Simulation of the In Situ Spatially Varying Ground Motions and Nonlinear Seismic Response Analysis of the Cable-Stayed Bridge

To study the nonlinear seismic behavior and seismic resistance of the long-span cable-stayed bridges subjected to earthquakes, the multidimensional and multisupported artificial ground motions are synthesized first based on the in situ site conditions of the bridge considering the coherent and traveling wave effects. +en, considering the material nonlinearity of the cable-stayed bridge, a 3D finite element model is established based on the OpenSees platform, and the nonlinear seismic response analysis of the bridge is carried out under the synthetic artificial ground motions. +e nonlinear seismic response of main bridge components such as piers, towers, bearings, and cables is analyzed, and key conclusions and observations are drawn.


Introduction
Simulation of ground motions is always a critical issue in earthquake engineering. Using the uniform ground motions in seismic response analysis of large-span structures with large planar dimensions, such as the bridges, oil pipelines, and offshore platforms, is often not accurate. e dimension of the large-span structure cannot be ignored compared to the wavelength of the seismic waves. erefore, it is important to consider the spatial variability of ground motion in seismic analysis of such large-span structures [1]. When seismic analysis of large-span structures is carried out, different seismic excitations at each support should be considered to ensure the seismic response accuracy of longspan structures. Meanwhile, the multidimensional and nonstationary characteristics of ground motions should also be modelled. Unfortunately, due to the scarcity of measured spatial ground motions, engineering seismic analysis of large-span structure under natural ground motions has been greatly restricted. erefore, it is of prime significance for synthesis of spatially variable ground motions.
In terms of artificial ground motions simulation, worldwide scholars have made important contributions. Assuming that the phase angle is uniformly distributed within (0-2π), Hao et al. [2] used the Cholesky decomposition of the power spectral matrix for synthesis of multisupport artificial seismic waves. Ramadan and Novak [3] proposed and compared the computational efficiency of simulation methods in seismic analysis of single-span and multispan simply supported beams. Qu and Wang [4,5] used stochastic vibration theory to derive the relationship between amplitude, phase angle, and power spectrum matrix of spatially varying ground motion; however, multidimensionality of ground motions needs further research.
e new version of the program, RspMatch2005 proposed by Hancock et al. [6], enabled the accelerograms to be matched to the pseudoacceleration or displacement spectral ordinates as well as the spectrum of absolute acceleration, with matching to be performed simultaneously to a given spectrum at several damping ratios. An energybased envelope function which directly related to the Arias intensity of the ground motion was developed by Stafford et al. [7] for utilization in the stochastic simulation of earthquake ground motion. On the other hand, a fully nonstationary stochastic model employing filtering of a discretized white-noise process for strong earthquake ground motion was developed by Rezaeian and Kiureghian [8]; specially, nonstationarity was achieved by modulating the intensity and varying the filter properties in time. Bi and Hao [9] synthesized spatially varying ground motions based on one-dimensional wave propagation theory in soil layers. Konakli and Kiureghian [10] developed a simulation method by taking into account effects of coherence, traveling wave, and local site. Yamamoto and Baker [11] proposed a wavelet packet transform approach which can be used to characterize complex time-varying earthquake ground motions and illustrated the potential benefits of such an approach in a variety of earthquake engineering applications. Jia et al. [12,13] applied pseudoexcitation method (PEM) in seismic analysis of long-span structures under tridirectional spatially varying ground motions, and a mathematical scheme in simulating nonstationary tridirectional spatially correlated ground motions has been proposed.
e virtual real-time separation method proposed by Wu et al. [14] was used to decompose the power spectrum matrix, and the ground motion correlation characteristics generated by the decomposition method are compared with the traditional cholesky decomposition method. e results showed that the method was not only fast, accurate, and feasible but also had obvious advantages in many aspects. Furthermore, Dabaghi and Kiureghian [15] developed a parameterized stochastic model of nearfault ground motion in two orthogonal horizontal directions.
e major characteristics of recorded near-fault ground motions are represented to employ this model, which include near-fault effects of directivity and fling step; temporal and spectral nonstationarity; and intensity, duration, and frequency content characteristics.
In summary, a series of research results of artificial ground motion synthesis were given. Among them, the method proposed by Bi and Hao [9] fully considers the characteristics of the soil layer, and the traveling wave and coherence effect is widely used. erefore, considering the local site layer conditions of bridges, the theoretical method of Bi and Hao [9] is used in the subsequent artificial ground motion synthesis to complete this work in the paper.
