Journal of Constructional Steel Research 189 (2022) 107074

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Seismic performance of square concrete-filled steel tubular columns with

diagonal binding ribs
Zheng Zhou a, b, Dan Gan b, *, Mark D. Denavit c, Xuhong Zhou a, b
a
College of Civil Engineering, Hunan University, Changsha 410082, China
b
School of Civil Engineering, Chongqing University, Chongqing 400045, China
c
Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, TN 37996, United States

A R T I C L E I N F O A B S T R A C T

Keywords: The structural performance of square concrete-filled steel tubular (SCFST) columns can be improved by stiffening
Concrete-filled steel tube the tube. Stiffening delays local buckling and can improve composite action while also facilitates the use of
Local buckling thinner and higher-strength materials. A particularly efficient form of stiffened SCFST column using diagonal
Diagonal ribs
binding ribs was proposed recently. Initial studies on the axial behavior of these diagonal ribs stiffened SCFST
Seismic performance
Plastic stress distribution method
columns have shown excellent performance. However, further investigation is necessary to better understand
their seismic performance under combined loading. In this work, four diagonal ribs stiffened SCFST columns
were tested to failure under combined constant axial compression and cyclic lateral load. The main parameters in
this study were the presence of holes in the diagonal rib, the width-to-thickness ratio of the steel tube, and the
level of axial loading. Detailed analysis demonstrated that the effect of holes was slight, while the width-to-
thickness ratio and axial load ratio affected the cyclic behavior significantly. Furthermore, based on the veri­
fied ABAQUS model, the use of the plastic stress distribution method for computing the interaction strength of
short diagonal ribs stiffened SCFST columns was validated among a range of parameters. These parameters are
the hole diameter, axial load ratio, material strength and width-to-thickness ratio, including for columns that
would be classified as susceptible to local buckling by current design standards. These results provide new
knowledge on the behavior of diagonal ribs stiffened SCFST columns and will support the use of this efficient
form of composite construction.

1. Introduction [13].
The use of diagonal binding ribs was proposed by the authors [14].
Concrete-filled steel tubular columns possess high strength, high Diagonal ribs stiffened SCFST columns are simple and efficient to
stiffness, favorable ductility, and large energy dissipation capacity fabricate. First, two U-shaped sections are cold-formed from steel plate.
[1–3]. Square concrete-filled steel tubular (SCFST) columns have the Then, diagonal ribs are fillet welded to the inside of the U-shaped sec­
additional benefits of being relatively easy to fabricate and compatible tions at a third of the tube width. Finally, the two U-shaped sections are
with relatively simple connections. Countering these benefits are two butt welded together. This process is shown in Fig. 1.
key disadvantages of SCFST columns: 1) the steel tube is far less effective Previous research on the proposed diagonal ribs stiffened SCFST
at confining the concrete near the walls than it is at confining the con­ columns has focused on axial behavior [14]. Experimental tests of 38
crete near the corners, and 2) despite the presence of the concrete infill, short columns subjected to concentric axial compression were per­
the steel tube remains susceptible to local buckling. formed. Parameters of the study included the width-to-thickness ratio of
Adding stiffeners to the steel tube addresses both of these disad­ the steel tube, the thickness and height of the diagonal rib, the presence
vantages. A variety of types of stiffeners for SCFST columns have been of holes in the diagonal rib, the diameter and spacing of the holes, and
studied, including shear studs [4], longitudinal steel plates [5–7], the concrete strength. A finite element analysis model was developed
binding bars [8], tie bars [9], reinforcement cages [10], perforated rib and validated against the experimental results. Parametric studies per­
shear connectors [11], circular liner stiffener [12], and steel lining tubes formed with the model expanded the range of results and formed the

* Corresponding author.
E-mail addresses: zhouzhengchn@126.com (Z. Zhou), gandan@cqu.edu.cn (D. Gan), mdenavit@utk.edu (M.D. Denavit), zhouxuhong@126.com (X. Zhou).

https://doi.org/10.1016/j.jcsr.2021.107074
Received 24 April 2021; Received in revised form 22 October 2021; Accepted 25 November 2021
Available online 6 December 2021
0143-974X/© 2021 Elsevier Ltd. All rights reserved.
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

basis for proposed design recommendations [15,16].

Recently, Liu et al. [12] proposed SCFST columns with diagonal rib
stiffeners located in the potential plastic hinge regions, as shown in
Fig. 2. The diagonal rib stiffeners were mainly applied to improve the
seismic behavior of the plastic hinge regions, however, the strength of
the unstiffened region was lower than that of the stiffened region. As a
result, the weakest section was the unstiffened section under the axial
compressive load and two weak sections (i.e., sections A-A and B-B
shown in Fig. 2) would occur under the lateral load, and thus the width-
to-thickness ratio of steel tube of the column was governed by the
unstiffened sections.
In comparison to the SCFST columns with diagonal rib stiffeners
located in the potential plastic hinge region, the proposed diagonal
binding ribs can provide continuous confinement to the steel tube and
no obvious weak section exists. Moreover, additional merits of diagonal Fig. 2. SCFST column with diagonal rib stiffeners in potential plastic re­
gion [12].
ribs stiffened SCFST columns were identified in the previous work. The
diagonal ribs were shown to 1) enable greater and more uniform
confinement of the concrete, 2) delay local buckling of the steel tube, shown that local buckling can be neglected if the inequality of Eq. (3) is
and 3) contribute to the axial capacity directly by carrying axial forces. satisfied.
As a result, the ultimate strength and ductility of diagonal ribs stiffened / √̅̅̅̅̅̅̅̅̅̅̅̅̅̅
/ ̅
SCFST columns were significantly enhanced in comparison to unstiff­ b tt ≤ 52 235 fyt (3)
ened SCFST columns.
The delay in local buckling is particularly beneficial as it can allow a where, b is the width from the edge of the tube to the stiffener or be­
relaxation of design limits on the width-to-thickness ratio. Relaxed tween stiffeners.
limits would facilitate the use of thinner and higher-strength materials. The previous research has shown the promise of diagonal ribs stiff­
According to the Eurocode [17], local buckling can be neglected in the ened SCFST columns, however, only in the context of monotonic axial
design of SCFST columns if the inequality of Eq. (1) is satisfied. No compression. This work expands upon the previous research by inves­
specific provisions for accounting for local buckling are provided for tigating the behavior of diagonal ribs stiffened SCFST columns under
cases that do not satisfy the inequality. combined loading. First, an experimental program consisting of four
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅ diagonal ribs stiffened SCFST column specimens subjected to combined
/ / ̅
B tt ≤ 52 235 fyt (1) constant axial compression and cyclic lateral load is described and the
results are analyzed in detail. Then, methods of computing the strength
where, B is the outside width of the steel tube; tt is the thickness of the of diagonal ribs stiffened SCFST columns under combined axial
steel tube; and fyt is the yield strength of the steel tube in MPa. compression and bending moment are evaluated. Together, these
According to the AISC Specification for Structural Steel Buildings studies advance knowledge on the behavior of diagonal ribs stiffened
[18], a SCFST column is considered compact and local buckling can be SCFST columns and will support the use of this efficient form of com­
neglected if the inequality of Eq. (2) is satisfied. Eq. (2) is more posite construction.
permissive than Eq. (1) for all practical cases. Also, provisions are pro­
vided in the AISC Specification for cases that exceed the limit of Eq. (2). 2. Test program
/ √̅̅̅̅̅̅̅̅̅̅̅
/ ̅
B1 tt ≤ 2.26 Es fyt (2) 2.1. Specimens details

