Engineering Structures 189 (2019) 206–221

Contents lists available at ScienceDirect

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

Transverse seismic failure mechanism and ductility of reinforced concrete T

pylon for long span cable-stayed bridges: Model test and numerical analysis
⁎
Xiaowei Wanga,b, Jiaxin Fanga, Lianxu Zhoua, Aijun Yea,
a
State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai 200092, China
b
College of Civil and Transportation Engineering, Hohai University, Nanjing 210024, China

A R T I C LE I N FO A B S T R A C T

Keywords: Although often designed to behave elastically, seismic damage to reinforced concrete (RC) pylons of cable-stayed
Quasi-static test bridges have been witnessed in history such as the 1999 Chi-chi earthquake. This paper aims to assess the
inverted Y-shape reinforced concrete pylon transverse seismic failure mechanism and ductile properties of typical inverted Y-shape RC pylons for long span
Cable-stayed bridge cable-stayed bridges using quasi-static model tests and numerical analyses. To facilitate the limited laboratorial
Transverse seismic failure mechanism
loading system, a simplified displacement-controlled two-node load-pattern, one at the bifurcation-node and the
Finite element method
other at the crossbeam, is first proposed using numerical analyses. It is found the ratio of displacements at the
two loading nodes is correlated generally well with the ground motion parameter, bracketed duration. A dis-
placement ratio of 5.0 is then adopted in the test. Test results indicate a flexural damage mode with considerable
ductility: plastic hinges were detected first at bottom of the upper column (i.e., above the crossbeam), then at
bottom and top of the lower column, successively; multi-level displacement ductility factors are proposed to
associate with numbers of plastic hinges formed in the pylon. Moreover, an experimentally validated numerical
model is adopted to study the impact of loading displacement ratios on the failure mechanism and ductility. It is
found that the loading displacement ratios may significantly affect them. Smaller displacement ratios tend to
transfer the location of first plastic hinge from the bottom of the upper column to that of the lower column.

1. Introduction
During the past few decades, hundreds of cable-stayed bridges have
been constructed in the world, mostly in China. Among them, there are
over twenty long span cable-stayed bridges with main spans beyond
600 m, serving as pivotal joints in transportation networks. Due to their
high requirements for post-earthquake serviceability, seismic design
codes often give preference to elastic design strategies for pylons of
cable-stayed bridges under design-level earthquakes [1–3], and nearly
elastic design strategies even under occasionally strong events (e.g.,
return periods over 2,000 years) [1]. To this end, uneconomical high
demands on reinforcement ratios for columns and crossbeams are im-
perative for the design of reinforced concrete (RC) pylons. Nevertheless,
unexpected large earthquakes may cause damage to such bridges. A
documented evidence is the Chi-lu cable-stayed bridge in Taiwan
during 1999 Chi-chi Earthquake [4], where the pylon above the deck
experienced severe concrete spalling in the transverse direction and
formed an approximately 2 m-length plastic region [5]. Therefore,
several later constructed long span bridges (e.g., Stonecutters in China
[6], Rion-Antirion in Greece [7] and Third Tacoma Narrows in USA [8])
allowed limited damage in pylons to account for occasionally strong
earthquakes. On the other hand, seismic isolation devices and viscous
fluid dampers (VFDs) have been widely adopted in the longitudinal
direction to reduce deck-pylon relative displacements [9–14], which in
turn protects the pylons. In the transverse direction, although VFDs and
steel dampers have be adopted or experimentally/numerically studied
for a few bridges [15–19], the rigid connection between pylon and deck
is still the most practical solution to as-built cable-stayed bridges [20],
because it can provide sufficient stiffness for resistance of wind and
traffic loads. With this configuration, pylons may behave as elasto-
plastic components under unexpected strong earthquakes, which sti-
mulates the motivation of this study.
Numerical studies on modelling of RC pylons often use elastic ele-
ments [21–24]. Few studies focused on their inelastic behavior. Endo
et al. [25] studied the seismic yielding process of a H-shape steel pylon
with multi-struts (crossbeams) for a long span suspension bridge.
Pushover analysis was applied transversally to the fiber/shell-element-
based pylon model. Numerical results revealed that the plastic region

⁎
Corresponding author.
E-mail addresses: x.wang@hhu.edu.cn (X. Wang), yeaijun@tongji.edu.cn (A. Ye).
URL: https://orcid.org/0000-0002-4168-4328 (X. Wang).

https://doi.org/10.1016/j.engstruct.2019.03.045
Received 21 December 2018; Received in revised form 15 March 2019; Accepted 16 March 2019
Available online 28 March 2019
0141-0296/ © 2019 Elsevier Ltd. All rights reserved.
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Table 1
Summary of as-built long span cable-stayed bridges with RC pylons.
No. Bridge (Country, Year to open) Main span Lm (m) Pylon height Hp (m) Pylon shape Hp/Lm

1 Russky (Russia, 2012) 1104 320 A 0.290

2 Sutong (China, 2008) 1088 300 Inverted Y 0.276
3 E'dong (China, 2010) 926 242 Inverted Y 0.261
4 Normandie (France, 1995) 856 215 Inverted Y 0.251
5 Second Jiujiang (China, 2013) 818 242 H 0.296
6 Jingyue (China, 2010) 816 267 H 0.327
7 Yachi River (China, 2016) 800 258 H 0.323
8 Incheon (Korea, 2009) 800 231 Diamond 0.289
9 Shanghai Yangtze (China, 2009) 730 216 Inverted Y 0.296
10 Minpu (China, 2009) 708 210 H 0.297
11 Xiangshan Harbor (China, 2012) 688 226 Diamond 0.328
12 Second Nanjing Yangtze (China, 2001) 628 195 Inverted Y 0.311
13 Jintang (China, 2009) 620 204 Diamond 0.329
14 Baishazhou (China, 2000) 618 175 A 0.283
15 Erqi Yangtze (China, 2011) 616 209 Diamond 0.339
16 Qingzhou (China, 2001) 605 175 A 0.289
17 Yangpu (China, 1993) 602 200 Diamond 0.332

first developed at the lowest strut, and then gradually shifted to upper
struts, finally extended to pylon columns. Camara and Astiz [26] pro-
posed a coupled nonlinear static pushover analysis method to estimate
elastic and inelastic responses of inverted Y-shape and Diamond-shape
pylons subjected to multi-directional excitations. Pang et al. [27] and
Zhong et al. [28] assessed the seismic fragility of cable-stayed bridges
with diamond-shape pylons modeled using fiber elements. Although
sophisticated finite element (FE) models are popular for assessments of
structural inelasticity, their reliabilities are highly dependent on related
experimental validations [29].
Shake tables have been used extensively because they can provide
conditions representing real seismic excitations. Early attempts [30–32]
modeled cable-stayed bridge specimens using organic glass in small
scales (i.e., 1/200 ∼ 1/100). Although these studies can generally
capture the elastic behavior of such bridges, they are not capable of
revealing the inelastic behavior of the bridges, especially for those with
RC pylons, because the inelastic behavior of rebar and concrete cannot
be accurately represented by organic glass or other substituted mate-
rials. In addition, the mechanical property of rebar or concrete often has
a certain level of uncertainty. For these reasons, RC materials tend to be
the essential choice for assessing the inelastic behavior of RC pylons. To
the best knowledge of the authors, rare studies have been reported in
this area. Wang et al. [33] investigated the transverse failure process of
a 1/20-scale middle span cable-stayed bridge model supported by twin
H-shape RC pylons, each with two struts. Artificial waves with in-
creasing intensities (i.e., peak ground acceleration (PGA) from 0.1 g to
1.2 g) were transversally input to trigger a full failure process. Test
results exhibited that plastic regions occurred first at the bottom of
pylon, then developed at the lower and upper struts, successively, and
finally concentrated at the upper strut. Recently, Xu et al. [34] per-
formed a shake-table test to study the longitudinal performance of a
similar cable-stayed bridge model with H-shape RC pylons. It is found
that damage of the pylon appeared first at the bottom of pylon, then
extended to columns around the lower strut. In summary, inelastic
behavior of H-shape pylons have been studied experimentally and/or
numerically, while studies on other types of pylons (e.g., diamond-, A-
and inverted Y- shapes) have been rarely reported.
By contrast, another test technique, quasi-static test [35] and asso-
ciated hybrid simulation/test [36], has rarely been used in assessing
seismic behavior of RC pylons. Possible reasons include (1) challenges
of the structural equivalency between a single pylon model and the full
cable-stayed bridge prototype, and (2) complexities of multi-point
loading regimes for the pylon model due to its multi-degree-of-freedom
properties. Nevertheless, considering its cost effectiveness as compared
to shake-table test, the quasi-static test is still deemed to be an effective
way for seismic failure mechanism assessments of RC pylons.
This study presents results of a quasi-static test on a 1/35-scale in-
verted Y-shape RC pylon model. To facilitate the limited loading
equipment in laboratory, a displacement-controlled two-node load-
pattern for the single pylon model is first proposed through numerical
analyses on the full cable-stayed bridge model. Physical observations
recorded in the test are described and interpreted. Meanwhile, the
quasi-static test is simulated using FE method. Based on the validated
FE model as well as the test records, the seismic failure mechanism of
the RC pylon model is revealed. In addition, multi-level displacement
ductility factors are proposed to quantify ductile properties of the test
model. After that, based on the validated FE model, parametric studies
are performed to understand the impact of loading displacement ratios
for the two-node load pattern on the seismic failure mechanism and
ductility of the RC pylon model. Finally, conclusions and limitations are
addressed.
2. Design of the quasi-static model test
2.1. Selection of prototype for long span cable-stayed bridges
Table 1 summarizes long span cable-stayed bridges with RC pylons
(main spans beyond 600 m). Fig. 1 shows typical shapes for such py-
lons. As can be seen from Table 1, the inverted Y- and diamond-shape
are the most popular ones. They share the same inverted Y-shape por-
tion above the crossbeam, and the only minor difference is the portion
below the crossbeam. In fact, these two shapes have quite similar static
and dynamic characterizations [20]. Ratios of pylon height to main
span fall into a range of 0.251 ∼ 0.339. Among these bridges, the Su-
tong Bridge in China is the most widely studied case in literature, in-
cluding areas such as earthquakes, winds and structural health mon-
itoring [21,37–41], which provide sufficient information on numerical
modelling of the bridge that can be used to validate the FE model in this
study. In this regard, the Sutong Bridge is adopted as the prototype in
this study, which is a box-girder cable-stayed bridge with twin 300 m-
height inverted Y-shape pylons. More details on configurations of the
Sutong Bridge can be found in [18,21,37,39].
2.2. Simplified two-node lateral load-pattern
The Pushover analysis is an increasingly popular and generally di-
rect method for assessing the seismic failure mechanism of structures
that are subjected to monotonically increasing loads up to failure [29].
Recalling the basic theory of the Pushover analysis as well as the fact
that high order modes of pylons for long span cable-stayed bridges may
significantly contribute to the seismic responses of pylons [26], it is best
to “push” the pylons with load-patterns imitating shapes of the

