An Optimization Method for Coupled-Line Bandpass Filters Using Transformer-Based Estimator and Multilayer Perceptron

A novel method for estimation and optimizing structural parameters of coupled-line bandpass filters (BPFs) using transformer-based estimator (TBE) and multilayer perceptron (MLP) is proposed. Once trained, the TBE can quickly obtain the predicted values of the BPF structural parameters from desired S-parameters, while the trained MLP can replace the time-consuming electromagnetic (EM) simulation process, and establish the mapping from structural parameters to S-parameters. In order to obtain the optimal structural parameters, the trained MLP is combined with genetic algorithm (GA) for fast optimization. To demonstrate the effectiveness of the proposed method, a BPF using three pairs of coupled-line microstrip structure is designed and fabricated. The experimental results demonstrate that the proposed method can quickly and accurately obtain the optimal structural parameters from the desired frequency response of the coupled-line BPF.