where, B1 is the width of the flat portion of the steel tube; and Es is the Four specimens were subjected to combined constant axial
steel elastic modulus with a typical value of 200,000 MPa. compression and cyclic lateral load until failure. Each specimen con­
Neither of these design standards contains specific provisions sisted of a diagonal ribs stiffened SCFST column embedded into a rein­
regarding local buckling for stiffened SCFST columns, despite experi­ forced concrete base as shown in Fig. 3(a). All specimens had a square
mental observations that stiffening the tube delays local buckling. For cross section with outside width, B, of 200 mm and calculating length, L,
diagonal ribs stiffened SCFST columns, the previous research [16] has of 660 mm. The calculating length was taken as the vertical distance

Fig. 1. Fabrication process of diagonal ribs stiffened square steel tube [14].

2
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 3. Specimen details (mm).

from the top surface of the concrete base to the centroid of the loading (Fig. 3(a)).
pin (Fig. 3(b)).
As described in Table 1, the specimens varied in the width-to- 2.2. Material properties
thickness ratio of the steel tube, the axial load ratio, and whether or
not the diagonal ribs were constructed with holes. The specimens were All of the specimens were constructed with the same batch of con­
named as SCFST followed by a sequence of two numbers and a letter. crete. The average tested compressive strength fcu of three 150 mm ×
The first number indicates the width-to-thickness ratio of the steel tube, 150 mm × 150 mm cubes at testing time was 40 MPa. The corresponding
B/tt. The second number indicates the diameter of holes in the diagonal cylinder compressive strength fc was computed as 32 MPa [17]. All the
ribs, d (this number is taken as zero when the ribs were constructed specimens were fabricated from two heats of steel, one for each nominal
without holes). The letter indicates the level of axial loading with “A” plate thickness. Steel material properties are listed in Table 2 based on
indicating an axial load ratio of 0.23 or 0.25 and “B” indicating an axial the average of three coupon tests for each thickness of steel used.
load ratio of 0.46. The axial load ratio, n, is calculated as:

n=
N
(4) 2.3. Test setup and measurements
fc Ac + fyt At + fys As
The test setup is shown in Fig. 4. The steel reaction frame provided
where, N is the applied axial load; fc is the cylinder compressive strength the axial load and restrained out-of-plane movement of the rigid L-beam.
of the concrete; fyt and fys are the yield strength of the steel tube and The rigid L-beam was connected to the ground through a four-leg
diagonal ribs, respectively; and Ac, At, and As are the cross-sectional pantograph system, which ensured that the L-beam did not rotate.
areas of the concrete, steel tube, and diagonal ribs, respectively. Note Both the specimen and base were fixed on the strong floor through the
that As is the gross cross-sectional area of the diagonal ribs (i.e., high strength rods. Prior to each test, the horizontal hydraulic actuator
computed assuming no material removed for the holes). was adjusted to ensure that the three horizontal hinges (i.e., Points A, B,
The thickness of the diagonal ribs, ts, the spacing between two and E in Fig. 4(a)) were in a horizontal line and that the three vertical
adjacent holes, s, and the steel ratio, ρ, are also listed in Table 1. The
steel ratio is taken as the total volume of steel including the steel tube
Table 2
and diagonal ribs minus the material removed for the holes divided by
Material properties of steel.
the total gross volume of the column. The size of the fillet weld between
the steel tube and diagonal ribs was 2 mm. Type Yield strength Ultimate strength Modulus of
(MPa) (MPa) elasticity (GPa)
To avoid failure at the top of the specimen, a stiffener plate was
welded to the outside of each face of the tube as depicted in Fig. 3(b). To 2 mm thick steel 448.3 563.3 198
plate
ensure a fixed connection at the base, the steel tube was embedded about
3.44 mm thick 498.0 560.3 200
390 mm into the base and two stirrups with diameter of 18 mm were steel plate
fillet welded to the outside of the embedded portion of the steel tube

Table 1
Specimen information.
Specimen B(mm) L(mm) tt(mm) B/tt ts(mm) d(mm) s(mm) ρ(%) N(kN) n

SCFST-100-40-A 200 660 2.00 100 2.00 40 100 5.65 522 0.23
SCFST-100-0-A 200 660 2.00 100 2.00 n/a n/a 5.89 522 0.23
SCFST-58-40-A 200 660 3.44 58 2.00 40 100 8.53 737 0.25
SCFST-100-40-B 200 660 2.00 100 2.00 40 100 5.65 1044 0.46

Note: n/a indicates not applicable since specimen SCFST-100-0-A was constructed without holes in the ribs.