207
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 1. Typical pylon shapes for long span cable-stayed bridges: (a) A-shape; (b) Inverted Y-shape; (c) Diamond-shape; and (d) H-shape (more struts may be adopted).

predominant modes. However, for the quasi-static pushover test, due to
limited loading equipment and technique in laboratory, it is practically
impossible to consider all of those modes. In this regard, this study
explores a simplified two-node load-pattern for the quasi-static push-
over test of the inverted Y-shape pylon, one at the bifurcation-node and
the other at the crossbeam. To this end, a full bridge FE model is built
and subjected to a series of ground motions using incremental dynamic
analyses (IDA) [42]. The simplified two-node load-pattern is then
proposed based on the obtained displacement ratios across increasing
intensity measures such as PGA. A primary focus of this section is to
examine whether the obtained displacement ratios develop to generally
stable values from low to high intensities (i.e., from elastic to plastic
states).
2.2.1. Full bridge FE model and validation
A three-dimensional FE model of the Sutong Bridge is built in
OpenSees [43], an open source platform. Fig. 2 schematically shows the
FE model and pylon sections. Plasticity-distributed fiber models are
used to represent pylon sections accounting for material nonlinearity
and axial force-moment interaction. The Kent-Scott-Park model [44,45]
is adopted to model concrete fibers. For the concrete cover, peak
strength fc,cover = 32.40 MPa, corresponding to a strain εc,cover = 0.002;
ultimate strain εcu,cover = 0.004, corresponding to an ultimate strength
fcu,cover = 22.73 MPa. As for the concrete core, parameters for the fiber
models are listed in Table 2. The rebar is represented by a bilinear
model with kinematic hardening, with the following parameters:
yielding strength fy = 400 MPa, Young’s modulus Es = 200 GPa and
strain-hardening ratio bs = 0.01. It is worth noting that these values are
theoretical values from design documents.

Fig. 2. Schematic illustration of finite element modelling of the Sutong Bridge prototype.

208
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Table 2 Table 3
Model parameters for the confined concrete of the pylon in the full bridge FE Information of the adopted PEER-NGA ground motion records for IDA.
model.
# Earthquake, Year Station RSN Mw R (km)
Section # fc,core (MPa) εc,core fcu,core (MPa) εcu,core
1 Chi-Chi, 1999 TCU054 1494 7.62 5.28
I 35.53 0.0022 26.51 0.0134 2 Loma Prieta, 1989 Piedmont Jr. High School 788 6.93 73.00
II 35.65 0.0022 25.95 0.0137 3 Loma Prieta, 1989 San Francisco Sierra Pt. 804 6.93 63.15
III 35.14 0.0022 26.82 0.0122 4 Iwate, 2008 AOMH13 5536 6.90 159.62
5 Chi-Chi, 1999 TCU129 3507 6.30 36.81
6 Coyote Lake, 1979 Gilroy Array #6 150 5.74 9.12
7 Chi-Chi, 1999 TCU045 1485 7.62 77.91
Displacement-based beam-column elements with fiber sections dis-
8 Loma Prieta, 1989 UCSC 809 6.93 24.05
cretized into 15 m in depth (i.e., approximately one time of the section 9 Loma Prieta, 1989 Gilroy Array #1 765 6.93 33.55
length; 5 integration points for each element) are used to model in- 10 Helena, Montana, 1935 Carroll College 1 6.00 8.71
clined columns of the pylon. This modelling technique is empirically 11 Northridge, 1994 Vasquez Rocks Park 1091 6.69 41.90
12 Victoria, Mexico, 1980 Cerro Prieto 265 6.33 35.48
deemed to produce a stable result of plastic hinge region [46–48]. The
Corotational Transformation command in OpenSees [43] is used to ac-
count for the P-Δ effect [49]. Elastic beam-column elements are used to
motions and their mean as well as the site spectrum. In general, the
model (1) the crossbeam of pylon, (2) the upper vertical column of
adopted ground motions have a wide range of predominant periods
pylon, (3) the deck, and (4) piers. Reasons for such modelling are
(i.e., 0.15 ∼ 0.54 s), which generally covers the platform segment of the
below. The crossbeam is normally designed in elastic behavior (e.g.,
acceleration spectrum for the site of Sutong Bridge [55]. These records
allocated with plenty of high-strength prestressed rebars) to ensure a
are then scaled to 0.1 ∼ 1.0 g with a step of 0.1 g. It is worth noting that
reliable load-path under earthquakes, while the upper vertical column
the upper boundary of 1.0 g is found to trigger apparently inelastic
and the deck are often not controlled by earthquake loads. As for piers,
behavior of the bridge.
a preliminary analysis by the authors show that their nonlinearities
have almost negligible impacts on transverse seismic responses of py-
lons. Hence, elastic elements are used to model the piers for compu-
2.2.3. Displacement ratios for two-node load-pattern and correlations to
tational efficiency. A floating deck system is used in the longitudinal
motion parameters
direction, while rigid deck-pylon connections are used in the transverse
Fig. 5 shows the IDA-based development of displacement ratio (δ)
direction due to the application of wind-resistance bearings at the py-
between the bifurcation-node and crossbeam when the maximum
lons. Bearings for pier-deck connections are transversally fixed for the
bending moment along the pylon reached. From this figure, δ is found
purpose of transverse stability of the bridge. Cables are modeled as
to be motion-dependent (i.e., δ ranges from approximately 1 to 7).
large-displacement truss elements using the Ernst method (modified
Among them, most of the ground motions lead to generally stable re-
modulus of elasticity) [50] owing to its convenient usage and the
sults across the increasing PGA (mostly for those beyond 0.3 g). As for
capability to account for the sag effect. To simulate as-built tensional
the Iwate AOMH13 record (#4) that contains an extremely long dura-
forces in cables, initial strains are assigned in these elements without
tion (recall Fig. 3), δ keeps almost consistent for PGA = 0.1 ∼ 0.5 g and
masses. It is worth noting that local cable dynamics are neglected since
then gradually rises up to nearly 7 (being generally stable for those
Caetano et al. [51] suggested that only a narrow band of earthquake
beyond 0.9 g). This result may be attributed to the fact that the ground
excitation frequencies can cause dynamic interactions between local
motion duration often has a noted impact on inelastic behavior of RC
cables and the global system. Soil-structure interaction effects are ne-
components [58]. Besides, as highlighted in Fig. 5 for the Chi-chi
glected for simplicity (i.e., piers and pylons are base-fixed). Before
TCU054 record (#1), the stable δ value of nearly 5 is adopted as the
dynamic analyses, a gravity analysis is conducted to verify the as-built
two-node load-pattern in the quasi-static test described later, since this
condition for the FE model; that is, almost identical cable-forces and
earthquake witnessed the damage of RC pylon for the Chi-lu cable-
deck-deflections between the FE model and prototype (i.e., absolute
stayed bridge, which is the unique documentation of recorded damage
percentage errors are within 2%).
to RC pylon as yet.
Modal analyses show first seven modes of the OpenSees FE model
Since δ is found to be motion-dependent, correlation analyses are
with periods of 15.86, 9.89, 5.61, 4.52, 3.53, 3.22 and 2.68 s. On the
performed to formulate δ with ground motion parameters. Values of δ
other hand, Wang et al. [39] used ANSYS [52] to model the Sutong
in the correlation analyses are those at PGA = 1.0 g in Fig. 5. On the
Bridge with the same constraints (i.e., neglecting soil-pile interactions)
other hand, since IDA are performed, amplitude-related parameters
and predicts the first seven corresponding modes with periods of 16.07,
such as PGA and Spectrum Intensity are not appropriate for the corre-
10.01, 5.50, 4.47, 3.49, 3.08 and 2.68 s. It is seen that very close results
lation analyses. Therefore, this study focuses on frequency- and dura-
are achieved between the two FE models (i.e., absolute percentage er-
tion-related ground motion parameters, as listed in Table 4. Fig. 6
rors are within 5%). Ground motions are input uniformly in the
shows results of the correlation analyses, including coefficients of de-
transverse direction. Average-acceleration-based Newmark-β integrator
termination (R2) and Pearson’s linear correlation coefficients (ρ) [59],
[53] and Newton algorithm [54] are used to solve dynamic equilibrium
with respect to the studied parameters. It is found that the bracketed
equations.
duration (Db) has the highest correlation, closely followed by the
number of effective cycles and the significant and uniform durations,
2.2.2. Adopted ground motions for IDA while the frequently used predominant period shows the least corre-
Based on the site condition of Sutong Bridge [55], twelve non-pulse- lation. Fig. 7 shows δ-Db regression models in linear, logarithmic and
like ground motion records are selected from PEER-NGA strong motion logarithmically linear forms. These formations can be used to estimate
database [56]. Table 3 lists their information, including stations, record the displacement ratio of the two-node load-pattern for inverted Y-
sequence number (RSN), moment magnitude (Mw) and source distance shape RC pylon subjected to ground motions. It should be noted that
(R). Fig. 3 shows their acceleration time histories scaled to the same application ranges of these formulations are limited to the upper and
PGA, together with their durations. From this figure, it is seen that the lower boundaries of Db among the adopted ground motions as well as
adopted ground motions contain a wide range of duration (i.e., the studied object. In addition, Db can be estimated through ground
30 ∼ 120 s for the total duration and 2.3 ∼ 50 s for the significant motion prediction equations in [57] or related hazard analyses, which
duration [57]). Fig. 4 illustrates pseudo-acceleration spectra of these is out of the scope of this study.

