposite to Mj, will resist
S 00
 K C 1 ðzÞ C 002 ðzÞ SC 03 ðzÞ C 004 ðzÞ the distortional deformation and stresses. In IPM, Mpi is indicated
by a vector Zpi, given by
where S ¼ 2k2 þ2k
1
2, K ¼ EI
EIt
c
, C 1 ðzÞ ¼ uk31ðzÞ  uk12ðzÞ, C 2 ðzÞ ¼ u2 ðzÞ  T
1 2
Mpi
k21 k22
u4 ðzÞ, C 3 ðzÞ ¼ 3k21 k22
u3 ðzÞ  k21 3k22 u4 ðzÞ
u1 ðzÞ, C 4 ðzÞ ¼ 2k1 k2 . Z pi ¼ 0; 0; 0; ð12Þ
2k1 k2 k1 k2 EIt t pi
The jth distortional load in IPM is indicated by a vector Zj, given
by
 T
Mj (2) Compatibility condition between the girder and diaphragms
Zj ¼ 0; 0; 0; ð10Þ
EIt
In-plane shear strains of diaphragms are considered, given by
where Mj is the distortional moment for the jth distortional load, cpi = Mpi /(Gbhtpi). The compatibility condition is that the distor-
and Mj = Pjn1/2 [26]; n1 is the distance between the web D and tional angle at the mid line of diaphragms is opposite to the in-
point the original O (see Fig. 1b). plane shear strain of diaphragms. That is v(zpi) = cpi (0 5 i 5 n).
This is a key aspect for the distortion of girders with inner
4. Distortion of box girders with inner diaphragms diaphragms.
Combining Eq. (8) and Eq. (12), the state vector Z(z) can be
4.1. IPM solution expressed as
R Z
X zpi þt pi =2 X
S
For analysis, a statically indeterminate structure is modeled ZðzÞ ¼ PðzÞZð0Þ  Pðz  zi ÞZ pi dzi  Pðz  zj ÞZ j
with inner redundant forces acting along the junctions between i¼1 zpi t pi =2 j¼1
the girder and the diaphragms. The entire model is shown in
ð13Þ
Fig. 3a and the horizontal and vertical redundant forces Hej and
Vej are illustrated in the zoomed picture, where the subscript e where R and S are the numbers of diaphragms and moments Mj
indicates the webs and flanges and e = A,B,C,D (see Fig. 1b). The before point z, respectively. The transfer matrices P(z  zi) and P

Mpi

Mpi

Fig. 3. A statically indeterminate structure and the equivalent boundaries for the ith diaphragm.
48 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

Pn Pm
(z  zj) are obtained from P(z) by substituting the variable z by i¼1R
2
gi ðzÞMpi þ j¼1S
2
e24 ðz; zj ÞMj
‘z  zi’ and ‘z  zj’. WðzÞ ¼ ð16Þ
2k1 k2 EIt U24 ðl; lÞ
For z = l, Eq. (13) changes into
where
n Z
X zpi þt pi =2 X
m
     
ZðlÞ ¼ PðlÞZð0Þ  Pðl  zi ÞZ pi dzi  Pðl  zj ÞZ j ð14Þ 2 00 tpi t pi
i¼1 zpi tpi =2 j¼1 R
1
gi ðzÞ ¼ n31 l  zpi ; U3342 ðz; lÞ þ n 02
31 l  z pi ; U13
42 ðl; zÞ
tpi 2 2
   
where the vectors Z(l) and Z(0) are referred to Eq. (6); W(0) and 1 2U24 ðl; lÞ 02 t pi
H Rþ i n31 z  zpi ;
Md(0) in vector Z(0) can be calculated from the first and third simul- 2 t pi 2
taneous equations in Eq. (14). Then, substituting W(0) and Md(0)
     
into Eq. (13), the distortional angle and the warping function can 2 00 tpi t pi
be obtained as R
2
gi ðzÞ ¼ n31 l  zpi ; U30 02
42 ðl; zÞ  n31 l  zpi ; U10
42 ðl; zÞ
tpi 2 2
Pn Pm    
1
gi ðzÞMpi þ 1
e24 ðz; zj ÞMj 1 2U24 ðl; lÞ 0ð1Þ t pi
vðzÞ ¼ i¼1R j¼1S
ð15Þ þH Rþ i n24 z  zpi ;
2k1 k2 EIc U24 ðl; lÞ 2 t pi 2

z z
O O
x x
y y

(a) n=2 (b) n=5

z z Ph Ph
O O x
x
y y
Pv Pv
Pv Pv

Ph Ph
at section z1=0.45l at section z2=0.55l
(c) n=9 (d) loading conditions
Fig. 4. Meshing grid, ending boundaries and loading conditions in FEA model.

(a) Distortional warping displacement (b) Distortional warping stress (c) Frame deformation
Fig. 5. 3D contours of a girder with two diaphragms under distortional loads (tp = 0.01 m, amp = 3000).

(a) Distortional warping displacement (b) Distortional warping stress (c) Frame deformation
Fig. 6. 3D contours of a girder with five diaphragms under distortional loads (tp = 0.01 m, amp = 10,000).
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 49

 
1 uðiÞ ðxÞ uðiÞ ðxÞ uðiÞ ðxÞ uðiÞ ðxÞ
1
e ðz; z Þ ¼ K ðz; z Þ  H S þ  j u0004 ðz  zj ÞU24 ðl; lÞ m n ij m n
S 24 j 24 j
2 Umn ðx; yÞ ¼ ðjÞ
ij
; nmn ðx; yÞ ¼ ðjÞ
um ðyÞ uðjÞ
n ðyÞ
um ðyÞ uðjÞ
n ðyÞ

