Skip to Main Content
This paper introduces a new artificial neural networks (ANNs)-based reverse-modeling approach for efficient electromagnetic compatibility (EMC) analysis of printed circuit boards (PCBs) and shielding enclosures. The proposed approach improves the accuracy of conventional or standard neural models by reversing the input-output variables in a systematic manner, while keeping the model structures simple relative to complex knowledge-based ANNs (e.g., KBNNs). The approach facilitates accurate and fast neural network modeling of realistic EMC scenarios where training data are expensive and sparse. To establish accuracy, efficiency, and feasibility of the proposed reverse-modeling approach, PCB structures such as perforated surface-mount shields and partially shielded PCB traces are treated as proof-of-concept examples. Although the modeling examples presented in the paper are based on training data from EM simulations, the approach is generic and hence valid for EMC modeling based on the measurement data. The approach is particularly useful in the electronic manufacturing industry where PCB layouts are frequently reused with minor modifications to the existing time-tested designs.