A Machine-Learning-Based Method to Accelerate the Design of SAW Filters at Different Frequency Bands

Surface acoustic wave (SAW) filters play a critical role as radio frequency components in the front-end module of electronic equipment. The conventional design methods of SAW filters heavily rely on researchers’ experiences and extensive software simulations, thus requiring substantial computational expenses. To overcome these limitations, this work explores the machine learning techniques to speed up the design of ladder-type SAW filters. A novel approach is proposed by combining the convolutional neural network (CNN) and cuckoo search (CS) optimization method. The method utilizes neural networks as surrogate models, replacing simulation software to predict the performance of SAW filters. This substitution aims to reduce the time needed for the entire filter optimization process. The incorporation of CS optimization method avoids the reliance of the optimization process on the designer’s experiences. Numerical examples are provided to demonstrate that the proposed method is efficient to design SAW filters with different frequency bands. Additionally, the method is adaptable to designing filters with diverse topologies, design standards, and varying numbers of resonators.