 
1
S
2
e24 ðz; zj Þ ¼ K24 ðz; zj Þ þ H S þ  j u4 ðz  zj ÞU24 ðl; lÞ where um(i)(x) and un(i)(y) are the ith differentiation of functions
2
um(x) and un(y). un(1)(y) is the integral of function un(y), given by
H(x) is the unit step function. H(x) = 1 for x > 0 and H(x) = 0 for
x<0 k1 u3 ðyÞ þ k2 u1 ðyÞ k1 u1 ðyÞ  k2 u3 ðyÞ

uð1Þ
2 ðyÞ ¼
ð1Þ
; u4 ðyÞ ¼
u000 ðxÞ u000 ðyÞ u ðxÞ U4i ðl  y; lÞ k21 þ k22 k21 þ k22
i j i
Uij ðx; yÞ ¼ 0 ; Kij ðx; yÞ ¼ ; When the calculated point z is located in the thickness of (R + 1)
ui ðxÞ u0j ðyÞ uj ðxÞ U4j ðl  y; lÞ
3
th diaphragm (zp(R+1)–tp(R+1)/2 5 z 5 zp(R+1)+tp(R+1)/2), an additional
d Kij ðx; yÞ angle vadd and function Wadd should be involved,
Kij ðx; yÞ ¼ 3
dx

(a) Distortional warping displacement (b) Distortional warping stress (c) Frame deformation
Fig. 7. 3D contours of a girder with nine diaphragms under distortional loads (tp = 0.01 m, amp = 20,000).

-5 -6 6
x 10 x 10 x 10
2 4
IPM-tp/t=0.5
15 IPM-tp/t=0.5 IPM-tp/t=1 IPM-tp/t=0.5
3 IPM-tp/t=1
IPM-tp/t=1 IPM-tp/t=2
IPM-tp/t=2 1 IPM-tp/t=2
10 2 FEA-tp/t=0.5
FEA-tp/t=0.5
FEA-tp/t=1 FEA-tp/t=1
dN(z)/Pa
wdN(z)/m

1 FEA-tp/t=2
FEA-tp/t=2
(z)

0
5 0

-1
0 -1
FEA-tp/t=0.5
FEA-tp/t=1 -2
FEA-tp/t=2
-5 -2 -3
0 0.5 1 0 0.5 1 0 0.5 1
z/l z/l z/l
(a) distortional angle (z) (b) warping displacement wdN(z) (c) warping stress dN(z)

Fig. 8. The distortional angle, warping displacements and stresses between IPM and FEA for a simply supported girder with two diaphragms of different diaphragm
thicknesses.

-5 -7 6
x 10 x 10 x 10
5 5 2
IPM-tp/t=0.5
IPM-tp/t=0.5 IPM-tp/t=1 IPM-tp/t=0.5
4 IPM-tp/t=1 1.5
IPM-tp/t=2 IPM-tp/t=1
IPM-tp/t=2 IPM-tp/t=2
1
3 FEA-tp/t=0.5 FEA-tp/t=0.5
wdN(z)/m

σdN(z)/Pa

FEA-tp/t=1 FEA-tp/t=1
0.5
FEA-tp/t=2 FEA-tp/t=2
χ(z)

2 0
0
1
-0.5
FEA-tp/t=0.5
0 FEA-tp/t=1 -1
FEA-tp/t=2
-1 -5 -1.5
0 0.5 1 0 0.5 1 0 0.5 1
z/l z/l z/l
(a) distortional angle χ(z) (b) warping displacement wdN(z) (c) warping stress σdN(z)
Fig. 9. The distortional angle, warping displacements and stresses between IPM and FEA for simply supported girders with five diaphragms varying with three diaphragm
thicknesses.
50 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

h

 i
M pðRþ1Þ ½2k1 k2  u004 ðz  zpðRþ1Þ þ tpðRþ1Þ =2Þ t pi tpi
31 l  zpi ; 2 U42 ðzpT ; lÞ þ n31 l  zpi ; 2 U42 ðl; zpT Þ
33 13
vadd ¼ ð17Þ T
1
gi ðzpT Þ ¼ t2pi n00 02

2k1 k2 EIc t pðRþ1Þ 


H T  12  i 2U24 ðl;lÞ 02 t
t pi
n31 zpT  zpi ; 2pi
ð1Þ 
M pðRþ1Þ u4 ðz  zpðRþ1Þ þ t pðRþ1Þ =2Þ T e24 ðzpT ; zj Þ ¼ K24 ðzpT ; zj Þ  H kT þ 2  j U24 ðl; lÞu4 ðzpT  zj Þ
1 1 000
W add ¼ ð18Þ
2k1 k2 EIt t pðRþ1Þ
kT is the number of distortional loads before the Tth diaphragm.
where zp(R+1), tp(R+1) and Mp(R+1) are the mid-line location, the thick- The matrix equation system for Eq. (19) is
ness and the distortional moment of the (R + 1)th diaphragm,
g  Mp þ e ¼ 0 ð20Þ
respectively.
Based on Eqs. (15)–(18), both the angle v(z) and the function W where
2 3
(z) are related to Mj and Mpi. Since Mj has been given in Eq. (10), R11 T
1
g2 ðzp1 Þ T gn ðzp1 Þ
1

solutions rest in Mpi. 6 1 g ðz Þ 7

T gn ðzp2 Þ 7
1
6 R22
M p ¼ fMp1 ; Mp2 ; . . . ; Mpn gT ; g ¼ 6 T 1
p2
7
4 ... ... ... ... 5
4.2. Derivation of Mpi
T g1 ðzpn Þ T g2 ðzpn Þ
1 1
Rnn
The compatibility condition gives the equation
" #T
Xm X
m X
m
X
n X
m
e¼ 1
e24 ðzp1 ; zj ÞMj ; 1
e24 ðzp2 ; zj ÞMj ; :::; 1
e24 ðzpn ; zj ÞMj
T
1
gi ðzpT ÞMpi þ T
1
e24 ðzpT ; zj ÞMj j¼1
T
j¼1
T
j¼1
T

i¼1 j¼1
    
U24 ðl; lÞ tpT EIc and the diagonal elements in matrix g is
 u004  2k1 k2 1 þ M pT ¼ 0 ð19Þ
tpT 2 Gbh     
U24 ðl; lÞ tpi EIc
Rii ¼ T 1 gi ðzpi Þ  u004  2k1 k2 1 þ
for the Tth diaphragm (T = 1,2,. . .,n), where tpi 2 Gbh

