Physics-Informed Neural Network for Elastic Wave-Mode Separation

TL;DR

This paper introduces a physics-informed neural network (PINN) that effectively separates P and S wave modes in elastic media, reducing computational costs and improving accuracy over traditional methods.

Contribution

The paper presents a novel scalar PINN approach for elastic wave-mode separation, offering a scalable and efficient alternative to vector-based techniques.

Findings

Modes closely match conventional numerical methods

Reduced transverse wave leakage observed

Effective in both homogeneous and non-homogeneous models

Abstract

Mode conversion in non-homogeneous elastic media makes it challenging to interpret physical properties accurately. Decomposing these modes correctly is crucial across various scientific areas. Recent machine learning approaches have been proposed to address this problem, utilizing the Helmholtz decomposition technique. In this paper, we investigate the capabilities of a physics-informed neural network (PINN) in separating P and S modes by solving a scalar Poisson equation. This scalar formulation offers a dimensionally scalable reduction in computational cost compared to the traditional vector formulation. We verify the proposed method in both homogeneous and realistic non-homogeneous elastic models as showcases. The obtained separated modes closely match those from conventional numerical techniques, while exhibiting reduced transverse wave leakage.