Seismic response of subway station in soft soil: Shaking table testing versus numerical analysis

As revealed by the collapse of the Daikai Metro station during the 1995 Kobe earthquake, underground structures are not immune to seismic loading. Shanghai Metro operates 16 lines of 676 km length, comprising 413 underground stations. An additional 1000 km with 600 underground stations are planned for the next 20 years, calling for improved understanding of their seismic response. This paper studies the seismic performance of a typical 2-storey, 3-span Shanghai Metro station in soft soil, combining shaking table testing and numerical modelling. Notwithstanding scale effects, shaking table testing is performed to allow detailed simulation of the complex structural system of the station. The structure is modelled using granular concrete and galvanized steel wires to simulate the RC prototype. To remedy the problem of scale effects, synthetic model soil (a mixture of sand and sawdust) is used, along with similitude relations derived considering dynamic equilibrium. The properties of the synthetic model soil are adjusted to satisfy similitude; target stiffness and density are attained by adjusting the mixture proportions. To quantify the transferability of the results to prototype scale, the experiments are simulated with nonlinear finite elements (FE), modelling the synthetic model soil with a kinematic hardening constitutive model, calibrated against resonant column and direct shear tests. The FE model is shown to compare adequately well with the shaking table tests. The validated FE model is used to predict the seismic response of the prototype, thus allowing indirect transfer of the results from model to prototype scale. The model in prototype scale is calibrated for the real soil layers against in situ (down-hole) and laboratory (resonant column) tests. Moving from model to prototype scale, the racking deformation remains qualitatively similar. The racking drift is reduced by 50% going from model to prototype scale, which is partly due to scale effects, but also related to differences between the idealized soil of the experiments and the multiple soil layers encountered in reality. The maximum bending moment also reduces by 30% going from model to prototype scale. The base of the lower-storey columns is proven to be the most vulnerable section, as was the case for Daikai.