Identification of elastic basin properties by large-scale inverse earthquake wave propagation

The importance of the study of earthquake response, from a social and economical standpoint, is a major motivation for the current study. The severe uncertainties involved in the analysis of elastic wave propagation in the interior of the earth increase the difficulty in estimating earthquake impact in seismically active areas. The need for recovery of information about the geological and mechanical properties of underlying soils motivates the attempt to apply inverse analysis on earthquake wave propagation problems.; Inversion for elastic properties of soils is formulated as an constrained optimization problem. A series of trial mechanical soil models is tested against a limited-size set of dynamic response measurements, given partial knowledge of the target model and complete information on source characteristics, both temporal and geometric. This inverse analysis gives rise to a powerful method for recovery of a material model that produces the given response.; The goal of the current study is the development of a robust and efficient computational inversion methodology for material model identification. Solution methods for gradient-based local optimization combine with robustification and globalization techniques to build an effective inversion framework. A Newton-based approach deals with the complications of the highly nonlinear systems generated in the inversion solution process. Moreover, a key addition to the inversion methodology is the application of regularization techniques for obtaining admissible soil models. Most importantly, the development and use of a multiscale strategy offers globalizing and robustifying advantages to the inversion process.; In this study, a collection of results of inversion for different three-dimensional Lame moduli models is presented. The results demonstrate the effectiveness of the inversion methodology proposed and provide evidence for its capabilities. They also show the path for further study of elastic property identification inverse problems.