Implementation of Maxwell's equations solving algorithm based on PINN

In this paper, we use a physically informed neural network (PINN) to solve the Maxwell's equations in time domain. PINN is a kind of physically constraint network, which can be used to solve various problems related to partial differential equations, and provide a way for solving multi-scale problems in multi-physical. When the network is trained, the Maxwell's equations and the associated initial and boundary conditions are set as constraint conditions, and the iterative calculation of electromagnetic (EM) field is transformed into the optimal training of the model, which can realize high-precision prediction of electromagnetic field with fewer neural network layers and neurons. To verify the implementation of the algorithm, the results of the proposed method are compared with those of the analytical solution. The experimental results show that the proposed method can predict the complex behavior of Maxwell's equations accurately, which proves the effectiveness and practicability of the proposed method.