Structures 33 (2021) 2166–2176

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Comparisons of design methods for beam string structure based on

reliability and progressive collapse analysis
Hao Zhou a, Youbao Jiang a, *, Sondipon Adhikari b, Qianqian Yin a, Jianguo Cai c
a
School of Civil Engineering, Changsha University of Science and Technology, Changsha 410114, China
b
College of Engineering, Swansea University, Swansea SA1 8EN, United Kingdom
c
National Prestress Engineering Research Center, Southeast University, Nanjing 210096, China

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

Keywords: The current design method of ultimate capacity of beam string structure (BSS) is mainly based on the fixed load
Beam string structure ratio (FLR) criterion, seldom considers the effect of the random load ratio (RLR) on bearing capacity. This paper
Random load ratio compared the coefficient of variation (COV) of bearing capacity obtained by RLR criterion with that obtained by
Reliability analysis
the FLR criterion. It indicates that the random properties of load ratio have a significant impact on COV and
Load and resistance factors
Progressive collapse
should be accounted for. A more realistic limit state function is built with a practical simplification, which is
proved to have a better accuracy by reliability verifications for typical cases. Parametric reliability analyses are
also carried out with Monte Carlo simulations. The results show that the reliability with FLR criterion is larger
than that with RLR criterion. Thus the reliability of BSS would be overestimated following the current design
method, and an unsafe design would be resulted in, too. Three targeted reliability indexes are selected for
representative cases. Two improved design methods with optimum load and resistance factors are obtained
according to minimum differences between the calculated reliability indexes and targeted ones among cases.
Finally, the performance of anti-progressive collapse of BSS designed by the two improved methods is compared
when the strut or cable fails. The results show that the representative BSS designed by improved design method 2
with fixed load partial factors and optimum resistance factor, which varies with cases, has better performance of
anti-progressive collapse.

1. Introduction algorithms for calculating the stiffness matrix and internal force vector;
Jiang et al. [13] derived the formulas of geometric nonlinear FEM for
As a self-balanced system, a beam string structure (BSS) is usually spatial beam element, cable element and truss element, respectively. Wu
consisting of the upper chord (e g. rigid steel arch), the lower chord (e g. et al. [14,15] investigated the variation of the lateral buckling of the
flexible cable), and struts in the middle. In recent years, the BSS has been struts in the BSS for different string layouts, and deduced the formulas
widely applied in engineering practice due to its light weight, high for calculating the critical buckling load of struts in the BSS. Ye et al.
bearing capacity, good space utilization and beautiful and smooth [16] and Cao et al. [17] conducted a study on the structural properties of
architectural image (e.g. Dong et al. [1]; Zhao et al. [2]; Cai et al. [3]; the beam string structures, and performed numerical simulations and
Luo et al. [4]; Han et al. [5]). experimental research on the form-finding of beam string structures.
So far, many scholars have carried out works on the structural Jiang et al. [18] adopted the force method to study the stiffness for­
analysis of BSS. As early as in the 1980s, Satioh et al. [6–8] began to mulations for cable-arch structures, and proposed an efficient method
study the basic mechanical principles of the prestressed BSS. Afterwards, for stiffness calculation of the concave cable-arch structure. Xue et al.
Kato et al. [9,10] conducted a theoretical analysis and experimental [19,20] used the ANSYS program to perform a design optimization for
study on the BSS. To improve the calculation efficiency, many analysis the BSS of the Shanghai Yuanshen Arena and investigated its bearing
methods for BSS have also been proposed. For example, Thai et al. [11] capacity through experimental testing.
and Abad et al. [12] proposed new elements for nonlinear finite element The wind resistance performance and seismic performance of BSS
analysis of cables under static and dynamic loads, and also presented have attracted significant attentions. Chen et al. [21] studied wind

* Corresponding author.
E-mail address: [email protected] (Y. Jiang).

https://doi.org/10.1016/j.istruc.2021.05.085
Received 26 August 2020; Received in revised form 25 May 2021; Accepted 29 May 2021
Available online 9 June 2021
2352-0124/© 2021 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
H. Zhou et al. Structures 33 (2021) 2166–2176

