Feature-based Surrogate-assisted Harris Hawks Optimization Algorithm for Microwave Filters

This paper proposes a feature-based surrogate-assisted Harris hawks optimization (FBSAHHO) algorithm for microwave filters. The method is constructed in three steps. Firstly, the vector fitting technology is used to extract the features (parameters of a transfer function) of the electromagnetic (EM) responses. Secondly, artificial neural networks are trained to establish the relationship between design parameters and the extracted features. Thirdly, a surrogate is generated using the filter frequency responses recovered by substituting obtained features into the transfer function. Lastly, Harris hawks optimization algorithm is adopted to find optimal solutions using the surrogate,. A fourth-order coaxial cavity filter is used to verify the proposed method. Combining Harris hawks optimization with surrogate-assisted modeling algorithm demonstrates faster and better convergence than other surrogate-assisted and direct optimization algorithms for filter design.