Thin-Walled Structures 116 (2017) 212–224

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Experiment on seismic performance of concrete filled steel tubular arch-rib MARK

under multi-shaking-tables
⁎
Fu-yun Huanga,b, , Cui Fua, Yi-zhou Zhuanga, Zheng-hui Xionga
a
College of Civil Engineering, Fuzhou University, 2 Xueyuan Road, University Town, Fuzhou, Fujian 350108, China
b
College of Civil and Environmental Engineering, University of Nevada, Reno, NV 89557, USA

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

Keywords: During the last two decades, a large number of concrete filled steel tubular (CFST) arch bridges had been built in
CFST arch-rib model China. However, the experiment on seismic performance of CFST structural arch rib was seldom conducted
Experiment especially under spatial wave excitation. In order to understand the seismic behaviors of CFST arch bridges, a
Multi-shaking-tables scaled 1:10 model of CFST arch-rib was constructed with fully compensative mass. The test was carried out
Uniform and non-uniform excitation
uniform and non-uniform excitations under a Multi-shaking-tables array system in Fuzhou University, China.
Seismic performance
Artificial waves and several recorded ground motions were employed to identify the basically dynamic behavior
and to evaluate the elastic and plastic seismic performance of this arch-rib model.
The testing results indicate that the seismic responses of CFST arch-rib are highly related to the predominant
frequencies, ranges as well as peak magnitudes of input waves’ spectrum. The testing results also show that the
transverse seismic response is more distinct than that longitudinal response, and the bi-directional excitation has
much larger responses than the single excitation. Furthermore, the non-uniform excitation has more significant
effect than the uniform wave especially on the displacement responses. Therefore, the earthquake effect on the
CFST arch bridge need necessary to take account of bi-directional excitation and non-uniform excitation.
In addition, the testing indicates that the CFST arch-rib structure has performed the nonlinear and plastic
behavior especially under intensive excitations of El-centro wave and Hanshin wave. However, the natural
frequencies or stiffness of the CFST arch-rib model don’t have significant decrease and there is no distinct crack
or damage on the steel surface because of the hooping effect on the core concrete. This indicates that the CFST
arch-rib structure has a favorable seismic behavior and ductility.

1. Introduction CFST arch-rib had a reasonable mechanical performance and load-

The main shortcomings for conventional large-span reinforced
concrete arch bridge are its huge self-weight and difficult construction
as well as poor ductility [1]. Compared to reinforced concrete arch
bridge, concrete filled steel tubular (CFST) arch bridge is more
convenient and lightweight in construction as well as high cost-
efficiency of interaction between the steel tube and the core concrete
[2].
Concrete filled steel tubular arch bridges have been widely used in
China. The largest one is the Hechuan Yangtze River Bridge located in
Sichuan province, Southwest of China, with clear span of 530 m.
A large number of static tests on mechanical behavior of CFST arch-
rib model had been conducted to identify the basic static mechanical
behavior and ultimate carrying-load capacity. The tests showed that
carrying capacity [3–5]. However, comprehensively seismic experiment
on CFST arch structure under Shaking Tables had seldom been carried
out especially under spatial excitations [6,7]. Previously, some re-
searches mainly studied the basically dynamic behavior and mode
analysis based on a practical CFST arch bridge [8–11]. Most recently,
researchers have analyzed that the CFST arch bridge possesses a
reasonable seismic performance [12–14,15]. Nevertheless, almost all
of these analyses and results were gained by the finite element models
(FEM) rather than based on the experiments [16,17]. Therefore, the
experimental study on seismic performance of CFST arch bridge model
under shaking tables is essential and significant.
This paper described a Multi-shaking-tables experiment on a scaled
1:10 CFST arch-rib model with fully slave mass that were conducted in
Fuzhou University, Fujian Province, China. The testing CFST arch-rib

Abbreviations: AR, Acceleration response; CFST, Concrete filled steel tube; E1 & E2, Artificial waves for two level design; L-S, Left arch springing; MF, Magnification factor; PFP,
Predominant frequency platform; PGA, Peak ground acceleration; R-S, Right arch springing
⁎
Correspondence to: School of Civil Engineering, Fuzhou University, Fuzhou 350108, China.
E-mail address: Huangfuyun@fzu.edu.cn (F.-y. Huang).

http://dx.doi.org/10.1016/j.tws.2017.03.020
Received 14 November 2015; Received in revised form 1 March 2017; Accepted 14 March 2017
0263-8231/ © 2017 Elsevier Ltd. All rights reserved.
F.-y. Huang et al. Thin-Walled Structures 116 (2017) 212–224

Nomenclature Ci Seismic essentiality coefficient (see Eq. (2))

Cs Field factor (see Eq. (2))
l & Sl Quantity and scale of length (see Eq. (1)) D Time-history displacement (see Eq. (3))
δ & Sl Quantity and scale of deformation (see Eq. (1)) D1 Time-history displacement relative to side small table (see
E & SE Quantity and scale of elastic modulus (see Eq. (1)) Eq. (3))
σ &Sσ Quantity and scale of stress (see Eq. (1)) D2 Time-history displacement relative to mid large table (see
ε&Sε Quantity and scale of strain (see Eq. (1)) Eq. (4))
ρ&Sρ Quantity and scale of density (see Eq. (1)) Dr Relative peak displacement (see Eq. (4))
ρ′&Sρ′ Quantity and scale of density (see Eq. (3)) Es Young's modulus of steel
a&Sa Quantity and scale of acceleration (see Eq. (1)) Ec Young's modulus of concrete
g&Sg Quantity and scale of gravity acceleration (see Eq. (1)) f Natural frequency of structure and sample frequency of
P &SP Quantity and scale of load input excitation
M &SM Quantity and scale of moment fcu 28-day cubic strength of concrete
T & ST Quantity and scale of period (see Eq. (1)) fs Yield stress of steel
ω&Sω Quantity and scale of natural circular frequency (see Eq. fu Strength stress of steel
(1)) L Arch spans’ length (see Fig. 5)
ζ & Sξ Quantity and scale of damp (see Eq. (1)) H Arch rise (see Fig. 5)
v&Sv Quantity and scale of velocity (see Eq. (1)) Smax Predominant frequency platform (see Eq. (2))
k &Sk Quantity and scale of stiffness t Time
m m &Smm Quantity and scale of model mass Tg Characteristic period (see Eq. (2))
ma &Sma Quantity and scale of artificial mass μc Poisson's ratio of concrete
A Designed horizontal peak acceleration (see Eq. (2)) μs Poisson's ration of steel
Cd Adjustment factor of damp (see Eq. (2))

