Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Journal of Constructional Steel Research 168 (2020) 106005

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Experimental and numerical study on the nonlinear performance


of single-box multi-cell composite box-girder with corrugated
steel webs under pure torsion
Yingbo Zhu a, Shui Wan a,⁎, Kongjian Shen b, Qiang Su a, Muyun Huang a
a
School of Transportation, Southeast University, Nanjing 210096, China
b
Jiangsu Provincial Transportation Engineering Construction Bureau, Nanjing 210004, China

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

Article history: Currently, an increasing number of single-box multi-cell composite box-girders with corrugated steel webs
Received 21 December 2019 (SBMC-CBGCSWs) have been built in China and Japan. However, the torsional stiffness of SBMC-CBGCSW is
Received in revised form 22 February 2020 dramatically decreased due to the employment of thin corrugated steel webs (CSWs). Investigations of SBMC-
Accepted 24 February 2020
CBGCSW under pure torsion are relatively limited, therefore, additional research is needed to investigated to
Available online 5 March 2020
guarantee the safety of such bridges. In this paper, an experimental study on four scaled specimens subjected
Keywords:
to pure torsion is conducted to understand the effect of the number of cells and spacing of CSWs on the full tor-
Single-box multi-cell sional response of SBMC-CBGCSW. The failure mode, cracking patterns, and buckling patterns of specimens are
Corrugated steel webs presented, and a new shear strain relationship between CSWs obtained from experimental results is proposed
Torsional response to evaluate the contribution of inner CSWs to the torsional behavior of SBMC-CBGCSW. Then, finite element anal-
Experiment ysis (FEA) models are created to compare with experimental results, such as torque-twist curves, shear strains of
FEA CSWs and shear strains of concrete slabs. Finally, a parametric study is performed to examine the effect of geo-
Parametric study metric factors, such as the ratio of width to height, ratio of width to span length, ratio of width to thickness of
slab, number of cells and spacing of CSWs, on such bridges under torsion.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction CSWs, respectively, while the torsion needs to be focused by designers,


due to the small torsional stiffness. Even worse, when the wide or
Replacing concrete webs with corrugated steel webs (CSWs), com- curved CBGCSWs are subjected to eccentric loading, the torsion will
posite box-girder with corrugated steel webs (CBGCSW) shows excel- be more prominent and dominant, so the torsional response should be
lent performance over traditional concrete box-girder, with benefits fairly investigated to guarantee their safety. It should be noted that the
such as lighter self-weight, non-cracking of webs, less prestressing full response of CBGCSWs under complex loading should be studied,
loss and so on [1–3]. Hence, since first built in 1986, CBGCSW has however due to the complexity of the torsional effects of such bridges,
been widely constructed worldwide, especially in Japan and China [4]. this paper mainly focuses on the nonlinear behavior of SBMC-
Currently, to meet the rapidly growing traffic demand, more and more CBGCSW under pure torsion, providing a basis for further research.
CBGCSWs employ multi-cell form to provide more traffic lanes. For in- In the past two decades, many researchers [3,6–13] have investi-
stance, more than 30 single-box multi-cell CBGCSWs (SBMC- gated the bending performance and shear performance of such bridges,
CBGCSWs), including 5-cell, 6-cell, 9-cell and 10-cell, have been or are as well as the buckling performance of CSWs, while little attention has
being built in China. Previous literature [5] suggests that the employ- been paid on the torsional behavior of CBGCSW, especially its nonlinear
ment of thin CSWs will lead to a remarkable reduction in torsional stiff- behavior. As a result of the complexity of the torsional response, most of
ness of CBGCSW, with a decrease of approximately 60%–70%. In previous literature mainly concentrated on the linear performance of
practical engineering, bridges are usually subjected to complex stress CBGCSW under antisymmetric loading or eccentric loading. Li et al.
conditions, for example, bending-shear-torsion stress. It is believed [14], Wang [15] and Ma et al. [16] conducted experimental research
that the bending and shear are well resisted by concrete slabs and on scaled single-cell CBGCSWs under eccentric loads, and they mainly
analyzed deflections, twists and shear strains of specimens at the elastic
phase. Meanwhile, they also extended the Second Theory of A. A.
Abbreviations: CSW, corrugated steel web; SBMC-CBGCSW, single-box multi-cell com-
posite box-girder with corrugated steel webs; FEA, finite element analysis.
Umanskii for concrete member to CBGCSW by converting the composite
⁎ Corresponding author. cross section into an equivalent section with the same material. Com-
E-mail address: lanyu421@163.com (S. Wan). pared with experimental results, these improved theories could well

https://doi.org/10.1016/j.jcsr.2020.106005
0143-974X/© 2020 Elsevier Ltd. All rights reserved.
2 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

