Physics-Informed Neural Networks for Transient Response

Project Details

Student(s): Reem Shehayib
Advisor(s): Dr. Dani Tannir
Department: Electrical & Computer
Academic Year(s): 2024-2025

Abstract

The advancement and race of integrating Artificial Intelligence and its branches of Machine Learning and Deep Learning have indubitably shaped the etiquette in approaching the design of several engineering systems. Considered to be multilevel complexity systems, electrical engineering systems demonstrated by various designs of electrical and electronic circuits can introduce several levels of complexities that hinder the path to analysis and study of their solutions. Specifically, complex circuits demonstrated by multiple differential equations can reasonably lead to complexities in time and accuracy while solved analytically in the frequency or time domain. Nevertheless, with the presence of Physics-Informed Neural Networks (PINNs), a scientific supervised learning architecture that deals with differential equations, we decided to use these PINNs to study a circuit’s transient response. Our PINN architecture allows the representation of the system of differential equations governing the relative circuit