Artificial neural networks modeling for computer-aided design of microwave filter

Abstract

Artificial Neural Networks (ANN) have recently gained attention as a fast and flexible vehicle for microwave modeling, simulation and optimization. This paper presents an example of ANN usefulness in microwave filter design process. The filter consists of four segments. Every segment is represented by electromagnetically trained neural network model (EM-ANN). Furthermore EM-ANN model maps a set of scattering parameters of coupled microstrip lines (CML) in terms of its geometric dimensions. A comparison of results from different simulation methods and measurements has been presented.