Influence of input motion selection and soil variability on nonlinear ground response analyses

Abstract

This thesis studies the influence of input motion selection strategies on nonlinear site response analyses of soft clay soil deposits. It also investigates the role of elastic and nonlinear soil properties in nonlinear site response predictions. The research adopts a fullycoupled finite element (FE) procedure, employing an advanced kinematic hardening soil model. In addition, the results of equivalent linear site response analyses are considered for comparison purposes. Firstly, the research validates the performance of the advanced soil constitutive model as implemented in the FE code. For this purpose, the free-field response at the Lotung LargeScale Seismic Test (LSST) site in Taiwan is modelled in two dimensions. The best-fit shear modulus values according to in-situ data are used and the remaining soil model parameters are calibrated against experimental data from samples retrieved at different depths at the site. One weak (LSST11) and one strong (LSST7) input motion recorded along the downhole array are simulated. The predictions from the nonlinear FE approach at 17 m, 11 m, 6 m and at ground surface are compared with the corresponding recordings in terms of acceleration time histories, spectral response and maximum acceleration profiles. The accumulation of excess pore water pressures obtained from the FE nonlinear analyses is also compared to the recorded data. Secondly, the research investigates the influence of five input motion selection strategies on nonlinear site response predictions. In particular, the effect of Peak Ground Acceleration (PGA) scaling, spectral acceleration scaling at T1 (Sa(T1)), where T1 is the natural period of the soil column,, 0.2T1-2T1 scaling, Mean Squared Error (MSE) scaling and spectral matching selection strategies is studied by modelling an ideal soft clay soil deposit with soil class D properties according to Eurocode 8 (EC8). Sets of seven input motions are formed for each selection strategy with lower (0.15g) and higher (0.35g) seismic intensity levels. The target response spectra are constructed based on the EC8 prescription. The selection strategies are evaluated with respect to Sa(T1), PGA profile and amplification factors. Moreover, they are assessed based on relative displacement, PGA and Sa(T1) at the ground surface (referred to as Engineering Demand Parameters, EDPs). Finally, the research investigates the impact of the variability of the elastic (shear wave velocity, Vs) and dynamic (shear modulus reduction, G/Go, and damping, D, curves) soil ii properties on nonlinear site response predictions. The same soil model and input motions used for the deterministic analysis of the LSST site are adopted for nonlinear Monte Carlo Simulations (MCS) of the site. The results are evaluated in terms of spectral response, PGA profile and maximum shear strain profile, and the median values are compared with the recorded data. Overall, the research verifies the capacity of the advanced constitutive soil model within a fully-coupled nonlinear FE procedure to correctly predict the free-field ground response at a given site, including the accumulation of excess pore water pressures. It also concludes that the spectral matching method is the best choice amongst the investigated selection strategies, as it leads to the least scatter in the EDP response. Moreover, seven input motions are found to be sufficient to obtain a stable response at ground surface. Lastly, the impact of variability of the Vs profile and G/Go and D curves is shown to be dependent on both seismic intensity level and the adopted numerical approach. (e.g. equivalent linear code vs fullycoupled nonlinear code).