Fuzzy local linearization and local basis function expansion innonlinear system modeling

摘要

Fuzzy local linearization is compared with local basis functionnexpansion for modeling unknown nonlinear processes. First-ordernTakagi-Sugeno fuzzy model and the analysis of variance (ANOVA)ndecomposition are combined for the fuzzy local linearization ofnnonlinear systems, in which B-splines are used as membership functionsnof the fuzzy sets for input space partition. A modified algorithm fornadaptive spline modeling of observation data (MASMOD) is developed forndetermining the number of necessary B-splines and their knot positionsnto achieve parsimonious models. This paper illustrates that fuzzy localnlinearization models have several advantages over local basis functionnexpansion based models in nonlinear system modeling