Automatic MoM Source Integral Quadrature Selection via a Machine Learning Approach

Abstract

In this paper, a new technique, based on machine learning (ML) and dimensionality reduction, is proposed for drastically improving the performance in the evaluation of the singular and near singular potential integrals in the method of moments (MoM). The MoM source surface integral is first reduced to a line integral via a dimensionality reduction method, and, then, an ML algorithm is trained on a set of line integrals evaluated with Gauss-Legendre (GL) quadrature schemes of different orders. Finally, the trained ML algorithm is used to determine the minimum number of GL sample points and weights required for each potential line integral to get the requested accuracy.