Microwave applicators modeling — Alternative approaches based on neural networks incorporating domain knowledge

Abstract

Alternative and more efficient approaches for microwave applicators modeling, in respect to conventional numerical techniques, are presented in the paper. These approaches are based on artificial neural networks incorporating previously acquired partially knowledge about problem domain. Neural models, created using the suggested approaches, provide the similarly accuracy as numerical methods but performing much faster and less memory demanding simulations. A short review of neural models developed for microwave cavity applicators will be given along with their comparison through an example referring to the experimental cylindrical metallic cavity with a circular cross-section.