Design Of GA-Based Neural Fuzzy Controller And Its Application

In the process of modern industrial control, fuzzy control is used more and more widely in the system with uncertain models. But for different control objects, the choices of control rules, membership function parameters and scaling factors, should be dissimilar too. They are very crucial for fuzzy control system's performance. And the central problem is how to acquire the parameters and rules in the design of fuzzy controller.Firstly, the paper studies neural fuzzy technology. Combining the two technology's merits, here we put forward a kind of neural fuzzy controller with full-net structure, which uses a five-layer neural network to realize all functions of fuzzy controller. Secondly, based on the study of advantages and disadvantages of standard genetic algorithms, the paper brings forward an improved genetic algorithm which adopts many improving strategies, such as multi-point crossover, regeneration operator, optimum maintaining, fuzzy control method used to achieve the values of probabilities of crossover and mutation. Then the improved algorithm is used to synthetically optimize the structure and parameters of the controller. At the same time, the corresponding neural fuzzy controller generates. The results of simulations and datum of experiments show that the optimized controller has good dynamic capabilities and strong robust performance.Finally, based on the mathematical model of single-wheel vehicle, the optimization method as well as the fuzzy neural controller is used in designing the automobile anti-lock brake system to fulfill the control of the slip ratio in the brake process. The results of simulations validate the feasibility and efficiency of the method in practice.