A Deep-Learning-Assisted Integral Equation Solver for Cost-Effective em Analysis of 3-D PEC Targets

Surface integral equation solvers are an attractive approach for analyzing scattering problems involving metallic targets. However, when applied to large-scale targets, they still face the challenge of high memory and computational resource consumption. In this work, we propose a deep-learning-assisted combined field integral equation solver (DL-CFIE) for cost-effective electromagnetic (EM) analysis of PEC targets. In this method, a deep-learning-based neural network (DNN) is designed and trained to replace the traditional preconditioners within the standard CFIE solvers. Numerical experiments are conducted to show the performance of the DL-CFIE, which show that the DL-CFIE solver achieves the same accuracy and convergence as standard CFIE solver while reducing the memory cost significantly.