Using Generative Adversarial Networks for Design of Passive Microstrip Filters
Chen Chen Nie, Qing Chen, Zhong Ren, Haipeng Wang, Yun Bo Li · 2024
For the microstrip filter design, traditional iterative optimization methods are time-consuming and resource-intensive. In this paper, a passive microstrip filter inverse design method based on generative adversarial networks (GAN) is presented. It can automatically design passive filters under given single-band and dual-band magnitude responses. The overall framework includes a GAN-based graphic generator, a convolution neural network (CNN)-based electromagnetic response predictor, and a genetic algorithm optimizer. By constructing a hybrid model of the filter unit structure and electromagnetic response, a solution that meets the requirements of the target frequency response is finally inversely designed. The inverse-designed filters are then verified by simulations and comparisons with the targets. The proposed method enables the one-time solution of the patch structure and parameters of microstrip filters, with the advantages of automation, fast speed, and low computational resource consumption.