ファジィ重み係数多目的非線形制御システムの一般化学習ネットワークによる構成 [in Japanese] Multi-Objective Nonlinear Control System by Fuzzy Weighting Factor Method based on Universal Learning Network [in Japanese]
Access this Article
Search this Article
Control probrems on the large-scale complicated systems such as optimal distributed control and optimal hierarchical control have been extensively studied in the 1960's. On the other hand, recently, multi-objective optimization problems have been studied in the field of Linear Programming and Non Linear Programming problems. In this paper, a new control methodology called fuzzy weighting factor method is presented that automatically determines the weighting factors of the criterion function of the large-scale multi-objective non linear systems which are made up of plural subsystems with their own criterion functions. Hitherto, weighting factor method, ε constraint method and fuzzy Min Max method etc. have been presented as the method of optimizing the multiobjective systems. Most of them are applied to static optimization probrems, and they have the problem to be solved such that as the methods are converted to linear programming problems, the membership functions of the plural criterion functions should be approximated linearly and the methods need a fairly amount of computations. In this paper, the controller's parameter variables in the dynamic multi-objective non-linear control systems are determined by using Universal Learning Network (ULN) and a new fuzzy weighting factor method. And the new method can solve the above mentioned problems and is also different from the usual fuzzy Min Max method in the point that the new fuzzy weighting factor method can arbitrarily adopt the membership functions. The Universal Learning Network is a super set of all kinds of neural network paradigms with supervised learning capability and in this paper both control objects and its controllers of the large-scale complicated systems are represented by ULN. Finally, characteristics of the fuzzy weighting factor method are studied by simulations on a nonlinear tank network.
- IEEJ Transactions on Sensors and Micromachines
IEEJ Transactions on Sensors and Micromachines 117(3), 298-305, 1997-03
The Institute of Electrical Engineers of Japan