In this paper, a variable-structure Fuzzy Neural Network is designed and applied as the decoupling compensator, to cope with the problem that the structure of a usual FNN is too complicated and a FNN is too difficult to be constructed. To design the decoupling compensator, a methodology called subnet model is employed and the corresponding recursion algoritm is brought forward which effectively handles the above problem. Finally, the approach is illustrated and tested to application to a two-in and two-out system for simulation and decoupling with satisfing results. |