-5 -7 5
x 10 x 10 x 10
4 15
IPM-tp/t=0.5
2.5 IPM-tp/t=0.5 IPM-tp/t=1 IPM-tp/t=0.5
IPM-tp/t=1 IPM-tp/t=1
IPM-tp/t=2
IPM-tp/t=2 2 10 IPM-tp/t=2
2 FEA-tp/t=0.5
σdN(z)/Pa
FEA-tp/t=0.5
wdN(z)/m

FEA-tp/t=1 FEA-tp/t=1
1.5 FEA-tp/t=2
χ(z)

FEA-tp/t=2
0 5
1

-2 0
0.5 FEA-tp/t=0.5
FEA-tp/t=1
0 FEA-tp/t=2
-4 -5
0 0.5 1 0 0.5 1 0 0.5 1
z/l z/l z/l
(a) distortional angle χ(z) (b) warping displacement wdN(z) (c) warping stress σdN(z)
Fig. 10. The distortional angle, warping displacements and stresses between IPM and FEA for simply supported girders with nine diaphragms varying with three diaphragm
thicknesses.

n=2, with shear strain of the cross section n=5, with shear strain of the cross section
n=2, without shear strain of the cross section n=5, without shear strain of the cross section
-5 -6 6
x 10 13.3% x 10 x 10
10 1.5 4
14.9%
8 1 3

6 0.5 2
wd

σd

4 0 1
χ

17.8%
2 -0.5 0

0 -1 -1

-2 -1.5 -2
0 0.5 1 0 0.5 1 0 0.5 1
z/l z/l z/l
(a) (b) (c)
Fig. 11. Comparison of girders with and without the shear strain of cross section for (a) the distortional angle, (b) warping displacements and (c) stresses of simply supported
girders with 2 and 5 diaphragms.
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 51

0 1 0 1
Consequently, Mpi can be obtained by w11 w12 w1n a11 a12 a1n
B w21 w22 w2n C B a21 a22 a2n C
X
m B
where w ¼ @ C B
;a¼@ C and
M pi ¼  Q ij Mj ð21Þ ... ... ... ... A ... ... ... ... A
j¼1 wn1 wn2 wnn an1 an2 ann
the elements in matrix a are
according to the Cramer rule. Qij = |gi|/|g|, where the |g| indicates 8 Xm
the determinant of g, and gi is the same of g except for the ith col- < ½S 1 e24 ðz;zj ÞT 1 gk ðzpg Þ=n  R 1 gk ðzÞT 1 e24 ðzpg ;zj ÞM j ðg–kÞ
umn [T 1 e24 ðzp1 ; zj Þ, T 1 e24 ðzp2 ; zj Þ; . . . ; T 1 e24 ðzpn ; zj Þ]T agk ¼ Xj¼1
: m
½ 1 e ðz;zj ÞRgg =n  R 1 gg ðzÞT 1 e24 ðzpg ; zj ÞMj ðg ¼ kÞ
j¼1 S 24

4.3. Simplification of v(z) and W(z)

In this approach, the distortional angle v(z) can be simplified as
Substituting Eq. (21) into Eq. (15), and the angle v(z) changes 1 X n

into vðzÞ ¼ w ð24Þ

2k1 k2 EIc U24 ðl; lÞ i¼1 ii
Pn Pm
S 1 e24 ðz;zj Þ 
i¼1 j¼1 n
 R 1 gi ðzÞQ ij M j For the warping function W(z), a matrix equation can also be
vðzÞ ¼ ð22Þ
2k1 k2 EIc U24 ðl; lÞ established similar to Eq. (23), and the elements agk (g, k = 1,2,. . .
n) in matrix a are
where n and m are the total numbers of diaphragms and distor- 8 Xm
tional loads, respectively. < ½S 2 e24 ðz;zj ÞT 1 gk ðzpg Þ=n  R 2 gk ðzÞT 1 e24 ðzpg ;zj ÞM j ðg–kÞ
The number of calculation steps is m  n in Eq. (22) and it agk ¼ Xj¼1
: m
½ 2 e ðz;zj ÞRgg =n  R 2 gk ðzÞT 1 e24 ðzpg ; zj ÞM j ðg ¼ kÞ
would be time-consuming for girders with many diaphragms j¼1 S 24
under many distortional loads. For solution, a matrix equation is
established, given by Therefore, the function W(z) can be simplified as

1 X n
gw¼a ð23Þ WðzÞ ¼ w ð25Þ
2k1 k2 EIt U24 ðl; lÞ i¼1 ii

Fig. 12. anti-symmetrical loading design.

20 Welded field
150

Stiff unit on webs Stiff unit on flanges

1500
(a) stiff unit on flanges (mm)
20 Welded field
350

Welded field
1500

(b) stiff unit on webs (mm) (c) actual picture of stiff unit
Fig. 13. Stiff units.
52 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

