Space-mapped neuro-fuzzy optimization for microwave device modeling

Four modeling methods of microwave devices using multiple neuro-fuzzy inference systems (MANFIS) based on space-mapping (SM) approach are presented. In contrast with previous works using conventional mathematical tools and artificial neural networks, the proposed SM-based neuro-fuzzy modeling methods can construct nonlinear multidimensional mapping functions based on both human knowledge in the form of fuzzy if-then rules easy to understand by human beings and stipulated input-output data pairs, without employing precise quantitative analyzes. Optimization by microgenetic algorithm is used to find nonlinear multidimensional mapping functions for the modeling methods that present a singular system. These methods are applied to a shielded microstrip line and the obtained models are discussed. These space-mapped neuro-fuzzy models could be integrated in a toolbox of any commercially available computer-aided design tools for radio frequency/microwave circuits. © 2007 Wiley Periodicals, Inc. 10.1002/mop.22460.