model was subjected to uniform and non-uniform earthquake excita-
tions at low intensity or high intensity to evaluate its elastic and plastic
behaviors. The sine wave, and the recorded earthquakes of Taft wave,
EL-centro wave, Hanshin wave as well as the artificial waves of E1 and
E2 that established by the Guidelines Specification for Seismic Design of
Highway Bridge, China [18], are utilized. Then, the basically findings
were presented including the dynamic behavior of arch-rib, the seismic
responses of acceleration, displacement and strain, et al. At last,
comments and discussions are presented based on the damage and
failure modes.
to fasten the model upon the Shaking-tables by high strength bolts.
The cross-section of the arch-rib model consisted of single tube and
core concrete. According to the similarity law, the outer diameter and
wall thickness of the steel tube were 80 mm and 1.4 mm. Nevertheless,
they could not be available in steel tube market. Finally, the dimension
of 76 mm×3.8 mm was preferred instead of theoretical size. By the
way, both of two cross-sections have the same relatively composite
stiffness of EscIsc basically. That is to say the 76 mm×3.8 mm tube has a

2. Prototype bridge

The prototype of Qunyi Bridge shown in Fig. 1(a), similar to the

New Damen Ave. Bridge in Chicago City, locates at Fu-an County,
Southeast China. It is a half-through CFST arch bridge with clear span
and arch-rise of 46 m and 15.33 m, respectively. The total width of deck
is 18 m, which includes the clear lane width of 9 m+2×2.5 m, the
footway of 2×1.75 m and so on.
The main arch-ribs are composed of single steel tube and core
concrete with outside diameter of 800 mm and thickness of 14 mm. The
steel tube is Q345 that is similar to the A572 Grade 50 steel in the ASTM
and core concrete is C30. The bridge has a gravity abutment located
upon the enlarged rigid foundation to resist the reaction-force at arch
springing. The ground site belongs to the Classification II that is similar
to stiff soil conditions (Soil Type II) in AASHTO (2014), and its seismic
intensity is VII according to Chinese Specification [18].

3. Model

3.1. Model design

The CFST arch-rib model was designed to be a scale of 1:10 that

illustrated in Fig. 1. The arch-rib axis was the parabola that similar to
prototype, and the corresponding clear span and rise were 4.6 m and
1.53 m, respectively. In order to implement the slave mass for
compensating gravity weight due to reduced scale, total of 24 steel
bars were perpendicularly welded on the model along the arch-rib axis
with a spacing of 0.2 m. The steel plate with the size of
1.0 m×1.0 m×0.035 m were welded on each arch springing in order Fig. 1. CFST arch-rib.

213
F.-y. Huang et al. Thin-Walled Structures 116 (2017) 212–224

higher hooping effect while a lower section, whereas 80 mm×1.4 mm
has a larger section while a smaller hooping effect (or vice versa).
Therefore, their whole EscIsc are similar, which can minimize the
influence of size on the seismic behavior of CFST members.
3.2. Model material
The material of arch rib, steel tube and core concrete, were almost
identical to the original arch bridge. Its properties for steel were:
ultimate strength stress fu =540 MPa, yield strength stress fs =320 MPa,
Young's modulus Es=2.00×105 MPa and Poisson's ratio μs =0.283. The
properties for core concrete were: the cubic compression strength
fcu =45.6 MPa, Young's modulus Ec=3.65×104 MPa and Poisson's ratio
μc =0.245.
3.3. Model similarity and compensation mass
For the seismic response analysis on almost all of structures, it can
be formulated by the equation of (1).
σ = f (l , E , ρ , t , δ , v , a , g , ω )
(1)
where, σ is dynamic response stress of structure; l, E and ρ are
geometrical dimension, elastic modulus and density, respectively; t is
time, and δ is the deformation; v, a and g are structural response of
velocity, acceleration and gravity; ω is the natural circular frequency of
structure.
Generally, the absolute similitude ratios of dimensions are very
difficult to achieve in dynamic model design based on similarity law.
For example, if taking the quantities of l, E and ρ as the basic dimension
in this arch-rib model, thus the similarity coefficients SL=0.1(for the
scale is 1:10), SE and Sρ =1.0(due to the same materials). As the
similarity coefficient of gravity acceleration Sg cannot be varied and
always equals to 1. However, by calculating the acceleration coeffi-
cients of Sg (or Sa ) from formula (2), it will be equal to 10.0 rather than
the constant 1.0 if substituting the three basic coefficients SL=0.1,
SE=1.0 and Sρ =1.0 just as formula (2).
Sg (or Sa ) = E ρ−1L−1 = SE /(Sρ SL ) = 1.0/(1.0 × 0.1) = 10.0 ≠ 1.0 (2)
Actually it is very difficult to meet the requirement of Sg =Sa =1.
Therefore, most of scaled models under Shaking-table are only by
adding partly man-made mass and magnifying the input PGA to meet
the requirement of response acceleration (Sa ), while neglecting gravity
Sg . However, this will be induced the distortion of seismic response.
In this paper, a pseudo-density coefficient S′ρ is assumed and equal
to 10.0 by subjected to enough man-made (slave) mass on the arch-rib
model to meet the requirement of Sg =Sa =1. In other words, the whole
weight of compensation mass should be 9 times of self-weight of the
[18], were utilized to test the basically dynamic behavior and elastic
response. For the E2 (500-year return-period, the second level seismic
fortifications in Chinese Code) and the rare earthquakes of Taft, El-
centro as well as Hanshin wave were selected to test plastic response,
damage and failure mode. The peak acceleration of E1 wave was 0.1g
based on the Chinese design; both E2 wave and Taft wave were 0.26g;
while El-centro wave and Hanshin wave were 0.34g and 0.78g,
respectively. Here the similarity coefficients of acceleration are satisfied
by subjected to fully compensative mass, thus the excitations don’t need
to magnify, just only to compress the time with scale factor.
Otherwise, due to the device's precision and deviation of Shaking-
tables, the input excitations would be a little different to the actually
acquired values on the shakers [19]. Therefore, the acquired data listed
in Table 2 were the real input excitations.
The sample frequency f for all input excitations should be trans-
formed to 2 n, and also must be at least over 2 times of the maximum
working frequency of Shaking-tables, 50 Hz. In this paper, the sample
frequency f=256 Hz is used for all input waves except Taft wave, which
is 512 Hz.
In addition, E1 and E2 waves were generated by employing the
standard response spectrum of the Seismic Design of Highway Bridge
[18] based on the local site. The function of the response spectrum is
defined as following:
⎧ Smax (5.5T + 0.45) T < 0.1s
⎪
S = ⎨ Smax 0.1s ≤ T ≤ Tg
⎪ S (T / T )
⎩ max g T > Tg (2a)
Smax = 2.25Ci Cs Cd A (2b)
Where, S, denotes the maximum horizontal acceleration response
spectrum (ARS) and Smax is the predominant frequency platform
(PFP); Tg is the characteristic period (unit: s); T is the model's vibration
period; Ci is the seismic essentiality coefficient; Cs is the field factor; Cd
is the adjustment factor of damping; A is the design horizontal peak
ground acceleration (PGA) of basic earthquake intensity.
Taking the case of E2 artificial wave as an example, these factors of
Tg , Ci , Cs , Cd and A are 0.35 s, 1.3, 1.0, 1.0 and 0.2 g respectively. E1 and
E2 waves were illustrated in Fig. 2.
4.2. Analysis on the response spectrum of excitations
The designing and fitting curves of acceleration response spectrum
(ARS) to frequency are shown in Fig. 3, including artificial E2 wave,
Table 1
Similarity coefficients.

