Parametric Modeling of Four-Pole Waveguide Filter Integrating Neural Networks and MOR-Based Rational Transfer Functions

Neuro-transfer function (neuro-TF) methods are increasingly applied in the parametric modeling of microwave passive devices. Current neuro-TF methods based on poles/residues have the problem of poles/residues discontinuity when the geometry varies widely, and neuro-TF methods based on poles/zeros form require specific algorithms for ordering or poles/zeros matching. The neuro-TF method based on rational transfer functions proposed in recent years solves the above three problems simultaneously. The purpose of this paper is to apply the MOR-based neuro-TF method with rational transfer function to model a four-pole waveguide filter.