209
X. Wang, et al.
Fig. 3. Acceleration time histories of the adopted ground motions scaled to the same PGA.
Fig. 4. Pseudo-acceleration spectra of the adopted ground motions and their
mean spectrum as well as its comparison to the site spectrum (Note: All are
scaled to 0.5 g with 5% damping ratio; GM = ground motion).
2.2.4. Validation of the two-node load-pattern
To validate the two-node load-pattern, a single pylon FE model is
extracted from the full bridge FE model and then subjected to the
proposed two-node load pattern. Note that the cable-induced initial
axial loads in the pylon are considered in the single pylon model. Fig. 8
compares envelope responses of axial force, shear force, bending mo-
ment and deflection along the pylon that are obtained from the dy-
namic analysis of the full bridge FE model under 1.0 g Chi-chi TCU054
wave and the corresponding two-node load-pattern pushover analysis
of the single pylon FE model with the displacement ratio of 5, together
with illustrations of plastic regions and their occurrence orders.
From Fig. 8, it is seen that plastic regions appear first at the bottom
of upper column, then at the bottom of lower column and finally at the
top of lower column. These sections are deemed to be critical sections.
In general, peak internal forces and lengths of plastic regions near these
critical sections agree reasonably well between the dynamic and two-
node pushover analyses. Considering the inherent simplicity of the two-
node load-pattern itself, it is reasonably deemed to be capable of cap-
turing the characteristics of inverted Y-shape RC pylons under later
loads. In addition, it is worth noting that the proposed two-node load-
210
Engineering Structures 189 (2019) 206–221
Fig. 5. IDA-based development of displacement ratio between the bifurcation-
node and crossbeam when the maximum bending moment along the pylon is
derived: Results from Chi-chi TCU054 (#1) and Iwate AOMH13 (#4) are
highlighted in red and blue, respectively. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this
article.)
pattern for pushover analysis can be used for efficient estimates of
seismic responses of as-built bridges that were not designed for occa-
sionally large earthquakes, which can assist in seismic retrofits and
maintenance. On the other hand, for bridges under design at early
stages of projects that dimensions are often changed, the two-node
pushover analysis can efficiently estimate responses and consequently
assist in optimal seismic designs.
2.3. Single pylon model with equivalent dead load
2.3.1. Similarity coefficients and test model simplification
Due to the limited space of Earthquake Engineering Hall of Tongji
University at Jiading campus where this test took place, the
Buckingham π theorem of dimensional analysis [64] is followed to scale
the single pylon. Table 5 lists the adopted similarity coefficients, where
the scale factor for length is set as 1/35, whereas scale factors for
material properties of stress, strain and elastic modulus are all set as
1.0. It is worth noting that in this manner, real steel and concrete
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Table 4
Studied frequency- and duration-related ground motion parameters.
Parameter Name (Unit) Definition Ref.

V/A Peak velocity and acceleration ratio (s) Peak ground velocity (PGV) divided by PGA (i.e., PGV/PGA). /
Tp Predominant period (s) Period where the maximum spectral acceleration occurs. /
Tm Mean period (s) Tm = ∑ (Ci2/ fi )/ ∑ Ci2 , where Ci are Fourier amplitudes; fi are discrete Fourier transform frequencies between 0.25 [60]
and 20 Hz.
Nc Number of effective cycles (/) Sum of ratios of cycle amplitudes in accelerogram divided by maximum cycle amplitude. [61]
Du Uniform duration (s) Total time during which the acceleration is larger than the limit value of 5% of PGA. [62]
Db Bracketed duration (s) Total time elapse between the first and the last excursions of the 5% of PGA. [62]
Ds Significant duration (s) Interval of time over which the 5% and 95% of the total Arias Intensity [63] is accumulated. [57]
Dt Total duration (s) The total time of the input ground motion. /

actuator and prestressed tendon, respectively, and they remained con-

sistent during the test. Fig. 10 shows comparisons of initial axial com-
pressive force between the ideally scaled prototype and the test model.
It is seen that the initial axial loads at bottoms of upper and lower
columns are perfectly equivalent between the prototype and test model,
while that at the top of lower column is somewhat apart with a varia-
tion of approximately 60 kN. The effect of such a variation is discussed
later in this paper.

2.3.2. Section and reinforcement modifications for construction feasibility

Fig. 6. Coefficient of determination (R2) and Pearson’s correlation coefficient
(ρ) between the displacement ratios and the studied ground motion parameters.
materials can be used to reduce uncertainties from the construction and
measurement of micro-concrete and iron-wire that are often used in
tests where the scale factors for material properties are not equal to 1.0.
Fig. 9 shows schematic of the scaled test model.
On the other hand, for quasi-static tests, the role of the similarity of
mass provides equivalent initial axial forces in sections, rather than
inertial force as that for shake-table tests. In this regard, the self-weight
of vertical portion above the bifurcation-node as well as vertical com-
ponents of cable forces are provided by a vertical actuator for simpli-
fication, as illustrated in Fig. 9. As for the portion below the bifurcation-
node, additional mass blocks that are normally discretely assigned
along the pylon are not considered for testing feasibility. Another
reason is that assigning such mass blocks often require steel/iron boxes
embedded into the pylon model, which inevitably affects the sectional
properties. In this way, initial axial loads for the test model cannot
perfectly follow the similarity law. Instead, more loads were provided
by the vertical actuator to meet the equivalence of axial loads for cri-
tical sections as perfectly as possible. In addition, a prestressed tendon
was adopted between the crossbeam and the base to provide additional
axial loads for the lower columns of the pylon model, as displayed in
Fig. 9. To this end, 570 kN and 220 kN were assigned in the vertical
Since a relatively small scale factor of 1/35 is used, wall thicknesses
of ideally scaled hollow sections are quite small (i.e., 3.43 ∼ 4.29 cm),
which are very difficult for construction using real rebar and concrete
materials. Therefore, the ideally scaled dimensions are not perfectly
followed in the test. Instead, this study re-designed the RC hollow
sections to meet prototype-model equivalence on critical limit states of
the sections, including the first-yield and ultimate limit states. Note that
the first-yield limit state is defined by the first rebar reaching its
yielding stress while the ultimate limit state is represented by the
concrete core crushing or rebar snapping, whichever occurs first. Fig. 9
shows the re-designed sections and their reinforcements. For column
sections, φ8 mm rebars and φ6 mm bars are adopted as the longitudinal
and transverse reinforcements, respectively. It should be noted that the
ideally scaled stirrup-spacing of 6 cm is empirically deemed to induce
buckling issue with respect to the φ6 mm stirrups [65]. For this matter,
the stirrup-spacing is designed as 3 cm for the test. As for the crossbeam
that is normally allocated with plenty of prestressed tendons in proto-
type to ensure an elastic behavior, a solid section with sufficient re-
inforcements is adopted in this test for construction feasibility as well as
for ensuring a practically elastic behavior during the test.
Section moment-curvature curves are then obtained using pure
bending analyses in OpenSees. Fig. 11 displays moment-curvature re-
lationships for the ideally scaled prototype and test model under initial
axial loads, together with their first-yield and ultimate limit states.
From Fig. 11, they coincide generally well for individual sections. In
addition, Table 6 lists the first-yield and ultimate curvatures for these
sections. It is seen that errors are within 8% for the first-yield limit state

Fig. 7. Regression models between displacement ratios for the two-node load-pattern and bracketed duration: (a) linear; (b) logarithmic; and (c) logarithmically
linear.