resistance performance of a beam-truss roof structure by means of wind forms of upper chord sections.
tunnel test, field test and numerical simulation. Han et al. [22] analyzed For the BSP under the full-span and half-span load combinations, it is
dynamic stability of beam string structures under earthquake loads, and assumed that there is sufficient support out of the plane. The basic pa­
proposed some suggestions on selecting a proper structural model in rameters of the members in this model are shown in Table 1.
project design. Chen et al. [23] studied the dynamic characteristics and Herein, four BSP models with different spans, structural heights and
wind-induced displacement response of BSS by the finite element upper chord sections are selected as shown in Table 2.
method. Lee et al. [24,25] developed a novel two-way beam string For the TSS under the action of full-span and half-span load combi­
structure. The structure is equipped with two types of cables which are nation, as shown in Fig. 1, in the same way, it is also assumed that there
arch-shaped and sagging to resist bi-directional loads. Among them, the is sufficient support outside the plane. Jiang et al. [34] reported the
arched cable mainly resists negative wind pressure. basic parameters of this structure. The span is 128 m. The rise-to-span
BSS has been widely used in public buildings because of its strong ratio of the arch is 0.08, while the sag-to-span ratio of the cable is
spanning ability. But compared with the frame structure, the redun­ 0.03. In addition, the steel is considered to be ideal elastic–plastic with
dancy of BSS is lower, and it is more prone to progressive collapse due to yielding strength 345 MPa, and the elastic modulus of the upper chord
local failures. Therefore, the anti-progressive collapse performance of and struts are 2.0 × 105 MPa. The elastic modulus and prestress of the
BSS has been paid more attention by many researchers. Malla et al. [26] cables are 1.95 × 105 MPa and 400 MPa, respectively; the spacing be­
analyzed the structural response caused by transient local damage and tween vertical Strut 4 is 9.2 m. The sections of all members are shown in
thought that the risk of collapse of the spatial structure was high. Table 3, of which t1 and t2 are the thickness of the Chord 1 and Chord 2,
Murtha-Smith et al. [27] analyzed the causes of progressive collapse respectively. The truss height is 2600 mm, and the width between Chord
accidents of a large-span stadium according to the alternate load path 1 is 3000 mm.
method. Hu [28] analyzed the collapse law of BSS under local failure or Herein, four different schemes of the upper chord section are
strong earthquake and put forward the anti-progressive collapse mea­ selected, which nearly has the same steel weight, as shown in Table 4.
sures that can be applied to design. Cai et al. [29–30] studied the in­
fluence of cables or struts failure on BSS based on major engineering 2.2. Verification of finite element analysis model for BSS
projects such as cable-arch structure of the New Guangzhou Railway
Station and truss-string structure of the Meijiang Exhibition Center, and In this paper, the finite element models of BSS are established by
proposed some strengthening measures. ANSYS12.0 software. Geometric nonlinearity and material nonlinearity
Based on the researches above, this structure has been widely used in are considered in the structural analysis. In order to check the finite
engineering practices as its design method developed. The conventional element models, experimental results are introduced to make compari­
design methods mainly follow the FLR criterion, which usually adopt an sons. Taking BSP model as an example, the finite element model of the
assumption that ultimate capacity is only affected by the stochasticity of BSP is shown in Fig. 2, in which the upper chord was simulated by
the resistance variables, e.g. steel or concrete strength, section di­ BEAM188 element, and the struts and cables were modeled by LINK8
mensions, neglecting the effects of random properties of load ratio. The element and LINK10 element, respectively.
structural bearing capacity varies largely with different load ratios (e.g. The test data of mid-span deflection for a scale model of BSP in the
ultimate capacity of beam string structures under full-span load and literature (the BSS-3 model reported by Xue and Liu [19]) was selected
half-span load combination, strength of reinforced concrete columns for comparisons, and the results are shown in Fig. 3.
under vertical load and horizontal load combination), and thus the In Fig. 3, it can be seen that the maximum mid-span deflection of
random properties of load ratio have a significant impact on the bearing experimental model is about 140 mm, while that of analytical model is
capacity (see [31,32]). It is reported that the adverse effect on bearing about 130 mm, and the results between two models are close. Moreover,
capacity caused by non-uniform snow load may lead to low safety of the the ultimate bearing capacity of experimental model and analytical
BSS designed according to the current load partial factors (see Takahashi model are about 7.0kN and 7.2kN, respectively, and the results between
et al. [33]). two models are close, too. It shows that the finite element analysis model
The previous experimental and theoretical studies mainly focus on adopted in this paper has a better accuracy.
the mechanical performance of BSS under the fixed load ratio criterion,
while the research on the reliability of BSS under random load ratio is
seldom. This paper analyzed the uncertainties of bearing capacity
through combining the finite element method with the Monte Carlo
simulation, and proposed a simplified approach to establish a more Table 1
Basic parameters for Members.
realistic limit state equation of the BSS under both full-span load and
half-span load, and carried out the bearing capacity reliability calibra­ Members Section types(mm) Material strength/ Elasticity modulus
MPa /105MPa
tion considering the random properties of load ratio. Representative
cases are established by selecting three targeted reliability indexes, and Upper 345 2.04
the optimum design factors are obtained accordingly to minimum the chord

reliability differences between the calculated reliability and targeted

one among cases. The calibration results show that the recommended
design factors can achieve the goal better and has better performance of
anti-progressive collapse. The results obtained in this paper will enrich
the reliability design and Anti-progressive collapse performance of BSS.
Lower 163D5 1670 1.90
string
2. Ultimate capacity of BSS with different load ratios struts 345 2.03

2.1. Beam string structural analysis model

In practical engineering problems, two types of BSS, namely, beam

string pipeline crossing (BSP, e.g. Shanghai Yuanshen Arena, China) and
truss string structure (TSS, e.g. Harbin International Exhibition Center,
China) are popular. They differ from the fact that they have different Note: 163D5 means that lower string has 163 wires with diameter 5 mm.

2167
H. Zhou et al. Structures 33 (2021) 2166–2176

Table 2
Four Models of BSP.
BSP Span/m Structural height/m Upper chord section/mm

Model 0 63 6.65 600 × 400 × 18

Model 1 42 3.36 500 × 350 × 12
Model 2 63 7.88 600 × 400 × 12
Model 3 77 9.63 700 × 500 × 15

Fig. 2. Finite element model of BSP.

(a)

Fig. 3. Experimental and analytical mid-span deflections of a BSP model.

Fu = gu + qu (1)

(b) Let load ratio be defined as r = q/g. Then, taking 8 beam string
structure models as examples, the variations of Fu with different values
Fig. 1. Model of a TSS:(a)3D view, (b) Truss configurations. of r are shown in Fig. 4.
It can be seen from Fig. 4 that with the increases of load ratio from
0.1 to 20, the ultimate capacities of the BSS models decrease dramati­
Table 3
cally. For example, the ultimate capacity of TSS Model 0 decreases from
Sectional dimensions of structure.
158.4kN/m to 73.8kN/m, by about 53%; while that of BSP Model 2
Member D/mm t/mm Member Area/mm2
decreases from 73.92kN/m to 25.03kN/m, by about 66%.
Chord 1, 2 480 t1, t2 Cable 16,895
Strut 1 168 6 Strut 3 3051
Strut 2 273 7 Strut 4 7961

Note: D refers to the outer diameter of the section, and t refers to the section
thickness.

Table 4
Four Models of TSS.
t/mm TSS
Model 0 Model 1 Model 2 Model 3

t1 18 16 15 13
t2 12 16 18 22

2.3. Variation of ultimate capacity for BSS with different load ratios

For BSS with full-span load g (e.g. dead load) and half-span load q (e.
g. snow load), the ultimate capacity Fu can generally be expressed as the
sum of the ultimate loads: gu and qu, and is given by: Fig. 4. Ultimate capacity of two beam string structures under different
load ratios.

2168
H. Zhou et al. Structures 33 (2021) 2166–2176

3. Analysis of ultimate capacity for BSS with random load ratio Table 6
Statistics of ultimate capacity for FLR case.
3.1. Statistics of variables for capacity analysis rn BSP TSS
Mean (kN/m) COV mean(kN/m) COV
As mentioned earlier, the current design method following the FLR 0.25 81.48 0.062 147.63 0.063
criterion. Following this, the load ratio adopts a fixed value. Usually, a 0.5 70.99 0.066 141.12 0.075
nominal load ratio rn is considered and given by 1 60.17 0.068 111.99 0.074
2 50.26 0.064 90.93 0.068
rn = qn /gn (2)

where qn and gn are nominal values of loads. Actually, the RLR criterion
is more realistic due to random properties of load g and load q. Herein,
the variation analysis of ultimate capacity is compared for these two
cases. Generally, three random variables: load g and q and steel strength
fy, were selected to be considered for their significant effects (see [34]),
as shown in Table 5.
For FLR case, the finite element analysis results of BSP Model 0 and
TSS Model 0 with rn = 0.25, 0.5, 1, and 2, respectively, are obtained by
sampling with 1000 runs and shown in Table 6. For typical case rn =
0.25, the frequency histograms of the ultimate capacity for the BSP
Model 0 and the TSS Model 0 are shown in Figs. 5 and 6, respectively.
From Table 6, it can be seen that the COV for ultimate capacity of two
models is about 0.07, which is close to the statistics of steel strength
shown in Table 5. The reason is that only random properties of steel
strength is involved in this case.
It can be seen from Table 7 that the COV of ultimate capacity in RLR Fig. 5. Frequency histogram of ultimate capacity for BSP Model 0 with rn
case is close to 0.1, which is about 42% larger than that in FLR case. = 0.25.
Therefore, it can be stated that the randomness of the load ratio has an
important influence on the variability of the ultimate capacity for the
BSS (Figs. 7 and 8).

