Fuzzy model structure of complex systems and its recursive identification method (Q2704896)

The paper deals with an identification procedure for fuzzy rule-based models composed of rules of the following form NEWLINE\[NEWLINE\text{if }z_1\text{ is }F_1\text{ and }\dots\;z_n\text{ is }F_n\text{ then }y(t+ 1)= A(q^{-1}) y(t)+ B(q^{-1}) u(t)+ CNEWLINE\]NEWLINE with NEWLINE\[NEWLINEA(q^{-1})= A_1+ A_2q^{-1}+\cdots+ A_n q^{-n+1},\quad B(q^{-1})= B_0+ B_1 q^{-1}+\cdots+ B_{n-1} q^{-n+1}.NEWLINE\]NEWLINE The membership functions in the antecedent part of the rules are bell-shaped and optimized by means of fuzzy clustering. A recursive LSE estimation procedure is discussed. Some numerical examples are also included.