APP下载

Generalized response displacement methods for seismic analysis of underground structures with complex cross section

2023-11-13XuZigangDingLinlingDuXiuliXuChengshunandZhuangHaiyang

Xu Zigang, Ding Linling, Du Xiuli, Xu Chengshun and Zhuang Haiyang

1. Jiangxi Key Laboratory of Infrastructure Safety Control in Geotechnical Engineering, East China Jiaotong University,Nanchang 330013, China

2. Key Laboratory of Urban Security and Disaster Engineering of the Ministry of Education, Beijing University of Technology,Beijing 100124, China

Abstract: The response displacement method (RDM) is recommended for the seismic analysis of underground structures in the transverse direction for many codes, including bases for design of structures-seismic actions for designing geotechnical works (ISO 23469) and code for seismic design of urban rail transit structures (GB 50909-2014). However, there are some obvious limitations in the application of RDM. Springs and the shear stress of the soil could be approximately evaluated for the structures having a simple cross section, such as rectangular and circular structures. It is necessary to propose simplified seismic analysis methods for structures with complex cross sections. This paper refers to the idea of RDM and proposes three generalized response displacement methods (GRDM). In GRDM1, a part of the soil surrounding a structure is selected to generate a generalized underground structure with a rectangular cross section, and the same analysis model as RDM is applied to analyze the responses of the structure. In GRDM2, a hollow soil model without a generalized structure is used to compute the equivalent load caused by the relative displacement of the soil, and the soil-structure interaction model is applied to calculate the responses of the structure. In GRDM3, a continuous soil model is applied to compute the equivalent load caused by the relative displacement and shear stress of the soil, and the soil-structure interaction model is applied to analyze the responses of the structure, which is the same as the model used in GRDM2. The time-history analysis method (THAM)is used to evaluate the accuracy of the proposed simplified methods. Results show that the error of GRDM1 is about 20%,while the error is only 5% for GRDM2 and GRDM3. Among the three proposed methods, GRDM3 has obvious advantages regarding calculation efficiency and accuracy. Therefore, it is recommended to use GRDM3 for the seismic response analysis of underground structures that have conventional simple or complex cross sections.

Keywords: underground structures; seismic analysis; response displacement method; equivalent seismic load; complex cross section

1 Introduction

For the past few years, urban underground spaces and resources are being continuously developed and utilized with the rapid development of urbanization (Broere,2016). With regard to urban rail transit, an increasing number of subway structures are being constructed to solve the problem of urban traffic congestion. The subway is one of the important future trends in the development of large cities.

The traditional idea was that the seismic performance of underground structures was better than that of aboveground structures. However, there has been some earthquake damage observed in underground structures,and some structures have even collapsed under strong seismic motions (Dowding and Rozan, 1978; Iidaet al., 1996; Wang and Litehiser, 1985; Sharma and Judd,1991). After the Hanshin earthquake caused the collapse of the Daikai station in 1995, the topic of earthquake resistance of underground structures began to attract the widespread attention of scholars and engineers. A series of studies has made progress in the research on seismic analysis of underground structures (Abuhajaret al., 2015; Chenet al., 2013; Yuet al., 2018; Yuanet al.,2018). Regarding the seismic model test of underground structures, there have been three main test methods,including the 1-g ordinary shaking table test (Guanet al.,2019; Taoet al., 2019), theN-g centrifuge shaking table test (Chen and Shen, 2014; Cilingir and Madabhushi,2011a and b) and the pseudo-static pushover test (Namet al., 2006). The research on seismic model tests of underground structures has focused on the first two test technologies and has thereby accumulated a wealth of data about seismic responses of underground structures.

