The Research Of Fault Detection Of Servo System Based On Parameter Estimating

Servo systems are the most essential parts of CNC machines. Their performance has direct impact on the machining accuracy. Therefore, it's necessary to adopt the fault detection technology for minimizing fault time of servo systems.One servo system consists of several subsystems such as position control, velocity control and feedback of position. This paper builds the model of single axis servo system, and analyzes the different influence of each subsystem's parameters to the system performance. Then it applies the method of parameter estimation based on the model of parameters-subsystem parameters function to the fault detection of servo system.The model parameters-subsystem parameters function is built in this thesis. According to the requirements of system performance, the tolerant errors of each subsystem parameter are selected to define the standard of fault. Combining the tolerant errors with sensitivity analysis, this paper has summarized the connection between the variety characters of model parameters and the faults of subsystems. A characteristic vectors table to identify the fault has been set up.The proposed process of fault detection is divided into two steps. Firstly, offline analysis is carried out by using the prior knowledge of the system. Then online diagnosis of faults is applied by parameters estimation with the sampled data. The order of model can be estimated by accumulating matrix at first. An extended least square method based on SVD is presented to solve the problem of singular or close to singular constructed matrix. The thesis also presents the constitution of a key matrix, or the observation matrix, with the extended least square method when there is noise in output. With the least distance method and the table of eigenvectors of subsystem faults, the corresponding subsystem parameter can be found which has varied in the system. Thereby, the location of fault can be detected.The results of simulation have demonstrated that the method of fault detection based on parameter estimating proposed in the paper is valid and thus is promising in practice.