model, and they were imposed on the model just as shown in Fig. 1. Quantities Scales Scaled relationships Scaled ratios
Therefore, the update acceleration coefficients and formula (3) are
following, Length/l SL [L] 1/10, SL=0.1
Deformation/δ SL
Sg (Sa ) = E ρ’−1 L−1 = SE /(S’ρ SL ) = 1.0/(10.0 × 0.1) = 1.0 (3)
Modulus/ E SE [E] SE =1.0
, and which meets the similarity relationship and also avoids the Stress/σ Sσ Sσ = SE Sε =1.0
Strain/ε Sε — Sε =1.0
distortion compared the general models without enough artificial mass.
Density/ρ Sρ [ρ] Sρ =10.0
Based on the three fundamental coefficients of SL, SE, and updated
pseudo-density coefficient S′ρ , all others coefficients were derived and Acceleration/a Sa [Eρ−1L−1] 1.0

listed in Table 1. Gravity acceleration/ g Sg

Load/P SP [EL2] SP = SE SL2 =0.01
4. Testing Moment/M SM [EL2] SM = Sσ SL3 =0.001
Time/T ST [E−0.5ρ0.5L] ST = 1/Sω =0.316
4.1. Input excitations Circular frequency/ω Sω [E0.5ρ−0.5L−1] Sω =3.16
Damp/ζ Sξ — Sξ =1.0
Velocity/v Sv [T−0.5] Sv = Sω =3.16
The input excitations included the white noise, the artificial waves Stiffness/k Sk [EL] Sk = SE SL =0.1
of E1 and E2 as well as the real earthquakes of Taft, El-centro and Model mass/mm Smm [ρL3] Sm = Sρ SL3 =0.001
Hanshin. For artificial wave of E1, which is corresponding to first level Artificial mass/ma Sma [EL−2]-[ρL3] Sa = 9Sm = 0.009
seismic fortifications (medium earthquake) based on Chinese Code

214
F.-y. Huang et al. Thin-Walled Structures 116 (2017) 212–224

Table 2
Cases and corresponding peak accelerations of excitation and response (Unit: m/s2).

Excitation Responses

Uniform excitation Non-uniform excitation

X Y X+0.85Y X-Traveling Y-Traveling

(at L/4) ( at crown, L/2) (at crown, L/2) (at L/4) (at crown, L/2)

Waves Peak values Case No. Peak values Case No. Peak values Case No. Peak values Case No. Peak values Case No. Peak values

WN 1. 0 Case0 1.25 Case0 1.69 / / / / / /

E1 1.04 Case1 4.08 Case2 4.60 Case3 6.65 / / / /
E2 2.39 Case4 9.31 Case5 10.98 Case6 14.82 / / / /
Taft 2.39 Case7 4.37 Case8 3.58 Case9 5.80 Case10 4.94 Case11 3.50
El-centro 3.40 Case12 8.74 Case13 13.86 Case14 16.96 / / / /
Hanshin 7.65 Case15 12.52 Case16 18.95 Case17 21.44 / / / /

Note: ‘X′ means longitudinal direction; ‘Y′ means transverse (lateral) direction to X; ‘/’ means no test.