Based on experimental and numerical studies regarding underground structures, some theories and methods have been put forward the case for seismic analysis, including analytic methods and time-history analysis methods (THAM). The latter methods were more rigorous, by considering the complexity of waveabsorbing boundaries and wave-motion input (Choiet al.,2002; Dinget al., 2006; Huo and Bobet, 2003; Liu and Li, 2005). In recent years a series of simplified analysis methods also have been developed. The simplified analysis methods approximately solve the dynamics problems by means of static theory, which can reflect the main dynamic features of underground structures under earthquake loads. Similar to an aboveground structure,the seismic coefficient method was also used to analyze the seismic response of underground structures in an early stage. Newmark (1968) pointed out that the responses of underground structures depended on the deformation of a given site and developed the free-field racking deformation method (Hashashet al., 2001; Wang, 1993).This method applied the deformation of the free-field on a structure to analyze its dynamic response. To consider the difference in stiffness between a structure and its surrounding soil, Penzien (2000) presented the flexible coefficient method by introducing a flexible coefficient.As shown in many experimental and theoretical studies(e.g., Bhallaet al., 2005; Ramazi and Jigheh, 2006),the responses of a structure were consistent with the surrounding soil, including displacement, velocity and acceleration. Since then, some scholars have proposed using the response displacement method (RDM) and introduced foundational springs to reflect soil-structure interaction (Hamadaet al., 1984a and 1984b; Leeet al., 2016; Tateishi, 2005). Xuet al. (2019) proposed the vertical inertia force-response displacement method and integral vertical inertia force-response displacement method, in which the vertical inertia force of overburdened soil was taken into account. Both the Chinese standard (GB 50909-2014, 2014) and the international standard (ISO 23469, 2015) recommend the use of the response displacement menthod (RDM)for seismic design. However, the calculation accuracy of the spring was very limited. A soil-structure model was proposed in response to the acceleration method(Liu and Shi, 2006; Tateishi, 2005). Liuet al. (2014)proposed utilizing the pushover analysis method and analyzed the nonlinear seismic response of underground structures. Zouet al. (2017) proposed a more reasonable load distribution and modified the pushover method.The validity of this method was verified by experimental comparison.

The RDM is the most commonly used simplified analysis method at present. This method establishes the structure-foundation springs model for the transverse seismic analysis of underground structures, and the equivalent seismic load consists of the inertial force of the structure, the shear stress of the soil, and a relative horizontal displacement of the soil. Compared with other simplified methods, the RDM is theoretically rigorous,taking into account the inertial force of the structure and the effect of soil-structure interaction. However,concentrated springs cannot fully reflect the restraint effect of soil on underground structures under actual conditions, according to the response displacement method. In addition, this particular method cannot provide the soil shear force around a structure with a complex cross section.

This study proposes a series of simplified equivalent static methods for the seismic analysis of underground structures with complex cross sections. First, this study introduces in detail the analysis model of the traditional response displacement method, and notes the limitations of this simplified method, which cannot effectively solve for the foundation springs and soil shear forces around the underground structure with complex cross sections.Secondly, a series of simplified analysis methods are proposed by selecting a part of the soil around the complex section structure as the new generalized underground structure. The proposed methods include one model that introduces foundation springs and two models that establish soil-structure integral systems.Finally, the dynamic time-history analysis method is applied to a subway station structure with a complex cross section in order to evaluate the accuracy of each simplified analysis method. The results verify the feasibility and effectiveness of the proposed methods.

2 Response displacement method (RDM)

2.1 Analysis model of RDM

The theoretical basis of RDM is the response characteristics pertaining to how an underground structure moves, together with the surrounding soil when it is subjected to an earthquake load. The analysis model of RDM consists of the underground structure and the foundation springs. The springs are introduced to consider the difference in the stiffness between the structure and the surrounding soil. The analysis model of RDM is shown in Fig. 1, which is recommended by the ISO 23469 (2015) and GB 50909-2014 (2014). The foundation springs are distributed around the structure,including the top slab, bottom slab, left wall, and right wall. The equivalent seismic load mainly consists of three parts, including the relative displacement of the soil (Load-I), the shear stress (Load-II) and the inertia force of the structure (Load-III).

When the relative displacement of the soil at the position of the top and bottom slabs of a structure reaches the maximum value, it can be regarded as the most unfavorable time. Therefore, the equivalent seismic loads in the simplified method are the displacement and acceleration distributions of the free-field at this special time: then Load-I and Load-III can be obtained. If the soil shear forces for the top and bottom slabs areτTandτBrespectively, the soil shear stress at the lateral wall can be calculated as follows:

in which,τLandτRare the soil shear stress for the left and right walls.