predict the linear behavior of such bridges. Furthermore, Yang [17] de- investigations on the full response of multi-cell CBGCSW under torsion,
veloped the Second Theory of A. A. Umanskii for curved CBGCSWs by mainly conducted by Shen et al. [4,27,28]. In addition, experimental re-
deducing the torsion differential equation for curved CBGCSWs, and search on the nonlinear performance of SBMC-CBGCSW with more cells
he also analyzed the influence of the radius of curvature and type of (such as 5-cell, 7-cell) and SBMC-CBGCSW with different spacing of
CSW on the torsional behavior of curved CBGCSWs. In addition, Li CSWs has not been conducted until now. Furthermore, whereas the ef-
[18], Wang [15] and Yang et al. [19] investigated parametric studies to fect of fc′, fwy, and tw has been determined by Shen et al (2018a), other
analyze the influence of the ratio of height to span length, ratio of geometric factors, such as the height, length and thickness of concrete
width to span length, thickness of CSW and shape of CSW on the re- slab, also play important roles in the torsional response of multi-cell
sponse of such bridges under eccentric loads. Nevertheless, the greatest CBGCSW, and relevant research has not been introduced. Moreover,
limitation of these studies is that they could only predict the linear per- the biggest issue in the shear strain relationship between CSWs pre-
formance of single-multi CBGCSW under torsion. sented by Shen et al. [4] is that the relationship cannot be applied to
For the nonlinear behavior of single-cell CBGCSW, Mo et al. [20] con- the analysis of the nonlinear performance after yielding of CSWs, as it
ducted an experiment on four scaled specimens subjected to cyclic tor- was obtained from numerical results at the elastic stage.
sion to examine the effect of the prestressing and concrete strength (fc′) The aims of this study are: (1) to understand the influence of the
on the mechanical response of CBGCSW. Considering the contributions number of cells and spacing of CSWs on the torsional response of
of slabs and CSWs to total torque of CBGCSW, he first extended SBMC-CBGCSW in terms of the failure mode, cracking patterns and
rotating-angle softened truss model (RA-STM) for concrete structure buckling patterns, torque-twist curves, shear strains of slabs and shear
to single-box single-cell CBGCSW, which could roughly predict the non- strains of CSWs; (2) to investigate the effect of geometric factors, such
linear torsional performance of the specimens. Then, Ding et al. [21] as the ratio of width to height (B/H), ratio of width to span length (B/
tested four scaled specimens to analyze the impact of fc′, thickness of L), ratio of width to thickness of slab (B/T), number of cells and layout
CSW (tw) and type of CSW on the yield torque and maximum torque of CSWs in the transverse direction, on the yield torque and ultimate
of CBGCSW. Ding et al. [22] also developed a fixed-angle softened torque of SBMC-CBGCSW; (3) to obtain a rational shear strain relation-
truss model (FA-STM) for single-cell CBGCSW under pure torsion, and ship between CSWs that can be used for analyzing the nonlinear perfor-
the comparison between theoretical results and experimental results mance of SBMC-CBGCSW under pure torsion.
showed that the proposed model had good precision. Subsequently,
Ko et al. [23] tested a scaled specimen under pure torsion and simulated
corresponding finite element analysis (FEA) model to analyze the tor- 2. Test program
sional response of CBGCSW. Taking the effect of concrete tension into
account, he also improved a RA-STM for CBGCSW under torsion, 2.1. Specimens
which was in good agreement with torque-twist curves obtained from
experiment and FEA. Next, Shen et al. [24] first extended softened mem- In previous literature, Shen et al. [4] indicated that the innermost
brane model (SMM) for concrete member to CBGCSW without CSW arranged on the symmetric axis of CBGCSW with even-cell (such
prestressing under pure torsion. Shen et al. [25] also developed a new as 2-cell, 4-cell, 6-cell …) has little influence on the ultimate torque;
SMM for CBGCSW with prestressing by considering the influence of ini- thus, CBGCSWs with even-cell (such as 2-cell, 4-cell, 6-cell …) could
tial prestressing on the constitutive laws of concrete and steel bars. The be regarded as CBGCSWs with odd-cell (such as 1-cell, 3-cell, 5-cell
two models were validated by a test on a 2-cell specimen without …). In view of this, this paper mainly studies the torsional performance
prestressing (He suggested that 2-cell CBGCSW could be considered as of odd-cell CBGCSWs. Four scaled specimens with odd-cell are designed,
single-cell CBGCSW) and previous experimental data, respectively. including a 1-cell CBGCSW, a 3-cell CBGCSW with equal spacing of
Later, Zhou et al. [26] improved a SMM by assuming a new relationship CSWs, a 5-cell CBGCSW with equal spacing of CSWs and a 5-cell
between CSW and concrete slab, and this improved model roughly CBGCSW with unequal spacing of CSWs. Corresponding to the number
agreed with the results from previous literature. of cells and spacing of CSWs, these specimens are named S-1, S-3, S-5,
For the nonlinear performance of multi-cell CBGCSW, Shen et al. [4] SU-5, respectively, where the character “S” refers to the specimen and
performed experimental and numerical investigations on two scaled the character “U” stands for unequal spacing of CSWs.
specimens (2-cell and 3-cell) under pure torsion, and the simulated All parameters of the specimens are designed to be the same. The
FEA models in that study compared well with experimental results, spe- size of the cross-section and thickness of CSWs of all specimens are
cifically for the torque-twist curves and shear strains. A parametric appropriately 1/10 of Chaoyang Bridge in Nanchang, China. As
study was also conducted to examine the impact of the fc′, tw, tensile shown in Fig. 1(a)–(d), specimens are 3600 mm long with a rectan-
yield stress of CSW (fwy), number of cells (n) and spacing of CSWs on gular cross section of 2000 × 450 mm, and the thickness of concrete
the torsional response of SBMC-CBGCSW. In addition, based on the slab is 80 mm. To fix the specimens and apply loads on the speci-
shear strains of CSWs obtained from FEA models, a shear strain relation- mens, diaphragms with a 300 mm thickness are arranged at both
ship between CSWs was proposed to estimate the contribution of inner ends of each specimen. ϕ8 mm double-layer steel bars with a spacing
CSWs to the torsional performance of SBMC-CBGCSW. Later, taking into of 150 mm are arranged along the longitudinal direction, while
account the proposed shear strain relationship and a new shear strain ϕ8 mm stirrups with a spacing of 160 mm are arranged along the
relationship between CSW and concrete slab, Shen et al. [27] first devel- transverse direction. In addition, prestressing force is introduced by
oped a unified RA-STM for multi-cell CBGCSW under pure torsion. The arranging ϕs15.2 mm unboned prestressing tendons with a spacing
investigation showed that the proposed unified RA-STM had good of 300 mm. Fig. 1(e)–(f) illustrates the geometric parameters of
agreement with experimental results and numerical results, including CSWs. Embedded connections are adopted into these specimens to
the shear strains and torque-twist curves. In addition, Shen [28] ex- better transmit force between CSWs and concrete slabs, meanwhile
tended SMM to SBMC-CBGCSW by taking into account the shear strain ϕ8 mm transverse steel bars go through the holes of CSWs to
relationship between CSWs and the influence of initial prestressing, enhance the embedded connection.
which could accurately predict the torsional response of multi-cell The properties of all the materials are tested in laboratory. While the
CBGCSW with prestressing. compressive strength of standard concrete cylinder is 28 MPa, the ten-
From the above, research related to the linear behavior of single-cell sile yield strength of CSWs, steel bars and unboned prestressing tendons
CBGCSW under eccentric loading has been broadly investigated, while are 260 MPa, 350 MPa and 1860 MPa, respectively. The effective
some investigators also have carried out experimental and theoretical prestressing force, measured by the pressure sensors arranged at the
studies on its nonlinear performance. By contrast, there are few end of prestressing tendons, is 105 kN (750 MPa).
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 3

Fig. 1. Dimensions of Specimens: (a) Cross Section of S-1; (b) Cross Section of S-3; (c) Cross Section of S-5; (d) Cross Section of SU-5; (e) Side View of CSW; (f) Top View of CSW;
(g) General View of CSW with Steel Bars (Unit: mm).

2.2. Test instrumentation arrangement rotate around the center of the arc shape of the rotating module, called
the rotating end. According to Fig. 2(b), when the center of the arc shape
Fig. 2 illustrates the test setup used for all the specimens. One end of of the rotating module is the same as the center point of the specimen,
the specimen is fixed by the reaction frames, namely, the fixed end, pure torsion of specimen can be achieved. Meanwhile, a hydraulic jack
whereas the other end, supported by the supporting pier, can freely is fixed to the reaction frame, and it can provide the vertical load F for

Fig. 2. Test Setup: (a) General View; (b) Loading Schematic Diagram of Rotating End; (c) Hydraulic Jack.
4 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

the specimen via a load stiffening girder. As illustrated in Fig. 2(b), pure At midspan (Section B), strain rosettes are employed to measure the
torque T applied on the specimen can be calculated by: shear strains of concrete slabs and CSWs. In addition, because the strain
rosettes arranged on the concrete slab may fracture as a result of the
T ¼ Fl ð1Þ wide and dense diagonal cracks, the linear variable differential trans-
formers (LVDTs) arranged in a strain-rosette shape are utilized to
where, F is the concentrated load provided by the hydraulic jack and can gauge the average shear strain of top concrete slab. For the convenience
be measured by the pressure sensor; l is the horizontal distance be- of analysis, the shear strains are defined as positive in this study.
tween the center of rotation and the concentrated load, l = 1650 mm.
As shown in Figs. 3–6, six high-precision tilt sensors (TSs) with an
accuracy of 0.005° are arranged on Section A and Section C of the top 2.3. Loading procedure
slab to gauge the angle of the top slab. In addition, eight displacement
sensors (DSs) with an accuracy of 0.005 times the length of full span As shown in Fig. 7, four specimens are tested at Southeast University.
are set on Section A and Section C of the bottom slab to measure the A force-controlled procedure is adopted for all the specimens. The con-
angle of the bottom slab. It should be noted that both the twists of the trol force is approximately 50 kN-m for each step in the pre-cracking
top slab and bottom slab are regarded as the twist of the specimen in stage, followed by a prescribed force of 30 kN-m for each step before
this study. According to the layout of TSs and DSs, the unit twist of the the external force reaches 75% of the expected ultimate torque, while
specimen can be expressed as: the control force is reduced to 20 kN-m for each step up to the failure
of specimen. When the monotonous force F drops rapidly and spontane-
1h i ously, it is considered to be the failure of the specimen. It should be
θTS ¼ ðβTS−6 −βTS−3 Þ þ ðβTS−5 −βTS−2 Þ þ ðβTS−4 −βTS−1 Þ ð2Þ
3L mentioned that the external force will decrease rapidly due to the
 
spalling of the concrete and the severe buckling of the CSWs, so it is
1 ðμ DS−5 þ μ DS−8 Þ ðμ DS−1 þ μ DS−4 Þ ðμ DS−6 þ μ DS−7 Þ ðμ DS−2 þ μ DS−3 Þ hard to attain experimental data after the failure of specimen occurs.
θDS ¼ − þ −
2L LDS−O LDS−O LDS−I LDS−I
ð3Þ
3. FEA model
where, θTS and θDS are the twists of the top slab and bottom slab gauged
by TSs and DSs, respectively; βTS-N (N = 1,2,3, …,6) is the angle of FEA is powerful for computing the mechanical performance of struc-
gauging-point N on the top slab measured by TS-N; μDS-N (N = 1,2,3, tures, especially in the post-cracking stage, and it is also an essential
…,8) is the displacement of gauging-point N on the bottom slab mea- method for parametric analysis. In this paper, FEA models correspond-
sured by DS-N; L is the distance between Section A and Section C; LDS-I ing to the specimens are simulated by ABAQUS to calculate the nonlin-
is the distance between DS-6 and DS-7 (or DS-2 and DS-3); and LDS-O ear performance of SBMC-CBGCSW under pure torsion. Different from
is the distance between DS-5 and DS-8 (or DS-1 and DS-4). the force-controlled procedure in experiments, the displacement