Taft wave, El-centro wave and Hanshin wave, respectively.
It can be clearly seen from Fig. 3 that the platform of Smax for Taft
wave is about from 15 Hz to 25 Hz, and E2 is about from 2.0 Hz
to10.0 Hz, El-centro from 2.0 Hz to 8.0 Hz and Hanshin from 2.0 Hz to
3.0 Hz, respectively. By Compared these PFP (Smax ), it shows that each
wave has a relatively lower initial frequency platform (about 2.0 Hz)
except Taft wave, which is over 15 Hz. Furthermore, for the Smax of
Hanshin wave, it is very large but narrow just like a shock wave,
whereas others Smax are far wider.
4.3. Setups
The responses of acceleration, strain and displacement in key
sections of this CFST arch-rib model were measured. The key sections
involved the left and right arch springing (L-S and R-S), arch crown (L/
2) and two quarter-spans (1 L/4 and 3 L/4), et al. 23 accelerometers
and 4 displacement transducers were set up in these key locations. In
addition, 36 strain gauges around these cross-sections with axial and
circumferential direction such as top and bottom edges, front and rear
edges were placed. Fig. 4 is the set-up of strain gauges, accelerometers
and displacement transducers, et al.
4.4. Testing cases
considering both wave velocity and time compressed.
5. Test results and analyses
5.1. Dynamic behavior
Under white noise excitation (Case 0) and Fast Fourier transform
function, the natural frequencies, mode-shapes and damping of CFST
arch-rib model were acquired and identified. The several previous
mode-shapes of the CFST arch structural model are the same as its
prototype bridge. These modes are in sequence of out-plane symme-
trically transverse buckling, in-plane anti-symmetrically longitudinal
buckling, out-plane anti-symmetrically transverse buckling and in-
plane symmetrically vertical buckling, and their corresponding natural
frequencies are 2.1 Hz, 5.5 Hz, 6.0 Hz, 12.0 Hz, respectively. Thus the
basically natural period is about 0.48 s.
Based on the similarity law, the extrapolations of natural frequen-
cies of prototype will be 0.66 Hz, 1.74 Hz, 1.90 Hz and 3.80 Hz as the
frequent scale of Sω is 3.16 from Table 1, which are slightly larger than
the testing results in field. Besides, the damping ratio of the model is
0.012 which is less than the standard magnitude of 0.05. Therefore, its
damping ratio is closer to the steel tubular arch rather than concrete
arch under elastic excitation.

Total of 18 testing cases from ‘Case0′ to ‘Case17′ that listed in
Table 2 were presented, which covered the basic dynamic behavior
tests, the elastic-plastic seismic response as well as failure tests. From
Table 2, we can see that the arch-rib model was excited by the uniform
and non-uniform seismic excitation. For the uniform direction, it
involved single longitudinal (X-) and single transverse (Y-) input as
well as bi-directional input of X+0.85Y. For the non-uniform direction,
the traveling wave was input in single X- or Y- direction, and it traveled
from left arch springing (L-S) to right springing (r-L) with 0.1 s delay by
5.2. Results of acceleration response
5.2.1. Results under E1 waves
The CFST arch model had been shaken with artificial waves of E1 to
test its elastic behavior under X- and Y-directional excitations. The
corresponding peak ground motions are 1.04 m/s2 just as the Case 1–3
in Table 2. Figs. 5 and 6 show the time history curves of acceleration
response in the key sections of arch crown, 1/4 L and 3/4 L, other
locations do not illustrate due to the limit space hereafter.

Fig. 2. Artificial waves of E1 and E2.

215
F.-y. Huang et al.
Fig. 3. Curves of ARS of input waves.
From the results, it can be seen that the responses of acceleration
time-history at crown and quarter-spans have a little different with the
input wave of E1 excitation wave (at springings). In Fig. 5, the peak
acceleration responses at the crown and the quarter spans are 3.86 m/s2
and 4.08 m/s2, respectively. The maximum magnification factor is close
to 4.0 times of excitation, which means that the E1 artificial wave has
significant effect on the model. Need to mention that the peak response
acceleration at quarter-spans is a little larger than that at crown under
X-longitudinal excitation, but do not have too much different.
In Fig. 6, their peak acceleration responses reach up to 4.60 m/s2
and 3.75 m/s2 at the crown and the quarter spans respectively, and
corresponding peak magnification are about 4.5 times. It can be seen
that transverse responses are a little larger than longitudinal one by
comparing the excitation of X direction and Y direction. After that, the
model decays with the same natural period of 0.48 s, and that means
the model is elastic.
5.2.2. Results under E2 waves
The model had also been shaken by E2 to test its plastic behavior
Fig. 4. Set-ups of strain gauges, accelerators and displacement transducers (Unit: mm).
216
Thin-Walled Structures 116 (2017) 212–224
under X- and Y-directional excitations. The corresponding peak ground
motions are 2.39 m/s2, just as the Case 4–6 in Table 2, and their results
show in Figs. 7 and 8.
In Figs. 7 and 8, the peak response accelerations at quarter span and
crown are 9.31 m/s2 and 10.98 m/s2 under E2, almost as the same
magnification factors as E1. Moreover, the model's decay period is
identical to 0.48 s after under bi-directional excitation (Case 6). That
means the CFST arch-rib model is still linear yet and doesn’t damage
under the high intensity earthquake. The CFST arch-rib has a favorable
seismic performance.
5.2.3. Results under Taft wave
5.2.3.1. The results of uniform excitations. Fig. 9 and Fig. 10 are the time
history curves of longitudinal and transverse acceleration responses of
the CFST arch structural model under corresponding directional
excitation of Taft wave with the peak acceleration of 2.39 m/s2,
which correspond to the Case 7 & 8 in Table 2.
From Fig. 9, the acceleration responses at quarter-spans of 1/4 L and
3/4 L as well as arch crown (1/2 L) are −4.37 m/s2, 3.56 m/s2 and
F.-y. Huang et al. Thin-Walled Structures 116 (2017) 212–224

Fig. 6. Transverse acceleration response under Y-directional E1 artificial wave.

Fig. 5. Longitudinal acceleration response under X-directional E1 artificial wave.

Fig. 7. Longitudinal acceleration response under X-directional E2 artificial wave.

Fig. 8. Transverse acceleration response under Y-directional E2 artificial wave.

