Full phase-velocity spectrum inversion of multimode Rayleigh waves

Abstract

Multi-channel analysis of surface wave (MASW) has attracted popularity in site investigation activities for retrieving the shear wave velocity profile of the ground. Although the workflow of processing MASW has developed for many years, manual interpretation of the multi-mode dispersion spectrum is still required, relying on experience and judgment. We propose a new surface wave inversion framework that does not require surface wave identification and multi-mode assignment. We use the finite element method (FEM) to simulate the seismic wave propagation as the forward modeling. Then, the phase shift method is employed on the multiple-channel records to obtain the dispersion spectrum. We propose a new way of using Wasserstein distance (WD) as objective function to evaluate the difference between experimental and theoretical spectra. We utilize a Bayesian optimization (BO) framework to invert the shear wave velocity profiles from the phase velocity spectra. We verify the framework with synthetic and field experimental data. Numerical results suggest that our method can automatically capture the dispersion signatures of multi-mode Rayleigh waves and provide additional interpretation, such as detecting the leaky body wave.