3
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 4. Test setup.

hinges (i.e., Points C, D, and E in Fig. 4(a)) were in a vertical line. of the horizontal component of the load from the vertical actuator is
Following this adjustment, the specimen was loaded to approximately small as the angle of the actuator was approximately 0.006 rad at peak
30% of the target axial load to test sensors and check alignment. This lateral capacity of the specimen and a maximum of 0.01 rad throughout
preload was removed prior to application of the main axial load. each test.
Lateral loading was applied in displacement control following the The instrumentation of the specimens is shown in Fig. 6. Two linear
procedure shown in Fig. 5. Both drift ratio and horizontal displacement variable differential transducers (LVDTs) were installed to measure
are shown in Fig. 5 noting that the calculating length was the same for all horizontal displacement of the loading pin. Two inclinometers were
specimens tested. Loading was applied at a rate ranging from 0.7 mm/ placed around the plastic hinge zones of the column to obtain the
min to 3.8 mm/min. Testing was concluded once a 20% drop in capacity rotation. A total of 20 strain gauges were placed on the steel tube and
after the peak was observed. diagonal ribs in the expected plastic hinge zones.
The vertical hydraulic actuator above the specimen maintained a
constant axial load ratio (i.e., 0.23 for SCFST-100-40-A and SCFST-100- 3. Test results
0-A, 0.25 for SCFST-58-40-A, and 0.46 for SCFST-100-40-B) throughout
each test. The horizontal component of the load in the vertical actuator 3.1. Failure modes and load-displacement relations
generated as the specimen deformed was neglected. Reported values of
lateral load, P, are that of the horizontal actuator only. The contribution All four specimens failed with the formation of a plastic hinge at the

4
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

For each of the specimens, local buckling is seen to initiate prior to

but near the peak lateral load. Welding failure occurred after the peak
lateral load was attained and precipitated additional degradation in
strength. All the hysteresis curves were relatively full, indicating
favorable energy dissipation capacity.

3.2. Discussion of test results

In this section, the effect of various structural parameters on the

load-deformation response is investigated. To aid this investigation,
several characteristics of the load-deformation response are quantified.
For each of the cyclic load-deformation curves, an envelope curve is
generated with the peak load from the first cycle of each drift ratio.
Yield, peak, and ultimate points are identified and determined from the
envelope curve as depicted in Fig. 11 following a standard procedure
[19], where Py is the yield load, Δy is the displacement at Py, Pp is the
peak load, Δp is the displacement at Pp, Pu is the ultimate load, and Δu is
Fig. 5. Lateral loading procedure. the displacement at Pu. The values for each of these parameters are listed
in Table 3, where the loading direction is defined as shown in Fig. 4. Also
column base. The following describes the progression of damage in shown in Table 3 are the ductility coefficient, μ, and ultimate story drift
specimen SCFST-100-40-A. Rules for naming the surface of steel tube are ratio, θu, which are defined as
shown in Fig. 7 and photos of the specimen at three stages of testing are /
μ = Δu Δy (5)
shown in Fig. 8. Buckling of the steel tube initiated during the first cycle
of 1% drift with bulges forming on Surface B centered approximately 25
θu = Δu /L (6)
mm above the base (Fig. 8(a)). Five individual bulges on Surfaces A, B
and C were noted, indicating that the diagonal ribs effectively restrained The strength degradation coefficient λj and secant stiffness Kj are
the tube where the two were connected. During the first cycle of 4% defined with Eqs. (7) and (8), respectively.
drift, a bulge across Surface B developed (Fig. 8(b)), indicating that the
welds between the steel tube and the diagonal ribs had failed. At the first Pj 2
λj = (7)
cycle of 6% drift, fracture occurred in the steel tube on Surface D (Fig. 8 Pj 1
(c)) and the test was halted. The progression of damage was similar for
the other three specimens with the exception that fracture of the steel Pj 1 + Pj 2
Kj = (8)
tube only occurred in specimen SCFST-100-40-A. 2Δj
After the test, the steel tube and some concrete were removed from
specimen SCFST-100-40-A. The concrete was crushed where the steel where Pj1 and Pj2 are the peak loads for the first and second cycles at the
tube buckled (Fig. 9(a) and (b)). The diagonal ribs buckled and partially same lateral displacement of Δj, as illustrated in Fig. 12.
fractured because of large plastic deformation (Fig. 9(c)). The weld
between the steel tube and the diagonal ribs also failed (Fig. 9(d)), which
agrees well with the phenomenon observed during the test.
The cyclic load-deformation response for all specimens is shown in
Fig. 10. These plots show the applied lateral load, P, as a function of the
horizontal displacement of the loading pin at the top of the specimen, Δ.
Superimposed on the load-deformation response are points representing
when tube buckling was first visually observed, when the tube buckling
mode changed from multiple individual bulges to one large bulge
(indicating failure of the weld between the steel tube and the diagonal
ribs), and when tube fracture was first observed. Fig. 7. Rules for naming the surface of steel tube.

Fig. 6. Instrumentation of the specimens.

5
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 8. Photos and buckling modes of failure process of specimen SCFST-100-40-A.

Fig. 9. Photos of failure modes of specimen SCFST-100-40-A.

The cumulative dissipated energy, Ej, is calculated as the summation (without holes), the lateral load capacity of specimen SCFST-100-40-A
of the area enclosed within the load-displacement curve from the start of (with holes) was 5.6% lower due to the weakened diagonal ribs, while
the test to the last cycle with a maximum lateral displacement of Δj. the ductility coefficient was 11.0% higher since the holes facilitate
As a general observation, all specimens exhibited good deformation concrete placement and load transfer between the steel and concrete to
capacity. For all four specimens, the ductility coefficient was greater achieve the better integrity. The ultimate story drift of both specimens
than 3.3 and the ultimate story drift ratio was greater than 1/30. The were almost the same. As shown in Fig. 13(b), the strength degradation
latter satisfies the ultimate story drift ratio limit of 1/50 required by the of specimen SCFST-100-0-A decreased faster than that of specimen
Chinese seismic code [20]. SCFST-100-40-A. No obvious difference was observed in the stiffness
degradation and the energy dissipation capacity (Fig. 13(c) and (d)).
3.2.1. Effect of the presence of holes Based on analyses of diagonal ribs stiffened SCFST columns subjected
Holes in the diagonal ribs help facilitate placement of the concrete, to axial load, it has been recommended that the diameter of holes in the
however, their impact on structural behavior is uncertain. While they diagonal ribs be between 20% and 50% of the width of the diagonal rib
may weaken the diagonal ribs, they may also improve load transfer [16]. For the specimens tested in this work, the diameter of the holes is
between the steel and concrete. Specimens SCFST-100-40-A and SCFST- 43% of the width of the diagonal rib. Given that observed differences in
100-0-A were nominally identical with the exception that the diagonal performance between specimens SCFST-100-40-A and SCFST-100-0-A
ribs of specimen SCFST-100-40-A were fabricated with holes. The effect are minor, the recommendation appears to be appropriate for diagonal
of the presence of holes is seen in a comparison between these two ribs stiffened SCFST columns subjected to combined loads as well.
specimens.
Key metrics from the load-deformation response of the two speci­ 3.2.2. Effect of the width-to-thickness ratio
mens are presented in Fig. 13. Compared with specimen SCFST-100-0-A Specimens SCFST-100-40-A and SCFST-58-40-A were nominally