211
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 8. Comparison of responses between the full

bridge model dynamic analysis subject to1.0 g
Chi-chi TCU054 record and the corresponding
two-node load-pattern pushover analysis: Plastic
regions and occurrence orders; and envelope
distributions of (a) axial force; (b) shear force; (c)
bending moment; and (d) deflection along the
pylon.

Table 5 and 15% for the ultimate limit state. Such results indicate that the re-
Similarity coefficients for the test model. designed RC hollow sections can generally reflect the bending behavior
Parameter (Notation) Similarity relationship Similarity coefficient
of the prototype sections. It should be noted that the re-designed sec-
tions are capacity-protection for shear failure as commonly adopted in
Length (l) Sl 0.0286 (i.e., 1/35) structural seismic design.
Displacement (δ) Sδ = Sl 0.0286
Rotation (θ) Sθ 1
Strain (ε) Sε 1 2.3.3. Construction of the pylon model
Stress (σ) Sσ 1 Fig. 12 displays the construction process of the test model, including
Elastic modulus (E) SE 1
Acceleration (A) SA 1
pasting of strain gauges, colligations of reinforcement cages, forma-
Mass (m) Sm = Sσ·Sl2/SA 0.000817 tions, concrete casting and curing. Meanwhile, concrete cubes and
Force (F) SF = Sσ·Sl2 0.000817 prisms were constructed and rebar/bars were collected and tested to
Curvature (φ) Sφ = 1/ Sl 35 determine the material properties. Skins of the test model were blan-
ched before test for convenience of crack observations.

Fig. 9. Schematic of the quasi-static test: overview, reinforcement and instrumentation.

212
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Table 6
Comparison of section-curvature at first-yield and ultimate limit states between
the ideally scaled prototype and the test model.
Ideally scaled prototype Test model (1/ Error (%)
(1/m) m)

Limit state Section # ① ② (②-①)/①

First-yield B-B 0.0145 0.0153 6
C-C 0.0144 0.0152 6
D-D 0.0168 0.0154 −8

Ultimate B-B 0.575 0.655 14

C-C 0.621 0.715 15
D-D 0.609 0.679 11

During the test, an unexpected outage occurred in the data acquisition

Fig. 10. Comparison of axial compressive force before loading between the
ideally scaled prototype and the test model.
2.4. Instrumentation and lateral loading protocol
Again seen in Fig. 9, lateral forces and displacements at two loading
points were obtained through two horizontal actuators. In addition, 38
pairs of strain gauges and 24 pairs of displacement gauges (i.e., linear
variable differential transformers (LVDTs), strokes of 10 cm and preci-
sions of 0.05 mm) were adopted to trace developments of rebar strains
and/or section curvatures. Details of the technique for converting re-
corded raw data to curvature results can be found in Wang et al.
[66,67]. It is worth noting that in the elastic range where a relatively
low amplitude of curvature involves, a pair of strain gauges in a section
can accurately capture strains in rebars, which are then used to derive
the section curvature. For the pair of LVDTs at the same section, by
contrast, the limited accuracy hinders their application in measuring
small deformations of the section, but they are capable of estimating
large deformations of the section such as those in high levels of in-
elasticity where the strain gauges often malfunction.
As mentioned before, a displacement ratio of 5 for the two-node
lateral load pattern is adopted in the quasi-static test. To this end, a
displacement-controlled monotonic lateral loading protocol with a ratio
of 5 at two actuators (i.e., H1 and H2, as depicted in Fig. 9) is applied
with a speed up to 2 mm/s. Both actuators were servo-controlled and
produced by the MTS Systems Corporation (Eden Prairie, MN). The
Actuator H1 has a force-capacity of 1000 kN and a stroke of ± 50 cm,
while the Actuator H2 has a force-capacity of 500 kN and a stroke
of ± 50 cm. Fig. 13 shows the loading protocol for the quasi-static test.
system when the Actuator H1 pushed to a displacement level of 70 mm.
All the recorded data were consequently missed. In this regard, the test
model was unloaded and pulled back to the original position. It should
be noted that some horizontal minor cracks were observed near bot-
toms of the upper and lower trailing columns during this unsuccessful
load/unload process, but they practically closed up after pulled back to
the original position. After nearly a half hour of reboot and evaluation
of the data acquisition system, the test model was reloaded and con-
tinued until an unexpected rebar snapping (explained later) occurred
when Actuators H1 and H2 pushed up to 320 mm and 64 mm, respec-
tively. For convenience, displacement values mentioned later in this
paper refer to the displacements at Actuator H1 (i.e., the bifurcation-
node).
3. Physical damage observation
Table 7 summarizes the observed critical damage to the test model.
Fig. 14 shows global and local damage as depicted in Table 7. It is found
that horizontal cracks are the most frequent damage in the test, in-
dicating a flexural damage mode for the pylon model. Plastic regions
were observed at bottoms of the upper and lower trailing columns and
lower leading column as well as the top of lower trailing column, which
are generally consistent to the critical sections derived from the dy-
namic analysis mentioned above (see Fig. 8).
4. Test and numerical analyses for failure mechanism assessment
In order to reveal the seismic failure mechanism of Y-shaped RC
pylon for long span cable-stayed bridges as well as due to the limited
test records, numerical modelling is adopted as an assistance. A refined
FE model for the test model is built in this section. Results from the test
and numerical model, including force-displacement relationships, cur-
vature and strain distributions, are compared to validate the refined FE
model, which is then used to capture variations of axial forces during
the test. The influence of such variations on the yield- and ultimate-

Fig. 11. Section moment-curvature comparison between the ideally scaled prototype and the test model for critical sections under initial axial loads: (a) B-B; (b) C-C
and (c) D-D.

213
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 12. Construction and preparation process of the test model at the first loading up to 70 mm.

Fig. 9). This configuration leads to nearly synchronous movement

and rotation for the joint and end-crossbeam nodes.
(2) The crossbeam is modeled using displacement-based beam-column
elements to check and confirm the practically elastic behavior of
crossbeam during the test.
(3) The bond-slip effect at bottoms of the pylon is modeled using zero-
length-section elements with the Bond SP01 material [68] in
OpenSees, in which the key parameters are as below: the yield slip,
Sy = 0.36 mm, determined using Eq. (1) [68]:

1/ k
⎡ d fy ⎤
Sy = 2.54 ⎢ b (2k + 1) ⎥ + 0.34
8437 fc (1)
⎣ ⎦

where db = 10 mm is the rebar diameter; f y = 450 MPa is the yielding

Fig. 13. Loading protocol in the quasi-static test: malfunction of the data ac-
quisition system at the first loading up to 70 mm.
states of sections are considered to reveal the seismic failure mechanism
of the test model.
4.1. Refined FE modelling of the quasi-static test
Fig. 15 schematically shows the refined FE modelling of the quasi-
static test using OpenSees [43]. Compared to the full FE model of Sutong
Bridge (see Fig. 2), the refinements of this FE model are reflected in the
following aspects.
(1) Column-crossbeam joints are modeled using rigid elastic beam-
column elements, together with equal degree-of-freedom
(EqualDOF) technique, as illustrated in Fig. 15. The reason to use
the EqualDOF technique is due to the fact that anchor rods were
connected to Actuator H2 and extended into the crossbeam (see
stress of rebar obtained from the material test; f c = 41 MPa is the
concrete compressive strength; and k = 0.4 is the parameter used for
the local bond-slip relation [69]. Accordingly, the ultimate slip,
Su = 38Sy = 13.8 mm [68]. It is worth noting that in the test model, the
anchorage length of rebars is about 40 cm (see Fig. 9), which satisfies
the minimum requirement of anchorage length (approximate 11 cm) for
the adopted bond-slip model [68]. In addition, to assess the influence of
this bond-slip effect, another FE model with conventional base-fixed
constraints is built and analyzed for comparison.
Table 8 lists constitutive parameters of this refined FE model, which
are obtained through a series of material tests as well as the related
concrete and steel models in literature [44,45]. Meshes of the fiber
sections are shown in Fig. 15 as well. The displacement-controlled
loading technique in OpenSees [43] is used to simulate the two-node
load-pattern pushover in the test. To solve the nonlinear equilibrium
equations, algorithms such as Newton, Newton-with-line-search and
Krylov-Newton [70], are tried in turn to reach convergence results as
soon as possible. The convergence tolerance on the norm of the dis-
placement residuals is set to 10-4.

Table 7
Observed physical damage to pylon components during the test.
Disp. Description of observed damage Figure

25 mm Horizontal crack at the outer transverse-normal surface of the bottom of upper trailing column 14(a)
35 mm Horizontal crack at the outer transverse-normal surface of the bottom of lower trailing column 14(b)
45 mm Horizontal crack at the inner transverse-normal surface of the top of lower trailing column 14(c)
50 mm Horizontal crack at the outer transverse-normal surface of the top of upper trailing column (just below the bifurcation-node) 14(d)
140 mm Diagonal crack at the transverse-parallel surface of trailing column-crossbeam intersection 14(e)
220 mm Concrete cover began to spall at the bottom of lower leading column /
320 mm Rebar snapped at the bottom of upper trailing column; Concrete cover spalling extended along the elevation; Few minor cracks along the top of upper trailing 14(f)
column. Test ended.

