Microwave loaded cylindrical cavity modeling using knowledge-based neural networks

Abstract

A knowledge-based neural (KBN) model of a microwave-loaded cylindrical metallic cavity is presented in this paper. The considered cavity is loaded with a homogeneous dielectric layer located on the cavity bottom. Unlike the model based on a classical multilayer perceptron (MLP) network, the proposed KBN model includes an existing partial knowledge about the resonant frequency behavior of the cavity, yielding more accurate determination of the resonant frequencies. A comparison of MLP and KBN models, as well as an advantage of using the KBN model, is given through an example referring to the experimental cylindrical metallic cavity with a circular cross section. © 2005 Wiley Periodicals, Inc. Microwave Opt Technol Lett 46: 580–585, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop.21057