Research Of Control Strategies Of AC Servo System Based On Neural Networks

In this paper,AC servo system with high performance was studied. According to the characteristic of this kind of system,it has a wide range and time-varing the rotating inertia and very strong interference moment,all these can change the parameters of the system in wide range. To the traditional control methods,it's difficult to obtain the satisfied dynamic characteristic。In the second chapter, the AC motor control system based on the Vector Control is mainly presented. The excellent designs of DC control system is also used to make up a simulation model of the AC servo system decoupling.To enhance the static and dynamic performances, single-neuron adaptive PID control system is introduced in the third chapter. A single-neuron adaptive PID intelligent controller is obtained by combing neural network with traditional PID regulator. By detecting the expected value and actual value online, the adaptive control algorithm can be developed,which are not affected by the accuracy of the identified model. This adaptive control algorithm needs not to identify the parameters of the system and the controller has simpler structure and control algorithms and can be realized online easily.In the fourth chapter, an improved mixed fuzzy-neural network switching control is put forward with much better control performance. Fuzzy control is one of most important intelligent controls, it has been developing rapidly and been used widely with remarkable achievement due to its strong robustness, succinct algorithm and real-time characteristic. But it is a pity .that traditional fuzzy control has disadvantages such as low precision and limited adaptability at the same time.In the fifth chapter, a new multi-controller coordinating control method is presented. Unlike the controller switched in some point, this method make it possible to switch one controller to another more smoothly with a coordinator. Satisfactory results are obtained as expect.