Fig. 3. Instrumentation of S-1: (a) Layout of TSs and DSs on Section A or C; (b) Strain Rosettes on Section B (Midspan); (c) Layout of TSs, LVDTs and Strain Rosettes on Top Slab; (d) Layout of
DSs, LVDTs and Strain Rosettes on Bottom Slab (Unit: mm).
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 5

Fig. 4. Instrumentation of S-3: (a) Layout of TSs and DSs on Section A or C; (b) Strain Rosettes on Section B (Midspan); (c) Layout of TSs, LVDTs and Strain Rosettes on Top Slab; (d) Layout of
DSs, LVDTs and Strain Rosettes on Bottom Slab (Unit: mm).

Fig. 5. Instrumentation of S-5: (a) Layout of TSs and DSs on Section A or C; (b) Strain Rosettes on Section B (Midspan); (c) Layout of TSs, LVDTs and Strain Rosettes on Top Slab; (d) Layout of
DSs, LVDTs and Strain Rosettes on Bottom Slab (Unit: mm).
6 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Fig. 6. Instrumentation of SU-5: (a) Layout of TSs and DSs on Section A or C; (b) Strain Rosettes on Section B (Midspan); (c) Layout of TSs, LVDTs and Strain Rosettes on Top Slab; (d) Layout
of DSs, LVDTs and Strain Rosettes on Bottom Slab (Unit: mm).

loading mode is adopted in FEA models to conveniently determine the principle of equivalent complementary energy [30]. It is also worth not-
ultimate loading stage. ing that the cracking angle of concrete can be roughly derived from the
value of tensile damaged variable, which will be illustrated later.
For concrete in compression, the Hognestad model [31] for standard
3.1. Material models
concrete cylinder is employed in the study. However, the concrete
stress-strain relationships for CDP model must be linear in the initial
3.1.1. Concrete
stage. In this study, it is assumed that the performance of concrete in
The concrete damaged plasticity (CDP) model, which utilizes the iso-
compression is linear until stress is up to 0.45 fc′, which is within 1/3 fc′
tropic elastic damage theory and isotropic plastic theory to characterize
and 1/2 fc′ recommended by the relevant literature [32]. Thus, the
the behavior of damaged concrete [29], is employed in this study to sim-
Hognestad model employed in this study can be expressed as follows
ulate the nonlinear performance of concrete. In CDP model, damaged
in Eq. (4).
variables, including compressive damaged variable and tensile dam-
aged variable, are defined to quantitatively describe the damaged de-
gree of concrete. In this study, damaged variables are calculated by the 8
>
>
> f0
> Ec ε;
> ε ≤ 0:45 c
>
> Ec
>
>
>
> "     #
>
< ε ε 2 f0 ε
σ c ¼ f 0c 2 ε − ε ; 0:45 c b ≤1 ð4Þ
>
> o o Ec ε o
>
>
> "
>
>
>   #
>
> ε−εo 2 ε
>
: f 0c 1−0:15 εcu −ε o
> ; N1
εo

where, fc′ is the maximum compressive stress of standard concrete cyl-


inder; Ec is the Young's modulus of concrete, Ec = 4730 (fc′)1/2 MPa [33];
εo is the strain corresponding to fc′; and εcu is the ultimate strain of con-
crete, εcu = 0.0038.
For the concrete in tension, a two-stage relationship proposed by
Belarbi and Hsu [34] is utilized in this study. It includes a linear relation-
ship before cracking and a nonlinear curve after cracking, as illustrated
Fig. 7. Overall View of Tested Specimen. in Fig. 8(a), giving:
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 7

3.2. Interactions between concrete and steel

Referring to previous literature [35], the bond-slip relationship be-


tween steel bar and concrete has little influence on the torsional perfor-
mance of concrete structures. Hence, the embedded region constraint is
employed to simulate the connection between concrete slab and steel
bar. In addition, considering the fact that there is no obvious slip phe-
nomenon between slab and CSW during testing, the ends of CSWs at
both length and height directions are also embedded in diaphragms
and concrete slabs by utilizing embedded region constraint.

3.3. Boundary conditions and loading

As shown in Fig. 9, reference point A (RP-A) is created at the center


of the outside cross section at the rotating end; then, the coupling con-
straint in ABAQUS is employed to connect the nodes of the outside sur-
face to RP-A. That means, the motion of the rotating end is the same as
that of RP-A; thus, free rotation of the rotating end can be achieved by
imposing an angle-load on RP-A. Similarly, the coupling constraint is
also employed to the fixed end; and the boundary condition of the
fixed end can be imposed by restraining all degrees of reference point
B (RP-B). In addition, the decreasing temperature method is utilized to
apply the initial prestressing force to all specimens.

3.4. Meshing

In this study, the 3D solid elements and truss elements are employed
to model concrete and steel bars, respectively, while CSWs are simu-
lated by the shell elements due to their small thickness. ABAQUS soft-
ware provides two main integration schemes [29], full integration and
reduced integration, and both of them have benefits and drawbacks.
Using less Gauss points, the elements with reduced integration, such
as C3D8R, C3D20R, S4R and S8R, can improve the computational effi-
ciency. In addition, C3D20, C3D20R, S8R and T3D3 are the quadratic ele-
ments with more nodes, which can improve computational accuracy
but at a cost of computational efficiency.
Based on S-5 specimen, this study compares the effect of different el-
ement types on the torsional performance of SBMC-CBGCSW calculated
by FEA. S-5-B and S-5-D employ the reduced integration elements to
model concrete and CSWs, while S-5-A and S-5-C employ the full inte-
Fig. 8. Constitutive Relationship: (a) Concrete; (b) Steel Bar and CSW.
gration elements. Meanwhile, the quadratic elements with more
nodes are used in S-5-C and S-5-D. The global size of elements in all
FEA models are set to 25 mm. As illustrated in Fig. 10, overall, there is
8 no big difference among torque-twist curves simulated by different in-
< Ec ε;
> ε ≤ εcr tegration schemes. The ultimate torques calculated by S-5-C and S-5-D
σt ¼ ε 0:4 ð5Þ with quadratic elements are relatively conservative, but are almost the
>
: f cr cr ; ε N ε cr same as those of S-5-A and S-5-B. Thus, considering the efficiency re-
ε
lated to computational cost, the concrete and CSWs are modeled by

where, εcr is the cracking strain of concrete, and fcr is the cracking
strength of concrete, fcr = Ec εcr.

3.1.2. Steel
According to the behavior of steel bar and CSW tested in laboratory,
the ideal elastic-plastic model and the elastic strengthened model are
assumed to describe the stress-strain relationships of steel bar and
CSW, respectively. The constitutive laws for steel bar and CSW are illus-
trated in Fig. 8(b), in which fy, fwy and fwu are the yield strength of steel
bar, the tensile yield strength of CSW and the ultimate strength of CSW,
respectively; and εy, εwy and εwu are the strains corresponding to fy, fwy
and fwu, respectively. The Poisson's ratio of both steel bar and CSW is
0.3, and the Young's modulus is taken as 200 GPa in this study. In addi-
tion, the unboned prestressing tendons are assumed to be in the elastic
stage during the whole loading process. Fig. 9. FEA Model for S-5: (a) Overall View; (b) View of the Fixed End.
8 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Fig. 10. The Influence of Different Element Types. Fig. 11. The Mesh Sensitivity Analysis.