2.2 Limitations of RDM

With the rapid development of underground space technology, the cross section of an underground structure has become increasingly diversified. In addition to the commonly found rectangular and circular cross sections,there also are many other cross sections in subway stations and tunnels, including the horseshoe shape, the arch shape, and the composite shape. The RDM specified in the current seismic design codes for rail transit structures in China is only applicable to the seismic design of underground structures that have simple(rectangular or circular) sections. For underground structures with complex cross sections, the traditional RDM method cannot be directly applied for seismic response analysis. There are two main reasons for this, as shown in Fig. 2. First, it is more difficult to calculate the coefficients of the surrounding springs around structures that have complex sections. Even if the static finite element method recommended by the code is adopted,the spring stiffness coefficient cannot be accurately obtained. Second, for underground structures with rectangular cross sections, the stress of the soil around the structure is in a pure shear state. For underground structures with non-rectangular cross sections, the stress of the soil around the structures includes not only the shear stress of the soil but also the normal stress of the soil. As for underground structures with complex cross sections, the stress of the soil around the structure is in a complex stress state. When using conventional methods to analyze the soil-structure interface, as shown in Fig. 2,the shear stress at the interface with an inclination will be decomposed and composed multiple times.

3 Generalized response displacement method(GRDM)

3.1 Method 1

The GRDM1 is derived by the use of the substructure method. The soil-structure interaction model is shown in Fig. 3, and the soil-structure system is decomposed into two substructures. One is the generalized structure subsystem; the other is the generalized soil subsystem.S,F, and,Irepresent the generalized structure, generalized soil, and the interface nodes between them. If the effect of damping is ignored, the dynamic equation of the generalized structural subsystem as a substructure can be expressed as:in whichMandKare the mass matrix and the stiffness matrix, respectively;KIFis the stiffness matrix that considers the interaction between the generalized soil and the structure subsystem;uis the displacement vector;uIFis the displacement vector of the free-field; andqIFis the load at the interface between the generalized soil and the structure′s subsystem.

Fig. 1 Analysis model of RDM

Fig. 2 RDM for the structure with a complex cross section

The equilibrium equation of the generalized underground structure is obtained through Eq. (2), which can be expressed as:

in which the left side represents the response of the generalized structure. The first term on the right side in Eq. (4) is the load at the interface between the generalized soil and the structure′s subsystem, which consists of two parts. One is the equivalent load caused by the relative displacement of the soil (Load-I); the other is the force generated by the free-field at the interface (Load-II). The second term on the right sides of Eq. (4) and Eq. (3) are the inertia forces of the generalized structure (Load-III).

The GRDM1 can be used to solve the seismic response of an underground structure with a rectangular or non-rectangular cross section, and the foundation springs must be arranged around the generalized structure subsystem. Equivalent seismic loads must be applied to the generalized structure at the same time,including the equivalent loads caused by the relative displacement of the soil, the shear stress of the soil, and the inertial force of the generalized structure. It should be noted that the inertial force includes not only the inertia force of the structure but also the inertia force of the selected soil. The analysis model of GRDM1 is shown in Fig. 4. Figure 5 pictures the calculation models of the spring coefficients arranged around the generalized structure by the use of the static finite element method.Xuet al. (2019) applied the same method in solving the coefficients of the foundation springs and proved its validity. In the static finite element method, a tangent and normal uniform load (q) is applied to the soil corresponding to the lateral wall or top and bottom slab,respectively, and the deformation (δ) of the soil will be recorded. For convenience, the average deformation(δ) of the corresponding nodes is used to calculate the coefficient of the subgrade reaction, as follows:

Fig. 3 Diagram of substructure analysis method

Fig. 4 Analysis model of GRDM1

Fig. 5 Calculation model of springs coefficients

It is assumed that the coefficient for the springs at the same interface in the same direction is considered as a constant value. The average deformation of the nodes on the same side is regarded as the deformation under the corresponding uniform load.