Taking the node N (see Fig. 1b) of the cross section as an exam-
ple, the distortional warping displacement wdN and the stress rdN
are given by
WðzÞbd bh EW0ðzÞbd bh
wdN ðzÞ ¼  ; rdN ðzÞ ¼ 
4ðbd þ 1Þ 4ðbd þ 1Þ
where bd is the ratio of the distortional warping stresses between
3 þht 1
nodes J and N, and bd ¼ 3bt
3bt 3 þht 2
5. Verifications of IPM
5.1. Verifications with FEA
To verify the proposed IPM, simply supported girders with 2, 5
and 9 diaphragms are investigated for three diaphragm thick-
nesses, respectively by FEA using a commercial code ANSYS. All
girders are modeled with the Young’s modulus E = 2.1  1011 Pa,
the Poisson’s ratio t = 0.3, the span l = 1 m, the width b = 0.1 m,
the height h = 0.2 m and the flanges and webs thicknesses
t = 0.01 m. Diaphragms are evenly distanced along the span with
the thicknesses tp = 0.005 m, 0.01 m and 0.02 m, respectively.
Fig.4a, b and c show the meshing grids of the girders and dia-
phragms using Shell63 element in FEA model, where translations
in the x- and y- axial directions and rotations about the y- and
z-axes on both ends are restrained. A convergent test shows that
1650 to 2026 elements are adequate in terms of the diaphragm
number. Fig. 4d shows the loading conditions, where two concen-
trated distortional loads are applied along the flange and web in
cross sections z1 = 0.45l and z2 = 0.55l, and Ph = 1.25 kN and
Pv = 2.5 kN, respectively.
Figs. 5–7 show 3D contours of the distortional warping defor-
mations and stresses for simply supported girders with 2, 5 and
9 diaphragms, respectively. The ‘amp’ indicates the amplified fac-
ð26Þ
tor for deformations. It is seen that the largest displacement and
stress occur at the junction between webs and flanges at the load-
ing sections. With the increment of diaphragm number, the largest
stress reduces from 5.51 MPa to 1.55 MPa and the displacement
from 1.61 lm to 0.268 lm, and frame deformations at the loading
sections clearly become small.
Figs. 8–10 show the distortional angle, warping displacements
and stresses at node N from IPM and FEA for simply supported
girders with 2, 5 and 9 diaphragms, respectively, in the relative
diaphragm thickness tp /t = 0.5, 1 and 2. The distortional angle in
FEA results can be calculated byFig. 9
UXN  UXM UYN  UYJ
vðzÞ ¼ þ ð27Þ
h b
where UXN and UXM are x-axial displacements at nodes N and M
(see Fig. 1b), respectively; UYJ and UYN are y-axial displacements
at nodes J and N, respectively.
Good agreements are observed between IPM and FEA in
Figs. 8–10 for the distortional angle, warping displacements and
stresses for simply supported girders with inner diaphragms. Com-
pared the girders braced by 2 diaphragms with those by 5 and 9
diaphragms, it’s worth noting that the mid-span diaphragm effec-
tively restrains the transversal deformation.

Hydraulic cylinder Loading head Support Connection beam

FCS hydraulic servo

loading system

Hydraulic jack

Wheel units Tested girders Stiff unit

(a) entire scheme (b) loading equipment

Fig. 14. Entire experimental scheme.

B11 A11
A
B10 10A
9
B9 A8
B8 A7
B7 Rectangular tube
A6
B6 A5 Magnetic base
B5 A4
B4 A3 Dial gages
B3 A2
B2
A
B1 N 1

z x
J O M
y

K (b) actual picture

(a) scheme of tested points
Fig. 15. The scheme of tested points.
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59
For the distortional angle, the largest error between IPM and
FEA occurs at the loading sections, where the FEA result is
23.68% higher than the IPM one for girders with two diaphragms,
and reduces to 13.86% for those with five diaphragms and 10.18%
for those with nine diaphragms. Since there is no diaphragms or
stiffeners at the loading sections, the error between IPM and FEA
can be attributed to the local stress concentration. So the distor-
tional angle obtained from IPM is susceptible to the influence of
stress concentration.
In addition, the influence of shear strains of the cross section on
distortional deformations and stresses are examined in Fig. 11 for
simply supported girders with 2 and 5 diaphragms, where the
compatibility condition between the girder and diaphragms is con-
sidered. It is seen that the shear strain of the cross section makes
little effect on warping displacements and stresses, but a large
influence on the distortional angle. The largest difference occurs
at the loading sections z = 0.45l and z = 0.55l, which is 14.9% for
girders with 2 diaphragms and 17.8% with 5 diaphragms. Also,
the error at the mid span for girders with 2 diaphragms is 13.3%.
Thus shear strains of the cross section cannot be ignored when
the transversal deformation of girders is considered.
5.2. Verifications with experiments
For further verification of the IPM, a series of experiments were
performed using four groups of samples – girders with no dia-
phragms, one, two and three diaphragms, subjected to distortional
loads. Diaphragms are equally distanced along the span. Both gird-
ers and diaphragms were fabricated from carbon structural steel
plates (yield strength 235 MPa) of 8 mm thickness. All girders are
1 2.5
0 2
UXN/mm
UYN/mm
-1
IPM 1.5
Quadratic fitting
Experiments
-2
Singularity point 1
-3
0.5
A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11
Tested points
(a) x-axial displacement
Fig. 16. Comparison of transversal displacements and angle between IPM and experiments under 5t loading.
Fig. 17. Strain measurements.
53
3 meters long, with height h = 0.6 m and width b = 0.346 m, giving
a 30° angle between the diagonal and the web. All girders are
sealed by a steel plate of 6 mm thickness on both ends.
To simultaneously produce two concentrated distortional loads,
as in Fig. 4d, two steps were taken as follows
(1) For distortional loads, as shown in Fig. 12, units were
designed with two groups of wheels anti-symmetrically
about the shear center O of the cross section. This is to
decompose the horizontal power force Psource produced by
FCS hydraulic servo system into two orthogonal loading
components Ph and Pv.
(2) For concentrated loads, as shown in Fig. 13, four stiff units of
1.5 m in length and 0.15 m and 0.350 m in width, respec-
tively, were welded onto flanges and webs symmetrically
over the mid span, and the loading sections were hence
located at z1 = 0.75 m and z2 = 2.25 m, respectively.
During the experiments, compressive loadings were applied
along the diagonal of the cross section by two anti-symmetrical
wheel units, as shown in Fig. 12a. The entire experimental setup
is shown in Fig. 14a, including two FCS hydraulic servo loading sys-
tems, two connection beams, two wheel units, four stiff units, the
tested girder and two supports. The maximum loading was set to
5t with a loading speed of 2 mm/s, controlled by a loading equip-
ment in Fig. 14b.
Fig. 15a shows measured points on the girder, with twelve
equally distanced points A1–A11 on webs and B1–B11 on flanges.
The transversal deformations on measured points were measured
by dial gages, magnetically attached to a rectangular tube of
-3
x 10
15
10
χ/rad
5
IPM
IPM 0 Quadratic fitting
Quadratic fitting Experiments
Experiments Singularity point
-5
Singularity point
0.5 1 1.5 2 2.5
Tested points z/m
(b) y-axial displacement (c) distortional angle
54 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

3.5 m length, as shown in Fig. 15b. Measurements at each point
gave the x- and y- axial deformations UXN and UYN (N = 1,. . ., 11),
and the distortional angle is calculated as
2UXN 2UYN
vðzÞ ¼ þ ð28Þ
h b
Compared with those from IPM, the transversal displacements
and distortional angle are depicted in Fig. 16 for girders without
diaphragms under the source loading of 5t. There exist some errors
between the IPM and experimental results at the singularity points
A1 for x-axial displacement and B10 and B11 for y-axial displace-
ment, which are mainly caused by the residual strains of welding
and manufacturing. Eliminating the influences induced by singu-
larity points, fitting lines were obtained from experimental results
by applying the quadratic fitting method provided in the software
MATLAB. Fig. 16 shows reasonable agreements between two

Fig. 18. Comparisons of distortional warping stresses between IPM and experiments (n being the total diaphragm number).

