This paper presents a hybrid quasi-ARMAX modeling and identification scheme for nonlinear systems. It is shown that a general nonlinear ARMAX system can be represented by using a constrained time-variant ARMAX model, whose coefficients are nonlinear functions of input-output variables of system. A hybrid quasi-ARMAX model are then obtained by embedding a group of fuzzy models in the coefficients of the constrained time-variant ARMAX model to describe the time-variant coefficients. One of the distinctive features of the hybrid quasi-ARMAX model is that its parameters to be estimated have useful explicit meanings which can be taken as advantages for setting initial values of parameter estimation and for improving the generalization ability of the model.
|ジャーナル||Research Reports on Information Science and Electrical Engineering of Kyushu University|
|出版物ステータス||Published - 1997 9 1|
ASJC Scopus subject areas
- Computer Science(all)
- Electrical and Electronic Engineering