Neural Network Control And Application On A Class Of Non-Linear System

In production process there is a class of non-linear system with long time delay andits character are the long lag time and slow response speed. The typical representative ofthis class system is the temperature control system of industrial resistance heating furnace.To solve the control problem in this system, the method adopted artificial neural networkhas been proposed, and it has also been used to the practical control for resistance heatingfurnace.It introduces the main character and development progress of neural network firstlyin the paper, and a new kind of associative memory neural network and its algorithm isproposed to solve the difficult problem how to establish the model of non-linear systemwith long time delay. Through adopting the decaying factor of associative memory, it hasraised the ability of identification to the non-linear system. And it provides the analysis ofalgorithm convergence property and how to get the decaying factor of associative memory.At last, compared with the neural network of Elman, associative memory neural networkhas good identification and extensive ability. Then it gives the introduction of the situationof neural network in control filed. It points out that fuzzy neural network control isdevelopment trend of neural network control. It proposes a new fuzzy neural network withinverse identification structure and introduces its algorithm. In this scheme it has thethree-layer fuzzy neural network controller to satisfy the early control fast speed request;In order to solve the steady-state error in the later stage, adopting the new type associativememory neural network, through the inverse identification method, as the controller tocompensate the control output. Adopted the coordinated control factor, adjusting theoutputs of the fuzzy neural network and the neural network inverse controller, thecontrolled system is under the best control state. In order to solve the model and controlproblem of multi-variable system, it also adopts associative memory neural network andfuzzy neural network with inverse identification structure to realize the identification anddecoupling control for multi-input and multi-output system in the paper. It has introducedthe multi-input and multi-output system decoupling method with inverse identificationstructure detailly. The principle and quality of neuron decoupling method are alsoanalyzed in detail in the paper and identification simulation research and decouplingsimulation research on the two inputs and two outputs non-linear system are also introduced.In the paper it analyzes the characteristic in details of temperature control system ofindustrial resistance heating furnace and introduces its structure. Adopting the methods ofidentification and control proposed in the paper to realize the practical system'sidentification has a good result.