In addition, compared with other specimens, specimen S-1 quickly


reduced integration elements, C3D8R and S4R, respectively, whereas the reaches the ultimate torque with marked spalling of bottom concrete
steel bars and unboned prestressing tendons are simulated by 3D truss slab after yielding of CSWs. By contrast, under a specific condition
element T3D2. where the outside CSWs reach their yield strength earlier than steel
In general, the size of mesh has a direct effect on the accuracy of bars, specimens with multi-cell (such as S-3, S-5…) can resist more ex-
FEA, for example, the finer the mesh is, the closer the simulated result ternal torque due to the inner CSWs and steel bars until the failure of
will be to the true value, but it will reduce the computational efficiency. concrete slab occurs.
So a mesh sensitivity analysis is carried out in this study to obtain ap-
propriate global sizes of C3D8R and S4R. Based on S-5 specimen, the
4.1.2. Cracking patterns
global sizes are set to 10 mm, 25 mm, 40 mm, 80 mm, 120 mm and
Fig. 14 presents the cracking patterns of top slabs after testing. When
160 mm in FEA models. As shown in Fig. 11, FEA models with different
external torques reach 66.12%, 59.26%, 54.63% and 46.52% of the ulti-
mesh sizes have the same torsional performance at the elastic stage,
mate torques of S-1, S-3, S-5 and SU-5, respectively, diagonal cracks
while the mesh size has a significant effect on the torsional perfor-
first occur on the surface of bottom slabs. As external torque increases,
mance at the nonlinear phase. It can be clearly seen that the ultimate
diagonal cracks expand in length and width; at the same time, new di-
torque reduces significantly as the size decreases from 160 mm to
agonal cracks continue to appear, which gradually shape a helical-form
40 mm, whereas the torque-twist curves of 25 mm and 10 mm com-
of space truss. After yielding of outside CSWs, almost no new cracks ap-
pare well with each other. That is to say, the simulated results are not
pear, whereas the width of cracks increases dramatically, resulting in a
sensitive to the mesh size when the mesh size is not greater than
significant degradation of concrete stiffness. Per the measurement, the
25 mm. Thus, taking the simulated accuracy and computational effi-
angle of diagonal cracks of top slabs ranges from approximately 32° to
ciency into account, the global sizes of C3D8R and S4R are set to
42°. As mentioned in Section 3.1.1, the tensile damaged variable
25 mm in FEA models.
employed in CDP model can quantitate the degree of the tensile damage
of concrete, and the direction of diagonal cracks can be detected by the
contour lines of the tensile damaged variable. As shown in Fig. 14, the
4. Comparison and discussion
cracking angle of top slabs obtained from tensile damage diagrams in
FEA models is 33°-48°, which is within the range of experimental
4.1. Test phenomenon
results.
4.1.1. General behavior and failure mode
Fig. 12 reflects the torque-twist curves for four specimens. Initially
all specimens perform well without cracks, whereas when reaching
the cracking point, diagonal cracks with an inclination of approximately
35–40° are first observed on the bottom concrete slab, and subsequently
on the top concrete slab. Then, with the increase of external load, these
diagonal cracks develop rapidly in width and length, while more cracks
continue to emerge. Soon the outside CSWs reach the yield point, by
contrast, the steel bars and inner CSWs are still in the elastic stage and
can continue to resist more external loading. At the same time, several
main diagonal cracks are observed on the concrete surface, shaping a
clear helical-form of space truss. As the external torque increases, the
tensile strains of steel bars increase significantly after reaching their
yield strength, while the width of diagonal cracks expands dramatically,
resulting in the spalling of concrete slabs, as shown in Fig. 13. It should
be noted that no obvious spalling of concrete is observed S-3, due to the
irregular prefabrication of the bottom slab in this specimen. The average
thickness of the bottom slab in S-3 is measured to be 85 mm rather than
the designed 80 mm. Fig. 12. Force-Displacement Curves from Experimental Results (TSs).
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 9

Fig. 13. Failure Mode of Specimens: (a) S-1; (b) S-5; (c) SU-5.

4.1.3. Buckling patterns of CSW with a standard deviation of 1.1%. For the torques, the averages of
Fig. 15 presents the buckling patterns of the outside CSWs (CSW-1) Tcr-FEA/Tcr-test, Ty-FEA/Ty-test and Tu-FEA/Tu-test are 1.047, 1.048 and 1.002,
after testing. The angle of buckling on all outside CSWs ranges from respectively, with standard deviations of 0.7%, 2.1% and 2.4%, respec-
approximately 23° to 27°. Obviously, global shear buckling, which tively. For the twists, the averages of θcr-FEA/θcr-test, θy-FEA/θy-test and
is defined as “buckling occurs in multiple wavelengths with the θu-FEA/θu-test are 1.046, 1.177 and 1.091, respectively, with standard devi-
buckling-shape extending over the height of CSW” [36], appears on all ations of 9.5%, 8.8% and 6.4%, respectively. From above, the FEA models
the specimens. As shown in Fig. 15, buckling occurs within almost all simulated in this study are in good agreement with experimental re-
the wavelengths of specimen S-1, whereas buckling only develops sults, which indicates that these FEA models are capable of simulating
within 5, 4 and 3 wavelengths of S-3, S-5 and SU-5. This is because, com- the torsional performance of SBMC-CBGCSW.
pared with single-cell specimen, inner CSWs in multi-cell specimen can Fig. 17 illustrates the torque increments of S-3 and S-5 compared to
share more external torque after yielding of outside CSWs, which can S-1 at key points. It can be clearly seen that the increase in cracking
reduce the buckling degree of outside CSWs. In addition, the larger spac- torque is the slightest, by only approximately 5% growth for S-5 versus
ing between inner CSWs and symmetric axis of box-girder in multi- S-1. By contrast, the ultimate torque rises dramatically with the increase
CBGCSW can greatly improve the torsional resistance of inner CSWs, of the number of cells, by approximately 27% growth for S-5 to S-1. Ac-
hence only slight buckling phenomenon is observed on SU-5 compared cording to the previous study [4], when softened truss model is
to that of S-5. employed to calculate the torsional performance, SBMC-CBGCSW can
be decomposed into several single-box single-cell CBGCSWs with a
4.2. Torque-twist responses common torsional center, as illustrated in Fig. 18. Compared with S-1,
in addition to the torque resisted by the outer box (box-1), the inner
As illustrated in Fig. 12, the relationship between torque (T) and boxes formed by inner CSWs and concrete slabs (such as box-2, box-3)
twist (θ) can generally be divided into four stages: (a) the pre- also contribute to the torque resistance. However, in the pre-cracking
cracking stage, where T is linearly proportionate to θ, and no cracks stage, the shear strains of inner CSWs are rather smaller than those of
are observed in this stage; (b) the elastic-plastic stage, where when T outside CSWs, plus the geometric parameter (that is, the width) of
reaches approximately 46%–66% of the ultimate torque, the concrete inner boxes is also smaller; therefore, the contribution of inner CSWs
stiffness begins to degrade due to the occurrence of diagonal cracks, to the cracking torque is negligible. By contrast, after cracking of con-
whereas CSWs and steel bars are still in the elastic stage. In this stage, crete, especially after yielding of outside CSWs, the inner CSWs are
external torque is mainly resisted by CSWs and space truss formed by still in the elastic stage and can continue to bear more external force,
steel bars and concrete struts; (c) the yield stage, where when external thus leading to a remarkable increase of the yield torque and ultimate
torques reach 80.78%, 74.63%, 72.86% and 70.72% of the ultimate torques torque.
of S-1, S-3, S-5 and SU-5, outside CSWs first reach their yield strength In addition, compared with S-5, the cracking torque, yield torque
compared to steel bars. After yielding of outside CSWs, steel bars and and ultimate torque of SU-5 increase by 4.98%, 16.69% and 20.22%, re-
inner CSWs can continue to bear more external torque until the failure spectively. This is because, a greater distance between inner CSW and
of concrete; (d) the decline stage, where after reaching the peak, the symmetric axis will form a larger inner box and also lead to a greater
curves drop sharply due to the failure of concrete. shear strain of inner CSW, so it can contribute more to the torsional per-
Fig. 16 compares the T-θ curves obtained from experiments and FEA formance of SBMC-CBGCSW. Hence, in practice, setting the proper spac-
models. It can be clearly seen that the T-θ curve of each specimen com- ing of CSWs can achieve better torsional behavior, especially improving
pares well with that of FEA model. In addition, the precision of FEA the yield torque and ultimate torque strikingly.
models in terms of the initial torsional rigidity, the torques and the
twists at key points is indicated in Tables 1–3. In this study, the initial
torsional rigidity K is defined as follows in Eq. (6). 4.3. Shear strain of concrete

