An Efficient FDTD-CNN Method for Analyzing the Time-Domain Response of Objects above Layered Half-Space under HPEMP Illumination

In the analysis of high-power electromagnetic pulses (HPEMP) effects on objects above layered half-space, the traditional half-space computational electromagnetics (CEM) algorithms are often limited by high computational cost and the complexity of half-space Green's function. This study proposes a finite-difference time-domain (FDTD) method assisted by a convolution neural network (CNN), which can significantly reduce computational cost and complexity through a redivision of the computational region and the CNN-assisted prediction of the reflected wave. In this FDTD-CNN method, the FDTD region only contains the object, whereas the CNN prediction models are established for the reflected wave of the multilayered planar and the rough surfaces, avoiding the complex half-space Green's functions for characterizing the reflection effect. Numerical examples from typical half-space models are performed to illustrate the accuracy and efficiency of the proposed method. This article provides an accurate and efficient scheme that combines machine learning and traditional CEM to solve the electromagnetic coupling problem of objects over a layered half-space under an HPEMP illumination.