6
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 10. Load-displacement curves of the specimens.

specimens is presented in Fig. 14.

The lateral load capacity was 41% greater for the specimen with the
thicker tube. Also for the specimen with the thicker tube, the ductility
coefficient was about 30% lower, the ultimate story drift ratio was about
11% lower, and the post-peak strength degradation was greater (Fig. 14
(b)). A potential reason for observing lower ductility may be that the
larger axial load carried by the thicker steel tube transferred to the
concrete when the steel tubes buckled, accelerating the concrete
crushing and leading to worse ductility.
With the thicker tube, the initial secant stiffness increased, the
stiffness degradation was slower (Fig. 14(c)), and the energy dissipation
capacity was enhanced (Fig. 14(d)).

3.2.3. Effect of axial load

Specimens SCFST-100-40-A and SCFST-100-40-B were nominally
Fig. 11. Determination of yield, peak and ultimate points. identical with the exception of the axial load applied to them during the
test. The axial load ratios (Eq. (4)) for these two specimens were 0.23
identical with the exception of the thickness of the steel tube. The and 0.46, respectively. A comparison between the results of these two
thickness of the steel tube impacts both the width-to-thickness ratio and specimens is presented in Fig. 15.
the steel ratio. For these two specimens, the width-to-thickness ratios The lateral load capacity was 10% greater for the specimen with
were 100 and 58, respectively; and the steel ratios were 5.65% and higher axial load (i.e., SCFST-100-40-B). The ductility coefficient and
8.53%, respectively. A comparison between the results of these two ultimate story drift ratio, however, were about 14% and 25% less,

7
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Table 3
Test results of all specimens.
Specimen SCFST-100-40-A SCFST-100-0-A SCFST-58-40-A SCFST-100-40-B

Loading direction + – + – + – + –

Py (kN) 138.6 − 141.5 150.9 − 146.0 201.5 − 199.1 145.8 − 165.6

△y (mm) 6.5 − 6.3 7.0 − 7.6 8.4 − 8.0 5.7 − 5.5
Pp (kN) 162.5 − 164.1 173.6 − 171.5 233.0 − 228.0 168.0 − 191.5
Average Pp (kN) 163.3 172.5 230.5 179.8
△p (mm) 13.2 − 12.1 18.9 − 17.4 19.6 − 19.7 9.2 − 9.5
Pu (kN) 138.1 − 139.5 147.5 − 145.7 198.1 − 193.8 142.8 − 162.8
△u (mm) 29.9 − 31.0 29.9 − 31.6 28.4 − 26.0 22.5 − 23.2
μ 4.53 4.70 4.53 4.79 4.30 3.94 3.41 3.51
Average μ 4.73 4.21 3.30 4.08
θu (%) 4.58% 4.89% 4.25% 4.18% 3.37% 3.23% 3.98% 4.19%
Average θu (%) 4.61% 4.66% 4.12% 3.46%
Ej (kN⋅m) 63.65 54.51 77.32 44.80

In this work, the application of the PSD method to diagonal ribs

stiffened SCFST columns is explored. Expanded cross sectional slen­
derness limits are investigated along with how best to address the
increased concrete confinement and the effect of holes in the diagonal
ribs.

4.1. Comparison to experimental results

Interaction curves constructed using the PSD method for the tested
specimens are shown in Fig. 16 along with the experimentally obtained
peak strengths. The maximum moment at the base from the experiment,
Mu, was computed as:
Mu = Pp L + NΔp (9)