In terms of the nonlinear structural analysis, most of the previous studies focus on the high-pier continuous bridges or high-pier steel bridges and the failure location was usually predefined. However, there were very limited research literature studies on nonlinear seismic response analysis of cable-stayed bridges. Considering both the uniform and multiple-support seismic excitations, Nazmy and Abdel-Ghaffar [16] developed novel nonlinear dynamic formulation, in which the tangent stiffness and iterative scheme were utilized to capture the nonlinear seismic response. In addition, Nazmy and Abdel-Ghaffar [17] studied the nonlinear dynamic performance of a 3-D long-span cable-stayed bridge under earthquake and revealed that the multiplesupport seismic excitations can have a significant effect on structural response. Ren and Obata [18] studied the plastic behavior of longspan cable-stayed bridge structures based on a 2-D finite element model. e results indicated that the deformation of the cable-stayed bridge structure caused by static load has a significant influence on its dynamic seismic response. Li et al. [19] derived the incremental dynamic equations of multidegrees-of-freedom structure under earthquake loading, and the finite element model of the long-span cablestayed bridge was established. Furthermore, the seismic performance of a horizontally curved highway bridge was examined by Vamvatsikos and Sigalas [20] using incremental dynamic analysis (IDA) in 3D, in which IDA has been successfully applied to two-dimensional structures with a single horizontal ground motion component. Kaviani et al. [21] focused on identifying trends in seismic behavior of reinforced concrete bridges with seat-type abutments under earthquake loading, especially with respect to the abutment skew angle. In the study by Huang et al. [22], two kinds of finite element models based on implicit and explicit integral are established, respectively. e seismic response characteristics and failure modes of the single tower cable-stayed bridge are tested. Li et al. [23] investigated the seismic vulnerability of a five-span cable-stayed bridge in the longitudinal direction. e geometry and material nonlinearity of the cable-stayed bridge structure were considered comprehensively, and a 3-D finite element model was established based on the OpenSees platform. Omrani et al. [24] proposed a probabilistic framework to achieve the sensitivity of the seismic performance of a typical reinforced concrete overpass bridge to variations in the parameters of its abutment model. Han et al. [25] studied the seismic behavior of a reinforced concrete single-tower cable-stayed bridge equipped with sliding friction bearings under 2-D seismic excitations. Based on the OpenSees platform, a three-dimensional numerical model was established, and the nonlinear seismic response and seismic performance of the single-tower cable-stayed bridge were systematically studied. Recently, Wang et al. [26,27] evaluated the potential benefits of using buckling-restrained braces (BRBs) to seismically rehabilitate straight bridges and assessed the collapse capacity and failure modes of skewed bridges retrofitted with buckling-restrained braces (BRBs) at the column bent.
e OpenSees software, led by the Pacific Earthquake Engineering Research Center, is the dominant one in the current structural nonlinear time history analysis. Compared with large commercial finite element software ANSYS, ABAQUS, etc., OpenSees can easily call a variety of complex nonlinear stress-strain materials in a wider range. At present, the software has been widely used in nonlinear seismic response analysis of long-span bridges. Since the last century, many scholars have done a lot of research on the constitutive relations of common confined concrete and proposed different constitutive relation models. Some constitutive relation models are based on theoretical derivation, other models are based on statistical analysis of numerical calculations, and few models are based on the semiempirical and semitheoretical results. At present, there are mainly popular models: Mander model [28]; Kent-Park model [29]; and Hoshikuma model [30]. 2 Shock and Vibration For the seismic response analysis of reinforced concrete structures, it is especially important to accurately simulate the hysteretic behavior of reinforced bars under earthquake loading. In view of this, scholars from various countries had proposed a variety of mechanical calculation models, such as Ramberg-Osgood steel constitutive model [31]; Richard steel constitutive model [32]; and modified Giuffre-Menegotto-Pinto steel hysteresis constitutive model [32]. ese models are widely used in nonlinear seismic response analysis.
Most of the study focused on the nonlinear seismic response of high-pier continuous girder bridges with predetermined failure locations. Generally, plastic hinges are assumed to achieve the shape deformation capability of the components and to obtain the nonlinear seismic response of the structural members. For the complex structure, such as long span cable-stayed bridges, it is rare to consider the material nonlinearity in the nonlinear seismic analysis considering the in situ site condition.
Following above discussion, the objective of this paper is to study the seismic behavior of long-span cable-stayed bridges under spatial earthquake ground motions. In Section 2, the multidimensional and multisupport artificial ground motions are synthesized first based on the in situ site conditions of the bridges considering coherent and traveling wave effects. en, considering the material nonlinearity of the cable-stayed bridge, a 3-D finite element model is established based on OpenSees in Section 3. e nonlinear seismic response analysis of the cable-stayed bridge is carried out, and the seismic response of main structural components such as piers, towers, bearings, and cables is analyzed in Section 4. e major conclusions and findings are drawn in Section 5.