4. Capacity failure function and reliability analysis of BSS

4.1. Simplified capacity model with different load ratio

As early as in 2011, in order to obtain the variation law of capacity

under different load ratio, a practical capacity model of some typical
structures (e.g. arch structures and beam string structures) under the
combination of full-span and half-span load is proposed by Jiang et al.
[31], and this model is proved to be well applied through examples
verifications. With this model, a relative coefficient of ultimate capacity
λ is introduced, and is given by
Fu (r) p1 r + p2 Fig. 6. Frequency histogram of ultimate capacity for TSS Model 0 with rn
λ(r) = = (3)
Fu (r = 0.1) r + p3 = 0.25.

where p1, p2 and p3 are related parameters. If r=∞ (nearly half-span load
applied only), λ is close to p1 in this equation. For this sake, p1 can be Table 7
selected to denote the ratio of the capacity with only half-span load to Statistics of ultimate capacity for RLR case.
that with r = 0.1. For four BSP models and four TSS models above, the Models rn mean(kN/m) COV
values of these parameters are shown in Table 8. The accuracy of model BSP Model 0 1.0 59.21 0.094
fitting was measured by analyzing the determination coefficient (R2). TSS Model 0 1.0 110.38 0.104
Since there is no intercept term in the fitting capacity model, R2 may be
greater than 1. From Table 8, it is seen that the capacity model can be
applied well for BSS models. The fitting results of representative models Fu = Fu (r, fy ) (4)
are shown in Fig. 9.
As well known, the ultimate bearing capacity Fu is dependent of not Then, the more realistic failure function following the RLR criterion
only steel strength fy but also load ratio r, thus it is given by is expressed as
Fu (r, fy ) − g − q = 0 (5)

Table 5 As mentioned earlier, the current design method for bearing capacity
Distributions of three kinds of random variables. of BSS is mainly based on the FLR criterion. According to this criterion, it
Variable Distribution Mean COV Reference is assumed that the load variables does not affect the structural bearing
capacity, only the resistance variables (e.g. steel strength) causes the
g/gn Normal 1.06 0.07 [35]
q/qn Type I largest 1.14 0.256 [35] changes of the structural bearing capacity. If Fu1 is defined as the ulti­
fy/fyn Normal 1.09 0.07 [35] mate capacity under the FLR with r = rn, which is given by
Note: terms with subscript ‘n’ refers to the nominal value of this term.

2169
H. Zhou et al. Structures 33 (2021) 2166–2176

to its small COV (less than 0.05 reported in [35]), its random properties
are neglected for simplification. For BSS in engineering practice, cables
in BSS are usually designed with high safety level, and upper chord
failure usually dominates the significant failure modes. Thus, it can be
regarded as a structure with only single steel material for failure un­
certainty analysis. Then, the ultimate capacity with FLR for uncertainty
analysis is given by
fy c
Fu (rn , fy ) = ΩM F (rn , fyn ) (10)
fyn u

where Fcu(rn, fyn) is the nominal value of ultimate capacity calculated

with the nominal load ratio rn and the nominal strength fyn, and denoted
Fig. 7. Frequency histogram of ultimate capacity for BSP in RLR case. by Fcun for simplification; ΩM is the model uncertainty of resistance
calculation. Then the normalized ultimate capacity is expressed by
Fu1 fy
c
= ΩM (11)
Fun fyn

If upper chord failure is considered for the structure, which is sub­

jected to bending and compression, then the uncertainty of resistance
calculation model of BSS can be selected as that of steel members sub­
jected to bending and compression. Zhang [35] reported that for engi­
neering practices in China, the mean and COV of fy/fyn are 1.09 and
0.07, and the mean and COV of ΩM is 1.12 and 0.10, respectively. On the
basis of Eq. (10), the mean and COV for the normalized ultimate ca­
pacity Fu1/Fcun can be calculated as 1.09 × 1.12 ≈ 1.22
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
and 0.072 + 0.102 ≈ 0.12, respectively, which is also assumed to be
normal variable for simplification.

Fig. 8. Frequency histogram of ultimate capacity for TSS in RLR case. 4.3. Reliability analysis of structural ultimate capacity

Fu1 = Fu (rn , fy ) (6) To satisfy a required target reliability level, the nominal resistance is
often determined by magnifying the nominal load effects K times, and is
then the corresponding limit state equation is expressed by given by
Fu (rn , fy ) − g − q = 0 (7) Fun = K(gn + qn ) (12)
In Eq. (3), it is assumed that the coefficient of ultimate capacity λ is
where K is a safety factor. If load and resistance factors (e.g. dead load
only dependent of load ratio but independent of steel strength. Based on
partial factor γg, live load partial factor γ q, resistance partial factor γR)
this assumption, the ultimate capacity satisfies the following equation
are used, then Eq. (12) can be rewritten as
Fu (r, fy ) λ(r)
= (8) Fun = γ R (γ g gn + γq qn ) (13)
Fu (rn , fy ) λ(rn )
In order to verify the accuracy of the simplified method, two typical
Then, substitute Eq. (8) into Eq. (5), and the limit state equation with
cases (K = 1.7, rn = 4.0; K = 2.0, rn = 1.0) are selected to perform
RLR is easily built as
reliability analysis with the simplified method and Monte Carlo method
λ(rn ) (4000; 10,000 runs, respectively) for the TSS model 0. Herein, the Monte
Fu (rn , fy ) − (g + q) =0 (9)
λ(r) Carlo method is performed by direct finite element sampling, consid­
ering the uncertainty of steel strength fy, random loads including
Comparing Eq. (7) with Eq. (9), it is found that they are largely
random ratio r, and resistance calculation model uncertainty ΩM. The
different and the limit state equation with RLR is more complex.
reliability indexes obtained by these two methods are shown in Table 9.
It is seen that the reliability indexes with the simplified method are
4.2. Uncertainty of resistance calculation model close to those with the Monte Carlo method, but need much less
computational cost. Thus, the simplified method is used to efficiently
With regards to a BSS under a given load ratio, the stochastic char­ perform the following parametric reliability analysis.
acteristics of ultimate capacity are related to the uncertainties of steel The reliability indexes of BSP model 0 and TSS model 0 in different
strength, section dimension and resistance calculation model. Taking cases are shown in Tables 10 and 11, respectively.
into account the small influences of section dimension on reliability due From Tables 10 and 11, it can be seen that when K and rn given, the