0:5T cr Fig. 19 presents the relationship between torque (T) and shear strain
K¼ ð6Þ
θ0:5−cr of concrete slabs (γc) obtained from experiments and FEA models. A lin-
ear behavior of concrete with high torsional rigidity in the initial stage
where, θ0.5-cr is the twist corresponding to 0.5Tcr, and Tcr is the cracking can be observed, whereas with the further increase of load, γc suffers a re-
torque. For the initial torsional rigidity, the average of KFEA/Ktest is 1.059, markable increase as a result of concrete degradation caused by diagonal
10 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Fig. 14. Cracking Pattern of Top Slabs: (a) S-1; (b) S-3; (c) S-5; (d) SU-5.

cracks, rising to about 9000 με when the external torque reaches the peak. 4.4. Shear strain of CSW
In the elastic stage, the shear strains gauged by strain rosettes compare
well with the average shear strains simulated by FEA models (corre- Fig. 20 shows the curves of torque (T) versus shear strain of CSW
sponding to the positions of the strain rosettes arranged on concrete (γw) obtained from experiments and FEA models. The average shear
slabs), whereas there is no available experimental data after cracking strains calculated by FEA models (corresponding to the positions of
due to the fracture of the strain rosettes. Moreover, the average shear the strain rosettes set on CSWs) agree well with that measured by strain
strains measured by the LVDTs are in good agreement with the simulated rosettes in the elastic stage, whereas the simulated shear strains are
results both in the pre-cracking stage and post-cracking stage. This indi- generally larger than those measured in the experiments after yielding
cates that the CPD model of concrete employed in this study can satisfac- of CSWs, resulting in deviations between the curves of experiments
torily simulate the nonlinear behavior of concrete. and FEA models.
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 11

Fig. 15. Buckling Pattern of CSW-1(a) S-1; (b) S-3; (c) S-5; (d) SU-5.

pffiffiffi
In this study, the yield of outside CSWs is considered as the yield where, τwy is the yield strength of CSW, τwy = fwy/ 3, in which fwy is
point of the specimen, so the yield torque can be determined by the the tensile yield stress of CSW; Ge is the effective shear modulus of
value of the shear strains of outside CSWs. According to the orthotropic CSW, Ge = Gs(aw + bw)/(aw + cw), in which Gs is the shear modulus
plate theory, the yield strain of CSW (γwy) can be computed by Eq. (7): of steel, aw, bw and cw are the length of the flat plate, the length of hor-
izontal projection of the inclined panel and the length of the inclined
τwy panel, respectively. Thus, it can be calculated that when γwy reaches
γ wy ¼ ð7Þ
Ge 2097 με, specimens reach their yield points. In addition, Fig. 20 shows

Fig. 16. Comparisons of T-θ Curves Obtained from Experiments and FEA Models.
12 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Table 1
Comparison of initial torsional rigidity between experiment and FEA.

Specimens Ktest (kN-m2) KFEA (kN-m2) K FEA


K test
S-1 175,383 187,062 1.067
S-3 187,966 199,873 1.063
S-5 200,394 208,179 1.039
SU-5 214,397 228,414 1.065
Average 1.059
Standard deviation 0.011

that when S-5 and SU-5 reach their ultimate points, CSW-2 and CSW-5
in S-5 and SU-5 have reached their yield strains, whereas the innermost
CSWs, CSW-3 and CSW-4, are still in the elastic stage, only at approxi-
mately 800 με and 1200 με, respectively.
Compared with the outside CSWs, the shear strains of inner CSWs,
especially the innermost CSWs, are rather small. For instance, the aver-
Fig. 17. Percentage Change in Cracking, Yield and Ultimate Torques of S-3 and S-5
age shear strain ratios of CSW-3, CSW-2 and CSW-1 in specimens S-5 and
Compared to S-1.
SU-5 at the yield point are approximately 6:25:100 and 16:40:100, re-
spectively. That is to say, before reaching the yield torque, the external
force is mainly resisted by slabs and outside CSWs, whereas the contri-
bution of inner CSWs is limited. Meanwhile, compared with that of S-5, the average shear strains of outside CSWs at the ultimate point decrease
the higher shear strain ratio of inner CSWs to outside CSWs in SU-5 with the sequences of S-1, S-3, S-5 and SU-5, which is consistent with
(such as, 16:100 versus 6:100, and 40:100 versus 25:100) indicates the buckling degree of the specimens (Fig. 15).
that the farther the distance between inner CSW and symmetric axis
is, the greater the contribution of inner CSWs to torsional resistance of 4.5. A new shear strain relationship between CSWs
SBMC-CBGCSW is, which is consistent with the previous analysis in
Section 4.2. Thus, setting a larger distance between inner CSW and sym- As shown in Fig. 18, according to Shen et al. [27], the torsional perfor-
metric axis can fully utilize the innermost CSW and achieve better tor- mance of the outer box (box-1) is first calculated by softened truss
sional behavior. model; then, the contribution of inner CSWs can be evaluated by assum-
Furthermore, Fig. 20 also reflects that the shear strains of inner CSWs ing a shear strain relationship between CSWs. Hence, a key to evaluating
increase significantly after yielding of outside CSWs. This can be ex- the torsional performance of multi-cell CBGCSW is the accurate shear
plained by the fact that when reaching the yield strength, the outside strain relationship between CSWs. However, the relationship recom-
CSWs cannot bear more external load, whereas the inner CSWs are mended by Shen et al. [4] is obtained from the numerical results at the
still in the elastic stage and begin to play an important role in improving elastic phase, which means that it cannot be used in nonlinear analysis.
the torsional resistance of SBMC-CBGCSW. It is also worth noting that Moreover, the relationship proposed by him is related only to the

Table 2
Comparison of torques at key points between experiment and FEA.

Specimens Cracking state Yielding of CSW Ultimate state

Tcr-test (kN-m) Tcr-FEA (kN-m) T cr‐ FEA Ty-test (kN-m) Ty-FEA (kN-m) T y‐ FEA Tu-test (kN-m) Tu-FEA (kN-m) T u‐ FEA
T cr‐test T y‐test T u‐test

S-1 307.89 319.80 1.039 376.10 392.51 1.044 465.61 478.27 1.027
S-3 314.23 332.59 1.058 395.71 402.91 1.018 530.25 516.77 0.975
S-5 323.79 339.02 1.047 431.87 464.19 1.075 592.70 607.95 1.026
SU-5 331.55 346.04 1.044 503.93 532.03 1.056 712.56 698.88 0.981
Average 1.047 1.048 1.002
Standard deviation 0.007 0.021 0.024

cr: cracking of concrete; y: yielding of CSW; u: ultimate stage.

Table 3
Comparison of twists at key points between experiment and FEA.