Simulation of In Situ Spatially Varying Ground Motions
where ∇ 2 e and ∇ 2 Ω { } are the Laplace operator of the volumetric strain amplitude and the rotational strain vector, respectively, and c p and c s are the P wave and S wave propagation velocity, respectively. e stiffness matrices [S SH ] and [S P−SV ] can be expressed as where μ SH and P SH are the out-of-plane displacements and load vector and μ P−SV and P P−SV are the in-plane displacements and load vector.

e soil layer transfer function H(ω) is
where p � t R G * R /t L G * L .
In the linear elastic range, the auto-power spectrum of the soil layer at any position and the cross-power spectrum between any two points can be expressed by the following formula: where H j (iω) and H k (iω) are the site transfer functions at supports j and k, respectively. S g (ω) is the ground motion power spectral density at the base rock; c j′k′ (d j′k′ , iω) is the coherency loss function of spatial ground motions at the base rock,where the bedrock power spectrum values are based on the modified Tajimi-Kanai power spectrum model; the formula is as follows: where: |H P | is a high-pass filter function. S 0 (ω) is the Tajimi-Kanai power spectral density. ω g and ξ g are the central frequency and damping ratio of the Tajimi-Kanai power spectral density function. ω f and ξ f are the central frequency and damping ratio of the high-pass filter.

Multidimensional Multisupport Power Spectrum Matrix.
e acceleration power spectrum matrix of m ground support point excitations is expressed as [12] S(iω) � S 11 (iω) S 12 (iω) S 13 (iω) · · · S 1m (iω) Here, where, S kk (iω) and S ll (iω) are the auto-power spectral density function at supports k and l, respectively, S kl (iω) is the cross power spectral density function, and p kl (iω) is the coherent function and can be expressed as

Shock and Vibration
where |p kl (iω)| is the modulus for coherent function; ωd L kl /υ app is the angle of p kl (iω); and υ app is the apparent wave velocity.
Considering the multidimensional characteristic of ground motions, the power spectral density matrix of each points can be expressed as [12] S kl (iω) � rough equations (9)-(11), the matrix of the crosspower spectral density function is dimensionally extended.

Multidimensional and Multipoint Nonstationary Ground Motion Synthesis.
e (3 m × 3m)-dimensional matrix is decomposed into the product of the lower triangular matrix and the upper triangular matrix: where and [S € x ] can be expressed as R is the lagged coherence matrix given by where N � 3 m and R is a definite or semidefinite symmetric matrix. e rank of R is greater than 1, and it can be decomposed as the summation of nonzero eigenvalues α j and the corresponding normalized eigenvectors φ j (j � 1, 2, . . . , r, r ≤ 3 m), which is given by where where a mb and θ mb indicates the magnitude and phase angle of the kth frequency component at points m and b. φ bk is the random phase angle which follows the uniform distribution over the interval (0, 2π). l mn (iω) represents the element of the P matrix in any direction. I and Re represent the imaginary part and the real part, respectively. e nonstationary ground motion is mainly characterized by nonstationary amplitude and frequency. Specially, the nonstationary amplitude is simulated by a smooth process ground motion multiplied by a deterministic intensity envelope function, so nonstationary ground motion in any direction at any point can be expressed as where u j (t) represents the stationary process of the j time period with different frequency components in different time periods; d j (t) represents a rectangular function with a value of 1 in the j time period and 0 in the other two time periods; and f(t) represents an intensity envelope function and describes the nonstationarity of the amplitude, which can be expressed as In this paper, t 0 � 5 s and t n � 13 s. Because the actual vibration is a process varying from weak to strong, then from strong to weak, the trapezoidal window function in equation (20), in which T1, T2, and T3 are the preshock period, strong earthquake period, and the post-weak period, respectively. Taking 1/10 of the strong earthquake period as the transition time as shown in Figure 1 (in this example, the values of T1, T2, and T3 are 5 s, 8 s, and 7 s, respectively).

