Identification Of Magnetic Bearing System Based On Neural Network

The Magnetic Bearing, a new type of bearing with high performance, is widely used in the fields of mechanism, turbine, vacuum and so on, and recognized for its brilliant application foreground. At present, the control to the Magnetic Bearing is mostly based on the linear control theory, and its mathematic model is achieved by linearization and approximation. But the Magnetic Bearing is a complicated nonlinear system. Linearization of the bearing makes it hard to express the behavior of the system and to control it in higher precision.Neural network is proved superior in approximating the arbitrary nonlinear continuous function, without any experiential knowledge of the object. As a result, it has become a significant tool in dealing with nonlinear systems, and plays a more and more important role in system identification. In this paper, a kind of network is designed as a mathematical model of the actual Magnetic Bearing, based on the knowledge of the mechanical analysis of the Magnetic Bearing Rotor and the fundamental of neural network. It is proved by experimental results that the designed network, named by DTNN in this paper, can represents the feature of the input-output relationship of the actual Magnetic Bearing System, and the range of the errors is acceptable. Besides, by contrast, DTNN is more suitable to simulate the system than both BP and ELMAN networks. The research in this paper probes for a new way to model the Magnetic Bearing System. The main research work in this paper has been listed below: (1) Review of the fundamental of the system identification and its basic methods.(2) Collect data in experiment and wavelet de-noising technique is taken to remove the noise from the data collected from experiment, in order to make them appropriate for network training. (3)Design a kind of time-delay feedback network(DTNN), according to the character of input-output relationship of the Magnetic Bearing System. (4) Train the DTNN and two other kinds of network, BP and ELMAN, which have same number of layers and nodes with DTNN, using the new data. Experimental results based on new data provide a vehicle to compare the three kinds of...