Research On Performance Degradation Assessment For Train Rolling Bearings

The rail traffic, which has the characteristic of large capacity, high speed, safety, efficiency and environmental protection, has played a crucial role in an integrated transport network all over the world. As a key rotating component of the urban rail vehicle, rolling bearing have been increasingly significant for the rail vehicle’s stable and reliable operation. The rolling bearing always goes through a series of different stages of degradation from initial degradation to final failure. Therefore, if the rolling bearing performance can be accurate detected and assessed, it would be important to make the bearing maintenance plan well targeted and prevent the failure. In this paper, performance degradation assessment based on the above thoughts of proactive equipment maintenance is researched the main contents are as follows:1. Based on the research status about rolling bearing infrastructure, degraded mode, signal processing methods, and existing achievements in related fields were referenced. For varying degrees of completeness of the rolling bearing in rail vehicles, performance degradation assessment is systematically proposed.2. The extracted feature vector which should fully describe the degradation condition is difficult to realize. Therefore, in this paper, LMD(Local Mean Decomposition) is used to decompose the vibration signal, and the experiment shows better efficiency and identification compared with EMD (Empirical Mode Decomposition). Then feature vector is extracted form time-frequency. Principal component analysis can help reduce dimension. Finally, in order to reduce the dimensions and integrate the PF features, PCA is applied, and the principal components will be used as the final feature.3. For the situation of incomplete data, online performance degradation assessment method based on SVDD(Support Vector Data Description)is proposed, and self-adapting alarm technology is discussed. With this method, normal operation data of the rolling bearing is applied to train the SVDD to find a minimum super-sphere in the feature space, and quantify the degree of degradation according to the distance between the points to be tested and the center of the sphere. In the process of SVDD, kernel parameter and penalty value are chosen by DPSO. By using actual data of two rolling bearings, the above methods are tested. The experiment confirm that it can accurately quantify performance degradation and identify state changes even the data is incomplete, and make a timely alarm.4. For the situation of complete data, online performance degradation assessment method based on the ideological of segmentation vote is proposed for the first time. And segmentation vote and LSSVM and LSSVR are combined to realize identification degradation mode, performance degradation assessment and residual life prediction.The above methods can accurately identify and describe the performance degradation mode, and achieve good results in remaining life prediction by the whole life of rolling bearings verification.Finally, the method is generally verified by the other bearing of the same type,and functional structure of the train rolling bearing degradation performance assessment software developed based on above methods is briefly introduced.