Research On Fuzzy System And Neural Network And Its Applications In The Thermal Process

As to the widely applied PID control system, we usually use the linear transformation functions that are builded on a specific condition when designing the control system, they are hard to describe the whole character of the nonlinear thermal processes. It's basis to further improve the performance of the system that building a model fits the object in all working conditions. Fuzzy system and Neural Network are suit for such situation because of their special traits, and the control strategies based on them also show their advantages.This paper discussed the applications of the fuzzy system and NN in the thermal field. It is made up of four parts: the first part introduced the basic knowledge of the fuzzy system, interpreted its mechanism and the designing process, and do some melioration to the traditional fuzzy controller, so that its parameters should adapt to the changing conditions, there also is a instance cited to convince its effectivity; the second part of the paper researched the thermal system identification based on NN. Firstly, we illuminate the basic theory of the neural network, and the way using it to identify the dynamic system. The analysis emphasized on the NN's structure, working style, the hidden nods and their distribution's influence to the precision of the identification. Secondly, we united the NN models into one whole conditional model through fuzzy system. Discussed the performance of the united model, made out that its has higher precise near the sampling conditions, moreover lower when the conditions is relatively far from the sampling ones, indicate that method to improve the NN's effect is two: increasing the numbers of the sampling conditions, or changing its working style; The third part discussed the sameness and difference between the fuzzy system and NN, illustrate the possibility and necessity of their combination. Referred to an example for the fuzzy-neural network, explained its structure, and applied it in identifying the thermal object; The fourth part researched several control strategies based on NN, such as direct inverse control, internal model control, model reference control and model predictive control. Interpreted their applying conditions, working and design method. This paper stressed in the thermal control system designed by model predictive method. This method found on the NN predictive model, charactering by rolling optimizing and feedback revise, it offered a linear quadratic function including the adjusted variables's errors and the adjusting variables's increases, and to find the optimized control signal in the predictive field and control field. The simulation results show that this method has many advantages comparing to the traditional PID control.The modeling method applied in the paper has well generalization and precision, the control system realized to the model also shows its fineness, they are very valuable for engineering application.