1. INTRODUCTION
Underground structures have been prominently used to convey life lines such as water, gas, electricity and communication cables. The buried structures were less affected during earthquakes, however, few incidents of major earthquakes in the history, such as the wellknown Kobe earthquake(Japan) in 1995, Chichi earthquake (Taiwan) in 1999 and Gujarat earthquake (India) in 2001 happened to be the underground structures. When underground structures were damaged, the supplies of life lines are significantly affected and disasters may occur. Researchers have been interested in investigating the soilstructure interaction, such as on various kinds of buried structures under seismic loadings. Abuhajar et al.(2015) conducted both experimental and numerical tests regarding the seismic soilstructure interaction involving box culverts, Byrne et al.(1996) have studied a numerical analysis of dynamic soilstructure interaction regarding a threehinged large buried culvert structures using finite elements, and Wen et al.(2017) extended the dynamic soilstructure interaction with the inclusion of an electrical substation equipment using infinite elements.
However, these literatures only focused in the behavior and performance of underground structure under seismic excitation and did not consider possible methods in reduction of loads experienced by the buried structures with dynamic loading. Kim et al.(2018) presented a construction technique called Embedded Trench Installation (ETI) which was used to reduce the earth pressure acting on the underground structure. The ETI technique used soft compressible material of the expanded polystyrene (EPS) geofoam which encapsulated the whole structure. In recent years, studies regarding the use EPS geofoam for ETI technique had been proven to induce positive soil arching and reduce earth loads acting on the buried structure. Numerous studies had been conducted by Kang and his colleagues for the application of EPS Geofoam in pipes on different materials such as, corrugated PVC pipe (Kang et al, 2007a), concrete pipes (Kang et al, 2007b), concrete box tunnel (Kang et al, 2007c), and corrugated steel pipe (Kang et al, 2007d, 2007e). Jeong et al.(2018) extended the study to ecological bridges. Kim et al.(2018) determined the optimum size of geofoam in load reduction for underground arch structures. However, these studies were limited to static loading and did not include seismic loadings. This study presents the behavior of underground structure with the application of the EPS geofoam subjected to seismic loading.
2. BACKGROUND OF RELATED STUDIES
2.1 Mechanism of soil with installed structures
Soil arching happens when there is a difference stiffness between the culvert and the surrounding soil. Marston and Anderson(1913) investigated that buried pipes at great depth affects the relative settlements which play a vital role in earth pressure acting on the structure. The relative settlement created friction forces that could be either added or subtracted to resultant load acting on the pipe. Based from the study, the soil was divided into three rectangular prisms by projecting imaginary vertical lines which extend from both sides of the pipe up to the ground. If the relative settlement in the adjacent soil prisms was more than the central soil prisms, the earth load acting on the conduit was increased by the downward friction forces generated by the adjacent prisms (negative soil arching). On the other hand, when the relative settlement of the central soil prism was greater than of the adjacent, the generated frictional forces acted oppositely to the resultant load, thus, was subtracted resulting to decrease in earth pressure acting on the pipe (positive soil arching).
In order to induce a positive arching, Marston (1922) established a construction technique which was called as Imperfect Trench Installation (ITI). The ITI technique was a method that placed a soft compressible layer above the structure shown in Fig. 1(b), that was proven effective to reduce earth pressures at the top of the underground structure.
2.2 Embedded trench installation technique
The Embedded Trench Installation(ETI) technique came from the principles and ideas of the Imperfect Trench Installation. Kim et al.(2018) had studied that the reduced earth pressure on the top part of the structure moved to the side of the structure. Thus, in order to counter the effect of ITI technique, soft zone or compressible lightweight materials was also applied at the side of structure, hence, encapsulating the whole structure with soft compressible material as shown in Fig. 1(a) and Fig. 2(a).
2.3 Infinite elements
Nielsen(2006) studied that the fixed boundary conditions in static problems could be used without sacrificing much in terms of accuracy. However, the fixed boundary conditions caused the reflection of waves which effectively trapped the energy inside the dynamic model. Therefore, the simulation of the unbounded soil medium in numerical methods has been a challenging topic in dynamic soilstructure interaction(SSI). The simplest solution to this problem was to create a large enough field so that waves reflected from the boundaries did not have time to return to the region of interest. In order to simulate the soil medium in a simple yet realistic manner, numerous researches were conducted since Alterman and Kara(1968) first put forward the concept of placing the artificial truncated boundary which was far away from the domain of interest.
However, the simulation of far field boundary was not a practical solution since it created too many unnecessary elements and generated longer time for finite element analysis. Lysmer and Kuhlemeyer(1969) replaced the far field boundary conditions with viscous damping to simulate the elastic resilience of the soil, and provided the algorithm for first and secondorder infinite elements in ABAQUS for dynamic response.
Some studies were made regarding the infinite element in ABAQUS like Ismail and Mullen(2000) and Yue et al.(2006). The studies presented the comparisons in the viscous boundary and the infinite element in ABAQUS and provided that the use of infinite element in ABAQUS had many advantages over the viscous boundary.
3. FINITE ELEMENT MODELING
The ABAQUS was chosen to simulate the model due to its wide material modelling capability and the provided algorithm for first and secondorder infinite elements based on the work of Zienkiewicz et al.(1983) for static response and of Lysmer and Kuhlemeyer(1969) for dynamic response. The simulated structure was buried at 3m depth and was analyzed under clay soil medium, then further more investigated by varying the type of EPS geofoam, namely EPS 12, EPS 15 and EPS 19 which significantly differs in density.