Fig. 12. Definition of strength and stiffness degradation. Three different interaction curves are shown in Fig. 16. Each was
computed in accordance with variations of the PSD method. The first
respectively, and the strength and stiffness degradation were greater interaction curve was developed using basic and conservative assump­
(Fig. 15(b-c)), as the higher axial load accelerated the concrete crushing tions, specifically, using plastic stress in the concrete equal to 0.85fc and
and tube buckling. With higher axial load ratio, the dissipated energy using a cross section through the holes in the diagonal ribs (i.e., the net
was larger due to the higher lateral load capacity (Fig. 15(d)). section). This interaction curve is seen to be conservative in comparison
to the experimental results. Defining Mp as the moment strength from
4. Interaction strength the PSD method at the same axial load as from the experiments, the ratio
Mp/Mu quantifies the accuracy of the method. The mean value of this
Evaluation of available strength under axial load and bending ratio is 0.75 with a standard deviation of 0.09 for this first interaction
moment is an essential aspect of design. The plastic stress distribution curve.
(PSD) method is a common method to define the strength of steel- The concrete strength enhancement due to the confinement effect of
concrete composite column cross sections. To compute strength ac­ the unstiffened square steel tubes was not considered in typical codes. In
cording to the PSD method, steel components of the cross section are contrast, the diagonal ribs in stiffened SCFST columns provide more
assumed to have attained their yield stress in either tension or confinement and strength increases in the concrete have been observed
compression and the concrete is assumed to have attained a specified experimentally [14]. Strength enhancement due to confinement can be
stress, typically 0.85fc, in compression. Tension strength of the concrete accounted for in the PSD method by increasing the plastic stress used for
is neglected. Pairs of axial load and bending moment are then computed the concrete. For example, the AISC Specification [18] permits the use of
through integration of the assumed stress over the cross-sectional area. 0.95fc for circular concrete filled steel tube columns. An expression for
An interaction diagram is generated from multiple pairs of axial load the confined concrete strength, fcc, in diagonal ribs stiffened SCFST
and bending moment computed in this manner using different plastic columns was developed in previous work [14,23,24].
neutral axis locations. For unstiffened SCFST columns, closed-form ⎧ ( √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ )
⎪ ke fr ke fr
equations have been developed to compute key points on the interac­ ⎪
⎪
⎪
⎪ fc − 1.254 + 2.254 1 + 7.94 − 2 fc ≤ 50 MPa
⎨ fc fc
tion diagram [21]. In this work, more continuous interaction diagrams
fcc = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
are generated by computing many points and integration of the stresses ⎪
⎪
( )
⎪
⎪ f − 0.413 + 1.413 1 + 11.4 ke fr − 2 ke fr
over the cross section is performed numerically using a fiber dis­ ⎪
⎩ c fc > 50 MPa
fc fc
cretization approach.
The PSD method is described in several codes, including the Euro­ (10)
code [17], the AISC Specification for Structural Steel Buildings [18], and ( )
( ) ( ) d
the Chinese code [22]. However, in these codes, the use of the PSD 2 0.19fyt tt + 2 0.19fys cos45◦ ts 1 − = fr (B − 2tt ) (11)
method is limited to cross sections that meet slenderness requirements s
(e.g., Eqs. (1–2)). The slenderness requirements in these codes are based
where ke is the confinement effectiveness coefficient equal to 0.78; d is
on the outside dimensions of the tube and do not recognize the potential
the diameter of the holes; s is the spacing of the holes; and fr is the
for stiffeners to delay local buckling.
confining stress.

8
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 13. The effect of the presence of holes in the diagonal rib.

The second interaction curve in Fig. 16 was developed using a plastic testing is limited. To confirm the results over a broader range of pa­
stress for the concrete equal to fcc and the net section. This interaction rameters, an expanded comparison against supplemental results ob­
curve provides an improved, yet still conservative, prediction with mean tained from finite element analyses was performed.
value of Mp/Mu equal to 0.87 with a standard deviation of 0.05.
One potential reason for the remaining conservatism is that the use of
the net section is conservative. The net section, with the entire diameter 4.2. Finite element model
of the holes removed from the diagonal ribs only occurs for short lengths
along the column. Yielding in these short lengths would not necessarily 4.2.1. Model description
lead to significant member deformations. Similar to the way in which Numerical analyses can help expand the range of investigation
structural steel tension members with bolted connections are treated, it beyond what was tested physically. Three-dimensional models using
may be appropriate to consider two separate limit states, yielding on the shell and continuum finite elements are most appropriate for this study
gross section and rupture on the net section. To investigate this possi­ as the detailed geometry of diagonal ribs stiffened SCFST columns,
bility, the third interaction curve in Fig. 16 was developed using a plastic confinement stresses, and local buckling can all be captured explicitly.
stress for the concrete equal to fcc and the gross section (i.e., without The finite element model was developed using ABAQUS [25]. Shell
holes in the diagonal ribs). It is noted that for the cross sections inves­ elements (type S4R) were used for the steel tube and diagonal ribs, solid
tigated, the interaction curves computed in this way always show a elements (type C3D8R) were used for the concrete and truss elements
lower strength than interaction curves constructed using the net section (type T3D2) were used for the rebars. As shown in Fig. 17(a), the mesh
and the ultimate strength of the steel components. The interaction curve size of the column and RC base was selected as 0.1 and 0.2 times the
with fcc and the gross section provides a further improved prediction column width, respectively, and the reference point, where the loading
with mean value of Mp/Mu equal to 0.92 with a standard deviation of pin existed, was connected to the column top surface using the
0.04. While these results are promising, the range of experimental constraint “rigid body” which was used to simulate the rigid hinge joint.
In this model, the steel tube and diagonal ribs were merged together,

9
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 14. The effect of the width-to-thickness ratio B/tt.

the rebars were also merged together, and the rebars and diagonal ribs ε r
were embedded in the concrete. A surface-to-surface contact interaction σ = fc ( / ) (12)
εfem r − 1 + ε εfem r
was applied at the interface of the steel tube and concrete by specifying
hard contact in the direction normal to the interface plane and using a where σ is the stress, ε is the strain, εfem is the equivalent peak strain at
Mohr-Coulomb friction model with friction coefficient of 0.6 and peak stress in ABAQUS, and r is a parameter that controls the shape of
maximum interfacial shear stress of 0.6 MPa for the tangential behavior. the curve.
Noting that it was observed that the steel tubes tended to be pulled out According to the research conducted by Zhou et al. [16], the model
during the test process and the surrounding concrete of column was proposed by Han et al. [27] could well predict the peak strain of the
destroyed partially (Fig. 17(b)), indicating that the RC base was not diagonal ribs stiffened SCFST columns. Therefore, εfem can be calculated
perfectly rigid. Therefore, the RC base should be included in the model. as:
As shown in Fig. 17(c), the highlighted concrete surfaces of the column ( )
and RC base were tied together to simulate the integral concreting, and εfem = 1300 + 12.5fc + 800ξ0.2 × 10− 6 (13)
the highlighted concrete surfaces of the RC base and the steel tube
surfaces of the column were modeled by surface-to-surface contact At ⋅fyt
ξ= (14)
interaction. 0.67fc ⋅Ac
Elastic-perfectly plastic constitutive relations using the von Mises
The diagonal ribs can improve the ductility and thus the slope in the
yield criterion were adopted to model the steel components including
descending branch of the envelope curves decreased. Therefore, it is
the steel tube, diagonal ribs, and rebars. A damaged plasticity model was
necessary to include the beneficial effect of the diagonal ribs into the
used for the concrete. The damaged plasticity model in ABAQUS takes as
parameter r. According to Mander’s model [23], r was defined as:
input a stress-strain relationship, which in this work was defined based
on Popovics’ equation [26] and shown as follows: r=
Ec
(15)
Ec − fcc /εcc