217
F.-y. Huang et al.
Fig. 9. Longitudinal acceleration response under X-directional Taft wave excitation.
2.61 m/s2 respectively under the excitation of longitudinal X-direction.
The magnifications of response acceleration per excitation are about
1.83, 1.49 and 1.09, respectively. The absolute magnitudes of response
at 1/4 L and 3/4 L are much larger than that at arch crown. That reason
is the Taft wave mainly excites the second natural mode of in-plane
longitudinal unsymmetrical instability under longitudinal excitation.
Moreover, after the termination of excitation, the model will be
attenuated with a period of 0.182 s, which corresponds to the second
natural frequency of 5.5 Hz.
From Fig. 10, all the peak values of acceleration response at arch
quarter spans and crown are nearly same under the excitation of
transverse Y-direction, and they are 3.59 m/s2, 3.58 m/s2 and 3.56 m/
s2 respectively. The magnification factors are just about 1.5 times and
after the termination of excitation, the model will be attenuated with
period of 0.48 s, which corresponds to the first natural frequency of
2.1 Hz.
Generally the acceleration responses of CFST arch-rib model under
Y-transverse excitation are larger than X-longitudinal excitation, just as
E1 and E2 waves. However, for Taft wave, its X-longitudinal response is
larger than that of Y-. Furthermore, even thought the input peak
acceleration of E2 is identical to Taft, but the responses of E2 are much
larger than that of Taft. The reason is that the structural response is
relevant to the predominant period and frequency of response spectrum
besides the peak ground motion. The predominant response spectrum
of Taft wave is closer to the second natural in-plane structural
frequency than the first natural out-plane structural frequency, there-
fore, the in-plane response will be dominated and be larger than that of
out-plane.
Fig. 10. Transverse acceleration response under Y-directional Taft wave excitation.
218
Thin-Walled Structures 116 (2017) 212–224
5.2.3.2. Bi-directional and traveling excitations. While being excited with
bi-axial seismic waves (Case 9), the peak responses of acceleration at
quarts (1/4 L and 3/4 L) and arch crown(1/2 L) are 5.8 m/s2, 5.6 m/s2
and 4.8 m/s2, respectively just as listed in Table 2, which are far larger
than single excitation. It can be concluded that the CFST arch bridge
needs to consider bi-axial excitation when calculating the earthquake
effect.
Under the longitudinal excitation of X-directional traveling wave
(Case 10), the arch acceleration responses of left half-span (beside small
shaker) is slightly 15% larger than the uniform wave excitation, while
right half-span (beside large shaker) do not have much change.
Otherwise, under the transverse excitation of Y-directional traveling
wave (Case 11), the acceleration response does not have significant
difference to the effect of uniform excitation. Due to the limited space,
all of these figures are not presented in here.
5.2.4. Results under El-Centro wave
Figs. 11 and 12 are the curves of time-history vs. response
acceleration at crown and quarter spans of CFST arch-rib model under
X and Y directional excitation of El-centro wave.
It can be seen from Fig. 11 that the peak acceleration responses in
the section of 1 L/2, 1 L/4 and 3 L/4 are 7.54 m/s2, 7.33 m/s2, 8.74 m/
s2, respectively, whose magnification factors are 2.22, 2.16 and 2.57 as
the peak ground acceleration of the excitation is 3.40 m/s2 (Case 12).
The magnification factor of El-centro wave is also larger than that of
Taft wave, but lower than that of artificial wave. Therefore, its
acceleration responses are lower than E2 even though its excitation is
larger than E2 just as listed in Table 2. The times reaching the peak
acceleration responses at the crown and quarter are about 2.64 s and
F.-y. Huang et al.
Fig. 11. Longitudinal acceleration response under X-directional EL-Centro wave excitation.
2.43 s, respectively, which has a slightly delay compared to the peak
excitation time of 2.25 s.
From Fig. 12, it can be seen that the peak acceleration responses in
the sections of 1 L/2, 1 L/4 and 3 L/4 are very large, which are
13.86 m/s2, 8.66 m/s2 and 8.25 m/s2, respectively. Their correspond-
ing magnification factors are 4.08, 2.56 and 2.51 as the peak ground
acceleration of the excitation is 3.40 m/s2 (Case 13).
The times reaching peak acceleration response for the crown is
about 5.35 s, which obviously delays over 3 s compared to the peak
excitation time of 2.25 s.
After the excitation of Case 14 is over, the model declines with the
period of 0.49 s that has a little increase compared to the previous
period of 0.48 s, which means the stiffness of model dose not decrease
significantly even though the peak acceleration response reaches up to
16.96 m/s2.
5.2.5. Results under Hanshin Wave
Figs. 13 and 14 are the model's results of acceleration response time-
history curves at crown and quarter spans under X- and Y- directional
excitation of Hanshin wave (Case 15 & 16).
From Fig. 13, the acceleration responses at crown and quarters are
magnified a little to 11.35 m/s2, 12.52 m/s2 and 11.17 m/s2, while the
excitation is 7.65 m/s2 just as lists in Table 2. It can be seen that
Hanshin wave don’t magnify too much in longitudinal X-direction.
Otherwise, the delay time for reaching the peak acceleration response is
about 0.28 s.
From Fig. 14, it can be seen that the response has a little different at
crown and quarters span under excitation of Y-direction. The PGA is
18.95 m/s2, which is far larger than X-direction and the time delays
about 1.2 s. After the completion of excitation of Case 17, the model's
natural period will be declined a little to 0.50 s from the previous
period of 0.49 s. It means that the CFST arch-rib model has a favorable
seismic performance even though the peak acceleration response
Fig. 12. Transverse acceleration response under Y-directional EL-Centro wave excitation.
219
Thin-Walled Structures 116 (2017) 212–224
reaches up to 21.44 m/s2.
5.2.6. Magnification factors
Fig. 15 is the comparison of magnification factors from left spring-
ing (L-S) to right springing (r-S) along the arch span under different
directions and excitations. As the artificial waves of E1 and E2 have
almost the same factors, thus only E2 is selected to demonstrate in
Fig. 15 to compare with other earthquake motions.
From Fig. 15(a), it can be seen that the artificial wave of E2 has the
biggest magnification factors and more distinct response than other
waves. The maximum factors for Hanshin and Taft are basically less
than 2 times. Besides, the maximum factors are generally appeared at
quarter-span rather than at arch crown especially for Taft, which
appears around 1L/8. Even thought the arch-rib model is fully
symmetric, but the responses don’t be so symmetric. The failure mode
is the in-plane asymmetrically longitudinal buckling.
From Fig. 15(b), it is clear that the peak magnification factors are
appeared at arch crown especially for E2 wave and El-centro wave.
Besides, compared to the artificial waves and El-centro wave, their
magnification factors are very close to each other. While for Taft wave,
its factors are very small and nearly can be ignored to consider the
seismic influence. Besides, the failure mode is out-plane symmetrically
transverse buckling.
5.3. Displacement response
The measured displacement of the testing CFST arch-rib model is
the absolute value involved the Shaking-tables’ motion. The relative
displacement of the testing model is the measured absolute value
subtracting the Shaker's motion. If take the left side of small Shaking
table (Fig. 4) as the reference point, the formula of relative displace-
ment in the uniform excitation test can be expressed as:
Dr = max | D − D1 (3)
F.-y. Huang et al.
Fig. 13. Longitudinal acceleration response under X-directional Hanshin wave excitation.
Where Dr denotes the relative peak displacement; D is the time-history
displacement; D1 is the excitation displacement of small side shaker.
While for the traveling excitation, it is:
Dr = max | D − (D1 − D2 ) |
Where D2 is the relative displacement of the middle large Shaker; others
signs have same explanation as in Eq. (3).
Generally, the displacement in the section of arch crown is relatively
larger than that in other sections. Besides, transverse and longitudinal