214
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 14. Development of the global deformation and corresponding local damage to the test model for increasing displacement levels at the bifurcation-node: (a)
25 mm; (b) 35 mm; (c) 45 mm; (d) 50 mm; (e) 140 mm; (f) 320 mm (ending displacement).

4.2. Validation of the numerical model Table 8

Model parameters for the concrete cover and core and rebar of the test model.
4.2.1. Global force-displacement relationships Section
Fig. 16 shows recorded and predicted results of global force-dis-
placement relationships at two loading points. The predicted result is Material Parameter (Unit) Upper column Lower column Crossbeam
based on the full loading protocol shown in Fig. 13 to account for the
Concrete cover fc,cover (MPa) 41.00 41.00 41.00
degradation of stiffness due to the initial unsuccessful load/unload εc,cover 0.002 0.002 0.002
process, while the recorded counterpart represents the latter successful fcu,cover (MPa) 14.88 14.88 14.88
reload process. εcu,cover 0.006 0.006 0.006
From Fig. 16(a), it is seen that the predicted results at the bifurca- Concrete core fc,core (MPa) 53.66 54.82 50.91
εc,core 0.0026 0.0027 0.0025
tion-node coincide quite well with the recorded counterpart, regardless
fcu,core (MPa) 40.08 41.24 41.25
of whether the bond-slip effect is considered or not. As for the results at εcu,core 0.042 0.045 0.034
the crossbeam (Fig. 16(b)), some dispersions are observed between the Rebar fy (MPa) 450 450 450
recorded and predicted results for displacement levels beyond 30 mm. Es (GPa) 169 169 169
Nevertheless, the FE model with the bond-slip effect produces a rela- bs 0.02 0.02 0.02

tively smaller lateral stiffness and consequently a better prediction as

compared to the model without the bond-slip effect.

Fig. 15. Refined FE modelling of the quasi-static test: schematic and section mesh.

215
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 16. Validation of numerical model in terms of global force-displacement responses at two loading points: (a) bifurcation-node; and (b) crossbeam.

Fig. 17. Validation of numerical model in terms of curvature developments at increasing displacement levels at trailing and leading columns with or without bond-
slip effect consideration: (a) and (b) with bond-slip effect; (c) and (d) without bond-slip effect.

4.2.2. Curvature developments
Fig. 17 shows recorded and predicted curvature developments at the
trailing and leading columns with increasing indicative displacement
levels from 10 to 300 mm. It is worth noting that in this figure, the
recoded results at relatively small displacement levels of 10 and 50 mm
were derived from the strain gauges on rebars, whereas the rest of re-
corded results were derived from the displacement gauges. Note that
some gauges were impaired during the concrete casting process or
malfunctioned during the test. The impact of bond-slip effect is assessed
as well.
From Fig. 17(a) and (b), in general terms, good agreements are
achieved between the recorded and predicted results, thereby in-
dicating the FE model with the bond-slip consideration can generally
capture the curvature distributions across the considered displacement
levels. In addition, comparative inspections of Fig. 17(c) and (d) to (a)
and (b), respectively, imply that the bond-slip effect has a minor impact
on curvature distributions of the trailing column, while it significantly
affects those of the leading column, especially for those at the pylon
bottom. More specifically, the FE model without the bond-slip effect
may produce curvature responses three times as much as the model
considering the bond-slip effect. Therefore, it can be concluded that the
refined FE model with bond-slip effect is generally capable of esti-
mating curvature responses along pylons.
4.2.3. Rebar-strain developments
Fig. 18 shows recorded and predicted developments of rebar-strain
envelopes along the trailing column across relatively small displace-
ment levels from 5 to 40 mm. Note that data at elevations of 1.2 and
1.6 m represent strain responses of the rebar located in the inner corner
of the section of trailing column (refer to Fig. 9), while the rest of data
represent responses of the rebar in the outer corner (i.e., rebar near the
outer surface). The predicted results are based on the refined FE model
with the bond-slip effect. From Fig. 18, generally good agreements are
observed between the predicted and recorded results, which further
validates the refined FE model.
It is worth noting that someone may argue that the bond-slip effect
could also be involved in the column-crossbeam joints. For this issue,
the beamColumnJoint element proposed by Lowes et al. [71] in the

216
X. Wang, et al.
Fig. 18. Validation of numerical model in terms of rebar-strain-envelope de-
velopments at increasing displacement levels at the trailing column.
framework of OpenSees may be a practical solution. However, this
element falls into a limited application range where the test model in
the present study is not completely included. More specifically, (1) each
spring in the beamColumnJoint element is characterized by deformation
behavior including effects of both bond-slip and cracking; this task
often requires adequate FE modelling experience on parameter de-
terminations as well as reliable test results related to such effects, which
are absolutely beyond the obtained information in this test; and (2)
springs in the beamColumnJoint element are placed symmetrically with
respect to the centroid of the joint and thus may not be amenable to
treating unsymmetrical situations such as the sloping column-cross-
beam joints in this test. Therefore, the beamColumnJoint element as
currently available in OpenSees is not well suitable for the present study.
Nevertheless, considering the above validation results for the refined FE
model that accounts for the bond-slip effect at bottoms as well as its
relatively easy implementation, it is reasonable to conclude that the
refined FE model is capable of capturing the primary characteristics and
responses of the RC pylon.
4.3. Failure mechanism assessment considering variations of axial loads
4.3.1. Indicative process to determine the yielding and ultimate limit states
It is well known that for framed structures such as the studied pylon,
section axial loads shall vary under lateral loads, which inevitably af-
fects yield- and ultimate-states of the section [66,72]. In other words,
the variation in axial loads affects the “capacity” of the section. In this
regard, the validated FE model is used to trace the axial loads during
the test, which are then used in moment–curvature analyses of sections
to determine the curvatures at the yield- and ultimate-states (i.e., “ca-
pacity”). On the other hand, developments of curvature responses (i.e.,
“demand”) are predicted using the FE model as well as recorded
through strain/displacement gauges. The “demand” results across var-
ious displacement levels are plotted together with the “capacity”
counterparts to identify intersection points that present the yield and
ultimate limit states of the section.
Fig. 19 shows the indicative process to determine the yield-state of
bottom section of lower trailing column and the ultimate-state of
bottom section of lower leading column. Variations of the axial loads
are plotted together in this figure. From Fig. 19(a), it is seen that the
bottom of lower trailing column yields at displacement levels of 38 mm
for the predicted result and 43 mm for the recorded result. They are
reasonably close. It is worth noting that the section yield-state can also
be determined directly through the development of strain distribution
shown in Fig. 18. In this case, the “capacity” represents the yielding
strain of rebar obtained from the material test (i.e., 0.00267 = 450 MPa
217
Engineering Structures 189 (2019) 206–221
(yielding stress)/169 GPa (elastic modulus), refer to Table 8). In such a
way, the indicative section is found to yield at a displacement level
around 40 mm, which is quite close to the above curvature-based
method. As for Fig. 19(b), the predicted result indicates the fail (con-
crete core crushing) of bottom section of lower leading column at a
displacement level of 553 mm, which is extremely larger than the
ending displacement of 320 mm in the test.
In addition, it is worth recalling the variation of initial axial com-
pressive force (60 kN) at the top section of the lower column between
the ideally scaled prototype and the test model (Fig. 10). Based on the
process to determine the yield-state of the bottom section of lower
trailing column as shown in Fig. 19(a), it is reasonable to infer that the
relatively small variation (60 kN) should trigger quite small changes to
curvature demand and capacity that are used to determine the yield-
state of this section. Thus, the effect of this variation is deemed to be
generally neglectable from the view of practice.
4.3.2. Failure mechanism summary
Following the above indicative process, Fig. 20 summarizes the
predicted and recorded failure mechanism of the pylon model (i.e., the
occurrence order of yield- and ultimate-states for the critical sections).
Note that in the FE model, the first section reaching its ultimate-state is
defined as the ultimate limit state of the pylon. From Fig. 20, the re-
corded and predicted results indicate the same yielding order for the
critical sections. More specifically, for yield-states, plastic regions are
detected first at the bottom of upper trailing column (① of triangles in
Fig. 20), then successively at the bottom and top of lower trailing
column (③ and ② of triangles); and finally at the bottom of lower
leading column (④ of triangles). Fig. 21 displays the recorded and
predicted displacement levels for the yield-states of critical sections.
Obviously, they show quite good agreements in amplitudes with dis-
persions less than 15%. In addition, it is worth noting that the recorded
and predicted occurrence orders of plastic regions are consistent to that
obtained from the dynamic analysis mentioned above (Fig. 8). This
result further validates the refined FE model that can be used for
parametric analyses.
As for the ultimate-state, the predicted result shows that the bottom
of lower leading column fails first (④ of squares in Fig. 20), whereas the
recorded result indicates the failure of pylon at a much smaller dis-
placement level of 320 mm where an unexpected rebar snapping oc-
curred. This is a noted limitation of the quasi-static test. To explain this
unexpected issue, a plane-hypothesis-based preliminary analysis using
the recorded curvature and the dimension of this section implies that
the estimated rebar strain at this level (i.e., 0.0284 = approximately
0.2 m−1 (recorded curvature) × 0.142 m (distance of the rebar to the
section neutral axis)) is much less than its ultimate strain from the
material test (i.e., 0.15). The reason for this unexpected failure may be
attributed to the impact of the anchor rods that connected Actuator H2
with the pylon model, which inevitably inter-acted with the snapped
rebar at this section. In other words, the configurations for loading
points in the test may be somewhat unreasonable. Although sophisti-
cated FE modelling using continuum media may be a possible solution
to the mechanism of such a complex interaction, the limited test records
related to this issue probably hinder this solution. Moreover, the un-
reasonable configurations are not realistic for engineering practices.
Therefore, for the purpose of conciseness, such a tough investigation is
not performed in this study.
It should be noted that for inverted Y-shape RC pylons under
earthquakes, the two columns act as leading and trailing ones by turns.
Therefore, in a view of generalization, the terms “leading” and
“trailing” can be removed when summarizing the failure mechanism of
the pylon.
In addition, it is worth noting that the observed plastic hinges at the
columns for the crossbeam-column joints follow with current practices
that often design the crossbeams sufficiently strong to assure reliable
load transferring paths for pylons and/or for bearings installed on the
X. Wang, et al. Engineering Structures 189 (2019) 206–221

