Research On Model-Based Nonlinear Fault Detection Approach For Servo System

The model-based approach has been widely used in condition monitoring and fault detection of mechatronics servo system. As there is strong nonlinear property in the system, the linear approach at the state point has defect and can bring negative influences on detection and diagnosis accuracy. The research on nonlinear fault detection and diagnosis has been paied more and more attention.The main topics of this dissertation are focused on the model-based fault detection of the nonlinear mechatronics servo system. The main contents and results are as follows:1. The fault detection and diagnosis nonlinear methods have been summarized. The problems required to be researched deeply have been put forward.2. As to shortage of EKF (Extended Kalman Filter), the method of fault detection based on UKF (Unscented Kalman Filter) has been studied, and the arithmetic of fault diagnosis has been presented. The simulation on the permanent magnet synchronous motor proves the effect of the method.3. For the strong nonlinear system which is difficult to build analytical model, a fault detection and diagnosis method has been presented based on LS-SVM (Least Squares Support Vector Machine).The nonlinear relation between inputs and outputs is modeled by LS-SVM approximation. The residual by comparing the forecast output of the model and the actual output of the system is to detect fault. The simulation proves the effect.4. In order to evaluate the efficiency of the techniques proposed above, a BIT system of one mechatronics system is implemented. The electrical-control subsystem is chosen to demonstrate. The experimental result shows that BIT performance is improved after using the method presented in this paper.