来源:《安全与电磁兼容》2020年    作者:Lijun Jiang, Heming Yao, Huanhuan Zhang, Wei E.I. Sha

Even though machine learning (ML) has demonstrated attractive capabilities in both regression and categorization applications, it is still a question if it could become a new power house for scientific computations in either speed or accuracy. We are particularly interested in the possibility of using ML for computational electromagnetics (CEM) in signal integrity (SI), power integrity (PI), electromagnetic compatibility (EMC), and electromagnetic interference (EMI) analyses. Through our recent efforts in developing novel ML based CEM algorithms, we found that there could be three major directions in using ML methods for SI/PI, EMC/EMI related modeling: assisted model generation, black box modelling, and algorithm renovation. In this paper, we discuss these three ways in developing machine learning based computational electromagnetics algorithms for SI/PI, EMC/EMI related modeling. Pros and cons of each method are provided based on our insight.

看看这些对你有没有帮助...