Safety Region Modeling Without Fault Data For Wheel Set Bearings Of High-Speed Train

Security issues have become the focus of railway, the high speed train plays an important role in the work of ensuring the safety and the efficiency of the whole rail system, and the safety of trains depends on the state of the important equipment. However, it is very difficult to obtain samples with the fault diagnosis from the important equipment. For this problem, the state detection can‘t be achieved using a traditional method without fault samples. It becomes an urgent problem that how to detect the security without the fault data.This thesis proposes a method to detect the wheel set bearings state based on safety region estimation. Firstly, the literatures on diagnosis of wheel set bearings and safety region estimation theory are surveyed, based on relevant work, the method of safety region estimation for wheel set bearings is proposed using support vector data description. Then, the safety region fundamental theories and concepts are presented. The method of state detection for wheel set bearings is proposed based on safety region. The detailed steps to realize this method are provided in this thesis. Finally, the boundary of safety region for wheel set bearings is estimated by support vector data description of single-valued classification tool. The key research are following:In order to select the trade-off parameter of bearing’s safety region based on support vector data description and consider the efficiency of selecting trade-off parameter, a parameter selection algorithm is proposed which can deal with the various complicated data. The main idea of this algorithm is to calculate the distance of each sample to the center of kernel space. The entropies of all samples are obtained. Finally, optimal parameters are selected. Experimental results show that the method can keep the high accuracy and improve the computational efficiency of selection for the parameter.The accuracy of safety region of wheel set bearings on high speed train depends on the effects of feature selection. Considering the effectiveness and redundancy of the features for a comprehensive result, an algorithm based on multi-metric fusion is proposed. The experimental results show that this method can reduce feature redundancy and the model complexity for improving the fault identification of safety region.For wheel bearing of high speed train, experiment has been designed to obtain data of normal condition, outer race fault, inner race fault, rolling element fault. The features are extracted from these collected data. The data without the fault from wheel set bearings are investigated using the proposed algorithm of safety region estimation. Experimental results show that the proposed method has a good performance in the detection of wheel bearing without fault data.