10
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 15. The effect of the axial load ratio n.

where, Ec is the modulus of elasticity of concrete given by [28]: be 1.16, the ratio of compressive to tensile meridians of the yield surface
√̅̅̅̅ in deviatoric stress space Kc was assumed to be 0.6667 and the viscosity
Ec = 4700 fc (16) parameter was selected as 0.0001. The damage factors were calculated
based on the energy loss [32].
The peak stress of confined concrete fcc was given by Eqs. (10)–(11).
Note that residual stresses and initial imperfections of the steel tubes
The peak strain of confined concrete εcc can be calculated as [23,29]:
were not considered in the model, as these are typically minor effects for
( )( ( ))
f fcc thin-walled composite columns [33]. Weld fracture between the steel
εcc = 0.0015 + c 1+5 − 1 (17)
70000 fc tube and the diagonal ribs was also neglected in the model despite being
observed very late in the loading of the tested specimens. Also, the
For concrete under tension, a fracture energy model was used with corners of CFST columns were assumed to be exact right angles.
the tensile stress ft and fracture energy Gf defined as [30,31]:

ft = 0.375fcu 0.55 (18) 4.2.2. Model verification

To confirm that the model is able to capture the strength of diagonal
Gf = α(0.1fc )0.7 (19) ribs stiffened SCFST columns accurately, numerical results are compared
to experimental results from this work as well as previous work. Thirty-
where, the parameter α depends on the maximum diameter of concrete eight axially loaded short diagonally ribs stiffened SCFST columns were
aggregate Dmax, namely, Dmax = 8 mm, α = 0.025; Dmax = 16 mm, α = tested in previous work [16]. Fifteen of them are investigated here. The
0.03; and Dmax = 32 mm, α = 0.058. other specimens constructed with reinforcing bars, longitudinal steel
Additionally, the detailed parameters which defined the concrete plates, octagonal shape, or discontinuous diagonal ribs were excluded
damaged plasticity model were as follows: the dilation angle was from investigation in this work. The peak load from the experiment, Np,
selected as 38◦ , the eccentricity ratio was chosen as 0.1, the ratio of is compared to the peak load from the analysis, NABAQUS, in Fig. 18(d).
biaxial compressive strength to uniaxial strength fb0/fc was assumed to The ratio of these two strengths (i.e., NABAQUS/Np) has a mean value of

11
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 16. Comparison of N-M interaction curves.

0.99 and standard deviation of 0.04. The load-deformation response stiffness are also captured well, albeit to a lesser degree. Local buckling
from both the experiment and the analysis of four representative spec­ of the sub-plates is captured well as shown visually in Fig. 19(b). The
imens is shown in Fig. 18(c). Geometric and material properties for these model does not capture the fracture of the weld between the diagonal
specimens are listed in Table 4. The initial stiffness, peak strength, and ribs and the steel tubes that was observed experimentally. However, this
strain at peak strength are all captured with good accuracy. The extent of behavior was observed at a relatively large drift ratio and after the peak
local buckling is also captured well as seen visually in Fig. 18(a). A dip is strength was achieved. Based on these comparisons to experimental
seen in the post-peak load-deformation response from several of the results, it was determined that this model is suitable for use in estimating
experiments, Fig. 18(b). This feature is not captured by the analysis and the peak strength of diagonal ribs stiffened SCFST columns.
is likely caused by weld fracture between the steel tube and diagonal rib.
However, since this behavior occurs after the peak strength has been 4.3. Effect of holes on interaction strength
observed, it has no effect on the estimation of the peak strength, which is
the primary purpose of the model. An initial parametric study was conducted to investigate the effect of
The ABAQUS model and element mesh of the specimens tested in this holes in the diagonal ribs on the strength of the columns. A total of 45
work is shown in Fig. 19(a). The column base was included in the model specimens with varying tube width-to-thickness ratio, hole diameter,
as it was not perfectly rigid and impacted the initial stiffness. The load- and axial load level were analyzed. Three width-to-thickness ratios were
deformation response from both the experiment and the analysis is selected: B/tt = 33.3, 66.7, and 128.7. The latter two values do not meet
shown in Fig. 19(c). The initial stiffness and peak strength are captured the slenderness limits from the Eurocode [17] nor would they be clas­
with good accuracy. The ratio of peak lateral capacity from the model to sified as compact per the AISC Specification [18] for the selected steel
that from the experiment (PABAQUS/Pp) has a mean value of 1.03 and a yield stress, fyt = 345 MPa. The largest of the selected width-to-thickness
standard deviation of 0.09. The post-peak response and unloading ratios was taken as the limit from Eq. (3). The diagonal ribs were

12
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 17. ABAQUS model.

assumed to have the same thickness and yield stress as the tube. Ana­ The strength of diagonal ribs stiffened SCFST columns decreases with
lyses were performed for the case of no holes as well as four different increasing hole diameter. The reductions in strength could be a direct
diameters of holes, specifically: d/bs = 0.2, 0.4, 0.6, and 0.8. The con­ result of the replacement of higher strength steel with lower strength
crete compressive strength was taken as fc = 50 MPa. Three different concrete. They could also be a result of the lower level of concrete
axial loading levels were investigated, n = 0.0, 0.3, and 0.7. confinement provided by the diagonal ribs with holes. These effects can
A finite element model was developed for each of the 45 cases. The be incorporated in the PSD calculations through definition of the cross
configuration of the model was the same as that used for the physical section and the value of the concrete stress, fcc (Eq. (10)). As seen in the
experiments, including the base. The analyses subjected the columns to results in Fig. 20, for higher axial loads, the effect of the holes on the
constant axial compression and monotonically increasing lateral load. concrete stress alone is sufficient to capture the strength reduction,
The main result from each analysis was the maximum bending moment without the need to use the net section. However, this is not true for
capacity determined using Eq. (9). The maximum bending moment ca­ lower axial loads, where the concrete stress has a minimal effect on the
pacity was also determined from the PSD method for each column. The resulting strength. While the effect of the holes is conservatively exag­
goal of this study is to determine how well the PSD method captures the gerated when using a concrete stress of fcc and the net section, this
effect of the holes on the strength. Therefore, for both the finite element approach is recommended to avoid potential unconservative errors.
analyses and the PSD method, the strength results are normalized by the
corresponding result without holes. The results from the analyses are
4.4. Comparison to numerical results
presented in Fig. 20. In this figure, the vertical axis is the ratio of the
calculated moment with holes, Mwith hole to the moment without holes
A second parametric study was performed to generate strength re­
Mwithout hole.
sults over a broad range of parameters for additional validation of the

13
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 18. Comparison of ABAQUS simulated and test results from [14].