1.2 0.8
h/l=0.1 tp/t=0.25
h/l=0.2 tp/t=0.5
1
h/l=0.3 tp/t=1
0.6 tp/t=2
h/l=0.4
0.8
h/l=0.5 tp/t=4
d/ d0

d/ d0

0.6 0.4

0.4
0.2
0.2

0 0
1 2 4 6 8 9 1 2 4 6 8 9
n n
(a) varying with h/l (b) varying with tp/t
Fig. 19. The non-dimensional warping stress rd/rd0.
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 55

results.For distortional warping stresses, strains (ed0 ,ed45 and
ed90 ) in three directions were measured at points A1–A6 using
strain gauges shown in Fig. 17a and recorded by the static strain
equipment in Fig. 17b. The warping stress is calculated as
E
rd ¼ ðed90 þ ted0 Þ
1  t2
where E = 2.1  1011Pa and t = 0.3.
Fig. 18 shows calculated distortional warping stresses on the
tested points A1–A6 for girders with (a) no diaphragm, (b, c) one
diaphragm, (d, e) two diaphragms and (f) three diaphragms,
respectively. It is seen that the warping stresses obtained from
the IPM and experiments have the same up-and-down trend and
with apexes at A3 for girders with diaphragms. The errors mainly
come from the drop of dh (Fig. 12a) between two horizontal source
forces, which are mainly caused by the residual strain during man-
ufacturing and welding. And this will produce torsional warping
stress besides the distortional one.
6. Parametric study
In this section, the effects of the ratio of height to span of the
girder h/l, the diaphragm thickness tp and number n are examined
on the distortion of simply supported girders with equally dis-
tanced inner diaphragms, where the loading and boundary condi-
tions are referred to those in Section 5.1.
Fig. 19 shows the non-dimensional warping stress rd/rd0 for
node N at the loading section z = 0.45l varying with the ratios h/l
and tp/t, where rd0 is the warping stress of girders without dia-
phragms. It is seen that except for the girder with two diaphragms
under h/l = 0.1, the non-dimensional stress rd/rd0 remains less
ð29Þ than 1 and becomes smaller with the increment of n, tp/t and h/l,
respectively. Besides, the rd/rd0 decreases substantially when the
number of diaphragm reduces from 2 to 1, or increases from 2 to
3, clearly indicating that the mid-span diaphragm plays a signifi-
cant role in the reduction of distortional warping stress.
Fig. 20 shows the non-dimensional distortional angle v/v0 at
the loading section z = 0.45l varying with h/l and tp/t, where v0 is
the distortional angle for girders without diaphragms. It is seen
that except for the girder with two diaphragms under h/l = 0.1,
the non-dimensional distortional angle v/v0 remains less than
0.25 and becomes smaller with the increment of n, tp/t and h/l,
respectively. This implies that the inner diaphragms are capable
of restraining the transversal deformation of the cross section,
especially when n > 3. Similarly, compared with the distortional
angle at n = 2, those at n = 1 and 3 decrease significantly due to
the restraint by the mid-span diaphragm.
Fig. 21 shows the non-dimensional warping displacement wd/
wd0 for node N at the loading section z = 0.45l varying with the
ratios h/l and tp/t, where wd0 is the warping displacement for gird-
ers without diaphragms. It is seen that except for the girder with
two diaphragms under h/l = 0.1, the non-dimensional displace-

0.6 0.25
h/l=0.1 tp/t=0.25
0.5 h/l=0.2 tp/t=0.5
h/l=0.3 0.2
tp/t=1
h/l=0.4 tp/t=2
0.4
h/l=0.5 0.15 tp/t=4
0

0.3
/

0.1
0.2

0.05
0.1

0 0
1 2 4 6 8 9 1 2 4 6 8 9
n n
(a) varying with h/l (b) varying with tp/t
Fig. 20. The non-dimensional distortional angle v/v0.

1.5 0.8
h/l=0.1 tp/t=0.25
h/l=0.2 tp/t=0.5
h/l=0.3 0.6 tp/t=1
1 h/l=0.4 tp/t=2
h/l=0.5 tp/t=4
wd/wd0
wd/wd0

0.4

0.5
0.2

0 0
1 2 4 6 8 9 1 2 4 6 8 9
n n
(a) varying with h/l (b) varying with tp/t
Fig. 21. The non-dimensional warping displacement wd/wd0.
56 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