deformations are much larger than the vertical one. Therefore, Figs. 16
and 17 only lists the relatively peak displacement response in the arch-
crown section under the uniform excitation, bi-dimensional excitation
and traveling wave excitation.
It can be seen from Fig. 16(a) that the sequence of displacements
from small to large is usually the Taft, the artificial, the El-centro and
the Hanshin, and from Fig. 16(b) the sequence of displacements is
similar only except E2 and El-centro. By comparing Fig. 16(a) and
Fig. 16(b), the displacements under Y-directional excitation are gen-
erally larger than under X-directional. The maximum relative displace-
ment Dr under rare earthquake of Hanshin wave is 25.0 mm under Y-
directional excitation, which is about 33.0% per diameter of cross-
section and 1.67% per arch-rise (=Dr/H=25 mm/1530 mm).
The comparisons of relatively peak displacement at crown under
varying excitations are shown in Fig. 17. Fig. 17(a) shows that E2 wave
causes the largest X-directional displacement, and the Dr is 8.0 mm;
while Taft wave only caused Dr=1.5 mm, just 1/5.33 of E2. Fig. 17(b)
shows that the relative Y-displacement caused by E2 artificial wave
exceeds 11.0 mm. Therefore, although the E2 artificial wave and Taft
wave have almost the same peak acceleration of excitation, the latter
generated only approximately displacement of 1.0 mm, just 1/11 of E2
and even smaller than E1.
Fig. 17 also indicates that the relative displacement caused by the
non-uniform excitation was significantly larger than uniform excitation.
Besides, the displacement under bi-directional excitation is also larger
Fig. 14. Transverse acceleration response under Y-directional Hanshin wave excitation.
Thin-Walled Structures 116 (2017) 212–224
than that of uniform excitation. Therefore, it is necessary to consider
both the bi-directional and traveling excitation for CFST arch bridge in
practical engineering.
(4)
5.4. Strain response
In this shaking test, the axial strains and circumferential strains in
the key sections of CFST arch-rib model were measured by attached
strain gauges along axial direction of arch rib (Fig. 4). The axial strains
attached on the top edge, bottom edge, front and rear edges around the
steel tube in key sections were arranged, while circumferential strain
gauges were mutually perpendicular with axial strain gauges.
Table 3 just lists the peak strain responses at top and bottom edges
in the key sections, while their front and rear edges’ results does not
present due to the limit space. The sign “+” represents the compressive
strain, while the sign “-” represents the tensile strain.
From Table 3, it indicates that the strain response under seismic
excitations in the arch springing is usually larger than that of any other
sections. Besides, static strains caused by both of self-weight and man-
made mass are far less than those by high intensity of E2 wave, El-
centro wave and Hanshin wave, but a little larger than Taft wave.
As for Taft wave in Table 3, the Taft causes much less strain
responses than other waves, just similar as the acceleration and
displacement responses as mentioned above. In addition, it generates
much larger strains under bi-directional excitation than single uniform
excitation, but there are no much difference between the traveling
excitation and the single uniform excitation.
The E2 imposes far larger strain responses than E1 and Taft in
Table 3. As for X-directional excitation, it will generate larger strains on
top & bottom points than that on front & rear points in the section of
arch springings. However, it will cause far less strains on top & bottom
points than that on front & rear points as for Y-directional excitation.
Such as case 10 (E2 artificial wave and Y-directional excitation), the
strain of front edge reached as large as −1400 με (×10−6) in the R-S
220
F.-y. Huang et al.
Fig. 15. The magnification of PGA in key section along arch span.
section, very close to the ultimate elastic strain of steel tube, while the
top edge is only −320 με (×10−6).
In terms of rare earthquakes of El-centro wave and Hanshin wave,
they generate very large response strains in almost all key sections,
where their strains were generally greater than the limit elastic strains
of steel as well as core concrete especially under Y-directional excita-
tion. The biggest strain that listed in Table 3 is 3180 με (×10−6) at the
bottom edge of crown. Thus, it shows that the arch structural model has
been in elastic-plastic state. However, it does not decrease the whole
stiffness because of hooping effect on the core concrete subjected by the
outside steel tube. It does also not appear cracks on the surface of steel
tube of arch-rib model, which means that the model has a favorable
seismic performance.
Figs. 18 and 19 present the peak axial response strains of bottom
edges at key sections along arch span under excitation of all four kinds
of earthquake waves, and as limited by space, the results of top edge,
front and rear edge do not show. It can be seen from Fig. 18 that the
axial response strains under X-directional excitation are generally less
than the limit elastic strain of steel tube and the comparatively larger
response strains are at the quarters of 1 L/4 and 3 L/4. Hanshin wave
generates the maximum response strains, and both E2 wave and El-
centro wave generate nearly the same response strains, but Taft wave
generates a very small response strains relatively to E2 wave even
thought they have the same exciting PGA.
In addition, it has a much larger response strain under Y-directional
excitation than that under X-excitation, especially for El-centre whose
response strains are larger than the limit elastic strain of steel. Actually,
the response strains at front and rear edges are far larger than that at
top and bottom edges that illustrate in Fig. 19.
6. Discussions
The failure mode of reinforcement concrete arch-rib model under
seismic effect is that the concrete in the springing are damaged or even
spalled, and the reinforcement bars are buckled [21]. It can be found
Fig. 16. The relatively peak displacement at crown under uniform excitation.
221
Thin-Walled Structures 116 (2017) 212–224
several mainly cracks in the springs just as the columns of RC frame
[20], and it is non-ductile that the deformation is about 1.5% of the rise
of arch while it fails. Therefore, compared to the failure mode of
reinforcement concrete arch rib, it has significant difference to the CFST
arch-rib model. Because of the existence of the hooping effect on core
concrete subjected by the outside steel tube, the CFST arch-rib model
will not be failed even though the concrete is cracked. Furthermore, the
stiffness of the CFST arch rib almost does not reduce significantly even
the strains are very large and the deformation of Dr =25 mm is about
1.67% of rise of arch (Dr/H), which shows a favorable ductility.
The arch rib structure can transfer the seismic shear force in arch
springings to the axial resistance under the excitation of longitudinal
direction (along the arch span direction), rather than to hardly resist
the base shear force compared to the frame-column structure or girder
bridge (just as Fig. 20). That means it is advantageous for the model to
resist axial compressive force rather than shear force and the CFST arch
rib has a favorable performance of axial compression. However, it needs
to study more comprehensively by FEM how to transfer for the seismic
effect force from frame-column bases to the springings of arch-rib if
supposing arch-rib to be equivalent to frame-column. Due to the limit of
space, only experimental results are presented in this paper.
7. Conclusions
Based on the Shaking-tables test of the scaled CFST arch-rib model
with fully man-made mass, some conclusions are drawn as follows:
(1) The primarily shape modes of the CFST arch-rib model are the same
as its prototype bridge. They are in sequence of out-plane symme-
trically transverse buckling, in-plane anti-symmetrically longitudi-
nal buckling, out-plane anti-symmetrically transverse buckling as
well as in-plane symmetrically vertical buckling, respectively.
(2) The arch-rib's responses has a significant relevance to the predo-
minant frequency (or period) of the input waves. The closer of
predominant frequency to the structural natural frequency, the
F.-y. Huang et al. Thin-Walled Structures 116 (2017) 212–224