14
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Table 4
Specimen information in the literature [16].
Specimen B(mm) L(mm) tt(mm) B/tt fyt(MPa) fys(MPa) fc(MPa)

SS-3-2-B 300 900 2.0 100.0 356.1 176.7 48.6

SS-3-3-B 300 900 3.0 100.0 356.1 356.1 48.6
SS-3-6-B 300 900 6.0 100.0 356.1 317.5 48.6
SS-3-2-B-c 300 900 2.0 100.0 356.1 176.7 81.4

Fig. 19. Comparison of ABAQUS simulated and test results in this work.

15
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

Fig. 20. The effect of the holes on interaction strength.

use of the PSD method. A total of 60 specimens with varying tube width- same axial load. Results from the 36 specimens of the first parametric
to-thickness ratio, steel yield strength, concrete compressive strength, study that do not overlap with this parametric study are also included in
and axial load level were analyzed. Three width-to-thickness ratios were Fig. 21. For the PSD method using fcc and net section, the mean value of
selected: B/tt = 33.3, 66.7, and the maximum ratio permitted by Eq. (3) MP/MABAQUS was 0.84 with a standard deviation of 0.06.
for the selected yield strength. Two values of steel yield strength were These results show that a conservative prediction of the strength of
chosen: fy = 345 MPa and 420 MPa. The third value of width-to- diagonal ribs stiffened SCFST columns can be obtained using the PSD
thickness ratio was 128.7 and 116.7 for the yield strengths of 345 method with a concrete stress of fcc and the net section. However, it is
MPa and 420 MPa, respectively. The diagonal ribs were assumed to have noted that the 0.85 factor on concrete strength typically used in the PSD
holes with diameter, d = 0.4bs and have the same thickness and yield method accounts for the difference in strength between concrete cylin­
stress as the tube. Two values of concrete compressive strength were ders and concrete in practical members due to loading rate, member
adopted, fc = 50 MPa and 70 MPa. Five different axial loading levels dimensions, and other effects that are expected in building structures
were investigated, n = 0.0, 0.15, 0.3, 0.5 and 0.7. The analyses were but are either not present or controlled in experimental tests and nu­
performed as described in the previous section. merical analyses. Therefore, the use of a concrete stress equal to 0.85fcc
In Fig. 21, the horizontal axis is the predicted moment based on is recommended for design.
ABAQUS and the vertical axis is the corresponding predicted moment This work has shown that the PSD method is suitable for use with
based on the proposed PSD method (i.e., using fcc and net section) at the diagonal ribs stiffened SCFST columns with thinner tubes than currently

16
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

350 accounted for by using 0.85fcc as the plastic stress in the concrete,
where fcc is given by Eq. (10).
PSD ( fcc, net section)
300
Declaration of Competing Interest
Predicted moment MP (kN·m)

250
We declare that there is no any conflict of interest.
200
Acknowledgement
150
-20%
The authors greatly appreciate the financial support provided by the
100 National Natural Science Foundation of China (No. 51878097), Natural
Science Foundation of Chongqing, China (No. cstc2019jcyj-xfkx0007),
50 and China scholarship council (No.201906050036). However, the
opinions expressed in this paper are solely the authors’ own.
0
0 50 100 150 200 250 300 350
Predicted moment MABAQUS (kN·m) References