Figure 6 shows the implementation steps of GRDM1.Those steps are similar to those of traditional RDM, but there are some differences. The specific implementation steps are as follows:

I. Select the appropriate region of the generalized underground structure. According to the actual size of the cross section of the underground structure, a part of the surrounding soil is selected to generate a generalized structure with a rectangular cross section. It should be pointed out that in this method the selected region of the generalized structure will affect calculation results, an aspect not considered in this study.

Ⅱ. Solve the horizontal seismic response of the free-field. Equivalent linearization programs such as SHAKE91 or EERA are applied to analyze the seismic response of the free-field under a horizontal earthquake load. The results consist of the relative deformation and absolute acceleration of the soil at the mentioned unfavorable time.

Ⅲ. Calculate the spring coefficients and Load-I. The equivalent elastic modulus of the soil are adopted as the input parameters to calculate the coefficients of the foundation springs, as shown in Fig. 6. Then, Load-I can be converted by using spring coefficients and relative displacement, according to Eq. (4).

Ⅳ. Calculate Load-II. Step Ⅱ illustrates the calculation method of the shear stress at the top and bottom of the generalized structure, as well as the shear stress to the right or left of the generalized structure, as calculated by Eq. (1).

Fig. 6 Flowchart of GRDM1

V. Calculate Load-III. As previously mentioned,the inertial force of the generalized structure includes not only the inertia force of the structure but also the inertia force of the selected soil. To obtain more accurate results, the inertial force of the soil or structure is equal to the mass of the soil or structure by the acceleration at the corresponding depth of the free-field, which is obtained from a seismic analysis of the 1D free-field done in Step Ⅱ.

Ⅵ. Establish the analysis model of GRDM1.According to the analysis model of GRDM1 shown in Fig. 5, the springs and equivalent seismic loads calculated from steps Ⅲ to V are applied to the generalized structure for static analysis.

3.2 Method 2

It can be seen from the above description that a series of assumptions are usually required in GRDM1,which uses concentrated foundation springs to simulate the soil around the generalized structure. However,discrete foundation springs are not related to one another and cannot truly reflect the interaction between the soil and generalized structure. This will cause the constraints of the generalized structure to be inconsistent with an actual situation. Moreover, there will be errors in the load distribution of the soil-generalized structure interface. The discrete foundation springs cannot form effective constraints, especially in the corners of the generalized structure, which may lead to an underestimation of the internal force response of the corners. GRDM1 uses foundation springs to simulate soil-structure interactions. It is found in the actual application process that the coefficients of the foundation springs have a significant impact on the calculation of the responses of the underground structure. The static finite element method shown in Fig. 5 is an approximate solution for the coefficients of the foundation springs.GRDM1 uses the static finite element method to obtain the coefficients of the foundation springs in different positions and directions. As previously mentioned, the spring coefficients are different around the generalized structure. In all, six models and six analyses should be carried out to calculate all spring coefficients.

The GRDM2 is further proposed based on GRDM1.In GRDM2, Load II and Load III are the same as those in GRDM1. There are no springs around the generalized structure in GRDM2. Therefore, the most significant difference between GRDM2 and GRDM1 is the calculation of Load I.

As mentioned before, foundation springs are introduced to consider the constraint of the hollow soil model, as shown in Fig. 5. In Eq. (4) the stiffness matrixKIFreflects the impedance of the hollow soil model, andKIF(uIF–uI) is the load caused by the relative displacement of the free-field. This load can be considered as the reaction force acting on the interface when the hollow soil causes a corresponding relative displacement of the free-field. In GRDM1, the relative displacement of the free-field is converted into a node force by the use of spring coefficients. While in GRDM2, the relative displacement of the free-field is imposed on the hollow soil model, as shown in Fig. 7. At this time there will be reaction forces at the nodes on the interface, which areKIF(uIF–uI). As shown in Fig. 7, the reaction forces areRDxiandRDyiat nodei, and there is Load I in GRDM2.

After Load I, Load II and Load III are determined according to the above method, the calculation model of GRDM2 can be established, as shown in Fig. 8. This model is actually similar to the model shown in Fig. 4.However, there are two main differences: the hollow soil(grey portion) replaces the discrete spring; the magnitude of Load I at the interface node is different. Load I in GRDM2 can be expressed as follows:

in which,F2xiandF2yiare the equivalent earthquake loads to be applied on nodeiin thexandydirections in GRDM2.RDxiandRDyiare the reaction force of nodeicaused by the relative displacement of the free-field in thexandydirections, respectively.