Fig. 17. Comparison of relatively peak displacement responses at crown under varying excitations.

Table 3
Peak strains in key cross-sections of arch rib model (με(×10−6)).

Earthquake input Sections

Left spring (L-S) 1 L/4 1 L/2 3 L/4 Right spring (r-S)

Top Bottom Top Bottom Top Bottom Top Bottom Top Bottom

X Taft wave Uniform － +72/−78 +71/−92 +82/−83 +39/−52 +46/−43 － +59/−78 +114/−80 +115/−71
Traveling － +76/−69 +78/−71 +78/−69 +65/−62 +74/−53 － +67/−61 +70/−66 +58/−51

Artificial wave E1 － +203/−112 +173/−238 +327/−388 +148/−163 +100/−115 － +208/−453 +167/−334 +196/−221
E2 － +212/−109 +425/−448 +458/−475 +147/−137 +146/−169 － +358/−394 +409/−425 703/−920
El-centro － +301/−421 +360/−411 +402/−408 +63/−205 +218/−120 － +424/−317 +288/−587 +173/−237
Hanshin － +456/−611 +519/−616 +612/−605 +104/−247 +264/−120 － +600/−538 +470/−551 +207/−284

Y Taft wave Uniform － +104/−154 +76/−134 +76/−159 +62/−121 +82/−135 － +98/−142 +123/−255 +141/−332
Traveling － +145/−160 +68/−137 +77/−142 +99/−103 +97/−125 － +100/−146 +111/−152 +130/−194

Artificial wave E1 － +68/−95 +30/−29 +26/−33 +29/−37 +59/−49 － +44/−83 +48/−82 +69/−97
E2 － +159/−221 +75/−164 +72/−231 +44/−245 +94/−254 － +77/−255 +84/−320 +113/−296
El-centro － +1057/−709 +2569/−44 +1815/−188 +976/−242 +3180/−11 － +352/−20 −252/−564 +53/−347
Hanshin － −43/−799 +116/−26 −26/−142 −82/−239 +202/−81 － +286/−62 +267/−736 −37/−212
X+Y Taft － +148/−188 +104/−113 +124/−120 +73/−56 +82/−61 － +152/−115 +117/−146 +111/−154
Static strains by weight −225 +245 +164 −144 +39 −19 － +150 +226 −215

Note：“－” indicates the strain gauge failed and nothing was recorded

Fig. 18. Peak axial response strains of bottom edges in key sections.

larger response it is. Therefore, the E2 artificial wave causes much
larger acceleration response, displacement response and strain
response than Taft wave even though they have the same PGA. It
also indicates that artificial waves and El-centro wave generate
greater magnification than that of Taft wave and Hanshin wave.
(3) The strains in the arch springing are larger than other sections
under seismic wave excitation and their front and rear edges in
springing section are relatively more adversely. The static strains
generated by self-weight and artificial mass are far less than those
by high intensity of E2 wave, El-centro wave and Hanshin wave, but
a little larger than Taft wave.
(4) For a CFST arch structure subjected to transverse seismic excitation,
the maximum acceleration response occurs at the arch crown, and
will be periodical faded with out-plane symmetric failure mode (the
first failure mode) after the termination of the excitation. However,
under the longitudinal excitation, the maximum acceleration
response occurs at the quarter-span and after the termination of
the excitation, the testing model attenuates periodically in the in-
plane asymmetric failure mode (the second failure mode).
(5) The non-uniform excitation has a more significant influence than