Table 8
Related parameters of capacity model for BSS.
Parameters BSP TSS
Model 0 Model 1 Model 2 Model 3 Model 0 Model 1 Model 2 Model 3

p1 0.39 0.57 0.32 0.33 0.41 0.60 0.71 0.82

p2 0.55 0.92 0.43 0.45 1.31 2.19 1.34 2.23
p3 0.49 0.89 0.36 0.39 1.23 2.15 1.32 2.23
R2 1.017 0.967 1.018 0.975 0.980 0.994 1.014 1.035

2170
H. Zhou et al. Structures 33 (2021) 2166–2176

load) is large, it may lead to an unsafe design, and the structure will
possibly collapse. For example, the structure of the roof of the ice-
skating rink in Bad Reichenhall, Germany collapsed due to blizzard
attack (Dietsch et al. [36]). In addition, Takahashi et al. [33] analyzed
the reliability of the steel roof members under snow disaster and found
that the reliability level of such members designed according to the
Japanese building code is low.

5. Research on values of design partial factors

5.1. Explanations on target reliability indexes

For the target design reliability index of structural member, it is

prescribed in Chinese code [37], as shown in Table 12.
As known, the ultimate capacity of BSS is usually controlled by the
(a) BSP Model 0
upper chord failure, which is flexural-compressive buckling and presents
a brittle failure. According to Table 12, the target reliability index can be
selected as 3.2, 3.7 and 4.2 for not important, normal, important safe
grades, respectively. However, for large load ratio cases (e.g. rn = 4.0),
the reliability of BSS models designed by the current design method with
K = 1.7 ~ 2.1 are much lower than the target one, and the reliability
differences is large among different cases. Thus, the current design
method can not be applied well for different demands and needs to be
improved. Herein, based on reliability calibration, two improved
methods: improved design 1 and 2 are proposed to try to achieve the
goal. The former uses three sets of fixed partial factors for three different
safe grades, and the latter uses variable partial factors with cases.
As mentioned earlier, p1 is a significant parameter. Among the
models studied, p1 is about from 0.32 to 0.82 as shown in Table 8.
Herein, 3 representative models are selected as the TSS model 0 (p1 =
0.41), the TSS model 3 (p1 = 0.82) and the BSP model 2 (p1 = 0.32).
(b) TSS Model 0 Moreover, the nominal load ratios are selected as 0.25, 0.5, 1, 2 and 4,
respectively. Thus, in the following analysis, 15 structural cases are
Fig. 9. Fitting results of representative models.
considered totally.

Table 9 5.2. Optimal partial factors for improved design 1

Analysis results with two methods for TSS model 0.
Two Typical cases Monte Carlo method Simplified method Generally, the optimal design partial factors are defined as those
β Time(s) β Time(s) which can make the design reliability agree with target reliability index
K = 1.7, rn = 4.0 2.27 41,040 2.08 5 well for different cases. To find them, multiple sets of tentative design
K = 2.0, rn = 1.0 2.79 101,602 3.05 5 partial factors are selected for analysis. For each set of design partial
factors, the summed reliability error I between design reliability indexes
and the target ones can be expressed as:
Table 10
∑
15
Reliability indexes of BSP model 0 under different cases. I = (β − [β])2 (14)
K FLR RLR i=1

rn = 1.0 rn = 4.0 rn = 1.0 rn = 4.0

where [β] is the selected target reliability index.
1.7 3.04 2.42 2.49 2.19
1.9 3.40 2.81 2.85 2.55
Herein, as many as 180 sets of tentative design partial factors are
2.1 3.87 3.13 3.19 2.85 selected for each target reliability. Through large number of calcula­
tions, the results show that the optimal design partial factors for
improved method 1 are γ g = 1.15, γ q = 2.3, γ R = 1.15; γ g = 1.15, γ q = 2.1,
Table 11 γ R = 1.40; and γg = 1.15, γq = 2.4, γR = 1.50 for the target reliability
Reliability indexes of TSS model 0 under different cases. indexes 3.2, 3.7 and 4.2, respectively, as shown in Table 13.
It is found that the optimum dead load partial factors γg are 1.15 for
K FLR RLR
rn = 1.0 rn = 4.0 rn = 1.0 rn = 4.0
all and these partial factors can achieve the mean reliability close to the
target one for the corresponding safe grade (e.g. βmean = 3.08 close to
1.7 3.03 2.43 2.51 2.09
[β] = 3.2 for normal safe grade). To illustrate it clearly, when γ g = 1.15,
1.9 3.49 2.81 2.88 2.42
2.1 3.80 3.13 3.20 2.72
Table 12
Target design reliability index of structural member.
reliability indexes with RLR are all lower than those with FLR.
Failure mode Safe grade
Furthermore, when the load ratio rn is 4.0, the maximum reliability Important Normal Not important
index with RLR is only about 2.85 (K = 2.1, BSP model 0) lower than 3.0.
Ductile 3.7 3.2 2.7
Therefore, in practical engineering, when the half-span load (e.g. snow
Brittle 4.2 3.7 3.2

2171
H. Zhou et al. Structures 33 (2021) 2166–2176

Table 13 reliability calibration from case to case, the obtained optimal values of
Optimal design partial factors for three target reliability indexes. γ R are shown in Fig. 11.
[β] γg γq γR Imin It can be seen from Fig. 11 that the optimal γR is not constant, which
varies from 1.1 to 1.25 for [β] = 3.2 cases, from 1.3 to 1.5 for [β] = 3.7
3.2 1.15 2.3 1.15 0.79
3.7 1.15 2.1 1.4 1.12 cases and from 1.4 to 1.75 for [β] = 4.2 cases, respectively.
4.2 1.15 2.4 1.5 1.2