ment wd/wd0 in Fig. 21a remains less than 1.0 and converges to a
fixed value between 0.1 and 0.2 for large diaphragm numbers.
7. Buckling of both the local web plate and the mid-span
diaphragm
Taking a simply supported girder with uniform three inner dia-
phragms as an example, Fig. 22 shows 3D contours for the warping
displacements and stresses, where the measurement, loading and
boundary conditions are referred to those in Section 5.1. It is seen
from Fig. 22 that the maximum displacement occurs at the local
web plate at the loading sections, and the maximum stress occurs
at the edge of the mid-span diaphragm, which may result in local
buckling. It is thus necessary to check the stabilities for both local
web plate and mid-span diaphragm.
In this section, the proposed IPM was carried to analyze the
critical buckling values Pcr1 and Pcr2 for the eccentric load Pj accord-
ing to the stabilities of the local web plate and the mid-span
diaphragm, respectively. Inner diaphragms include one fixed
(diaphragm II) at the mid span and two (diaphragms I and III) sym-
metrical to the mid-span one. The loading and boundary conditions
are referred to those in Section 5.1.
Fig. 23 shows the mixed stress boundary condition for the local
web plate between the loading section and the mid span, in which
a = 0.05l. Under distortional loads, the warping stress rd varies lin-
early from rd1 at the loading section to rd2 (=brd1) at the mid span
and from rd1(2) on the top to –rd1(2) at the bottom. There also exists
a constant shear stress sd1 (=grd1) on all four sides of the cross-
section, and a parabola sd2 on both lateral sides. Both shear stresses
can be obtained from [26]
H
M d ðzÞ Sd ðsÞqd ðsÞds
sd ðs; zÞ ¼  ðSd ðsÞ  F H Þ ð30Þ
It t q ðsÞds
F d
where Md(z) is the distortional moment in Fig. 2; It is the distor-
tional warping constant in Eq. (1); t is the thickness of local web
plate, given by t = 0.01 m; F is the cross-sectional area; qd(s) is the
distance between the distortional center [6] and the mid line of
the cross-sectional profile; Sd(s) is the second moment of area, given

Mid diaphragm Deformed frame

Diaphragm I (diaphragm II) Diaphragm III Original frame

Maximum warping displacement

(a) 3D contour of distortional warping displacement (b) deformation of mid-span diaphragm

Maximum
warping stress

(c) 3D contour of distortional warping stress (d) Stress of mid-span diaphragm

Fig. 22. 3D contours of warping displacement and stress of girders with three uniform inner diaphragms (tp = 5 mm).

Fig. 23. Mixed stress boundaries of local web plate.

 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 57

R
by Sd(s)= ŵds, ŵ is the distortional sectorial coordinate [7], s is the Table 1
Several definite integral items used in Eq. (35).
circumferential coordinate around the cross-sectional profile.
Under the mixed stress boundary condition, the out-of-plane Definite integrals u=p u – p and u + p :even u – p and u + p :odd
deformation wp for local web plate can be indicated by the combi- Ra a
sin upa z sin ppa z dz 2 0 0
nation of two orthogonal sinusoidal functions in z- and y-axial R0a upz ppz 2ap
0 cos a sin a dz
0 0
pðu2 p2 Þ
directions, given by Ra upz ppz
z sin sin a2 0 4a2 up
0 a a dz 4 p2 ðu2 p2 Þ2
X1 X 1
upz v py Ra upy
0 y cos a sin a dy
ppy
 4u
a 2
pa2 pa2
wp ¼ Auv sin sin ð31Þ Ra 2
p pðu2 p2 Þ pðu2 p2 Þ
u¼1 v ¼1
a h upy
0 y cos a sin a dy
ppy a3
 4u pa3 4pa3 ð3u2 þp2 Þ
 pðupa
3
p uðm2 p2 Þ p3 ðu2 p2 Þ3 2 p2 Þ

where u and v are the numbers of half waves in z- and y-axial direc-
tions; Auv is the undetermined coefficient, which is the maximum of
the combination of two sinusoidal functions. 2.5
For analysis, the distortional warping stress rd and the shear
stress sd for the local web plate can be presumed as [31]
  
time=12s
2 1b
rd ¼ rd1 1  y 1  z ð32Þ 2
h a
 
1b y2

d1)cr /MPa
sd ¼ sd1 þ sd2 ¼ rd1 g þ rd1 y ð33Þ time=508s
a h
1.5
where b is the ratio of the warping stresses rd1 to the rd2; g is the

(
ratio of the warping stress rd1 to the constant shear stress sd1. Both time=51220s
ratios are calculated from IPM.
To obtain the critical buckling stress (rd1)cr for the local web 1
plate, Galerkin equation [32] is applied, given by (p, q = 1,2,. . .,1) Dimension: 2×2
Z Z ! Dimension: 3×3
a h
@ 4 wp @ 4 wp @ 4 wp @ 2 wp @ 2 wp
þ 2 þ þ t rd þ 2t s d Dimension: 4×4
0 0 @z4 @z2 @y2 @y4 @z2 @z@y
0.5
ppz qpy 0 0.1 0.2 0.3 0.4
 sin sin dzdy ¼ 0 ð34Þ
a h zp1/l
Substitute Eqs. (31)–(33) into Eq. (34), the latter is changed into
Fig. 24. Convergence of critical distortional warping stress (rd1)cr.
8
 9
> v p 2 2 R a R h >
>
>
>
up 2
þ h sin upa z sin v hpy sin ppa z sin qph y dzdy >
>
>
X
1 X
1 < a 0 0
=
up 2 R a R h   warping stress on top of the web at the loading section z = 0.45l,
Auv
> t r d1 a 1  2
y 1  1b z sin upa z sin v hpy sin ppa z sin qph y dzdy > ¼ 0
u¼1 v ¼1 >
> R
0 0
R h h

i
a
>
> produced by the unit moment Me.
>
: þt rd1 2uv p2 a h g þ 1b y  y cos upz cos v npy sin ppz sin qpy dzdy >
2
;
ah 0 0 a h a h a h Fig. 25 gives the critical load Pcr1 varying with the location zp1 of
ð35Þ diaphragm I for various ratios h/b, based on the stability of the local
web plate. It is seen that the critical load Pcr1 increases remarkably
where the definite integrals on variables u and p are given in when diaphragm I is located close to the loading section z1 = 0.45l,
Table 1; and the integrals on v and q are obtained by replacing u implying that installation of a diaphragm at the loading section
and p. will enhance effectively the stability of the local web plate.
Eq. (35) can be translated into a matrix equation set and the While for the mid-span diaphragm (diaphragm II) subjected to
determinant of the coefficient matrix should be zero as the coeffi- pure shear boundary conditions, the critical value (Mp2)cr for
cient Auv cannot be zero. Therefore, the critical stress (rd1)cr can be moment Mp2 is given by [33,34]
obtained.
In addition, since the dimension of the coefficient matrix affects Ep4 t3p2
the calculation accuracy, the critical stresses (rd1)cr were examined ðM p2 Þcr ¼ ; ðh=b 6 2Þ
nð1  t2 Þ
first in Fig. 24 under different dimensions of the coefficient matrix.
In the calculation, the span l = 1 m, the height h = 0.25 m, the width   3
89h b Ep2 tp2
b = 0.1 m, the thicknesses for webs and flanges t = 0.01 m, and the ðM p2 Þcr ¼ þ ; ðh=b > 2Þ ð37Þ
200b 3h 1  t2
diaphragm thickness tp = 0.005 m. It is seen that the critical stress
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi



