Fast electromagnetic field simulation using a current-density- based physics-informed neural network.

Area of Science:

• Computational Electromagnetics
• Applied Physics
• Machine Learning in Physics

Background:

• Traditional numerical methods for electromagnetic field simulation and current density problems face challenges with efficiency and adaptability.
• Existing Poisson equation solvers can be computationally intensive and lack flexibility for complex scenarios.

Purpose of the Study:

• To introduce a novel physics-informed neural network (PINN) model for enhanced electromagnetic field simulation and current density analysis.
• To address the limitations of conventional methods by leveraging deep learning integrated with physical principles.
• To improve computational efficacy and flexibility in solving the Poisson equation.

Main Methods:

• Development of a PINN model incorporating a priori physical and mathematical knowledge.
• Application of the PINN model to two distinct scenarios: electromagnetic pulse simulation from laser-target interactions and electric field calculation for field-circuit coupling.
• Evaluation of the model's performance in terms of speed, accuracy, and adaptability.

Main Results:

• The PINN-based methodology demonstrated significant acceleration in computation speed compared to traditional solvers.
• The model achieved high accuracy, with a relative error of less than 1.4%.
• Enhanced adaptability to variations in current density was observed.

Conclusions:

• The developed PINN model offers a potent and efficient tool for electromagnetic field simulation and current density challenges.
• This research highlights the broad applicability of PINNs in electromagnetic field simulation and potential forecasting.
• The study validates the effectiveness of integrating deep learning with physical laws for complex computational problems.