Fig. 19. Indicative process to determine (a) the yield-state of the bottom section of lower trailing column; and (b) the ultimate-state of the bottom section of lower
leading column.

where Δy,1 is the displacement level for the first plastic hinge; and
Δy,i (i = 2, 3, 4, …) represents displacement levels for second, third
and fourth etc. plastic hinges, which are marked in Fig. 20. Moreover,
an ultimate displacement ductility factor, μΔu , that represents the full
ductile capacity of the test model is expressed in Eq. (3).
Δu
μΔu =
Δy,1 (3)

Fig. 20. Recorded and predicted failure mechanism of the test model.
where Δu represents the displacement level for the ultimate limit state
of the test model.
Table 9 lists recorded and predicted ductility factors for the pylon
y
model. Among them, the displacement ductility-limited factors ( μΔ,i )
fail into similar ranges of 1.15 ∼ 3.67 and 1.23 ∼ 3.00 for the predicted
and recorded results, respectively (i.e., the variations of 6.5– 22.3% are
generally small from the view of practice). As for the ultimate dis-
placement ductility factor ( μΔu ), relatively large values are obtained for
both recorded and predicted results. The apparent dispersion between
them is due to the noted distinction of their ultimate limit states as
mentioned before. In summary, the test pylon model exhibits a con-
siderable ductile capacity.

5. Parametric study of loading displacement ratio

This section performs a parametric study to further understand the

impact of displacement ratio for the two-node load-pattern on the
failure mechanism and associated ductility factors of the pylon. Various
loading displacement ratios ranging from 1 to 7 are assessed. These
values are selected based on the results of IDA previously shown in
Fig. 21. Comparison of recorded and predicted displacement levels for yield- Fig. 5. Note that the predicted result of the test model (loading dis-
states. placement ratio of 5) are presented hereinafter for comparisons. To
ensure the rest of conditions as identical as possible, the full loading
protocol shown in Fig. 13 is adopted in the parametric studies, but
crossbeams to support decks. Nevertheless, designing in a way that
merely the result of the reloading part is presented for conciseness.
plastic hinges are formed at the crossbeams, which scarifies the cross-
Fig. 22 compares the failure mechanism of the RC pylon model
beams to protect the columns, is an alternative solution for inverted Y-
under various loading displacement ratios and Table 10 lists the cor-
shape pylons where no bearings are installed on the crossbeam, e.g.,
responding ductility factors as defined above.
cable-stayed bridges with completely floating systems.
It can be seen from Fig. 22 that shapes of the global force-dis-
placement curves are dependent on the loading displacement ratios. In
4.4. Ductility factor definition and result general, a larger loading displacement ratio leads to a larger peak
strength. When measuring the secant stiffness between this peak
According to the obtained failure mechanism, multi-level ductility strength and the original point, it is found that a smaller loading dis-
factors are introduced as preliminary documents for performance-based placement ratio results in a smaller lateral stiffness of the pylon. This is
ductility design of bridge RC pylons in future.
A multi-level displacement ductility-limited factor (or called dis- Table 9
placement ductility factor for limited ductile design), Recorded and predicted ductility factors for the test model.
y
μΔ,i (i = 2, 3, 4,…) , that accounts for the number of plastic hinges (i) is y y y
μΔ,2 μΔ,3 μΔ,4 μΔu
defined using Eq. (2).
(1) Recorded 1.23 1.43 3.00 9.14
Δ y, i (2) Predicted 1.15 1.30 3.67 16.8
μΔ,y i = |(2)-(1)|/(1) (%) 6.50 9.09 22.3 /
Δy,1 (2)

218
X. Wang, et al. Engineering Structures 189 (2019) 206–221

for the case of loading displacement ratio of 1. This is because this case
leads to a quite small displacement level for the first plastic hinge (③ on
the curve) while a relatively larger displacement for the fourth plastic
hinge at the bottom of upper column (i.e., ① on the curve), as described
above. Besides, a larger loading displacement ratio tends to trigger a
larger ultimate ductility factor. In other words, inverted Y-shape RC
pylons subjected to long duration earthquakes tend to have larger
ductility.

6. Conclusions

This paper assesses the transverse seismic failure mechanism and

ductility of an inverted Y-shape RC pylon for long span cable-stayed
bridges using quasi-static tests combined with numerical analyses. A
two-node lateral load-pattern, one at the bifurcation-node and the other
at the crossbeam, is proposed based on IDA using recorded ground
motions to facilitate the limited loading equipment and technique in
Fig. 22. Impact of loading displacement ratio on the failure mechanism of test
laboratory. After test, a refined FE model is built and validated using
model.
the test results, which is then used to perform a parametric study for
further understanding the impact of loading displacement ratios on the
Table 10 failure mechanism and ductile properties. Considering the inherent
Impact of loading displacement ratio on ductility factors. limitations of the test, main findings are as follows.
y y y y
Loading displacement ratio μΔ,2 μΔ,3 μΔ,4 μΔ,5 μΔu
(1) The proposed two-node lateral load-patterns can be used for effi-
1 1.33 1.44 7.89 / 13.7 cient estimates of seismic responses of inverted Y-shape RC pylons.
3 1.14 1.47 2.14 / 13.6 Loading displacement ratios for the two-node load-pattern are
5 1.15 1.30 3.67 / 16.8
7 1.27 1.65 4.92 6.88 21.0
found to be correlated reasonably well with bracketed durations of
Mean 1.22 1.47 4.65 6.88 16.2 ground motions. Inverted RC pylons subjected to ground motions
Standard deviation (S.D.) 0.0938 0.144 2.43 / 3.61 with longer durations tend to trigger larger loading displacement
COV = S.D./Mean 7.67% 9.83% 52.4% / 22.3% ratios for the proposed two-node load patterns for pushover ana-
lyses. The range of the displacement ratios is found to fall into a
range of 1 to 7 in general.
because at a given displacement of the bifurcation-node, a smaller
(2) Test results indicate a flexural damage mode with considerable
displacement ratio triggers a larger deflection at the lower columns. In
ductility for inverted Y-shape RC pylon model in the test. Plastic
other words, the lower columns yield earlier while the upper columns
hinges were detected first at bottom of upper column, then at
yield somewhat later as compared to those under a larger loading dis-
bottom and top of lower column, successively, for the pylon model
placement ratio. This can be verified by an inspection of the yield-state
subjected to the two-node loading with a displacement ratio of 5.
points on the curves (e.g., comparing cases with loading displacement
These sections are deemed critical portions for seismic design of
ratios of 1 and 3). Another notable finding is that the loading dis-
inverted Y-shape RC pylons.
placement ratio can change the failure mechanism (i.e., occurrence
(3) Results from the refined numerical model of the inverted Y-shape
order of plastic hinges) of the RC pylon model. More specifically, cases
pylon with the bond-slip consideration at the base coincide rea-
with larger loading displacement ratios such as 5 and 7 trigger the first
sonably well with the test results in terms of global force-dis-
plastic hinge at the bottom of upper column, whereas those with rela-
placement curves, curvature and strain distributions along column
tively smaller loading displacement ratios (e.g. 1 and 3) yield first at the
heights, and displacements at multiple yield-states.
bottom of lower column. This is also due to the relatively larger de-
(4) A multi-level displacement ductility-limited factor related to the
flection for the lower column under a smaller loading displacement
number of plastic hinges is proposed in this study. The tested RC
ratio as compared to a larger one. In addition, it is interesting to note
pylon model under a displacement ratio of 5 shows ductility factors
that for cases with small loading displacement ratios, the bottom of
of 1.23, 1.43 and 3.00 for the second, third and fourth plastic hinges
upper column (① in Fig. 22) yields even later than the top of lower
formed in the test. Together with the further parametric study in-
column (②). Recalling the generally good correlation between loading
volving displacement ratios of 1, 3 and 7, mean ductility factors of
displacement ratios and bracketed durations of ground motions (Fig. 7),
1.22, 1.47 and 4.65 are detected for the second, third and fourth
it is reasonable to speculate that ground motions with relatively short
plastic hinges, respectively. Small dispersions are detected for
durations (e.g. those with bracketed durations less than 30 s, as referred
ductility factors for the second and third plastic hinges, while a
to Fig. 7) may transfer the location of first plastic hinge from the bottom
relatively larger dispersion is detected for that for the fourth plastic
of the upper column to that of the lower column, as compared to long
hinge.
duration ground motions. Moreover, it is worth noting that for the
(5) The parametric study indicate that the loading displacement ratio
studied largest loading displacement ratio of 7, the fifth plastic hinge
significantly affects the seismic failure mechanism and ductility of
occurs at the top of lower leading column (i.e., ⑤ in Fig. 22) before the
the studied RC pylon model. A relatively smaller displacement ratio
failure of any section.
generally transfers the location of first plastic hinge from the
Table 10 lists results of ductility factors, together with their statis-
bottom of the upper column to that of the lower column.
tical properties including mean, standard deviation and coefficient of
variance (COV). In general, the loading displacement ratio shows a
This study is limited to a scaled model of typical inverted Y-shape
generally slight impact on the displacement ductility-limited factors for
RC pylons for long-span cable-stayed bridges. Considering the length of
the second and third plastic hinges (i.e., COV less than 10%), while it
this paper, future studies will explore comprehensive parameter spaces
significantly affects those for the fourth plastic hinge and the ultimate
y
to assess the seismic failure mechanism and ductility factors of real
ductility factor. In particular, a larger value of μΔ,4 = 7.89 is obtained
inverted Y-shape RC pylons for various spans of cable-stayed bridges. In

