Inverse Design for WB-BGA Package Structure by Deep Learning

摘要

During the package design process, a lot of simulation even measurement need to be carried out to ensure the package can fulfill regulatory radiation standards. In general, full-wave simulation is time-consuming when adjusting the structural parameters of the package to obtain the far-field radiation. In this paper, an inverse design method based on a convolution neural network is proposed to quickly optimize and design the package structure. The cascade model combining an inverse design network and a forward predicted network can help the convergence of the inverse network. The feasibility and effectiveness of the inverse design method are finally demonstrated through a set of true radiation data.