Dual-Driven Solver for Reconstructing the Point Sources of Elastic Wave Based on Far-Field Data

Abstract

Aiming at the inverse source problem of an elastic wave, a dual-driver solver is considered to reconstruct the point sources. In this way, the number, location, and magnitude of the point sources can be reconstructed from far-field measurement data. The solver is composed of a data-driven module and a physical-driven module, which is coupled by a loss. The loss of the data-driven module and the physical-driven module are both the driving force of the solver evolution. The solver takes the far-field data as the input, and the number, location, and magnitude of the point sources as the output. It is trained by the Adam algorithm. Numerical experiments show that this method is effective for reconstructing the multi-sources.

Keywords

• inverse source problem
• dual-driven
• elastic wave
• multi-sources
• finite observation aperture