The Design Of Intelligent Controller For Continuous Stirred Tank Reactor

Continuous Stirred Tank Reactor (CSTR) is widely used as chemical reactors in the industrial fields of petrolic manufacture, chemical industry, ferment, biologic pharmacy and so on. In order to ensure the reaction on the rails, we need to control inner parameter of CSTR such as temperature, pressure, consistence etc. The benefit of manufacture and quality of production will be deeply affected by the control effect.The reaction process of CSTR is a highly nonlinear, time-varying process with time delay. It's hard to build accurate mathematic model. The conventional PHD control can only control the linear process with assured model. When the model and parameters of the controlled plant changes, the conventional PID control cannot control well. It has disadvantages as inconvenient parameters modification and cannot emendate itself; hence we should find new intelligent control strategy.At present, the amalgamation of fuzzy system, neural network and genetic algorithms (GA) is gradually showing enormous latent capacity in the control field. Fuzzy neural network (FNN) ,which combines fuzzy system and neural network, has the virtues of being easy to express the knowledge based on rules and self-study capability. It has become a research hotspot in the field of control. Due to the global optimization capability, genetic algorithms attract more and more attention. The combination of the previous three kinds of control technology is an important research aspect of the academic subject of computational intelligence.This thesis mainly researches the design of fuzzy neural network controller based on GA for the plant of CSTR system. Firstly, the paper studies the basis principle of simple genetic algorithms and analyzes the existent problem of simple genetic algorithms. Then make some modification to improve GA performance for controller parameters optimization. The modified method is demonstrated by a typical test function. It is shown that the proposed algorithm has the virtues of fast convergence. Secondly, the paper researches the development course of fuzzy neural network (FNN) and the combination of fuzzy system and neural network. The structure and learning algorithm of fuzzy neural network based on standard model and T-S model are separately given. At last, fuzzy neural network controller based on GA is designed. Both the structure and learning algorithm are also given. Then analyze the principle of CSTR system and simulate PID control, fuzzy control and FNN control based on GA separately for CSTR system. The results comparison shows the proposed method has better effect and stronger anti-jamming.