It should be noted that the two models shown in Fig. 7 and Fig. 8 are different. Particularly at the interface,the displacement of the two models is different. If the interface nodes shown in Fig. 8 are directly pulled to the relative displacement of the free field, it is inaccurate because the influence of the structure on site deformation is not reflected at this time.

Figure 9 shows the implementation steps of GRDM2.

I. Same as Step I in GRDM1.

Ⅱ. Same as Step Ⅱ in GRDM1.

Ⅲ. Calculate Load I: the soil-generalized structure interface is forcibly pulled to the free-field deformation position in the hollow soil model, and the reaction forces are recorded at the soil-generalized structure interface.

Ⅳ. Same as Step Ⅳ in GRDM1.

V. Same as Step V in GRDM1.

Ⅵ. Establish the analysis model for GRDM2.According to the analysis model of GRDM2 shown in Fig. 8, the equivalent seismic loads calculated from stepsⅢ to V are applied to the generalized structure for static analysis.

3.3 Method 3

As previously mentioned, GRDM1 considers the soil-structure interaction by introducing the foundation springs, and GRDM2 adopts the soil-structure integrated model. GRDM3 is proposed based on GRDM2 and still uses the soil-structure integrated model. Both GRDM1 and GRDM2 use the same assumptions for the calculation of shear stress around the generalized structure. Shear stress on the left and right sides are considered as the average value of the top and bottom surfaces. The shear stress changes with the height of the soil. In particular, when the height of the generalized structure is very high, and if the shear stress on the left and right sides is considered a constant, those factors will lead to serious calculation errors. GRDM3 attempts to improve the method for calculating Load II.

The shear stress around the structure is the stress of the soil in the free-field at a position corresponding to the soil-structure interface. To more accurately and effectively calculate shear stress around the generalized structure, Fig. 10 shows the force analysis of the soil in the area surrounded by the soil-generalized structure interface. During the earthquake the equilibrium equation of the soil surrounded by the interface is:

in which,m,c, andkrepresent the mass, damping, and stiffness of the soil, respectively, andqis the load at the interface. When the influence of the damping term is ignored, Eq. (8) can be simplified as follows:

Fig. 7 Equivalent loads caused by relative displacement

Fig. 8 Analysis model of GRDM2

Fig. 9 Flowchart of GRDM2

The exact Load II on the interface can be calculated as shown in Fig. 10. The displacement of the interface should be pulled to the relative displacement of the free field, and the corresponding inertia force should be applied inside the soil. The reaction forces (RSxiandRSyi)at the interface are Load II in GRDM3.

It is found that Load I is calculated by applying the relative displacement at interfaceIin the model illustrated in Fig. 7, and Load II also is calculated by applying the relative displacement at interfaceIas shown in the model represented in Fig. 10. The displacement conditions at interfaceIin these two models are consistent. Therefore,these models can be combined as shown in Fig. 11,which is equivalent to creating a continuous soil model.It should be noted that there is no need to remove the soil occupied by the generalized structure in the continuous model. The loads consist of the relative displacement of the free-field at the interface and the inertial force of the soil in the area surrounded by the interface. The sum of Load I and Load II are the reaction forces of the nodes at the interface, as shown in Fig. 11. According to the previous analysis, these reaction forces have the following relationship:

in whichF3xiandF3yiare the equivalent earthquake loads to be applied on nodeiin thexandydirections in GRDM3.RDxiandRDyiare the reaction force of nodeicaused by the relative displacement of the free-field in thexandydirections, respectively.RSxiandRSyiare the reaction force of nodeicaused by the shear stress of the free-field in thexandydirections, respectively, as shown in Fig. 10.

Figure 12 shows the analysis model of GRDM3. Both GRDM2 and GRDM3 use the soil-structure integrated model to reflect soil-structure interaction and analysis of the response of an underground structure. Figure 13 shows the implementation steps of GRDM3.