5.4. Comparisons between different design methods

the variations of the summed reliability error I with different partial
factors γ q and γ R for all 15 structural cases are shown in Fig. 10. It is seen By comparison, the robustness of the three methods: conventional
that the improved design 1 still lead to a large summed reliability error method (e.g. K = 1.9, 2.1), improved method 1, and improved method 2,
(e.g. larger than 1.2 for [β] = 4.2 for structural cases) is evaluated respectively, and the results are shown in Tables 14 and 15.
It is seen that the average COV with the improved method 2 is the
least (about 0.01), which is about 90% lower than that with the
5.3. Optimal partial factors for improved design 2 improved method 1, and about 95% lower than that with the conven­
tional method. This indicates that the design methods with fixed partial
As mentioned earlier, the improved design 1 may result in a large factors (including both the improved design method 1 and the conven­
reliability differences among different cases. Thus, the structural design tional design method) cannot achieve a robust design for the given
will be unsafe or conservative for cases. To overcome this shortcoming, target reliability level, because the reliability is scattered over a large
another method: the improved design 2 is proposed. It uses fixed load range among cases. The results also show that the average reliability
partial factors, γg = 1.15 and γq = 2.27 (average value of 2.3, 2.1 and 2.4 with the improved method 2 is much closer to the target reliability index
for 3 safety grades shown in Table 13) and a varing resistance partial value than that with the improved method 1 and the conventional
factor γR. The optimal γ R is the one with the design reliability closest to design method. Therefore, the improved method 2 can achieve a more
the targeted reliability index for total 45 different cases. Based on robust reliability design.

6. Progressive collapse analysis of BSS designed by improved

methods

6.1. Introduction for progressive collapse analysis

The aforementioned reliability analysis shows that the conventional

design method could overestimate the reliability of BSS with random
load ratio, resulting in a possible unsafe design, and two improved
design methods are proposed to address this phenomenon. It is known
that structural progressive collapse has increasingly drawn attentions of
researchers and engineers. Herein, a comparison of the improved
methods is further analyzed based on the progressive collapse resistance
(a) [ ]=3.2 of BSS with local failure. The safety factor should usually be required
more than 2.5 for cables according to CECS 212-2006 [38], which is
much larger than that for other members (e.g., strut) in BSS, therefore it
is more likely that the local failure occurred in rigid members or anchor
nodes of cable. In this section, taking BSP Model 2 with the target reli­
ability index of 3.7 as an example, the influences of strut failure or an­
chor failure on the anti-progressive collapse performance are discussed
for structures designed by the two improved methods.
At present, there are several methods for simulations of actions of
failure member in structural progressive collapse analysis, including
static analysis method considering dynamic increment factor, and
equivalent load transient unloading method considering initial

2.0
(b) [ ]=3.7
[ ]=3.2 BSP Model 2 [ ]=3.2 TSS Model 0 [ ]=3.2 TSS Model 3
[ ]=3.7 BSP Model 2 [ ]=3.7 TSS Model 0 [ ]=3.7 TSS Model 3
1.8 [ ]=4.2 BSP Model 2 [ ]=4.2 TSS Model 0 [ ]=4.2 TSS Model 3

1.6
R

1.4

1.2

1.0
(c) [ ]=4.2 0.25 0.5 1 2 4
rn
Fig. 10. Curves of summed reliability error with different design partial factors
for BSS. Fig. 11. Optimal values of γR for different cases.

2172
H. Zhou et al. Structures 33 (2021) 2166–2176

Table 14 Therefore, the constitutive model of the cable is assumed to be fracture

Robustness evaluation for BSS in conventional method. failure after reaching the ultimate strength, as shown in Fig. 13.
Conventional Method βmax βmean βmin COV The collapse of the structure is a transient process, in which the strain
rate of steel is very large, so the influence of the material strain rate
K = 1.9 4.20 3.13 2.42 0.18
K = 2.1 4.57 3.48 2.72 0.17 should be considered in the analysis. The Cowper-Symonds constitutive
equation is in good agreement with the experimental data and is widely
used. In this paper, the Cowper-Symonds constitutive equation is used to
conditions, and full dynamic equivalent load transient unloading consider the strain rate effect, it can be expressed as:
method. It is reported by Zhu et al. [30] that the full dynamic equivalent [ (ε )1p ]
load transient unloading method could effectively simulate initial con­ σ d /σ 0 = 1 + r (15)
ditions under the static load before local member failure, and eliminate C
the unnecessary dynamic influences of static load on the structure.
where σd is the dynamic yield stress and σ0 is the associated static yield
Herein, this method is also used to perform progressive collapse analysis
stress, εr is the strain rate. C and p are strain rate parameter, set as 40.4
for BSS with local failure of strut or anchor end of cable. The main steps
and 5.0 respectively (see [41]).
are as follows:
At present, there are many simulations aimed at the anti-progressive
collapse of steel structures, but the value of steel failure strain is selected
(1) Carry out a static analysis of the whole structure to extract the
differently. Xie et al. [42] used a failure strain of 3.7% for columns
internal force P of the local failure member (e.g., strut) under the
subjected to bending and compression to study the dynamic behavior of
given load case.
steel frames during progressive collapse, and the results are in good
(2) Remove the assumed local failure member as shown in Fig. 12(a),
agreement with the experimental results. Jiang et al. [43] established a
and apply the equivalent internal force P as shown in Fig. 12(b) to
fiber model failure simulation method based on FEMA 356, and stated
the remaining structure. In t0 period, the original static load (full-
that when the ultimate strain of the steel is set to 2.5%, the structural
span load and half-span load) and the equivalent load P increase
responses of failure member during the dynamic process can meet the
from zero to the maximum; t1 is the load duration and can be
determined by the complete attenuation of forced vibration of the
structure under the actions of both original static load and the
equivalent load P; TP is the local member failure stage and taken
as 0.00375 s (see [39]); T2 is the attenuation stage when vibration
amplitude continuously attenuates under the actions of damping
until reaching the final state of stability.

6.2. FEA modeling of BSS

Based on the techniques on ANSYS/LS-DYNA software in [40], the

nonlinear dynamic calculation model with both dynamic and nonlinear
effects considered was built for simulation following the alternative load
path method. It is known that the structure collapse behavior can be
simulated better by the nonlinear dynamic calculation model. The ma­
terial constitutive model of the upper chord and the strut is ideal elas­
toplastic model. The cable is composed of high-strength steel wire, with
Fig. 13. Stress–strain curve of the cable.
poor ductility, and the failure is characterized by brittle fracture.

Table 15
Robustness evaluation for three target reliability indexes in two methods.
Method [β] = 3.2 [β] = 3.7 [β] = 4.2
βmax βmean βmin COV βmax βmean βmin COV βmax βmean βmin COV

Improved Method 1 3.69 3.08 2.50 0.09 4.75 3.60 2.34 0.11 5.16 4.02 3.14 0.10
Improved Method 2 3.24 3.20 3.15 0.009 3.76 3.70 3.65 0.01 4.26 4.20 4.14 0.01

Fig. 12. The full dynamic equivalent load transient unloading method.