Multidimensional and Multisupport In Situ Ground Motion Synthesis.
e basic theory about multidimensional and multisupport nonstationary ground motion synthesis is given in Sections 2.1-2.3. Figure 2 shows the detailed soil parameters at the local site of the cable-stayed bridge, in which 1 and 1°, respectively, indicate the bottom position of the left tower and the position corresponding to the bedrock and 2 and 2°represent the bottom position of the right tower and the position corresponding to the bedrock. Among the soil layer parameters, h indicates the thickness of the soil layer, G indicates the shear modulus of the soil layer, ρ indicates the corresponding soil layer density, and ξ and ] are the damping ratio and Poisson's ratio, respectively.
In this paper, the peak ground acceleration of the bedrock is 2.94 m/s 2 (0.3 g), the corresponding bedrock acceleration (white noise) self-spectral density Γ is 0.0115 m 2 /s 4 , and the ground motion duration T is 40.96 s. After the parameters of the soil layer at the bridge site are determined, according to the multidimensional and multisupport nonstationary ground motion synthesis theory and considering the randomness of phase angle, three longitudinal, horizontal, and vertical artificial ground motions are synthesized, as shown in Figures 3-5, respectively.
It can be seen from Figure 3 that the peak acceleration ground motions at 1# site can reach 5.21 m/s 2 and that at 2# site can even reach 7.98 m/s 2 , which can be seen on the surface of the soil layer. e peak acceleration at ground can reach 0.81 g which is much larger than that at the bedrock, indicating the amplification effect of ground soil being evident.

Nonlinear Finite Element Simulation of the Cable-Stayed Bridge
A large railway cable-stayed bridge is located at the Wenzhou, China, with a span combination of (51 + 91 + 300 + 91 + 51) m. e ratio of the side span to middle span is 0.473, and the total length of the bridge is 584 m. e girder is a prestressed concrete box beam with a single box and two rooms. e tower above the bridge deck has an inverted Y shape, and the column below the bridge deck has a diamond shape. e height of the tower is 118 m, with the height of the bridge above the deck being 75 m. e longitudinal width of the tower is extended from 6.5 m at the top of the tower to 8 m at the bottom of the tower. e cable has the standard galvanized parallel wire having a tensile strength of 1860 MPa. e layout of main bridge is shown in Figure 6. e reinforcement arrangement of the pier is as follows: the main reinforcement bar is HRB400 steel with the nominal diameter of 16 mm and the basic spacing of 150 mm; the pier stirrup is also HRB335 steel with the nominal diameter of 12 mm, the spacing of the hoop of 100 mm in the 6 m range at bottom of the pier and the rest being 150 mm; and the concrete cover depth of the pier is 45 mm. To the towers, the main reinforced bar adopts the HRB400 steel bar with a diameter of the steel bar of 32 mm and the basic distance of 200 mm. e stirrup of the tower uses HRB335, with the nominal diameter of the steel bar of 20 mm, the spacing of the stirrup at the bottom of 100 mm, and the rest of 125 mm. e concrete cover depth of the tower is 45 mm, which is the same as that of the pier.
In order to obtain the nonlinear seismic response of the cable-stayed bridge, a 3-D nonlinear finite element model of cable-stayed bridge is established based on OpenSees platform.