219
X. Wang, et al.
addition, the research approach applied in this study can be used for
further investigations on other types of bridge pylons or multi-story
viaducts.
Acknowledgments
This study is supported by the Ministry of Science and Technology
of China (Grant No. 2013CB036302), National Natural Science
Foundation of China (Grant No. 51778469) and China Postdoctoral
Science Foundation (Grant No. 2018M640448). Special thanks to Mr.
Guang Chen and Dr. Xing Shen who provided great assistances to the
test.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.engstruct.2019.03.045.
References
[1] Ministry of Communications of the People’s Republic of China. Guidelines for
Seismic Design of Highway Bridges. Beijing, China: 2008.
[2] AASHTO. AASHTO LRFD bridge design specifications. 6th ed. Washington, D.C.:
American Association of State Highway and Transportation Officials (AASHTO);
2012.
[3] JSCE. Standard specifications for concrete structures-2002: seismic performance
verification. Japan Society of Civil Engineering; 2005.
[4] Loh C-H, Tsay C-Y. Responses of the earthquake engineering research community to
the Chi-Chi (Taiwan) earthquake. Earthq Spectra 2001;17(4):635–56. https://doi.
org/10.1193/1.1430680.
[5] Chang KC, Mo YL, Chen CC, Lai LC, Chou CC. Lessons learned from the damaged
Chi-Lu cable-stayed bridge. J Bridg Eng 2004;9(4):343–52. https://doi.org/10.
1061/(ASCE)1084-0702(2004) 9:4(343).
[6] Pappin JW, Kite S. Seismic design of Stonecutters Bridge, Hong Kong. In: Proc. 14th
World Conf. Earthq. Eng., Beijing, China: 2008.
[7] Combault J. The Rion-Antirion bridge-when a dream becomes reality. Front Archit
Civ Eng China 2011;5(4):415–26. https://doi.org/10.1007/s11709-011-0130-x.
[8] Arzoumanidis S, Shama AA, Marlow SJ, Orsolini GO. The New Tacoma Narrows
suspension bridge: critical issues in seismic analysis and design. Struct. Congr. 2005,
Reston, VA: American Society of Civil Engineers; 2005, p. 1–12. (171)21.
[9] Main JA, Jones NP. Evaluation of viscous dampers for stay-cable vibration miti-
gation. J Bridg Eng 2001;6(6):385–97. https://doi.org/10.1061/(ASCE)1084-
0702(2001) 6:6(385).
[10] Jung H-J, Spencer BF, Lee I-W. Control of seismically excited cable-stayed bridge
employing magnetorheological fluid dampers. J Struct Eng 2003;129(7):873–83.
https://doi.org/10.1061/(ASCE)0733-9445(2003) 129:7(873).
[11] Park K-S, Jung H-J, Lee I-W. Hybrid control strategy for seismic protection of a
benchmark cable-stayed bridge. Eng Struct 2003;25(4):405–17. https://doi.org/10.
1016/S0141-0296(02)00182-7.
[12] Soneji BB, Jangid RS. Passive hybrid systems for earthquake protection of cable-
stayed bridge. Eng Struct 2007;29(1):57–70. https://doi.org/10.1016/j.engstruct.
2006.03.034.
[13] Vader TS, McDaniel CC. Influence of dampers on seismic response of cable-sup-
ported bridge towers. J Bridg Eng 2007;12(3):373–9. https://doi.org/10.1061/
(ASCE)1084-0702(2007) 12:3(373).
[14] Zhu J, Zhang W, Zheng KF, Li HG. Seismic design of a long-span cable-stayed bridge
with fluid viscous dampers. Pract Period Struct Des Constr 2016;21(1):04015006.
https://doi.org/10.1061/(ASCE)SC.1943-5576.0000262.
[15] Combault J, Pecker A, Teyssandier J-P, Tourtois J-M. Rion-Antirion Bridge, Greece
– concept, design, and construction. Struct Eng Int 2005;15(1):22–7.
[16] Shen X, Wang X, Ye Q, Ye A. Seismic performance of Transverse Steel Damper
seismic system for long span bridges. Eng Struct 2017;141:14–28. https://doi.org/
10.1016/j.engstruct.2017.03.014.
[17] Camara A, Cristantielli R, Astiz MA, Málaga-Chuquitaype C. Design of hysteretic
dampers with optimal ductility for the transverse seismic control of cable-stayed
bridges. Earthq Eng Struct Dyn 2017;46(11):1811–33. https://doi.org/10.1002/
eqe.2884.
[18] Zhou L, Wang X, Ye A. Shake table test on transverse steel damper seismic system
for long span cable-stayed bridges. Eng Struct 2019;179:106–19. https://doi.org/
10.1016/j.engstruct.2018.10.073.
[19] Li J, Xiang N, Tang H, Guan Z. Shake-table tests and numerical simulation of an
innovative isolation system for highway bridges. Soil Dyn Earthq Eng
2016;86:55–70. https://doi.org/10.1016/j.soildyn.2016.05.002.
[20] Calvi GM, Sullivan TJ, Villani A. Conceptual seismic design of cable-stayed bridges.
J Earthq Eng 2010;14(8):1139–71. https://doi.org/10.1080/13632469.2010.
505275.
[21] Shen X, Camara A, Ye A. Effects of seismic devices on transverse responses of piers
in the Sutong Bridge. Earthq Eng Eng Vib 2015;14(4):611–23. https://doi.org/10.
1007/s11803-015-0049-7.
[22] Ren W-X, Obata M. Elastic-plastic seismic behavior of long span cable-stayed
220
Engineering Structures 189 (2019) 206–221
bridges. J Bridg Eng 1999;4(3):194–203. https://doi.org/10.1061/(ASCE)1084-
0702(1999) 4:3(194).
[23] Ren W-X, Lin Y-Q, Peng X-L. Field load tests and numerical analysis of Qingzhou
cable-stayed bridge. J Bridg Eng 2007;12(2):261–70. https://doi.org/10.1061/
(ASCE)1084-0702(2007) 12:2(261).
[24] Casciati F, Cimellaro GP, Domaneschi M. Seismic reliability of a cable-stayed bridge
retrofitted with hysteretic devices. Comput Struct 2008;86(17–18):1769–81.
https://doi.org/10.1016/j.compstruc.2008.01.012.
[25] Endo K, Kawatoh C, Unjoh S. Analytical study on seismic performance evaluation of
long-span suspension bridge steel tower. In: 13th World Conf. Earthq. Eng.,
Vancouver, Canada: 2004, p. No. 944.
[26] Camara A, Astiz MA. Pushover analysis for the seismic response prediction of cable-
stayed bridges under multi-directional excitation. Eng Struct 2012;41:444–55.
https://doi.org/10.1016/j.engstruct.2012.03.059.
[27] Pang Y, Wu X, Shen G, Yuan W. Seismic fragility analysis of cable-stayed bridges
considering different sources of uncertainties. J Bridg Eng 2014;19(4):04013015.
https://doi.org/10.1061/(ASCE)BE.1943-5592.0000565.
[28] Zhong J, Jeon J, Yuan W, DesRoches R. Impact of spatial variability parameters on
seismic fragilities of a cable-stayed bridge subjected to differential support motions.
J Bridg Eng 2017;22(6):04017013. https://doi.org/10.1061/(ASCE)BE.1943-5592.
0001046.
[29] Elnashai AS, Di Sarno L. Fundamentals of earthquake engineering. Chichester, UK:
Wiley; 2008. p. ..
[30] Godden WG, Aslam M. Dynamic model studies of Ruck-A-Chucky Bridge. J Struct
Div 1978;104(12):1827–44.
[31] Garevski MA, Brownjohn JMW, Blakeborough A, Severn RT. Resonance-search tests
on a small-scale model of a cable-stayed bridge. Eng Struct 1991;13(1):59–66.
https://doi.org/10.1016/0141-0296(91)90009-2.
[32] Yang C. Seismic analysis of long span bridges including the effects of spatial var-
iation of seismic waves on bridges. Ph.D. dissertation, Hong Kong. Hong Kong,
China: University of Science and Technology; 2007.
[33] Wang R, Xu Y, Li J. Transverse seismic behavior studies of a medium span cable-
stayed bridge model with two concrete towers. J Earthq Eng 2017;21(1):151–68.
https://doi.org/10.1080/13632469.2015.1118710.
[34] Xu Y, Zeng S, Duan X, Ji D. Seismic experimental study on a concrete pylon from a
typical medium span cable-stayed bridge. Front Struct Civ Eng 2018;12(3):401–11.
