Response prediction of earth structures to future earthquakes based on nonlinear system identification by the VAI method

Abstract

So far, the identification of the curvilinear hysteretic dynamic behavior of soils involved mainly synthetic numerical experiments that did not account for the realistic conditions emerging in the processing of real response data. Having learned the difficulties encountered in the only identification attempt from real response data that employed curvilinear hysteretic model [1], the authors have recently developed the VAI (Variable Amplitude Identification) scheme of identification which is simple and overcomes the delicate problem of stress reversals in the hysteretic systems [2]. The VAI method is essentially a multi-step strategy which determines one equivalent secant modulus corresponding to a certain displacement level. Hence, it employs those pairs of moduli and displacement amplitudes to construct the curvilinear backbone curve. In this paper, the method is reintroduced with idealizations that allow for directly identifying the material parameters rather than the structural ones. The assumption of the mirror symmetry of the hysteresis loop around the main secant modulus suggests dependence of the identification precision on both the identified material and the pertinent strain range. Therefore, the potential of the application of the new identification method to identifying the characteristic parameters for a wide range of sandy soils is evaluated. Together with that, the 24 inverse problems solved further confirmed the convergence and the stability of the utilized linear inverse solver. The identified parameters were successfully employed to predict the response of the analyzed soil deposits to a future earthquake yielding excellent accuracy. Numerically simulated noise-free observations were used to provide deterministic basis for judging the identification outcome.