Simulation of Electromagnetic Scattering from Rough Surface based on the Machine Learning

The Method of Moments (MOM) is widely applied to solving electromagnetic scattering from rough surfaces due to its high precision, yet it suffers from significant computational time in generating and solving the target impedance matrix. Traditional Conjugate Gradient Method (CGM), on the other hand, encounters high computational costs when dealing with complex rough surfaces, requiring numerous iterations to converge to a satisfactory solution. To address this issue, this paper integrates machine learning with CGM, utilizing the supervised learning algorithm of Multi-output Support Vector Regression (MSVR) to predict the gradient direction and current distribution within CGM iterations. Compared to traditional CGM, this approach reduces the number of iterations, significantly shortens the time required to solve the matrix equation, and maintains sufficient accuracy.