I. Same as Step I in GRDM1.

Ⅱ. Same as Step Ⅱ in GRDM1. The calculation results do not include the shear force of the soil at the top and bottom of the corresponding soil.

Ⅲ. Calculate Load I and Load II. The interface is forcibly pulled to the free-field deformation position in the continuous soil model, and the soil in the area surrounded by the interface is imposed on the horizontal inertial force at the same time. Thus, the reaction forces are obtained at the interface.

Ⅳ. Same as Step V in GRDM1.

V. Establish the analysis model of GRDM3.According to the analysis model of GRDM3 shown in Fig. 13, the equivalent seismic loads calculated from steps Ⅲ to Ⅳ are applied to the generalized structure for the static analysis.

4 Verification of the GRDM

4.1 Analysis model

A subway station with an arch cross section is chosen to verify the computational accuracy of the three proposed methods. The size of the soil and the structure are shown in Fig. 14. The height is 50 m from the surface to the bedrock at this site. The distance is 15 m from the ground to the top of the arch. There are two main types of soil, including sand and clay. Figure 14 also shows the material properties of the site, including unit height, unit weight, and shear velocity. The average shear wave velocity is about 338.76 m/s at this site.The soil is assumed to be homogeneous and isotropic.The nonlinearity behaviors of the soil are considered by the use of the equivalent linearization method, such as the EERA program (Bardetet al., 2000). Figure 14 shows the typical constitutive curves of sand and clay,including theG/Gmax-shear strain and damping ratioshear strain curves, which are used to analyze the dynamic response of the 1D free-field. The equivalent shear modulus is used to calculate the coefficients of the foundation springs or the equivalent earthquake loads for the proposed simplified methods. The equivalent shear modulus and damping ratios are applied in the time history analysis.

Fig. 10 Load II caused by shear stress

Fig. 11 Equivalent loads caused by relative displacement and shear stress

Fig. 12 Analysis model of GRDM3

Figure 14 also shows the model dimensions of the station. It is a typical framework structure, with two stories and two spans. The height and width of the structure are both 10 m. The top slab of the structure is composed of two arcs with a radius of about 3.39 m.The thickness is 0.8 m in all top/bottom slabs and left/right walls. In addition, the thickness of the middle slab is 0.6 m. The cross section of the central column is a square with a width of 0.8 m, and the column space is 5 m. Since this study only verifies the reliability of the proposed methods, the structures are assumed to be elastic and the influence of reinforcement is ignored in the numerical models. Therefore, the material of the structural beams and columns is assumed to be elastic,with an elastic young modulus of 30 GPa, a density of 2500 kg/m3and a Poisson ratio of 0.2.

Two earthquake records (Loma Prieta, 1989 and El Centro, 1940) are adopted for the seismic analysis of the subway station. Figure 15 shows the acceleration records and Fourier amplitudes of the earthquakes. The peak acceleration of the used earthquake motions are adjusted to 0.1 g in the numerical analysis. As mentioned before,this study conducts the equivalent linearization analysis of the free-field under the Loma Prieta and El Centro earthquakes, respectively, and obtains the equivalent shear modulus and damping ratio for each soil layer.The finite element code ABAQUS Program (Systèmes,2007) is used to carry out the analyses under plane strain conditions. The soil is simulated for the planestrain elements (CPE4R and CPE3), and the structure is simulated for the beam element (B21). The nodes of the soil and the structure at the interface are joined.Figure 16 shows the finite element model for the time history analysis of the soil-structure system. The width of the model is 70 m, which is seven times the width of the structure. Moreover, the height of the model is 50 m.

Fig. 13 Flowchart of GRDM3

The viscous-spring boundaries (Du and Zhao, 2009)and equivalent nodal forces (Liet al., 2018) are applied at the lateral truncated boundaries when using the time history analysis method. The diagonal spring and damping matrices can be calculated as follows:

in which the subscripts T and N represent the tangent and normal directions of the artificial boundary, respectively.Alis the area of all elements containing the boundary nodel. Additionally,ris the distance between the boundary and the wave source;Gis the shear modulus;λis the Lamé coefficient;ρis the mass density; andcsandcpare the velocity of the shear wave and the compression wave in the soil layer, respectively. Finally,AandBare modified coefficients that can be specified as 0.8 and 1.1,respectively.