(rd1)cr tends to a converged value as the dimension of the coeffi- 2 2
384b2 h2 h2 þb2 h2 þb2
cient matrix increases from 2  2 to 4  4. However, calculation where n ¼ 2
706
625
þ 81
25 h2 þ9b2
þ 81
25 9h2 þb2
; t is the Pois-
9ðh2 þb2 Þ
time increases significantly from 12 s to 51220 s. Considering both son’s ratio and equals to 0.3; tp2 is the thickness of mid-span dia-
the accuracy and time, the 3  3 coefficient matrix is regarded phragm and tp2 = 0.005 m.
most appropriate to calculate the critical stress (rd1)cr. Based on Eq. (21), the critical moment Mcr2 for external moment
In order to obtain Pcr1, a linear relationship is established Mj can be calculated by
between the critical moment Mcr1 and the critical stress (rd1)cr.
ðM p2 Þcr
M cr1 ðrd1 Þcr Mcr2 ¼  ð38Þ
¼ ð36Þ Q 21 þ Q 22
Me ðrd1 Þe
where Q21 and Q22 are defined in Eq. (21).
where Mcr1 is the critical value for external moment Mj based on the Furthermore, based on equation Mj = Pjn1/2 [26], the critical
stability of the local web plate, given by Mcr1 = Pcr1n1/2. Me is the load Pcr2 can be finally obtained from Eq. (38). For the stability of
unit external moment, i.e. Me = 1Nm. (rd1)e is the corresponding the mid-span diaphragm, Fig. 26 gives the critical load Pcr2 varying
58 Y. Ren et al. / Engineering Structures 145 (2017) 44–59

50 8. Conclusions
h/b=2.5
h/b=2 In this paper, the initial parameter method (IPM) is applied to
investigate the distortion of simply supported girders with inner
40
h/b=1.5 diaphragms subjected to concentrated eccentric loads, where in-
h/b=1 plane shear deformation of diaphragms is considered. The main
conclusions can be drawn as follows
15
30 (1) Compared with results from FEA and experiments, accurate
analyses can be obtained by IPM for the distortional angle,
Pcr1/kN

10 warping displacements and stresses for simply supported

0.3 0.4 girders with inner diaphragms. Both the in-plane shear
20 deformation of diaphragms and the compatibility condition
Zoomed picture
between the girder and diaphragms are taken into account
in IPM. Besides, comparison of results between considering
the shear strain of the cross section and not shows that
10 the shear strain of the cross section cannot be ignored when
calculating the distortional angle.
(2) Both the distortional angle and warping stresses decrease
0 0.2 0.4
with the increment of the ratio of height to span, the number
zp1/l
of diaphragms and their thickness. The warping displace-
Fig. 25. The critical load Pcr1 obtained from the stability of local web plate. ment converges to a fixed value between 0.1 and 0.2 for
large diaphragm numbers. And the mid-span diaphragm
plays a key role in reducing the distortional deformations
4 and stresses for girders under symmetrical loads.
x 10 (3) Stabilities of the local web plate and the mid-span
2.5
h/b=2.5 diaphragm were both investigated for box girders with sym-
metrical three inner diaphragms. Results show that both the
h/b=2 local web plate and the mid-span diaphragm increase their
h/b=1.5 resistance to buckling when the diaphragm I is located close
2
h/b=1 to the loading sections. Moreover, the local web plate will
buckle first as the primary failure mode. Therefore, atten-
tions are needed on the stabilities of the local web plate
for simply supported girders under eccentric loads.
Pcr2/kN

1.5
Based on the IPM, it is possible to improve the warping
displacements and stresses of simply supported girders through
optimizing the positions and wall thickness of diaphragms. Future
1 work are needed for (1) optimization of the distortion of girders
with diaphragms; (2) mechanical properties of girders with
perforated diaphragms.

0.5 Acknowledgements
0 0.1 0.2 0.3 0.4 0.5
zp1/l This work was supported by the National Natural Science Foun-
dation of China (NSFC) [grant number: 51175442 and 51675450].
Fig. 26. The critical load Pcr2 obtained from the stability of the mid diaphragm.

Appendix A
with the location zp1 of diaphragm I for various values of h/b. It is
seen that the critical load Pcr2 increases remarkably when dia- The relationships between ui(z) (i = 1,2,3,4) and their differenti-
phragm I is located close to the loading section z1 = 0.45l, indicat- ations is
ing increased resistance to buckling.
Back to the girder with equally-distanced three inner dia- u01 ¼ k1 u4 þ k2 u2 ; u02 ¼ k1 u3  k2 u1 u03 ¼ k1 u2  k2 u4 ; u04
phragms in Fig. 22, the girder will reach its yield strength limit
¼ k1 u1 þ k2 u3 ; ðA1Þ
of 235 MPa when the eccentric load Pj is 650kN, which is much
smaller than the critical buckling load Pcr2 of 8743kN at
zp1/l = 0.25 for the ratio h/b = 2 according to the stability of the u001 ¼ ðk21  k22 Þu1 þ 2k1 k2 u3 ; u002 ¼ ðk21  k22 Þu2  2k1 k2 u4 ;
mid-span diaphragm. Simultaneously, the plastic yield load of u003 ¼ ðk21  k22 Þu3  2k1 k2 u1 ; u004 ¼ ðk21  k22 Þu4 þ 2k1 k2 u2 ;
650kN is much larger than the critical load Pcr1 of 12.2kN at
ðA2Þ
zp1/l = 0.25 according to the stability of the local web plate. This
implies that buckling of the local web plate will be the primary
failure mode with the increment of eccentric loads Pj. Hence atten- u0001 ¼ ðk31  3k1 k22 Þu4 þ ð3k21 k2  k32 Þu2 ; u0002 ¼ ðk31  3k1 k22 Þu3  ð3k21 k2  k32 Þu1 ;
tions should be paid on the stability of the local web plate for the u0003 ¼ ðk31  3k1 k22 Þu2  ð3k21 k2  k32 Þu4 ; u0004 ¼ ðk31  3k1 k22 Þu1 þ ð3k21 k2  k32 Þu3 :
design of girders subjected to eccentric loads. ðA3Þ
 Y. Ren et al. / Engineering Structures 145 (2017) 44–59 59