Lou and Chen(1999) and Lou et al.(2000) suggested that the soil medium could be defined as five times the width of the structure in order to reduce the motion reflection on the boundary. The soil size of the finite element is 15m×15m, having a concrete tunnel structure of 3m×3m. Figs. 3 and 4 show the schematic diagram of the model and the crosssection of the used concrete tunnel model, respectively. The concrete tunnel and geofoam are modelled as C3D8R for hexahedral element, and the soil is composed of C3D8R and AC3D8 for hexahedral finite and infinite elements, respectively. In order to simulate the acoustic elements AC3D8, AC3D8 elements were manipulated into CIN3D8 elements by using ABAQUS input file, thus, creating infinite elements.
Infinite elements were used for lateral borderline of the soil structure in order to simulate the artificial boundary conditions for seismic loadings as shown in Fig. 3. The use of infinite elements has been proven to have built in damping constants to eliminate the reflection of waves back to the medium (Yue and Li, 2007). Fixed boundary conditions were applied at the bottom of the structure for static analysis and released the horizontal restraint during the dynamic analysis. Figs. 5 and 6 shows threedimensional model and all of the models were subjected to gravity loading. Timehistory data of ElCentro NorthSouth component, as shown in Fig. 7, were applied at the base of the structure which values were retrieved from vibrationdata.com.
Table 1 lists the properties of materials used in the model and soil values retrieved from finesoftware.com (Raymond and James, 1985;Nixon and Child, 1989: Liu et al., 2010), Table 2 shows the properties of concrete material where Kc is the ratio of the distances between the hydrostatic axis and respectively the compression meridian and the tension meridian in the deviatoric cross section. The f_{b0}/f_{c0} is the ratio of strength in biaxial state(f_{b0}) to strength in uniaxial state(f_{c0}). Table 3 presents geofoam materials from ASTM D6817(2002). DruckerPrager failure criteria were used to model the plasticity of clay soil medium, Concrete Damage Plasticity (CDP) was adopted to simulate the plasticity of concrete and the values from the stressstrain curve of geofoam were used to simulate the plasticity of the geofoam. A surfacetosurface contact type option was used for the interaction of the concrete structure and backfill and/or geofoam. The soilstructure interaction had a rough friction formulation for the tangential behavior and a hard contact for the pressureoverclosure section in normal behavior while restraining separation after contact. Dynamic, Implicit analysis was performed to simulate the seismic loading.
4. FINITE ELEMENT ANALYSIS RESULTS
Finite element analyses results were investigated at top slab, side wall and bottom slab in the middle part of the concrete structures. The solid lines represent the structure without geofoam and the broken lines represent all of the structures with geofoam. After applying gravity and dynamic loading, results show the static loading as well as the minimum and maximum loads throughout the earthquake simulation. Fig. 8
4.1 Effect of EPS geofoam
Fig. 9 shows the parabolic distributions of earth pressures on the top slab, side wall, and bottom slab similar with the works of previous researchers such as Dasgupta and Sengupta(1991). The reason behind the parabolic shape of earth pressures is due to the earth loads moving to the corners of the structure. Fig. 9 also shows the comparison between the earth pressures acting on the model with and without geofoam and presents that the use of geofoam generally reduces the earth loads experienced by the model with clay medium.
Fig. 10 shows the comparison for horizontal deflection of the structure with and without geofoam. It presents that the values for top slab and bottom slab of the structure are nearly forming a flat line and the values of the side walls showed deformation. The reason behind this behavior is that the horizontal deflection of the structure is mainly influenced by the earthquake excitation. The geofoam applied to the structure also reduces the horizontal deflection.
Fig. 11 for the vertical deflection of the structure shows that the values of the side walls almost formed a flat line and the same parabolic shape is observed for the top and bottom slab of the structure. The parabolic shape occurs before with the earth pressure experienced by the model and the reason behind this is that the vertical deflection is heavily influenced by static loading. Fig. 11 also indicates the same trend that the use of geofoam is effective in reduction of vertical deflection.
4.2 Percentage reduction considering geofoams types
The average percentage reduction was computed for the top slab, side walls and bottom slab considering the static, minimum and maximum values produced by dynamic loading (DN; Dynamic Minimum values, DM; Dynamic Maximum values). From the Tables 46, it displays that the values of percentage reduction for earth pressure decreases as the geofoam varies from EPS 12 to EPS 19. The same trends were also observed for percentage reduction of horizontal and vertical deflection and it shows that the best type of expanded polystyrene geofoam is the geofoam of EPS 12. Table 5
5. CONCLUSION
This study presents the behavior of underground structure with the application of EPS geofoam subjected to seismic loading using finite element analysis program, ABAQUS. The behavior of the structure was compared between the different types of geofoam applied to the model, in terms of load reduction and deflection. Based on this study, the following conclusions have been made:

1) The application of geofoam on underground structures generally reduces the earth loads acting on the structure, both in static and dynamic loading.

2) It was confirmed that the use of ETI construction technique with the application of EPS geofoam, which is a highly compressible material, reduces the stress on the buried structure up to 47% in static, and 50% in dynamic, compared to case of buried structures without EPS geofoam reinforcement.

3) EPS 12 geofoam type, which has the least value of density, provides the highest percentage reduction in earth pressure and deflection acting on the structure. The least density produced the highest compression in the soft zone, thus, the larger settlement in the central prism generated more frictional forces against the resultant load resulting to reduction of loads experienced by the structure.