The mesh generation for the simplified methods is identical to that of the time-history analysis method.Figure 17 shows all the finite element models used in the simplified methods, including the hollow soil model, the generalized structure model, the continuous soil model,and the soil-structure interaction model. In the analysis model shown in Fig. 17(d), the constraint condition is to tie the generalized structure and soil at the interface.

Fig. 14 Size of the soil and the structure

4.2 Results of the free-field

Figure 18 shows the time-history curves of the relative displacement between the soil layer corresponding to the top and bottom surfaces of the generalized structure under the effects of the Loma Prieta earthquake and El Centro earthquake. The maximum relative displacements are 8.512 mm and 11.014 mm for the two earthquake loads, and the corresponding times are 11.60 s and 2.22 s.Figure 19 shows the displacement and acceleration distributions of the 1D free-field at the corresponding time. The absolute accelerations are 2.062 m/s2and 2.339 m/s2at the center of the generalized structure.

The recorded results also include the soil shear stress at the top of the generalized structureτTand the soil shear stress at the top of the generalized structureτBfrom the seismic analysis of the free-field. As for the El Centro earthquake,τTandτBare 74.315 kPa and 144.226 kPa,respectively; with regard to the Loma Prieta earthquake,τTandτBare 58.524 kPa and 120.367 kPa, respectively.Additionally, the soil shear force at the left and right side of the generalized structure for GRDM1 and GRDM2 can be calculated as 109.271 kPa and 89.446 kPa for the two earthquakes, respectively.

Fig. 16 Finite element model for the time history analysis method

The equivalent modulus of different layers obtained from the seismic analysis of the free-field are used as the input parameters to establish the calculation models for the coefficients of foundation springs in the GRDM1. All the spring coefficients are listed in Table 1. For instance,kT1andkT2are the coefficients of the springs at the top of the generalized structure in the horizontal and vertical directions;kL1andkL2are the coefficients of the springs at the left of the generalized structure in the horizontal and vertical directions;kR1andkR2are the coefficients of the springs at the right side of the generalized structure in the horizontal and vertical directions; andkB1andkB2are the coefficients of the springs at the bottom of the generalized structure in the horizontal and vertical directions. As shown in Table 1, the spring coefficient of the El Centro earthquake is slightly smaller than the corresponding coefficient for the Loma Prieta earthquake, a finding consistent with the results of the previous analysis. As shown in Fig. 19(a), the horizontal displacement of the free-field is larger under the El Centro earthquake, and the corresponding shear strain for each soil layer is larger than that of the Loma Prieta earthquake. The equivalent shear modulus obtained by the equivalent linearization method is smaller, which results in smaller spring coefficients.

4.3 Results from the analysis of the structure

Figure 20 shows the internal force diagrams obtained by utilizing the time history analysis method and three generalized response displacement methods.The internal forces mainly include axial force, shear force, and bending moment. If only the horizontal earthquake load is considered in the numerical analyses,the subway station studied in this paper can be regarded as an antisymmetric structure. Therefore, Fig. 20 shows the left half of the structure when comparing the internal forces with regard to the structure. It also can be seen from the axial force diagrams that the axial force is 0 at the central column of such an antisymmetric structure.For different earthquake records, the magnitude and distribution of internal forces calculated by these four methods are relatively close. The internal forces of the structure under the influence of the El Centro earthquake are greater than those for the Loma Prieta earthquake,a finding consistent with previous analyses. As shown in the seismic analysis of the free-field, the equivalent seismic loads are larger under the effects of the El Centro earthquake.

