Neural Network Control Research Based On Extension Logic

Extenics is a new subject. Formalized tool is used to research the rules and methods of solving non-compatible problems from the point of view of the qualitation and quantification. The theory support of extenics is matter-element theory and extension set theory. Its logic cell is matter-element. Extenics is often used to research conflicting problems and can be widely used in many fields. It has the application possibility in the aspects of creative thinking, decision-making and artificial intelligence.Neurocontrol is one of the subjects in autocontrol field from 1980s. Its development bases on great breakthrough in artificial neural networks research. Neurocontrol is a branch of robot control. It presents a new approach to controlling complexly unlinear, dispersing and uncertainty systems. Neurocontrol is successfully applied in lots of control systems. But many incompatible problems appear in the process of its control. Because the research object of extenics is the incompatible problems in the objective world, combining neurocontrol with extenics is the certain thing.In this paper I have mainly completed the works as follow:(1) The related theories of extenics and neurocontrol are carefully researched. The thought of combining extenics theory with neurocontrol is brought forward.(2) Matter-element model of neural networks is constructed. Its extension character and matter-element transformation are detailedly analyzed.(3) Some neural networks in common use and their correlations are discussed.(4) Matter-element model of Neurocontrol system's performance index is constructed. According to the demand of performance index matter-element neural networks matter-element in neurocontrol system are optimized and selected applying matter-element transformation theory and rhombus-thinking method.(5) The reverse mode neurocontrol performance of one kind of reversible system is improved. And a simulation example is presented.(6) Extension control is studied in earnest. Through a simulation example extension control is compared with some kinds of neural networks control. Thesimulation result on slow time-varying nonlinear system shows that extension control is superior to neural networks control.