https://doi.org/10.1007/s11709-018-0464-8.
[35] Leon RT, Deierlein GG. Considerations for the use of quasi-static testing. Earthq
Spectra 1996;12(1):87–109. https://doi.org/10.1193/1.1585869.
[36] Abbiati G, Bursi OS, Caperan P, Di Sarno L, Molina FJ, Paolacci F, et al. Hybrid
simulation of a multi-span RC viaduct with plain bars and sliding bearings. Earthq
Eng Struct Dyn 2015;44(13):2221–40. https://doi.org/10.1002/eqe.2580.
[37] Bittner RB, Safaqah O, Zhang X, Jensen OJ. Design and construction of the Sutong
Bridge foundations. DFI J - J Deep Found Inst 2007;1(1):2–18. https://doi.org/10.
1179/dfi.2007.001.
[38] Ko JM, Ni YQ. Technology developments in structural health monitoring of large-
scale bridges. Eng Struct 2005;27(12):1715–25. https://doi.org/10.1016/j.
engstruct.2005.02.021.
[39] Wang H, Hu R, Xie J, Tong T, Li A. Comparative study on buffeting performance of
Sutong Bridge based on design and measured spectrum. J Bridg Eng
2013;18(7):587–600. https://doi.org/10.1061/(ASCE)BE.1943-5592.0000394.
[40] Wang H, Li A, Niu J, Zong Z, Li J. Long-term monitoring of wind characteristics at
Sutong Bridge site. J Wind Eng Ind Aerodyn 2013;115:39–47. https://doi.org/10.
1016/j.jweia.2013.01.006.
[41] Weber F, Distl H. Amplitude and frequency independent cable damping of Sutong
Bridge and Russky Bridge by magnetorheological dampers. Struct Control Heal
Monit 2015;22(2):237–54. https://doi.org/10.1002/stc.1671.
[42] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthq Eng Struct Dyn
2002;31(3):491–514. https://doi.org/10.1002/eqe.141.
[43] McKenna F. OpenSees: a framework for earthquake engineering simulation. Comput
Sci Eng 2011;13(4):58–66. https://doi.org/10.1109/MCSE.2011.66.
[44] Kent DC, Park R. Flextural members with confined concrete. J Struct Div
1971;97(7):1969–90.
[45] Scott BD, Park R, Priestley MJN. Stress-strain behavior of concrete confined by
overlapping hoops at low and high strain rates. J Am Concr Inst 1982;79(1):13–27.
https://doi.org/10.14359/10875.
[46] Wang X, Shafieezadeh A, Ye A. Optimal intensity measures for probabilistic seismic
demand modeling of extended pile-shaft-supported bridges in liquefied and lat-
erally spreading ground. Bull Earthq Eng 2018;16(1):229–57. https://doi.org/10.
1007/s10518-017-0199-2.
[47] Wang X, Ye A, Shafieezadeh A, Padgett JE. Fractional order optimal intensity
measures for probabilistic seismic demand modeling of extended pile-shaft-sup-
ported bridges in liquefiable and laterally spreading ground. Soil Dyn Earthq Eng
2019;120:301–15. https://doi.org/10.1016/j.soildyn.2019.02.012.
[48] He Z, Liu W, Wang X, Ye A. Optimal force-based beam-column element size for
reinforced-concrete piles in bridges. J Bridg Eng 2016;21(11):06016006. https://
doi.org/10.1061/(asce)be.1943-5592.0000926.
[49] MacRae GA. P-Δ effects on single-degree-of-freedom structures in earthquakes.
Earthq Spectra 1994;10(3):539–68.
[50] Ernst HJ. Der E-modul von seilen unter berucksichtigung des durchhanges. Der
Bauingenieur 1965;40(2):52–5.
[51] Caetano E, Cunha A, Taylor CA. Investigation of dynamic cable-deck interaction in
a physical model of a cable-stayed bridge. Part II: seismic response. Earthq Eng
Struct Dyn 2000;29(4):499–521. https://doi.org/10.1002/(SICI)1096-
9845(200004)29:4<499:: AID-EQE919>3.0.CO;2-A.
[52] Swanson Analysis Systems Inc. ANSYS User’s Manual, Version 8.0,. Houston, PA:
X. Wang, et al.
2004.
[53] Newmark NM. A method of computation for structural dynamics. J Eng Mech Div
1959;85(3):67–94.
[54] Crisfield MA. Nonlinear Finite Element Analysis of Solids and Structures, Volume 1:
Essentials. New York, NY: John Wiley; 1991.
[55] Earthquake Administration of Jiangsu Province. Seismic Design Parameters for
Ground Motion Inputs. Seismic Safety Evaluation of Site of Sutong Bridge, Nanjing,
China: Earthquake Administration of Jiangsu Province; 2000, p. 48–57 (in Chinese).
[56] Power M, Chiou B, Abrahamson N, Bozorgnia Y, Shantz T, Roblee C. An overview of
the NGA project. Earthq Spectra 2008;24(1):3–21. https://doi.org/10.1193/1.
2894833.
[57] Bommer JJ, Martinez-Pereira A. The effective duration of earthquake strong mo-
tion. J Earthq Eng 1999;3(2):127–72. https://doi.org/10.1080/
13632469909350343.
[58] Chandramohan R, Baker JW, Deierlein GG. Quantifying the influence of ground
motion duration on structural collapse capacity using spectrally equivalent records.
Earthq Spectra 2016;32(2):927–50. https://doi.org/10.1193/122813EQS298MR2.
[59] Tang WH, Ang A. Probability concepts in engineering: Emphasis on applications to
civil and environmental engineering. 2nd ed. Hoboken, NJ: Wiley; 2007.
[60] Rathje EM, Abrahamson NA, Bray JD. Simplified frequency content estimates of
earthquake ground motions. J Geotech Geoenviron Eng 1998;124(2):150–9.
https://doi.org/10.1061/(ASCE)1090-0241(1998) 124:2(150).
[61] Hancock J, Bommer JJ. The effective number of cycles of earthquake ground mo-
tion. Earthq Eng Struct Dyn 2005;34(6):637–64. https://doi.org/10.1002/eqe.437.
[62] Bolt BA. Duration of strong ground motion. 5th World Conf. Earthq. Eng., vol. 1,
Rome, Italy: 1974, p. 1304–13.
[63] Arias A. A measure of earthquake intensity. Santiago, Cambridge, MA: MIT,
221
Engineering Structures 189 (2019) 206–221
Cambridge and University of Chile; 1970.
[64] Buckingham E. On physically similar systems; Illustrations of the use of dimensional
equations. Phys Rev 1914;4(4):345–76. https://doi.org/10.1103/PhysRev. 4.345.
[65] Dhakal RP, Su J. Design of transverse reinforcement to avoid premature buckling of
main bars. Earthq Eng Struct Dyn 2018;47(1):147–68. https://doi.org/10.1002/
eqe.2944.
[66] Wang X, Ye A, He Z, Shang Y. Quasi-static cyclic testing of elevated RC pile-cap
foundation for bridge structures. J Bridg Eng 2016;21(2):04015042. https://doi.
org/10.1061/(ASCE)BE.1943-5592.0000797.
[67] Wang X, Ye A, Shafieezadeh A, Li J. Shallow-Layer p-y relationships for micropiles
embedded in saturated medium dense sand using quasi-static test. Geotech Test J
2018;41(1):20160289. https://doi.org/10.1520/GTJ20160289.
[68] Zhao J, Sritharan S. Modeling of strain penetration effects in fiber-based analysis of
reinforced concrete structures. ACI Struct J 2007;104(2):133–41. https://doi.org/
10.14359/18525.
[69] CEB-FIP. Bond of Reinforcement in Concrete. Bulletin 1. Lausanne, Switzerland:
CEB-FIP Model Code 1990; 2000.
[70] Scott MH, Fenves GL. Krylov subspace accelerated Newton algorithm: application to
dynamic progressive collapse simulation of frames. J Struct Eng
2010;136(5):473–80. https://doi.org/10.1061/(ASCE)ST.1943-541X.0000143.
[71] Lowes LN, Mitra N, Altoontash A. A beam-column joint model for simulating the
earthquake response of reinforced concrete frames a beam-column joint model for
simulating the earthquake response of reinforced concrete frames. PEER Rep 2003/
10 2003.
[72] Blanco G, Ye A, Wang X, Goicolea JM. Parametric pushover analysis on elevated RC
pile-cap foundations for bridges in cohesionless soils. J Bridg Eng
2019;24(1):04018104. https://doi.org/10.1061/(ASCE)BE.1943-5592.0001328.