A physics-informed machine learning framework for inverse multi-solution problems

Zexiong Wu

doi:10.1016/j.apm.2025.116618

A physics-informed machine learning framework for inverse multi-solution problems

2025-11-13

0

PRE

AI

DOI：10.1016/j.apm.2025.116618

Original

Original request for help

Save

Abstract

En 中文

• The proposed approach provides a unified framework for solving inverse multi-solution problems, without the need for complex modeling (e.g., meshing), iterative inverse-solving and multiple trial processes. • It is capable of learning a continuous functional relationship among the unknown parameters, rather than producing only discrete solution points. • It can be trained without label data and alleviates overfitting by continuously resampling training points throughout the training process.

A physics-informed machine learning framework for inverse multi-solution problems

Abstract

Journal

Organization

Cited Papers