A New Method For Rotating Machinery Fault Diagnosis Based On Blind System Identification

Blind system identification is a fundamental signal processing technology, which retrieved a system unknown information from its output only. The method is particularly driven by an unknown input. This thesis makes more advanced research on the application of blind system identification method to the mechanical fault diagnosis, which combines the National Natural Science Foundation of China(No. 50775208) and the Natural Science Foundation of He'nan Educational Committee, China (No: 2006460005, 2008C460003). The content of each chapter of the dissertation is as follows:In the first chapter, the subject and the meaning of this thesis is set forth. This chapter summarizes the development and application of blind system identification method in mechanical fault diagnosis up to now, and then states the main points and the innovation of this thesis.In the second chapter, the basis theory of blind system identification method is introduced, and then the basic conception and main algorithm of blind system identification are also illustrated. The content of this chapter is the basis of the dissertation.In the third chapter, based on the deficiency of traditional blind identification of time series model, i.e. it is only adaptive to the stationary signal. A new method of non-stationary signal processing based on empirical mode decomposition (EMD) and Blind identification of AR model is proposed. The proposed method not only retains the advantage of traditional blind identification of time series model, but also extends the range of traditional blind identification of time series model. The proposed method can process the non-stationary signal. At the same time, combined the proposed method with the correlation dimension, the vibration signal is firstly decomposed into a finite number of stationary intrinsic mode functions (IMF) by the EMD method, the AR coefficient of every IMF is obtained by the blind system identification (BSI). The obtained coefficient used as the feature vector, correlation dimension as a classifier. The proposed method is compared with the traditional blind identification of time series model The experiment result shows that the proposed method is very effective.The fourth chapter, combining the advantage of order analysis and blind identification of time series model, a new blind identification algorithm of AR model on order domain is proposed. This method can transform the non-stationary signal sampled at the time region into the stabilized signal at the order region with re-sampling of the constant angular interval, therefore the AR Model coefficient of the order region signal is obtained by the blind system identification. In the proposed method retain the advantage of traditional time series blind identification, and extend the range of traditional time series blind identification. It can process the stationary signal.Simultaneously, a new method of non-stationary signal processing based on order region and blind identification of auto-regressive (AR) model is proposed. The system parameters will be received by order region technologies and blind identification. On this condition , the method of parametric Bi-spectrum Analysis based on order region and blind identification of AR model is used to mechanical fault diagnosis. Simulation and experimental study results illustrate that the method provides accurate estimates.The fifth chapter elaborates the basic theory and algorithm of bilinear system of blind identification. The bilinear system which has the advantages of simple strcture and disciplinary can describe many nonlinear processes more accurate. Therefore, the algorithmic research of blind identification of bilinear system model is proposed in mechanical fault diagnosis, and the simulation experiment is done by this method. At the same time, Applying this method to analyze the different fault signal in the rotor system, the experiment result shows that the proposed method is very effective.In the sixth chapter, the conclusions of the dissertation are summarized. Future research of kernel methods is prospected .