Specimens Cracking state Yielding of CSW Ultimate state

θcr-test (°/m) θcr-FEA (°/m) θcr‐FEA θy-test (°/m) θy-FEA (°/m) θy‐FEA θu-test (°/m) θu-FEA (°/m) θu‐FEA
θcr‐test θy‐test θu‐test

S-1 0.135 0.152 1.125 0.308 0.407 1.322 1.821 2.072 1.138
S-3 0.144 0.164 1.138 0.371 0.411 1.106 1.609 1.758 1.093
S-5 0.134 0.137 1.017 0.430 0.506 1.175 1.657 1.634 0.986
SU-5 0.127 0.115 0.901 0.558 0.616 1.106 1.603 1.841 1.148
Average 1.046 1.177 1.091
Standard deviation 0.095 0.088 0.064

cr: cracking of concrete; y: yielding of CSW; u: ultimate stage; θcr-test, θy-test and θu-test: the average twist of the twists measured by TSs and DSs.
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 13

   2  3 !
1 Rdi R R
¼ 33:46 þ 1194:5 −23917:5 di þ 96700 di θ
Rγi n n n
   2  3
R R R
þ 44:2−1243:5 di þ 12057:5 di −37425 di ð10Þ
n n n

5. Parametric study

Based on an experiment on two specimens with 2-cell and 3-cell,


Shen et al. [4] conducted a parametric study on the torsional perfor-
mance of SBMC-CBGCSW affected by fc′, fwy, tw, number of cells (n)
and spacing of CSWs (Rdi). The results indicate that within a reasonable
range, the ultimate torque of SBMC-CBGCSW is linearly proportional to
the value of fc′, fwy, tw and n. Furthermore, in 3-cell CBGCSW, the ultimate
torque improves with the increase of Rd2 in the given range of 0 to 0.8,
then experiences a dramatic drop up to Rd2 = 1.
However, other influencing factors, such as the height (H), length (L)
and thickness of slab (T), also influence the torsional response of SBMC-
CBGCSW to some extent. Therefore, this paper mainly focuses on the
impact of geometrical parameters, including the ratio of width to height
(B/H), ratio of width to span length (B/L) and ratio of width to thickness
of slab (B/T). In addition, this study simulates 7-cell and 9-cell CBGCSWs
to better understand the influence of the number of cells, and also ex-
amines the impact of the spacing of CSWs on 5-cell CBGCSW in order
to attain the optimum positions of CSWs.
Fig. 18. Equivalent Schematic of SBMC-CBGCSW Under Pure Torsion.

5.1. Effect of the ratio of width to height

distance between CSW and symmetric axis, while in fact, it is also To examine the impact of the ratio of width to height (B/H) on the
closely related to the twist and the number of cells. Therefore, this torsional response of 5-cell CBGCSW, keeping the width (B =
study proposes a new shear strain relationship between CSWs based 1800 mm) unchanged, the height of H = 300 mm - 600 mm is selected
on experimental data in order to accurately predict the nonlinear be- in FEA models based on specimen S-5; that is, B/H = 6–3 is set in this
havior of SBMC-CBGCSW. study. From Fig. 22, it can be clearly observed that the yield torque
To quantify the shear strain relationship between CSWs, the relative and ultimate torque decrease with the increase of B/H. It is also worth
shear strain index is defined as: noting that the relationship between the change of torque and B/H is
roughly inversely proportional, that is to say, with the increase in
height, the yield torque and ultimate torque grow approximately
γ wi
Rγi ¼ ð8Þ linearly.
γw1

5.2. Effect of the ratio of width to span length


where, Rγi (i = 2,3…) is the relative shear strain index of CSW-i; γwi and
γw1 are the average shear strains of CSW-i (i = 2,3…) and CSW-1, re- Likewise, keeping the width (B = 1800 mm) unchanged, the span
spectively. To conveniently quantify the impact of the spacing of CSWs length is taken as L = 2800 mm–5200 mm in FEA model of specimen
on the torsional performance of SBMC-CBGCSW, the relative distance S-5 to study the effect of B/L on the torsional response of 5-cell CBGCSW;
index of CSW (Rdi) is defined by Eq. (9) that is, the ratio of width to span length at B/L = 0.64–0.35 is set in this
study. Fig. 23 illustrates that B/L has little impact on the yield torque,
whereas the ultimate torque improves linearly with increasing B/L in
ðBi =2Þ
Rdi ¼ ð9Þ the range of 0.38 to 0.6 and then remains roughly unchanged when B/
ðB1 =2Þ
L reaches a given degree.

where, Rdi (i = 2,3) is the relative distance index of CSW-i; Bi/2 (i = 5.3. Effect of the ratio of width to thickness of slab
2,3…) and B1/2 are the distances from the symmetric axis of CBGCSW
to CSW-i (i = 2,3…) and CSW-1, respectively. Based on the shear strains To better understand the effect of the ratio of width to thickness of
of CSWs obtained from specimens S-3, S-5 and SU-5, the new shear concrete slab (B/T), based on FEA model of specimen S-5, keeping the
strain relationship between CSWs is presented as Eq. (10). It is worth width (B = 1800 mm) unchanged, the thickness of concrete slab T
noting that the new Rγi is in terms of the distance between CSWs ranges from 60 mm to 120 mm; that is, B/T = 15–30 is set in this
(Rdi), twist (θ) and number of cells (n), which can be applied in both lin- study. As shown in Fig. 24, the yield torque improves slightly with in-
ear and nonlinear analysis. As indicated in Fig. 21, the new shear strain creasing B/T, whereas the ultimate torque decreases linearly as the in-
relationship agrees well with experimental results with an adjusted R- crease of B/T. This means that under the condition where the height
square value of 0.997. It should be mentioned that the proposed shear (H) and width (B) of CBGCSW are unchanged, the employment of a
strain relationship does not take into account the effect of material thinner concrete slab will lead to a decrease in ultimate torque; how-
properties of CSWs. ever, the height of CSW will increase in turn, improving the yield torque.
14 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Fig. 19. Comparisons of Shear Strain of Concrete Slabs between Experiments and FEA Models.

5.4. Effect of the number of cells the innermost CSW to the outside CSW, is chosen based on FEA model
of S-5 to understand its effect on the torsional performance of SBMC-
Based on specimens S-1, S-3 and S-5, FEA models of 7-cell and 9- CBGCSW. It is worth mentioning that when Rd2 = 0.2, the 5-cell
cell with an equal spacing of CSWs are simulated to examine the im- CBGCSW will become a 3-cell CBGCSW with 4 mm innermost CSWs,
pact of the number of cells on the performance of SBMC-CBGCSW whereas when Rd2 = 1, the 5-cell CBGCSW can be regarded as a 3-cell
under torsion. It can be clearly seen from Fig. 25 that the yield torque CBGCSW with 4 mm outside CWSs.
and ultimate torque improve linearly as the number of cells increases As shown in Fig. 26(a), it can be found that the ultimate torque is
within the range of 3-cell to 9-cell, whereas the increment trends in linear proportion to Rd2 until Rd2 reaches 0.9, while after that, the
of the yield torque and ultimate torque are smaller before 3-cell. ultimate torque undergoes a significant drop as a result of the change
Compared with 1-cell, the yield torque and ultimate torque of 9-cell in the number of cells. Compared with Rd2 = 0.2, the ultimate torque
improve by approximately 60% and 40%, respectively, which suggests of Rd2 = 0.9 increases by approximately 40%, which suggests that de-
that the number of cells has a significant impact on the torsional signing the transverse position of CSW-2 properly could obtain better
performance of CBGCSW. torsional resistance for 5-cell CBGCSW. In addition, Fig. 26(a) also
shows that the ultimate torque of Rd2 = 0.6 is about the same as that
5.5. Effect of spacing of CSWs in the transverse direction of Rd2 = 1. That is to say, when the total thickness of all CSWs is un-
changed, setting a higher thickness of outside CSW in 3-cell CBGCSW
5.5.1. Effect of the transverse position of CSW-2 in 5-cell CBGCSW can also achieve better torsional performance than 5-cell CBGCSW, if
Keeping Rd3 in S-5 (Rd3 = 0.2) unchanged, Rd2 = 0.2, 0.4, 0.6, 0.8, 0.9 the transverse positions of inner CSWs in 5-cell CBGCSW are not set
and 1, that is, the transverse position of CSW-2 (or CSW-5) varies from appropriately.
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 15

Fig. 20. Comparisons of Shear Strain of CSWs between Experiments and FEA Models.

