Accurate modeling of microwave devices using kriging-corrected space mapping surrogates

We present an accurate modeling technique that exploits standard space mapping (SM) as a trend function and the correction layer implemented with kriging. The kriging process allows us to efficiently utilize all available fine model data (not possible in the standard SM approach) and to obtain accuracy better than both the SM and recently published combinations of SM and radial basis function interpolation and fuzzy systems. We present application examples of design optimization and statistical analysis of several microwave filters. Copyright (C) 2011 John Wiley & Sons, Ltd.