Neuro-fuzzy Controller In A Servo System

The combination of fuzzy logic and neural network becomes an attractive research field in recent years. Neural network specializes in system identification and self adaptation, while fuzzy inference system is good at reasoning and decision making. Combination of them can solve some practical engineering problems effectively. This paper is to design a new type of neural fuzzy controller to meet the requirements of speediness, accuracy and stability of the alternating current (AC) servo system preferably, which works under complex circumstances such as large inertia, variety load and the moment disturb.The main contents of this paper are as follows:(1) Introduced the related knowledge of the hardware of the AC servo system, and collected the input and output data needed for the system identification.(2) Obtained an appropriate model of the AC servo system, through the method of system identification with neural network. Introduced the theory of BP network in the first instance, and then, identifies parameters of the system model with the algorithms of Gradient, conjugate gradient and Levenberg Marquardt (LM) respectively. Lastly, after analysis of the results, chose the LM algorithm to get the model of the system.(3)Under the circumstance of no enough experience knowledge, got the control rules with expressions method, and then completed the design of fuzzy controller for the AC servo system .Based on Matlab simulation, researched the influences of response characteristic of the system brought by the change of the parameters of fuzzy controller.(4) Proposed a method using Neural-fuzzy controller to improve the system performance. In the first instance, designed a BP Neural-fuzzy controller which improved the system performance obviously. Then, using GPFN to study the fuzzy "sample", combine the parameters adaptive method to control the system. The simulation results showed that the Neural-fuzzy controller satisfied all the requirements of AC servo system. The control method proposed in this dissertation is feasible for the control of AC servo system.