5.5.2. Effect of the transverse position of CSW-3 in 5-cell CBGCSW


Similarly, keeping Rd2 in SU-5 (Rd2 = 7/9) unchanged, Rd3 = 0, 2/9,
4/9, 1/2, 5/9, 6/9 and 7/9, that is, the transverse position of CSW-3 (or
CSW-4) varies from the symmetric axis to CSW-2 (or CSW-5), is involved
based on FEA model of SU-5. When Rd3 = 0 or 7/9, the 5-cell CBGCSW
can be considered as a 3-cell CBGCSW. As illustrated in Fig. 26(b), the ul-
timate torque increases as Rd3 increases in the initial stage, and then
reaches maximum when Rd3 is about in the range of 1/2 to 5/9. By con-
trast, the ultimate torque will experience a marked decrease of approx-
imately 10% when Rd3 changes from Rd3 = 5/9 to Rd3 = 7/9 due to the
change in the number of cells. Compared with Rd3 = 0, the ultimate
torque of Rd3 = 5/9 increases by 13.8%, which indicates that the contri-
bution of the transverse position of CSW-3 to the ultimate torque of 5-
cell CBGCSW is smaller in comparison with that of CSW-2.
Similar to Section 5.5.2, the ultimate torque of Rd3 = 7/9 (3-cell
CBGCSW with 4 mm inner CWSs) is approximately the same as that of
Rd3 = 0.2, which also indicates that when the total thickness of all
CSWs is unchanged, the pure torsional performance of 3-cell CBGCSW
with a higher thickness of inner CSWs can be preferable to 5-cell
CBGCSW in some cases. Thus, it is necessary to determine the best posi-
Fig. 21. Comparison of the Proposed Shear Strain Relationship between CSWs with
tions of the inner CSWs in 5-cell CBGCSW to achieve the best torsional
Experimental Results. behavior.
16 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

Fig. 22. Effect of Ratio of Width to Height: (a) T-θ Curves; (b) Percentage Change in Yield and Ultimate Torque.

Fig. 23. Effect of Ratio of Width to Span: (a) T-θ Curves; (b) Percentage Change in Yield and Ultimate Torque.

Fig. 24. Effect of Ratio of Width to Thickness: (a) T-θ Curves; (b) Percentage Change in Yield and Ultimate Torque.

Fig. 25. Effect of Number of Cells: (a) T-θ curves; (b) Percentage Change in Yield and Ultimate Torque.
Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005 17

Fig. 26. Effect of Spacing of CSW in 5-cell CBGCSW: (a) Effect of Transverse Position of CSW-2; (b) Effect of Transverse Position of CSW-3; (c) Optimum Positions of Inner CSWs.

5.5.3. Optimum positions of inner CSWs symmetric axis, which indicates that setting the proper spacing of
Referring to the effect of the transverse position of CSW-2 (that is, CSWs can improve the torsional capacity of SBMC-CBGCSW;
Fig. 26(a)), the distance index of CSW-2 is taken as Rd2 = 0.7, 0.8, 0.9 4. A new shear strain relationship between inner CSW and outside CSW
and 1 in FEA models based on SU-5, while the value of Rd3 is assumed to obtained from experimental results, which is in terms of the distance
be 0 to Rd2. The effect of the transverse positions of CSW-2 and CSW-3 between CSWs, twist and number of cells, is proposed in this study to
on the torsional performance of 5-cell CBGCSW is shown in Fig. 26(c). estimate the contribution of the inner CSWs to the torsional perfor-
It can be clearly seen that when Rd2 = 0.9 and Rd3 = 0.5–0.55, the mance of SBMC-CBGCSW. The new relationship can be applied in
ultimate torque of 5-cell CBGCSW reaches a maximum. Similar to the both linear and nonlinear analysis;
curves of Rd2 = 0.778 (Fig. 26(b)), the ultimate torques of Rd2 = 0.7, 0.8 5. The predictions of FEA agree well with experimental results, includ-
and 0.9 also reach a peak when Rd3 is between 1/2 and 5/9. Meanwhile, ing the torque-twist curves, shear strains of concrete slabs and shear
Fig. 26(c) also indicates that when Rd3 is in the given range of 0 and 0.7, strains of CSWs, which indicates that the FEA models simulated in
the ultimate torque improves as Rd2 increases from 0.7 to 0.9, then the ul- this study are capable of predicting the nonlinear performance of
timate torque will decrease significantly until Rd2 = 1, which shows a SBMC-CBGCSW under pure torsion;
similar trend as the curve of Rd3 = 0.2 (Fig. 26(a)). In addition, in the 6. The parametric study suggests: (a) the yield torque and ultimate
curve of Rd2 = 1, in which the 5-cell CBGCSW turns into a 3-cell torque of 5-cell CBGCSW are roughly inversely proportional to B/H;
CBGCSW, the ultimate torque reaches a maximum when Rd3 = 0.8, (b) the ultimate torque improves linearly with the increase of B/L in
which is consistent with the study conducted by Shen et al. [4]. a certain range, whereas B/L has little impact on the yield torque;
Overall, when Rd2 = 0.9 and Rd3 = 1/2–5/9, single-box 5-cell CBGCSW (c) while the ultimate torque decreases linearly with increasing B/T,
will achieve optimum torsional performance. However, when the spac- the yield torque improves slightly with increasing B/T; (d) the yield
ing between CSWs is narrow, the ultimate torque of 5-cell CBGCSW will torque and ultimate torque linearly improve with the increase of the
decrease as a result of the change in the number of cells. number of cells in given range; (e)when Rd2 = 0.9 and Rd3 = 1/2–
5/9, 5-cell CBGCSW will achieve optimum torsional performance.
6. Conclusions

To better understand the influence of the number of cells and the


spacing of CSWs on the response of SBMC-CBGCSW subjected to pure Declaration of Competing Interest
torsion, this paper has carried out experimental and numerical investiga-
tions on four scaled specimens: 1-cell, 3-cell with equal spacing of CSWs, None.
5-cell with equal spacing of CSWs, and 5-cell with unequal spacing of
CSWs. The FEA results, such as torque-twist curves, shear strains of Acknowledgements
slabs, and shear strains of CSWs, are examined by experimental results.
In addition, a parametric study is performed to investigate the effect of This work was supported by the National Natural Science Founda-
influencing factors, such as the ratio of width to height (B/H), ratio of tion of China [grant 51878151]. The authors wish to express apprecia-
width to span length (B/L), ratio of width to thickness of slab (B/T), num- tion for the research team for their help in the experiments.
ber of cells (n) and spacing of CSWs (Rdi), on the torsional performance of
SBMC-CBGCSW. The main conclusions are generalized as follows: References
1. All specimens exhibit ductile failure, and the failure of specimens is [1] M. Cheyrezy, J. Combault, Composite bridges with corrugated steel webs: achieve-
due to the spalling of concrete. The angle of diagonal cracks approx- ments and prospects, Proceedings of the IABSE Symposium on Mixed Structures in-
cluding New Materials, Brussels 1990, pp. 479–484.
imately ranges from 32° to 42°, while the angle of buckling of all
[2] M. Rosignoli, Prestressed concrete box girder bridges with folded steel plate webs,
specimens ranges from 23° to 27°. The degree of buckling of CSWs Proc. Inst. Civil Eng. Struct. Build. 134 (1999) 77–85, https://doi.org/10.1680/istbu.
is reduced with the increase of the number of cells and the distance 1999.31255.
[3] J. He, Y.Q. Liu, A.R. Chen, T. Yoda, Mechanical behavior and analysis of composite
between inner CSW and symmetric axis of box-girder;
bridges with corrugated steel webs: state-of-the-art, Int. J. Steel. Struct. 12 (2012)
2. Compared with single-cell CBGCSW, multi-cell CBGCSW can con- 321–338, https://doi.org/10.1007/s13296-012-3003-9.
tinue to resist more external torque after yielding of the outside [4] K.J. Shen, S. Wan, Y.L. Mo, A.M. Song, X.Y. Li, Behavior of single-box multi-cell box-
CSWs, as inner CSWs are still in the elastic stage. Inner CSWs have girders with corrugated steel webs under pure torsion. Part I: experimental and nu-
merical studies, Thin-Walled Struct. 129 (2018) 542–557, https://doi.org/10.1016/j.
limited contribution to the cracking torque, while the yield torque tws.2017.10.038.
and ultimate torque of SBMC-CBGCSW can be significantly improved [5] Prestressed Concrete Technology Association, Design and Construction Standards of
with an increase in the number of cells; Composite Bridges, Gihodo Shuppan Co. Ltd., Tokyo, 2005.
[6] M. Elgaaly, A. Seshadri, R.W. Hamilton, Bending strength of steel beams with corru-
3. Compared with S-5, the yield torque and ultimate torque of SU-5 im- gated webs, J. Struct. Eng. 123 (1997) 772–782, https://doi.org/10.1061/(ASCE)
prove strikingly due to the greater distance between inner CSW and 0733-9445(1997)123:6(772).
18 Y. Zhu et al. / Journal of Constructional Steel Research 168 (2020) 106005

