A neural network modeling of microwave circuits on PBG structures

Abstract

A new technique is proposed to analyze microwave circuits on uniplanar compact photonic bandgap (UC-PBG) structures. This structure is a two-dimensional square lattice with each element consisting of a metal pad and four connecting branches. The UC-PBG structure can be used in the development of high-performance and compact circuit components for microwave and millimeter-wave frequencies. Usually, accurate analyses of PBG structures are performed through complex techniques, like FDTD. Lately, a growing interest was observed in the use of neurocomputational techniques to study the properties of electromagnetic devices and circuits. The main goal of this work is to develop a new and efficient technique to analyze microwave circuits with UC-PBG based on neural networks. The proposed neural network analysis for UC-PBG structures gave results in excellent agreement with measured results available in the literature. Besides, this technique provided a very good generalization, allowing the comparison with a set of measured results that were not used in the training process.