222
F.-y. Huang et al.
Fig. 19. Maximum response compressive strain at L-S under the Y-directional excitation.
Fig. 20. Comparisons of damage and failure modes.
that uniform one. Besides, the bi-directional excitation has much
larger effect than the unidirectional excitation. The earthquake
effect on the CFST arch bridge should be taken account of both bi-
directional excitation and traveling wave excitation.
(6) This CFST arch structural model performs a little of nonlinear and
plastic behavior under the rare earthquakes of El-centro wave and
Hanshin wave. However, the whole stiffness and natural frequen-
cies of this model do not have significant decrease and there is no
distinct crack or damage on the model's surface because of the
hooping effect on the core concrete subjected by the outside steel
tube. It reveals that CFST arch rib structure has a favorable seismic
performance and ductility, and it is almost still elastic even the
deformation is up to 1.67% of arch rise.
Acknowledgement
The paper is partial financial supported from the Fundings of NSFC
(51208111), (51578161)and (51278126), China.
Appendix A. Supporting information
Supplementary data associated with this article can be found in the
online version at doi:10.1016/j.tws.2017.03.020.
223
Thin-Walled Structures 116 (2017) 212–224
References
[1] R. Bjorhovde, Current steel structures research [J], Eng. J. 48 (2) (2011)
p159–p168.
[2] F.Y. Huang, X.M. Yu, B.C. Chen, et al., The structural performance of axially loaded
CFST columns under various loading conditions[J], Steel Compos. Struct. 13 (5)
(2012) 451–471.
[3] J.G. Wei, F.Y. Huang, B.C. Chen, Research on the influence of initial stress to
ultimate load carrying capacity of concrete filled steel tubular (single tube) arches
[J], Eng. Mech. 27 (7) (2010) p103–p112 (in Chinese).
[4] Y.L. Pi, M.A. Bradford, W.L. Qu, Long-term non-linear behaviour and buckling of
shallow concrete-filled steel tubular arches [J], Int. J. Non-Linear Mech. 46 (9)
(2011) p1155–p1166.
[5] Y.L. Pi, C.Y. Liu, M.A. Bradford, et al., In-plane strength of concrete-filled steel
tubular circular arches [J], J. Constr. Steel Res. 69 (1) (2012) p77–p94.
[6] W.M. Yan, Y. Li, Y.J. Chen, Seismic testing of a long-span concrete filled steel
tubular arch bridge [J], Key Eng. Mater. 45 (6) (2011) p89–p102.
[7] W.M. Yan, X. Li, Y.J. Chen, et al., Shaking table research on a CFST arch bridge
model: effect of site condition [J], Eng. Mech. 30 (6) (2013) p116–p123.
[8] D.J. Haudhary, Vishal C. Shelare, Seismic analysis of concrete filled steel tube
composite bow-string arch bridge [J], Adv. Bridge Eng. 19 (3) (2006) p241–p247.
[9] Q.X. Wu, M. Yoshimura, K. Takahashi, et al., Nonlinear seismic properties of the
second Saikai bridge: a concrete filled tubular (CFS) arch bridge[J], Eng. Struct. 28
(2) (2006) p163–p182.
[10] B. Zhang, S.C. Li, X.Y. Yang, et al., Analysis of seismic response of deck type long-
span CFST arch bridge [J], J. Highw. Transp. Res. Dev. 26 (3) (2009) (p64-67, 73.)
(in Chinese).
[11] Y. Li, W.M. Yan, Y.J. Chen, Seismic response analysis of an arch bridge with
irregular concrete filled steel tube [J]. Technology for Earthquake Disaster,
Prevention 5 (3) (2010) p346–p351.
F.-y. Huang et al.
[12] J. Ma, Y.J. Chen, L.P. Liu, et al., Nonlinear seismic response analysis of half through
CFST arch bridge under 3-D earthquake waves [J], Key Eng. Mater. 45 (6) (2011)
p67–p76.
[13] C.Y. Liu, Y.Y. Wang, Q.H. Wang, Limit out-of-plane deformation of circular CFST
arches under severe earthquakes [J], J. Harbin Inst. Technol., 44, Sup 1 (2012)
p317–p320 (in Chinese).
[14] D.Y. Zhang, X. Li, W.M. Yan, et al., Stochastic seismic analysis of a concrete-filled
steel tubular (CFST) arch bridge under tridirectional multiple excitations [J], Eng.
Struct. 52 (2013) p355–p371.
[15] K.M. Bi, H. Hao, W.X. Ren, Seismic response of a concrete filled steel tubular arch
bridge to spatially varying ground motions including local site effect [J], Adv.
Struct. Eng. 16 (10) (2013) p1799–p1817.
[16] F. Xiong, Seismic Behavior of Concrete Filled Steel Tubular Arch Bridges [D],
224
Thin-Walled Structures 116 (2017) 212–224
Dissertation for the doctoral degree 2001 Sichuan University, 2001(in Chinese).
[17] C.Y. Liu, Y.Y. Wang, W. Wang, et al., Seismic performance and collapse prevention
of concrete-filled thin-walled steel tubular arches [J], Thin-walled Struct. 80 (2014)
91–102.
[18] JTG/T B02-01, Guidelines for Seismic Design of Highway Bridge [C], People
Transportation Press, Beijing, 2008(in Chinese).
[19] F.Y. Huang, Z.M. Fang, J.Z. Li, Performance of earthquake simulation three bi-axial
shaking tables [J], Appl. Mech. Mater. 518 (2014) (2014) p178–p183.
[20] K. Elwood, J. Moehle, Drift capacity of reinforced concrete columns with light
transverse reinforcement [J], Earthq. Spectra, Earthq. Eng. Res. Inst. 21 (1) (2005)
p71–p89.
[21] B.C. Chen, Concrete-Filled Steel Tubular Arch Bridge [M], Beijing People
Transportation Press, 2007(in Chinese).