A Deep Learning based Macro Circuit Modeling for Black-Box EMC Problems

In this paper, a deep learning-based macro circuit model approach for black-box electromagnetic compatibility (EMC) problems is proposed. The concept of the partial element equivalent circuit (PEEC) method is deployed in constructing the circuit topology in the full-space mesh of a black-box device. The mesh-based circuit model can serve as a powerful tool in solving the emission and immunity of the system-level EMC problems. A physics based deep neural network (DNN) is designed and optimized with the electromagnetic and circuit theories. The approach is validated by a proof-of-concept numerical example. The training and validation data are obtained by solving simplified PEEC models of randomly generated routes on a pre-defined mesh set of a black box problem. Good agreement and efficiency are observed.