[7] Y.A. Khalid, C.L. Chan, B.B. Sahari, A.M.S. Hamouda, Bending behaviour of corrugated [22] Y. Ding, K.B. Jiang, Y.Z. Zhou, J.K. Yang, Analytical model for torsional strength of
web beams, J. Mater. Process. Technol. 150 (2004) 242–254, https://doi.org/10. prestressed concrete box-girder with corrugated steel webs, China J. Mech. Mater.
1016/j.jmatprotec.2004.02.042. 30 (2013) 137–142.
[8] J. He, Y. Liu, A. Chen, T. Yoda, Shear behavior of partially encased composite I-girder [23] H.J. Ko, J. Moon, Y.W. Shin, H.E. Lee, Non-linear analyses model for composite box-
with corrugated steel web: experimental study, J. Constr. Steel Res. 77 (2012) girders with corrugated steel webs under torsion, Steel Compos. Struct. 14 (2013)
193–209, https://doi.org/10.1016/j.jcsr.2012.05.005. 409–429, https://doi.org/10.12989/scs.2013.14.5.409.
[9] J.G. Nie, L. Zhu, M.X. Tao, L. Tang, Shear strength of trapezoidal corrugated steel [24] K.J. Shen, S. Wan, Y.L. Mo, X.Y. Li, A.M. Song, A softened membrane model for com-
webs, J. Constr. Steel Res. 85 (2013) 105–115, https://doi.org/10.1016/j.jcsr.2013. posite box-girders with corrugated steel webs under pure torsion, Eng. Struct. 173
02.012. (2018) 357–371, https://doi.org/10.1016/j.engstruct.2018.07.021.
[10] T. Guo, R. Sause, Analysis of local elastic shear buckling of trapezoidal corrugated [25] K.J. Shen, S. Wan, Y.L. Mo, X.Y. Li, A softened membrane model for prestressed con-
steel webs, J. Constr. Steel Res. 102 (2014) 59–71, https://doi.org/10.1016/j.jcsr. crete composite box girders with corrugated steel webs under pure torsion, Adv.
2014.06.006. Struct. Eng. 22 (2018) 384–401, https://doi.org/10.1177/1369433218788597.
[11] B. Kövesdi, B. Jáger, L. Dunai, Bending and shear interaction behavior of girders with [26] C. Zhou, L.F. Li, L.H. Wang, Improved softened membrane model for prestressed
trapezoidally corrugated webs, J. Constr. Steel Res. 121 (2016) 383–397, https://doi. composite box girders with corrugated steel webs under pure torsion, J. Constr.
org/10.1016/j.jcsr.2016.03.002. Steel Res. 153 (2019) 372–384, https://doi.org/10.1016/j.jcsr.2018.10.023.
[12] M. Leblouba, S. Barakat, Z. Al-Saadon, Shear behavior of corrugated web panels and [27] K.J. Shen, S. Wan, Y.L. Mo, Z.W. Jiang, X.Y. Li, Behavior of single-box multi-cell box-
sensitivity analysis, J. Constr. Steel Res. 151 (2018) 94–107, https://doi.org/10.1016/ girders with corrugated steel webs under pure torsion. Part II: theoretical model
j.jcsr.2018.09.010. and analysis, Thin-Walled Struct. 129 (2018) 558–572.
[13] W. Zhang, M. Mahdavian, C. Yu, Lateral strength and deflection of cold-formed steel [28] K.J. Shen, Study on Torsional Behavior of Single-Box Multi-Cell Composite Box
shear walls using corrugated sheathing, J. Constr. Steel Res. 148 (2018) 399–408, Girders with Corrugated Steel Web, PhD Thesis, Southeast University, Nanjing,
https://doi.org/10.1016/j.jcsr.2018.06.009. 2018.
[14] H.J. Li, J.S. Ye, S. Wan, W.Q. Wu, Analysis and experimental study of torsion and dis- [29] ABAQUS, Software Package, ABAQUS Help System, Dassault Systèmes Simulia Co.,
tortion of box girder with corrugated steel webs, Bridge Constr. 33 (2003) 1–4, Providence, RI, 2016.
https://doi.org/10.3969/j.issn.1003-4722.2003.06.001. [30] F. Sidoroff, Description of anisotropic damage application to elasticity, in: J. Hult, J.
[15] W. Wang, Model Test Study and Analysis on Torsion Effect and Distortion Effect of Lemaitre (Eds.), Physical Non-Linearities in Structural Analysis, Springer, Berlin Hei-
Box-Girder with Corrugated Steel Webs, MSc Thesis, Hunan University, Hunan, delberg 1981, pp. 237–244, https://doi.org/10.1007/978-3-642-81582-9_35.
2008. [31] E. Hognestad, Study of Combined Bending and Axial Load in Reinforced Concrete
[16] L. Ma, S. Wan, Z.W. Jiang, Research on torsion and distortion performance of single Members, Champaign University of Illinois Engineering Experiment Station, Chi-
box double-cell girder with corrugated steel webs, J. Highw. Transp. Res. 29 (2016) cago, 1951.
77–85. [32] Y.L. Lu, L.P. Ye, Z.W. Miao, Seismic and Plastic Analysis of Buildings, Principles,
[17] B.W. Yang, Properties of the PC Curved Box-Girder Bridge with Corrugated Steel Models and Practices in ABAQUS, MSC, MARC and SAP2000, China Architecture &
Webs, PhD Thesis, Southeast University, Nanjing, 2013. Building Press, Beijing, 2009.
[18] H.J. Li, Experimental Study and Analysis On Torsion and Distortion of Box-Girder [33] A.C.I. Committee, Building Code Requirements for Structural Concrete (ACI 318–08)
with Corrugated Steel Webs, PhD Thesis, Southeast University, Nanjing, 2003. and Commentary, American Concrete Institute, 2008.
[19] B.W. Yang, Y.L. Li, S. Wan, Stress analysis of box girder with corrugate steel webs [34] A. Belarbi, T.T.C. Hsu, Constitutive laws of concrete in tension and reinforcing bars
under torsion, J. S. China Univ. of Technol. 40 (2012) 19–22, https://doi.org/10. stiffened by concrete, ACI Struct. J. 91 (1994) 465–474.
3969/j.issn.1000-565X.2012.02.004. [35] T.G. Mondal, S.S. Prakash, Nonlinear finite-element analysis of RC bridge columns
[20] Y.L. Mo, C.H. Jeng, Y.S. Chang, Torsional behavior of prestressed concrete box-girder under torsion with and without axial compression, J. Bridg. Eng. 21 (2016),
bridges with corrugated steel webs, ACI Struct. J. 97 (2000) 849–859. 04015037. https://doi.org/10.1061/(ASCE)BE.1943-5592.0000798.
[21] Y. Ding, K.B. Jiang, F. Shao, A.Z. Deng, Experimental study on ultimate torsional [36] J. Yi, H. Gil, K. Youm, H. Lee, Interactive shear buckling behavior of trapezoidally cor-
strength of PC composite box-girder with corrugated steel webs under pure torsion, rugated steel webs, Eng. Struct. 30 (2008) 1659–1666, https://doi.org/10.1016/j.
Struct. Eng. Mech. 46 (2013) 519–531, https://doi.org/10.12989/sem.2013.46.4.519. engstruct.2007.11.009.

You might also like