References [18] Park N, Choi S, Kang Y. Exact distortional behavior and practical distortional
analysis of multicell box girders using an expanded method. Comp Struct
2005;83:1607–26.
[1] Pezeshky P, Mohareb M. Distortional theory for the analysis of wide flange
[19] Park N, Kang YJ, Kim HJ. An independent distortional analysis method of thin-
steel beam. Eng Struct 2014;75:181–96.
walled multicell box girders. Struct Eng Mech 2005;21:275–93.
[2] Senjanovic I, Fan Y. On torsional and warping stiffness of thin-walled girders.
[20] Park N, Choi Y, Kang Y. Spacing of intermediate diaphragms in horizontally
Thin-Walled Struct 1991;11:233–76.
curved steel box girder bridges. Finite Elem Anal Des 2005;41:925–43.
[3] Alghamdi SA. Static and modal analysis of twin-cell box girder structures. AIAA
[21] Zhang L. Influences of diaphragm plate and geometric characteristics on
J 2001;39:1406–10.
distortion effect of steel box girder. J Railway Eng Soc 2013;8:68–73.
[4] Zhang H, Desroches R, Yang ZJ, et al. Experimental and analytical studies on a
[22] Li HF, Luo YF. Application of stiffness matrix of a beam element considering
streamlined steel box girder. J Constr Steel Res 2010;66:906–14.
section distortion effect. J Southeast Univ 2010;26:431–5.
[5] Dabrowski R. Curved thin-walled girders theory and analysis. Cement and
[23] Vlasov VZ. Thin walled elastic beams, 2nd ed. Jerusalem, Israel. Available from
Concrete Association, 1968.
Office of Technical Services, US Department of Commerce, Washington, D.C.:
[6] Boswell LF, Li Q. Consideration of the relationships between torsion, distortion
Israel Program for Scientific Translations; 1961.
and warping of thin-walled beams. Thin-Walled Struct 1995;21:147–61.
[24] Xu X, Qiang SZ. Research on distortion analysis theory of thin-walled box
[7] Li Q. Investigation of the warping functions of thin walled bars with closed
girder. Eng Mech 2013;30:192–201.
cross section. J Southwest Jiaotong Univ 1993;6:1–6.
[25] Xu X, Ye HW, Qiang SZ. Distortional analysis of thin-walled box girder taking
[8] Wright RN, Abdel-Samad SR, Robinson AR. BEF analogy for analysis of box-
account of shear deformation. Chin J Comp Mech 2013;30:860–6.
girders. J Struct Div 1968;94:1719–43.
[26] Harashima M, Usugi K. Distortional analysis of random box girders considering
[9] Hsu YT, Schelling DR. EBEF method for distortional analysis of steel box-girder
the shear deformation. Foreign Bridge 1980;1:1–29.
bridges. J Struct Eng 1995;121:557–66.
[27] Sakai F, Okumura T. Influence of diaphragms on behaviour of box girders with
[10] Hsu YT, Fu CC. Application of EBEF method for the distortional analysis of steel
deformable cross section. Intl Assoc Bridge Struct Eng 1972;9:285–98.
box girder bridge superstructures during construction. Adv Struct Eng
[28] Oleinik JC, Heins CP. Diaphragms for curved box beam bridges. J Struct Div,
2002;5:211–21.
ASCE 1975;101:2161–78.
[11] Zhao ZM, Fang ZZ, Guo JQ. Analysis of continuous box girders with diaphragms
[29] Park N, Kang YJ. Distortional analysis of multicell box girders using 3-
by finite strip method. Bridge Constr 1993;4:35–53.
dimensional shell elements-I. Proposal and application of an expanded
[12] Cheung YK. Finite strip method in structural analysis. Oxford: Pergamon Press;
method. J Korean Soc Civ Eng A 2004;24:557–63.
1976.
[30] Park N, Kang YJ. Distortional analysis of multicell box girders using 3-
[13] Zhao ZM. The calculating analysis for multiple span continuous curved box
dimensional shell elements-II. Distortional analysis based on the expanded
girder by the finite strip method. J Fuzhou Univ (N Sci) 1997;25:90–4.
method. J Korean Soc Civ Eng A 2004;24:565–72.
[14] Zhao ZM. Analysis of continuous curved box-girder bridge with flexible
[31] Lu YZ. Buckling of simply-supported rectangular plates under combined
transverse diaphragms by finite strip method. Comp Struct Mech Appl
actions of shear and bending stresses varying linearly in longitudinal direction.
1993;10:473–84.
J Tongji Univ 1991;19:481–7.
[15] Suetake Y, Hirashima M. Extended trigonometric series analysis of box girders
[32] Fletcher CA. Computational Galerkin methods. Berlin: Springer-Verlag, Berlin
with diaphragms. J Eng Mech. 1997;123:293–301.
Heidelberg; 1984.
[16] Park N, Lim N, Kang Y. A consideration on intermediate diaphragm spacing in
[33] Liu HW, Lin JX, Cao ML. The theory of plates and shells. China: Zhejiang
steel box girder bridges with a doubly symmetric section. Eng Struct
University Press; 1987.
2003;25:1665–74.
[34] Xiao MX. The stability of plates. China: Sichuan Science and Technology Press;
[17] Park N, Choi Y, Yi G, et al. Distortional analysis of steel box girders. Steel Struct
1993.
2002;2:51–8.

View publication stats