Development and Analysis of Chien-Physics-Informed Neural Networks for Singular Perturbation Problems

Abstract

In this article, we employ Chien-Physics Informed Neural Networks (C-PINNs) to obtain solutions for singularly perturbed convection-diffusion equations, reaction-diffusion equations, and their coupled forms in both one and two-dimensional settings. While PINNs have emerged as a powerful tool for solving various types of differential equations, their application to singular perturbation problems (SPPs) presents significant challenges. These challenges arise because a small perturbation parameter multiplies the highest-order derivatives, leading to sharp gradient changes near the boundary layer. To overcome these difficulties, we apply C-PINNs, a modified version of the standard PINNs framework, which is specifically designed to address singular perturbation problems. Our study shows that C-PINNs provide a more accurate solution for SPPs, demonstrating better performance than conventional methods.