[1] L.H. Han, W. Li, R. Bjorhovde, Developments and advanced applications of

Fig. 21. Comparison of results predicted by ABAQUS and proposed
concrete-filled steel tubular (CFST) structures: members, J. Constr. Steel Res. 100
PSD method. (2014) 211–228.
[2] B.C. Gourley, C. Tort, M.D. Denavit, et al., A synopsis of studies of the monotonic
and cyclic behavior of concrete-filled steel tube members, connections, and frames,
permitted by design standards. It is likely that the PSD method would be in: Report No. NSEL-008, NSEL Report Series, Department of Civil and
similarly suitable for other types of stiffened filled composite columns. Environmental Engineering, University of Illinois at Urbana-Champaign, 2008.
Therefore, a broader investigation aimed at developing more general [3] C.D. Goode, Composite columns - 1819 tests on concrete-filled steel tube columns
compared with Eurocode 4, Struct. Eng. 86 (16) (2008) 33–38.
design provisions that recognize the strength of stiffened filled com­ [4] T. Kitada, Ultimate strength and ductility of state-of-the-art concrete-filled steel
posite columns is recommended for future research. Further investiga­ bridge piers in Japan, Eng. Struct. 20 (4–6) (1998) 347–354.
tion on the use of the gross section in the PSD method for members with [5] H. Ge, T. Usami, Strength of concrete-filled thin-walled steel box columns:
experiment, J. Struct. Eng. 118 (11) (1992) 3036–3054.
material removed is also recommended for future research.
[6] Z.B. Wang, Z. Tao, Q. Yu, Axial compressive behaviour of concrete-filled double-
tube stub columns with stiffeners, Thin-Walled Struct. 120 (11) (2017) 91–104.
5. Conclusion [7] H. Dong, Y. Li, W. Cao, et al., Uniaxial compression performance of rectangular
CFST columns with different internal construction characteristics, Eng. Struct. 176
(2018) 763–775.
Four diagonal ribs stiffened square concrete-filled steel tubular [8] J. Cai, Z.Q. He, Axial load behavior of square CFT stub column with binding bars,
(SCFST) column specimens were tested under combined constant axial J. Constr. Steel Res. 62 (5) (2006) 472–483.
compression and cyclic lateral load until failure. The main parameters in [9] C.S. Huang, Y.K. Yeh, G.Y. Liu, et al., Axial load behavior of stiffened concrete-
filled steel columns, J. Struct. Eng. 128 (9) (2002) 1222–1230.
the experimental study include the presence of holes in the diagonal [10] Y. Wang, Y. Yang, S. Zhang, Static behaviors of reinforcement-stiffened square
ribs, the width-to-thickness ratio of the steel tube, and the axial load concrete-filled steel tubular columns, Thin-Walled Struct. 58 (18–31) (2012)
ratio. Additionally, based on the verified ABAQUS model, an investi­ 18–31.
[11] G. Cheng, Y.J. Liu, L. Jiang, Research on concrete-filled rectangular steel tube
gation aimed at expanding the range of applicability of the plastic stress stiffeners and axial compression bearing capacity, Adv. Mater. Res. 1065 (2015)
distribution (PSD) method to thinner diagonal ribs stiffened SCFST 1092–1096.
columns was conducted. The following observations and conclusions [12] J. Liu, Q. Zhao, Z. Xia, et al., Seismic behavior of thin-walled square CFST columns
with diagonal rib/circular liner stiffeners, J. Constr. Steel Res. 179 (2021), 106519.
were made:
[13] X. Zhou, J. Liu, X. Wang, et al., Structural performance and compression
resistances of thin-walled square CFST columns with steel lining tubes, J. Struct.
(1) All four specimens failed with the formation of a plastic hinge at Eng. 147 (7) (2021) 04021105.
[14] Z. Zhou, D. Gan, X.H. Zhou, Improved composite effect of square concrete-filled
the base and all four specimens exhibited favorable seismic per­
steel tubes with diagonal binding ribs, J. Struct. Eng. 145 (10) (2019) 04019112.
formance. The progression of damage included local buckling of [15] Z. Zhou, D. Gan, X.H. Zhou, Hysteretic behavior of square thin-walled CFST
the steel tube at approximately 2% drift, fracture of the weld column stiffened by diagonal ribs, China Civil Eng. J. 54 (6) (2021) 14–25 (in
between the steel tube and the diagonal ribs at approximately 4% Chinese).
[16] X. Zhou, Z. Zhou, D. Gan, Analysis and design of axially loaded SCFST columns
drift, and fracture of the steel tube, which was observed in one with diagonal ribs, J. Constr. Steel Res. 167 (2020), 105848.
specimen, at approximately 6% drift. [17] Eurocode 4, Design of Composite Steel and Concrete Structures Part 1–1: General
(2) The presence of holes in the diagonal ribs weakens the column Rules and Rules for Building, the European Committee for Standardization,
Brussels, Belgium, 2004.
somewhat but appears to slow strength degradation and enhance [18] ANSI/AISC 360–-16, Specification for Structural Steel Buildings, American
ductility to an extent. The effect of the presence of holes in the Institute of Steel Construction, Chicago, USA, 2016.
diagonal ribs on the ultimate story drift ratio, stiffness degrada­ [19] R. Park, State of the art report ductility evaluation from laboratory and analytical
testing, in: Proc., 9th World Conference on Earthquake Engineering, Tokyo, Kyoto,
tion, and energy dissipation capacity was minimal. The diameter Japan, 1988, pp. 605–616.
of holes in the diagonal ribs can be between 20% and 50% of the [20] GB50011-2010, Code for Seismic Design of Buildings, Architecture Industrial Press
width of the diagonal rib. of China, Beijing, 2002 [in Chinese].
[21] AISC, Steel Construction Manual, 15th edition, American Institute of Steel
(3) With decreased width-to-thickness ratio or increased axial load
Construction, Chicago, Illinois, 2017.
ratio, the load capacity and energy dissipation capacity [22] CECS, 159:2004, Technical Specification for Structures with Concrete-Filled
increased, while the ductility and ultimate story drift ratio Rectangular Steel Tube Members, China Association for Engineering Construction
Standardization, Beijing, China, 2004.
decreased. Post-peak strength and stiffness degradation also
[23] J.B. Mander, M.J.N. Priestley, R. Park, Theoretical stress-strain model for confined
increased. concrete, J. Struct. Eng. 114 (8) (1988) 1804–1826.
(4) The PSD method was validated for use in the design of diagonal [24] L. Bing, R. Park, H. Tanaka, Stress-strain behavior of high-strength concrete
ribs stiffened SCFST columns with tube width-to-thickness ratio confined by ultra-high-and normal-strength transverse reinforcements, ACI Struct.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ J. 98 (3) (2001) 395–406.
no larger than 156 235/fyt . It is recommended that the net [25] ABAQUS, ABAQUS Version 6.13 Analysis user’s Manuals, Dassault Systemes
section be used and that confinement of the concrete be Simulia Corporation, Providence, RI, USA, 2013.
[26] S. Popovics, A numerical approach to the complete stress-strain curves for
concrete, Cem. Concr. Res. 3 (5) (1973) 583–599.

17
Z. Zhou et al. Journal of Constructional Steel Research 189 (2022) 107074

[27] L.H. Han, G.H. Yao, Z. Tao, Performance of concrete-filled thin-walled steel tubes [31] CEB-FIP Model Code, Thomas Telford Services Ltd, Lausanne, Switzerland, 1990,
under pure torsion, Thin-Walled Struct. 45 (2007) 24–36. p. 1993.
[28] ACI, Building Code Requirements for Structural Concrete and Commentary, [32] Z.Q. Wang, Z.W. Yu, Concrete damage model based on energy loss, J. Build. Mater.
American Concrete Institute, Detroit (Michigan), 2011. 7 (4) (2004) 365–369 (in Chinese).
[29] M.P. Collins, D. Mitchell, Prestressed Concrete Structures, Prentice Hall, [33] Z. Tao, B. Uy, L.H. Han, et al., Analysis and design of concrete-filled stiffened thin-
Englewood Cliffs, NJ, 1991. walled steel tubular columns under axial compression, Thin-Walled Struct. 47 (12)
[30] J.G. Nie, Y.H. Wang, Comparison study of constitutive model of concrete in abaqus (2009) 1544–1556.
for static analysis of structures, Eng. Mech. 30 (004) (2013) 59–67 (in Chinese).

18