Pier and Tower.
As the main vulnerable members of the cable-stayed bridge under earthquake, the plastic hinge at the bottom section of the piers and towers is relatively easy to be developed, which are simulated by the nonlinear beamcolumn element. e fiber section of the piers and towers can be divided into three parts, including the confined No.5 silty clay, h = 9.5m, G = 25MPa, igure 2: Parameters of each soil layer at the bridge site. concrete, unconfined concrete, and reinforced bar section, with different materials.

Concrete Material
Model. e Concrete02 material model is based on the uniaxial constitutive relationship of plain concrete defined by Kent-Scott-Park [28]; the model changes the peak strain value of the concrete compression skeleton curve. e concrete tension capacity takes into account the tensile hardening of the material and the stiffness degradation effect after the initial cracking with the increase of the maximum tensile strain. e hysteresis curve of the material is determined according to the hysteresis rule defined by Karsan-Jirsa [34]. e stress-strain relationship and hysteresis curve of the concrete material model are shown in Figure 7.
In the Concrete02 material model, the ascending section is described by a quadratic polynomial, the descending section is described by a straight line, and finally a residual stress is retained. e model can be expressed as   Figure 6: Layout of the cable-stayed bridge (unit/m).

Shock and Vibration 7
where ξ 0 � 0.002K, where K represents the amplification factor for the hoop effect, f c ′ represents the maximum compressive stress of the concrete,ξ 0 is the strain corresponding to the maximum compressive stress of the concrete, and Z is the strain softening slope, which ξ u is the ultimate strain of the concrete. ξ u is the volume ratio of the stirrup, and f yv is the yield strength of the stirrup, h ′ is the height of the concrete core, and s h is the pitch of the stirrup. e ascending and descending sections of the concrete material are straight, and the maximum tensile stress f t ′ and softening stiffness E ts of the concrete are

Reinforcement Material Model.
e reinforcement material is defined by the Steel02 material in the OpenSees program. e Steel02 material uses the modified Giuffre-Menegotto-Pinto steel hysteresis constitutive model [31,32], which can consider two-way Bauschinger effect and the isotropic strengthening effect in the structures. It is widely used in nonlinear simulation of materials, and the model can be expressed as where where b is the strain hardening coefficient, R is the parameter considering the Bauschinger effect, σ 0 and ξ 0 are the stress and strain corresponding to the intersection of the asymptote and the postyield asymptote at the initial time; and σ r and ξ r are the stress and strain at the loading point of the reaction. e stress-strain relationship and hysteresis curve of the reinforced bar material model are shown in Figure 7.

Bearing Model.
e bearing is a critical component to connect the superstructure and the substructure. Under normal conditions, the constant load of the superstructure is effectively transmitted to the substructure. Under earthquake loading, the bearing transmits the inertial force of the substructure. As the first defense line for earthquakes, the bearing plays an essential role on the safety of the bridges. erefore, it is very important to accurately simulate the mechanical characteristics of the bearing in the seismic response analysis of the bridge structure. e GPZ series of pot-type rubber bearings are used in this bridge structure, which has a capacity ranging from 1000 kN to 50000 kN. GPZ series pot-type rubber bearings have low friction coefficient, large bearing capacity, and strong corrosion resistance, which can not only greatly improve the capacity of   Shock and Vibration the bearing and the life of the rubber but also ensure that the bearing has flexible rotation performance and good cushioning performance. e movable pot-type rubber bearing can be simulated by a bilinear ideal elastic spring element, and its resilience model is shown in Figure 8. Critical sliding friction force F max (KN) of movable pot rubber bearing can be expressed as e initial stiffness is where μ d is the sliding friction coefficient, generally taken as 0.02; R is the upper structure gravity (kN) assumed on the support; and x y is the movable pot-type bearing yield displacement, generally taken as 0.002-0.005 m. e hardening material model in the OpenSees is used to simulate the bilinear ideal elastic shape. e stress-strain relationship curve of the hardening material is presented in Figure 9.

Girder.
Under the earthquake loading, the girder is normally in the elastic state and is simulated by elastic beam column elements. e mechanical parameters of the elastic beam column element mainly include the sectional area, the moment of inertia, and the elastic modulus. ese parameters are calculated according to the material properties and the section size of the structure. And the second-stage load such as bridge decking is considered when simulating the superstructure.