2173
H. Zhou et al. Structures 33 (2021) 2166–2176

requirements of the member deformation limit in FEMA 356 well. Tian Under the two improved methods, the deformation law of the structure
et al. [44] used the failure strain of 2.5% to simulate the progressive before and after the strut failure is the same, but the deformation value
collapse of a large station structure, and stated that when the steel for the improved method 2 is smaller than that for the improved method
reached this strain value, the structure had excessive deformation and 1.
was not conducive to personnel escape and rescue operations. Therefore, The safety factor of the cable is usually large, but in engineering
the failure strain εf of steel is assumed as 0.025 herein for safety reasons. practices, the cable is still possibly broken due to accidental factors such
The element would fail if the strain ε > εf, and it is automatically deleted as material quality defects, maintenance defects, construction defects
from the FEA model. The automatic single-sided contact ASSC is [45,46]. Moreover, the failure of cable can lead to the overall collapse of
selected, and the structural damping is assumed as Rayleigh damping BSS. In order to disperse the risk, some researchers propose to split a
with the damping ratio 0.02. single cable into multiple cables. For this case, even if one cable fails, the
structure can still guarantee the existence of alternative load path [47].
In order to compare the cable safe margin of BSS designed by the two
6.3. Anti-progressive collapse performance of BSS designed by different
improved methods, assuming that part of the cable at the anchorage end
methods
of the cable fails, the critical value of cable area loss leading to BSS
collapse is expressed by introducing coefficient ρ, which can be
Taking the BSP Model 2 with target reliability index 3.7 as an
expressed as:
example, the optimum load partial factors and resistance partial factors
for the improve method 1 and method 2 are provided in Table 13 and Al
ρ= (17)
Fig. 11, respectively. If two representative load ratios 0.5 and 2.0 are Ai
selected, then the corresponding safety factor K is calculated with the
Eq. (12) and Eq. (13), and the dead load nominal values gn and the live where Ai is the area of intact cable, Al is the area of cable loss. The
load nominal values qn are obtained from the ultimate capacity of BSP calculation results are shown in Table 19.
Model 2 presented in Fig. 4. Finally, these parameters are all shown in It is seen that for the BSS designed with the given method, the cable
Table 16. safe margin with rn = 2.0 (e.g. 81.1%) is greater than that with rn = 0.5
It can be seen from Table 16 that for the given design method, the (e.g. 62.8%), which is also consistent with the rule of the safety factor in
safety factor with rn = 2.0 is larger than that when with rn = 0.5; four cases in Table 16. For these two improved methods, the cable safe
moreover for the given load ratio, the safety factor of improved method margin of BSS designed by improved method 2 is greater than that by
2 is larger than that of improved method 1. improved method 1. This indicates that the BSS designed by the
Due to the higher redundancy of the strut, this paper considers the improved method 2 shows a better anti-progressive collapse perfor­
strut failure in two different positions, namely the end Strut 1 and the mance than that designed by the improved method 1.
middle Strut 2. For simplification, they are denoted as ESF (end strut
failure) and MSF (middle strut failure) cases, respectively. Herein, the 7. Summary and conclusions
combined load is defined as 1.0gn + 0.5qn, and assumed as the equiva­
lent nodal forces acted at the upper chord nodes for analysis. When the This paper is aimed at investigating the bearing capacity reliability of
BSS which are designed by the two improved methods bear full-span beam string structures (BSS) considering the stochastic characteristics of
dead load and half-span live load , the overall deformation of the load ratio. The variations of the ultimate capacity for BSS with fixed load
intact structure is S-shaped with upward arch on the left side and ratio and random load ratio are compared and analyzed. A more realistic
concave downward on the right side, as shown in Fig. 14. limit state function based on some simplified ways is established to
The sensitivity index of any node in the structure corresponding to achieve an efficient analysis of the ultimate capacity reliability of BSS
the removal of member i can be expressed as: with a random load ratio. Finally, the optimal design partial factors are
searched accordingly to minimise the reliability differences between the
δi = (Δ − Δ’ )/Δ (16) design reliability and the target one. The performance of the anti-
Where Δ and Δ’ are the displacements of the same node of the intact progressive collapse of BSS designed by two improved methods is
structure and the damaged structure, respectively. When the strut fails, further compared. The main results can be summarised as below:
the dynamic response of the maximum upward displacement node and
the maximum downward displacement node are extracted, as well as the )1. With the increases in load ratio of half-span load to full-span load,
mid-span node. The node numbers are shown in Fig. 15. The calculation the ultimate capacity of BSS decreases dramatically. When the
results are shown in Tables 17, 18. load ratio increases from 0.1 to 20, the ultimate capacity de­
When rn = 0.5, due to the failure of Strut 1, the vertical displacement creases by as much as 66%.
of Node a-3 increases, the vertical displacement of node a-2 changes a )2. The randomness of load ratio has a significant influence on the
little, and the deformation of node a-1 decreases; When rn = 2.0, due to variability of the ultimate capacity for BSS, and the COV of ulti­
the failure of Strut 1, the vertical displacement of nodes b-2 and b-3 mate capacity for BSS with random load ratio is about from 38%
changes greatly, while the change of node b-1 is not significant. The to 65% larger than that with fixed load ratio within the param­
deformation law of BSS after Strut 2 failure is similar to the former, the eters in the analysis.
difference is that the vertical displacement of node a-2 and b-2 changes )3. With the simplified capacity model, a more realistic limit state
greatly, the maximum change is 63%. It shows that the local stiffness of equation can be established and applied efficiently in reliability
the upper chord near the strut decreases due to the failure of a strut, analysis. The results show that the reliability index with the
which leads to the increase of local deformation of the upper chord. random load ratio criterion is lower than that with the fixed load
ratio criterion.
)4. The design method with fixed partial factors can lead to large
Table 16
Parameters of BSS designed by different methods.
differences between the design reliability and target one among
cases and result in an unsafe design for large load ratio cases for
Design parameters Improved Method 1 Improved Method 2
BSS. However, the improved design method with fixed load
rn = 0.5 rn = 2.0 rn = 0.5 rn = 2.0
partial factors and optimum resistance factor, which varies with
K 2.053 2.497 2.285 2.845 cases, can dramatically decrease the reliability differences for
gn(kN/m) 16.52 4.46 14.85 3.91
qn(kN/m) 8.26 8.92 7.43 7.82

2174
H. Zhou et al. Structures 33 (2021) 2166–2176

Fig. 14. Deflections of the upper beam of BSS designed by improved method 1.

Fig. 15. Failed struts in BSS.

