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

In the realm of electromagnetic field simulation and in solving current density-related issues, traditional numerical methods are often hindered by inefficiencies and limited adaptability. This study introduces a physics-informed neural network (PINN) model predicated on current density, addressing the shortcomings of conventional Poisson equation solvers and enhancing computational efficacy and flexibility. By harnessing the synergies between physical-mathematical insights and deep learning, we have constructed a neural network model imbued with a priori knowledge of physics and mathematics, thereby facilitating an efficient resolution of the Poisson's equation. The model is evaluated in two distinct scenarios: simulating electromagnetic pulses generated by laser-target interactions and calculating the electric field for field-circuit coupling integration. The empirical results indicate that the PINN-based solution methodology not only achieves a remarkable acceleration in computation speed but also maintains commendable accuracy, and the relative error is less than 1.4%, while bolstering adaptability to variations in current density. This research not only presents a novel and potent tool for addressing electromagnetic field simulation and current density challenges but also underscores the broad applicative potential of PINN in the domains of electromagnetic field simulation and potential forecasting.