Cables.
In this section, two material models are used for the simulation of the mechanical properties of the cables.    e initial strain material model is used to model the initial strain of the cables. e MinMax material model is used to achieve the stress threshold of the cable, after which the cable element will automatically be withdrawn from the work state.
According to the aforementioned detailed descriptions of the elements and materials for piers, towers, bearings, girders, and cables, the materials and elements used in the simulation are listed in Table 1.
e values of the main parameters used in the finite elements are presented in Table 2.
It should be noted that (1) the influence of pile-soil interaction on its seismic response is not considered in the model, and the bottom positions of the towers and piers are assumed being fixed; (2) the seismic response analysis mainly considered the nonlinearity of the structural material, while the geometric nonlinearity of the cable-stayed bridge is not considered for brevity.

Dynamics Verification of the OpenSees Model.
In order to verify the correctness and rationality of the OpenSees dynamic analysis model, the finite element models of the cablestayed bridge was also established in ANSYS and MIDAS CIVIL platforms. Table 3     element models. It can be seen that the free vibration periods of the cable-stayed bridge obtained by OpenSees software are close to that derived from ANSYS and MIDAS CIVIL. e maximum relative error is less than 10%, which indicates that the finite model based on the established OpenSees platform is reasonable.

Nonlinear Seismic Response Analysis
e nonlinear seismic response analyses are conducted in this section based on the finite element model of the established cable stayed bridge. e seismic responses of the bearings, piers, towers, and cables are given below under three ground motions, and the response results are analyzed, respectively.

Bearing Response.
e longitudinal and transverse displacement responses of each bearing under 1# ground motion are presented in Figures 10(a) and 10(b), respectively. It should be noted that the bearing number starts from 1# at the left pier and is arranged to the right in turn and double bearings are set at the tower section. It can be seen from Figure 10 that the displacement of each bearing on the same tower is almost the same in longitudinal and transverse 5   directions, so only one of double bearings is represented in the next study. Specifically, it can be seen from Figure 10(a) that the longitudinal displacement responses at each bearing is basically consistent, with the peak response being 0.112 m, because the girder is longitudinally drifted under the ground motions and the displacement of each bearing is displaced by the girder and the substructure. As seen from Figure 10(b), it can be concluded that the transverse displacement response of 8# bearing is the largest, with the response value of 0.230 m, and response at 1# bearing being the second largest with the value of 0.123 m. e longitudinal and transverse displacement responses of each bearing under 2# and 3# ground motions are presented in Figures 11 and 12, respectively, from which it can be seen that the longitudinal displacement responses of each bearing are basically consistent and the transverse displacement responses at the 3# and 5# bearings are smaller than those of the others, as a result of the transverse bending of the girder.

Pier Response.
For brevity, the hysteresis curve of the bottom section of piers under 1# ground motion is presented in Figure 13.
It can be seen from Figures 13-16 that the longitudinal and transverse hysteresis curves of the bottom section of all  Due to the asymmetry of the ground motion input at the left and right piers and towers, the seismic demand of the right pier is presented herein. e bending curvature at 3# pier bottom is 2.01 × 10 − 4 m − 1 which is the largest in the longitudinal direction, with the corresponding bending moment of 1.52 × 10 8 N·m. e bending curvature in the transverse direction of the bridge is 7.52 × 10 − 5 m − 1 which is the largest, with the corresponding bending moment being 1.58 × 10 8 N·m. Similarly, the variation curve of the hysteresis curve at the bottom section of 4# pier is almost the same as that of the 3# pier.
It can be seen from Figures 13-16 that, under 1# ground motion, the energy and residual deformation of the 3# and 4# piers are very small, indicating that the right piers are basically in elastic working state, while the energy dissipation Shock and Vibration 13 of the left pier (1# and 2# piers) is obviously larger than that of the right side piers, which is mainly caused by the asymmetry of the ground motion input. In conclusion, there is no significant energy dissipation in each pier member, indicating that the piers are basically in elastic state with minor plastic deformation. Figures 17-19 show the time history curves of the longitudinal and transverse curvatures at the bottom section of each pier under the ground motions. e following observations can be made: (1) Under the excitation of 3-D ground motions, the curvature response at 1# and 2# piers (left side) is larger than that of 3# and 4# piers (right side) in both longitudinal and transverse directions. is is due to the asymmetry of ground motion inputs at the left and right sides, which takes the effects of the site, coherence, and traveling wave into account. (2) Longitudinal curvature response value is larger than that in the lateral direction in terms of the curvature of the bottom section of the same pier.
is is mainly due to the cross-sectional shape of the piers and the vertical input considered.    22(b) that, under 1# ground motion, the curvature response in the longitudinal direction at the bottom section of right tower is larger than that of the left tower. In the transverse direction, the law is exactly opposite. e curvature response at the bottom section of the left tower is larger than that of the right tower. e same conclusion can be observed in Figures 23 and 24. is is due to the asymmetrical ground motions input and the considered effects of the local site, coherence, and traveling wave.