Table 17
Node displacement before and after strut failure for Improved Method 1.
Node number rn = 0.5 Node number rn = 2.0
Δintact ΔESF δESF ΔMSF δMSF Δintact ΔESF δESF ΔMSF δMSF

a-1 0.025 0.014 44% 0.017 32% b-1 0.120 0.110 8% 0.101 16%
a-2 − 0.095 − 0.099 − 4% − 0.131 − 38% b-2 0.043 0.031 28% 0.016 63%
a-3 − 0.148 − 0.181 –22% − 0.167 − 13% b-3 − 0.048 − 0.073 − 52% − 0.074 − 54%

Table 18
Node displacement before and after strut failure for Improved Method 2.
Node number rn = 0.5 Node number rn = 2.0
Δintact ΔESF δESF ΔMSF δMSF Δintact ΔESF δESF ΔMSF δMSF

a-1 0.029 0.009 69% − 0.021 28% b-1 0.117 0.106 9% 0.094 20%
a-2 − 0.074 − 0.082 − 11% − 0.107 − 45% b-2 0.053 0.039 26% 0.031 42%
a-3 − 0.123 − 0.153 − 24% − 0.141 − 15% b-3 − 0.035 − 0.060 − 71% − 0.058 − 66%

interests or personal relationships that could have appeared to influence

Table 19
the work reported in this paper.
The critical value ρ of cable area loss leading to the collapse of the BSS.
Method ρ Acknowledgement
rn = 0.5 rn = 2.0

Improved Method 1 62.80% 81.10% The research is supported by the National Natural Science Founda­
Improved Method 2 64.50% 82.50%
tion of China (Grant No. 51678072), the Key Discipline Foundation of
Civil Engineering of Changsha University of Science and Technology
different target reliability levels and meet the target reliability (18ZDXK01), and the Hunan Provincial Innovation Foundation for
indexes well. Postgraduate (CX20190655). This support is gratefully acknowledged.
)5. The representative BSS designed by improved design method 2
with fixed load partial factors and optimum resistance factor, References
which varies with cases, has better performance of anti-
[1] Dong SL, Zhao Y, Xing D. Application and development of modern long-span space
progressive collapse than that designed by the improved structures in China. Front Struct Civ Eng 2012;6(3):224–39. https://doi.org/
method 1 with fixed partial factors. 10.1007/s11709-012-0166-6.
[2] Zhao XZ, Wu A, Xu Z, et al. Research and application of beam string structures.
Struct Eng Int 2015;25(1). https://doi.org/10.2749/
The reliability of BSS can be evaluated more reasonably by the 101686614X14043795570219. 26-33(8).
method proposed in this paper. To attain robust design results, it is [3] Cai JG, Feng J, Jiang C. Development and analysis of a long-span retractable roof
recommended to select optimum resistance factors according to the structure. J Constr Steel Res 2014;92(92):175–82. https://doi.org/10.1016/j.
jcsr.2013.09.006.
actual cases instead of using the fixed resistance factors. Further studies
[4] Luo YZ, Li Y, Shen YB, et al. Experimental study on cable-truss reinforced structure
are needed on how to enhance the redundancy of BSS and prevent system under heavy load. China Civ Eng Journal 2017;50(4):48–56. https://doi.
collapse after a local cable failure. org/10.15951/j.tmgcxb.2017.04.006 (in Chinese).
[5] Han Q, Wang L, Xu J. Test and numerical simulation of large angle wedge type of
anchorage using transverse enhanced CFRP tendons for beam string structure.
Constr Build Mater 2017;144:225–37. https://doi.org/10.1016/j.
Declaration of Competing Interest conbuildmat.2017.03.150.

