Reconstruction of structural anomalies out of seismic measurements by means of a non-deterministic full waveform inversion approach for application in mechanized tunneling

Abstract

A full waveform inversion may be utilized as a promising tool to invert the properties of a spatial model out of seismic waveforms, but in turn may claim a high computational effort if the inversion method is not effective. Unscented hybrid simulated annealing, which is developed for an application in mechanized tunneling, is an optimization method that combines the metaheuristic simulated annealing with the unscented Kalman filter. The method needs a priori knowledge in the form of a user-defined parametrization, but therefore may invert the structure effectively and precisely. In this work, the authors apply the novel method on laboratory-generated seismic data. In a first step, three concrete plates with different structural anomalies are constructed. Seismic data is acquired and implemented into the optimization algorithm to invert the positions, shapes, and the material parameters of the structural anomalies. The authors find that the algorithm performs well on the generated data, reconstructing the structural anomalies with a satisfying precision during a low number of forward simulations.