Enforcing hidden physics in physics-informed neural networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Nanxi, Wang, Sifan, Ma, Rujin, Chen, Airong, Cui, Chuanjie
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915667442139136
author Chen, Nanxi
Wang, Sifan
Ma, Rujin
Chen, Airong
Cui, Chuanjie
author_facet Chen, Nanxi
Wang, Sifan
Ma, Rujin
Chen, Airong
Cui, Chuanjie
contents Physics-informed neural networks (PINNs) represent a new paradigm for solving partial differential equations (PDEs) by integrating physical laws into the learning process of neural networks. However, ensuring that such frameworks fully reflect the physical structure embedded in the governing equations remains an open challenge, particularly for maintaining robustness across diverse scientific problems. In this work, we address this issue by introducing a simple, generalized, yet robust irreversibility-regularized strategy that enforces hidden physical laws as soft constraints during training, thereby recovering the missing physics associated with irreversible processes in the conventional PINN. This approach ensures that the learned solutions consistently respect the intrinsic one-way nature of irreversible physical processes. Across a wide range of benchmarks spanning traveling wave propagation, steady combustion, ice melting, corrosion evolution, and crack growth, we observe substantial performance improvements over the conventional PINN, demonstrating that our regularization scheme reduces predictive errors by more than an order of magnitude, while requiring only minimal modification to existing PINN frameworks.
format Preprint
id arxiv_https___arxiv_org_abs_2511_14348
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Enforcing hidden physics in physics-informed neural networks
Chen, Nanxi
Wang, Sifan
Ma, Rujin
Chen, Airong
Cui, Chuanjie
Machine Learning
Computational Physics
Physics-informed neural networks (PINNs) represent a new paradigm for solving partial differential equations (PDEs) by integrating physical laws into the learning process of neural networks. However, ensuring that such frameworks fully reflect the physical structure embedded in the governing equations remains an open challenge, particularly for maintaining robustness across diverse scientific problems. In this work, we address this issue by introducing a simple, generalized, yet robust irreversibility-regularized strategy that enforces hidden physical laws as soft constraints during training, thereby recovering the missing physics associated with irreversible processes in the conventional PINN. This approach ensures that the learned solutions consistently respect the intrinsic one-way nature of irreversible physical processes. Across a wide range of benchmarks spanning traveling wave propagation, steady combustion, ice melting, corrosion evolution, and crack growth, we observe substantial performance improvements over the conventional PINN, demonstrating that our regularization scheme reduces predictive errors by more than an order of magnitude, while requiring only minimal modification to existing PINN frameworks.
title Enforcing hidden physics in physics-informed neural networks
topic Machine Learning
Computational Physics
url https://arxiv.org/abs/2511.14348