The authors declare that they have no known competing financial

2175
H. Zhou et al. Structures 33 (2021) 2166–2176

[6] Saitoh M, Kuroki F. Structural planning of beam string structure. In: Space [26] Malla RB, Nalluri BB. Dynamic effects of member failure on response of truss-type
Structures for Sports Buildings-Proceedings of the International Colloquium on space structures. J Spacecraft Rockets 1995;32(3):545–51. https://doi.org/
Space Structures for Sports Buildings; 1995, 1987,. p. 693–4. 10.2514/3.26649.
[7] Saitoh M, Nakakawaji I. A Study on structural characteristic of beam string [27] Murtha-Smith EA. Alternate path analysis of space trusses for progressive collapse.
structure : Part 1 Prestressing for Dead Load. Summaries of Technical Papers of J Struct Eng ASCE 1988;114(9):1978–99. https://doi.org/10.1061/(ASCE)0733-
Meeting Architectural Institute of Japan. Architectural Institute of Japan, 1987. 9445(1988)114:9(1978).
[8] Saitoh M, Okada A. The role of string in hybrid string structure. Eng Struct 1999;21 [28] Hu SL. Progressive collapse analysis and anti-collapse design of beam string
(8):756–69. https://doi.org/10.1016/S0141-0296(98)00029-7. structure. Beijing: Beijing University of Technology,2010: 1–154 (in Chinese).
[9] Kato S, Nakazawa S, Matsue Y, et al. Active control of axial forces in beam string [29] Cai JG, Wang FL, Feng J, et al. Progressive collapse analysis of cable-arch
space frames. IASS-ASCE International Symposium, 1994:664–673. structures of the New Guangzhou Railway Station. J Build Struct 2010;31(7):
[10] Kato S, Nakazawa S, Okada Y. optimum locations for actuators to reduce the 103–9. https://doi.org/10.14006/j.jzjgxb.2010.07.013 (in Chinese).
responses of beam string struc structures to static and dynamic disturbance. Asia- [30] Zhu YF, Feng J, Cai JG, et al. Analysis on progressive collapse resistance of truss
pacific Conf Shell Spat Struct 1996:688–95. string structure of Meijiang Exhibition Center. J Build Struct 2013;34(3):45–53.
[11] Thai HT, Kim SE. Nonlinear static and dynamic analysis of cable structures. Finite https://doi.org/10.14006/j.jzjgxb.2013.03.003 (in Chinese).
Elem Anal Des 2011;47(3):237–46. https://doi.org/10.1016/j.finel.2010.10.005. [31] Jiang YB, He YH. Research on load effect function model under full-span and half-
[12] Abad MSA, Shooshtari A, Esmaeili V, et al. Nonlinear analysis of cable structures span loading. Eng Mech 2011;28(2):123–8 (in Chinese).
under general loadings. Finite Elem Anal Des 2013;73(73):11–9. https://doi.org/ [32] Jiang YB, Zhou H, Beer Michael, et al. Robustness of load and resistance design
10.1016/j.finel.2013.05.002. factors for RC columns with wind-dominated combination considering random
[13] Jiang ZR, Xu Mu, Duan WN, Shi KR, Cai J, Wang ST. Nonlinear finite element eccentricity. J Struct Eng 2017;143(4):1–9. https://doi.org/10.1061/(ASCE)
analysis of beam string structure. Adv Mater Res 2011;163–167:2124–30. https:// ST.1943-541X.0001720.
doi.org/10.4028/www.scientific.net/AMR.163-167. [33] Takahashi T, Ellingwood BR. Reliability-based assessment of roofs in Japan
[14] Wu M. Analytical method for the lateral buckling of the struts in beam string subjected to extreme snows: Incorporation of site-specific data. Eng Struct 2005;27
structures. Eng Struct 2008;30(9):2301–10. https://doi.org/10.1016/j. (1):89–95. https://doi.org/10.1016/j.engstruct.2004.09.001.
engstruct.2008.01.008. [34] Jiang YB, Zhou CY, Zhou H. Robustness of ultimate capacity for arch truss string
[15] Wu M, Hirai K. Lateral buckling of the struts in beam string structures considering structure with random ratio of half-span load to full-span load. 6th Asian-Pacific
the layout of strings. Int J Struct Stab Dyn 2012;12(03):1250015. https://doi.org/ Symposium on Structural reliability and its applications, 2016, Shanghai, China.
10.1142/S0219455412500150. [35] Zhang XP. Reliability analysis and design for building structures. Beijing: Science
[16] Ye J, Feng R-Q, Zhao X, Liu B. A form-finding method of beam string structures — Press; 2001:1–107 (in Chinese).
Offload by steps method. Int J Steel Struct 2012;12(2):267–83. https://doi.org/ [36] Dietsch P, Winter S. Structural failure in large-span timber structures: A
10.1007/s13296-012-2010-1. comprehensive analysis of 230 cases. Struct Saf 2018;71:41–6. https://doi.org/
[17] Cao ZG, Feng BS. Form finding analysis of large-span spindle Tensairity beam. J 10.1016/j.strusafe.2017.11. 004.
Harbin Inst Technol, 2016, 48(06): 25-29 (in Chinese), https://dx.doi.org/10 [37] GB50068-2001 Unified standard for reliability design of building structures,
.11918/j.issn. 0367-6234.2016.06.004. Beijing, China, Architecture & Building Press (in Chinese).
[18] Jiang YB, Cao Q, K X, et al. Stiffness study of inner concave cable–arch structure [38] CECS 212-2006, Technical specification for prestressed steel structures, Beijing,
based on an efficient method. Adv Struct Eng 2016;19(12):1927–39. https://doi. China, China Planning Press (in Chinese).
org/10.1177/1369433216649394. [39] Mozos CM, Aparicio AC. Numerical and experimental study on the interaction
[19] Xue WC, Liu S. Design optimization and experimental study on beam string cable structure during the failure of a stay in a cable stayed bridge. Eng Struct
structures. J Constr Steel Res 2009;65(1):70–80. https://doi.org/10.1016/j. 2011;33(8):2330–41. https://doi.org/10.1016/j.engstruct.2011.04.006.
jcsr.2008.08.009. [40] LS-DYNA® Keyword User’s Manual Volumn II: Material Models. California:
[20] Xue WC, Liu S. Studies on a large-span beam string pipeline crossing. J Struct Eng Livermore Software Technology Corporation, 2013.
2008;134(10):1657–67. https://doi.org/10.1061/(ASCE)0733-9445(2008) 134:10 [41] Lu Y, Liu K, Wang ZL, Tang WY. Dynamic behavior of scaled tubular K-joints
(1657). subjected to impact loads. Marine Struct 2020;69. https://doi.org/10.1016/j.
[21] Chen FB, Li QS, Wu JR, et al. Wind effects on a long-span beam string roof marstruc.2019. 102685).
structure: Wind tunnel test, field measurement and numerical analysis. J Constr [42] Xie Fuzhe, Gu Bin, Qian Hai. Experimental study on the dynamic behavior of steel
Steel Res 2011;67(10):1591–604. https://doi.org/10.1016/j.jcsr.2011.04.003. frames during progressive collapse - ScienceDirect. J Constr Steel Res 2021;177:
[22] Han QH, Ma CY, Zhang JY. Dynamic stability analysis of beam string structures 106459. https://doi.org/10.1016/j.jcsr.2020.106459.
under earthquake loads. Adv Steel Constr 2007;3(3):679–88. https://doi.org/ [43] Jiang L, Ni JG, Qu G, et al. Study on key issues of progressive collapse resistance
10.1016/B978-008044637-0/50179-7. capacity analysis for complex high-rise steel structure. J Build Struct 2019;40(6):
[23] Chen YJ, Feng ZF, Qi A, et al. The analysis of dynamic characteristics and wind- 155–65. https://doi.org/10.14006/j.jzjgxb.2018.c299 (in Chinese).
induced displacement response of space Beam String Structure, 2018 4th [44] Tian CH, Dong C, Liu M, et al. Simulation analysis of progressive collapse for a
International Conference on Energy Materials and Environment Engineering station. J Railway Eng Soc 2015;32(12):76–9 (in Chinese).
(ICEMEE 2018), 2018,38: 03032. [45] Shoghijavan M, Starossek U. An analytical study on the bending moment acting on
[24] Lee Seunghye, Seo Minhee, Baek Ki-Youl, Jeong Jinwoo, Kim Sun-Myung, the girder of a long-span cable-supported bridge suffering from cable failure. Eng
Lee Jaehong. Experimental study of two-way beam string structures. Eng Struct Struct 2018;167:166–74. https://doi.org/10.1016/j.engstruct.2018.04.017.
2019;191:563–74. https://doi.org/10.1016/j.engstruct.2019.04.033. [46] Guo J, Zhou G, Zhou D, et al. Cable fracture simulation and experiment of a
[25] Lee Seunghye, Seo Minhee, Park Sangeun, Kim Sun-Myung, Lee Kihak, negative Gaussian curvature cable dome. Aerosp Sci Technol 2018;78:342–53.
Lee Jaehong. Geometrical parametric study on two-way beam string structures. https://doi.org/10.1016/j.ast.2018.04.033.
J Korean Assoc Spatial Struct 2019;19(3):69–76. https://doi.org/10.9712/ [47] Liu GG, Wu ZW, Cai J. Study on static and dynamic behaviors of truss string
KASS.2019.19.3.69. structure with double cables. In: Proceedings of Seventh International Conference
on Advances in Steel Structure; 2012. p. 900–7.

2176