Cable Response.
As a critical component of the cablestayed bridge, it is necessary to study the axial force response of the cable under earthquake loading. Figure 25 shows the axial force time history of each cable under the 1# ground motion without considering the initial axial force. It can be seen that the axial force time-history curve of each cable is consistent. Furthermore, Figure 26 shows the maximum axial force of each cable under the ground motions considering the initial axial force of the cable. e maximum value is 694980 N, which occurs under the 2# earthquake. e corresponding cable area is 0.00858 m 2 , and the maximum stress of pull-down cable is 81 MPa, which is much smaller than the ultimate stress value indicating that the cable does not break.

Conclusion
To study the nonlinear seismic behavior and seismic resistance of the long-span cable-stayed bridges subjected to ground motions, the multidimensional and multisupported artificial ground motions are synthesized first based on in situ condition of the cable-stayed bridge considering the coherent and traveling wave effects. en, considering material nonlinearity, a three-dimensional finite element model was  (1) e longitudinal displacement responses at each bearing are basically consistent with each other because the girder is longitudinally drifted under the ground motions. e transverse displacement responses at the 3# and 5# bearings are smaller than that of others', which is due to the transverse bending of the girder. It can be seen that the maximum value of the displacement response of the bearings under three ground motions is greater than 0.10 m whether in the longitudinal or transverse directions, which is much larger than the critical value of the bearings sliding displacement equal to 0.005 m. erefore, the seismic performance of the bearings shows nonlinear characteristics, which means the linear elastic simulation of bearings cannot truly reflect the mechanical properties and nonlinear modelling is necessary.
(2) Under 3D ground motions action, the curvature seismic response of left side piers (1# and 2#) is larger than that of right side piers (3# and 4#), both in the longitudinal and transverse directions. is is due to the asymmetry of ground motion inputs at the left and right sides, which accounted for effects of the local site, coherence, and traveling wave. To the same pier, the longitudinal curvature response is larger than that in the transverse direction at the bottom section. is is mainly due to the cross-sectional shape of the piers  and the vertical input considered. It can be seen from the pier hysteresis curves that each pier shows different degrees of energy consumption, which means the piers enter nonlinear working stage. Generally speaking, the curvature response and energy consumption of the pier shows that the left side is larger than that of the right side and the longitudinal direction is greater than the transverse. (3) e curvature response in the longitudinal direction at the bottom section of the right tower is larger than that of the left tower. In the transverse direction, the law is opposite. e curvature response at the bottom section of the left tower is larger than that of the right tower. is is due to the asymmetrical ground motion input which takes into account the effects of the site, coherence, and traveling wave. It can be seen from the hysteresis curve of the bottom section of towers that the residual deformation of the towers after the earthquake is small and the energy consumption of the left and right towers in transverse direction is better than that in the longitudinal direction, which depend on the sectional characteristics of the left and right towers. (4) e maximum axial force of all the cables under the ground motions considering the initial axial force is 694980 N, which occurs under the 2# earthquake with the corresponding cable area 0.00858 m 2 , so maximum tensile stress of the cable is 81 MPa, which is very much smaller than the ultimate stress value indicating that the cable does not break. erefore, for the normal seismic analysis of cable-stayed bridges, the cables are not seismic vulnerable components. Furthermore, in general seismic response studies, the cables can be directly modelled with ideal elastic materials without considering their failure modes, which can improve the calculation efficiency while ensuring the calculation accuracy.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.