Fault Diagnosis Method Of Rolling Bearing Based On Time-frequency Analysis And VPMCD

As one of the most critical parts in mechanical system, rolling bearing may affect the whole mechanical system operation once the failure occurs.Therefore, fault diagnosis and condition monitoring of rolling bearing has become increasingly important. During the process of fault diagnosis of rolling bearing, mainly contains two aspects: signal processing and pattern recognition. Signal processing is that complex non-stationary nonlinear signal which is collected from rolling bearing is decomposed, and the parameters sensitive to the fault feature are extracted as the features from single component of the decomposed signal; A new method for pattern recognition based on Variable predictive model-based class discriminate(VPMCD) is proposed. In this method, intrinsic relationship between features is made full use in order to establish variable prediction model, and then classification is implemented through prediction model. Due to that the features extracted from rolling bearing tend to have certain inherent relationship, and have significantly difference between the different categories or systems, therefore, VPMCD method can be applied to fault diagnosis of rolling bearing. Based on this, in this paper, the fault diagnosis of rolling bearing based on time-frequency analysis and VPMCD is studied, the main research content is as follows:1. EMD, LCD, ASTFA method are studied and improved. Through the analysis of the simulation signal, the advantages and limitations of the above methods are analyzed, as to provide theoretical basis for the follow-up of rolling bearing fault diagnosis.2. The basic theory of VPMCD method is studied, the widely used pattern recognition methods, such as neural networks and support vector machine are compared, at the same time, ELCD method combined with VPMCD is applied into the rolling bearing fault diagnosis, and the experimental results verify the application feasibility of this method.3. Aiming at the defects of parameter estimation in original VPMCD, the unreasonable model selection method and the defect of lower recognition rate under the condition of smaller samples and multi-classification, BP neural network and WCPSO are used to improve VPMCD,and then BP- VPMCD and WCPSO- VPMCD method are put forward.4. The adaptive time-frequency analysis method combined with VPMCD and its improved method is applied to fault diagnosis of rolling bearing, effectiveness of this method is verified through experiment.