Fig. 18 Time-history curves of the relative displacement

Figure 21 shows a comparison of the peak internal forces and horizontal deformations of the members as calculated by using different methods. The name of each member is shown in Fig. 14. Compared with the time history analysis method, the error range for GRDM1 is the largest, especially in calculating the axial and shear forces of the corner components. The maximum error is the axial force of TLW under the effects of the El Centro earthquake, which is approximately 20%.The results are consistent with previous analyses: the GRDM1 uses the concentrated springs, and the corner constraints of the generalized structure are different from an actual situation. GRDM2 and GRDM3 establish the soil-structure integrated model to analyze the responses of the structure, and the calculation error of the internal forces and deformations are maintained within about 5%. Overall, the calculation results from GRDM3 are closer to the results of the time history analysis method,because GRDM3 can more accurately consider the shear force of the surrounding soil.

From the perspective of the peak value of the internal force, the calculation error between the proposed methods and the time history analysis method can only explain calculation accuracy for a certain element of the structure. In order to reflect whether the internal force is calculated accurately for all elements, the secondorder Euclidean Norm is applied to the analysis of the error values for all internal forces of the member. The calculation method is as follows:

in whichFGis the internal force calculated by the generalized response displacement method;FDis the internal force calculated by the dynamic time-history analysis method;iis the element number in one member;nis the total number of the element in one member; andFGiandFDiare the internal force of thei-th element, calculated by the use of the generalized response displacement method and the dynamic timehistory analysis method.

Table 2 shows the error for each member as calculated by the use of the three proposed method. GRDM1 shows the largest error and GRDM3 the smallest.This phenomenon is consistent with previous analysis results. The concentrated springs are unrelated to one another and reflect limited soil-structure interactions in GRDM1. Both GRDM2 and GRDM3 apply the soil-structure integral analysis model to consider the interaction between soil and a structure. The latter twoproposed methods have higher calculation accuracy.However, GRDM3 uses one model and one step to calculate the equivalent seismic load caused by relative displacement and the shear stress of the free-field. The calculation efficiency of GRDM3 is much higher than that of GRDM1 and GRDM2. Therefore, GRDM3 is more highly recommended for calculating the seismic response of underground structures.

Table 1 Spring coefficients for GRDM1 (Unit: 106 N/m)

Fig. 19 Displacement and acceleration distributions

Fig. 21 Comparison of peak internal force and deformation

Fig. 21 Continued

Table 2 Comparison of internal forces

5 Conclusions

To overcome the deficiency that the traditional response displacement method is only applicable to an underground structure with simple cross section, three new methods are proposed, and the details of calculating equivalent seismic load by employing different methods are introduced in this paper. In addition, the calculation accuracy of the proposed methods is evaluated by comparing them with the dynamic time history analysis method. The following conclusions can be drawn from this study:

(1) GRDM1 adopts the same analysis model as the traditional response displacement method, and the foundation springs are set around the generalized structure to reflect the soil-structure interaction.However, the restraint effect of the foundation springs on the generalized structure is different from the actual situation of the soil. Therefore, there is a certain error in the internal force and deformation of the component,which is compared to the accurate dynamic time history analysis method. In particular, the calculation error is about 20% when calculating the internal forces of the corner components.

(2) GRDM2 utilizes the soil-structure integral analysis model to fully consider the interaction between soil and structure. The most significant difference between GRDM2 and GRDM1 is the equivalent seismic load caused by the relative displacement of the soil.Equivalent seismic load can be obtained by applying the relative displacement to the hollow soil model without a generalized structure. The error rate for the internal force and deformation of the component is about 5%, which constitutes an effective simplified seismic analysis method.

(3) GRDM3 also uses the soil-structure integral analysis model to fully consider the interaction between soil and structure. The most significant difference between GRDM3 and GRDM2 is the equivalent seismic load caused by relative displacement and the shear stress of the soil. Equivalent seismic load can be obtained by applying relative displacement and the inertia force on the continuous soil model. The error rate for the internal force and deformation of the component is less than 5%. GRDM3 demonstrate the advantages of calculation efficiency and calculation accuracy, and is suitable for the seismic analysis of an underground structure that contains regular or complex cross sections.

Acknowledgement

This research is jointly funded by the National Natural Science Foundation of China (No.52108453), the Natural Science Foundation of Jiangxi Province in China (No. 20212BAB214014),the National Key R&D Program of China (No.2018YFC1504305), and Joint Funds of the National Natural Science Foundation of China (No. U1839201).


登录APP查看全文