Research On Fault Diagnosis Of Urban Rail Train Rolling Bearing Based On LCD And PSO-LSSVM

ABSTRACT:Nowadays. our country is moving into the period of rapid construction of urban rail transit, and a large number of new lines are being put into operation. As the carriers to carry passengers, the safe operation of urban rail trains has the direct relationship to the life safety and property safety of the passengers. Rolling bearing is one of the key components of urban rail train, and its running condition will directly influence the safe operation of urban rail train. Therefore, a real time and effective on-line monitoring and fault diagnosis method of rolling bearing can not only avoid train accidents, but also change the existing maintenance mechanism of rolling bearing: condition-based maintenance instead of preventive maintenance and fault maintenance. As a result, the operation costs can be reduced and the maintenance level of operation can be enhanced. The thesis takes the rolling bearing of the urban rail train as the research object and has a thorough research on rolling bearing’s fault diagnosis technology, including the processing analysis of vibration signal and feature extraction and fault pattern recognition. The main content of the thesis is as follows.1. For the nonlinear and non-stationary characteristics of rolling bearing’s vibration signal, the thesis builds a signal processing algorithm based on Wavelet Packet Analysis (WPA) and Local Characteristic-scale Decomposition (LCD) to realize the denoised signal and the analysis of failure mode. The WPA method can decompose the signal in multi-scale on the whole frequency band, and according to this speciality, it can effectively eliminate the noise of the signal and improve the signal-to-noise ratio. The LCD method adaptively can decompose the signal into several Intrinsic Scale Components (ISCs), and by analyzing the ISC’s envelope spectrum, the fault types of rolling bearing can be judged according to the apparent fault characteristic frequency on the envelope spectrum diagram. Compared with the commonly used Empirical Mode Decomposition method, the LCD method has fewer iteration times and faster computing speed, so it is more suitable for online analysis of vibration signal.2. Based on the time domain analysis, the frequency domain analysis and the time-frequency domain analysis of the vibration signal, the thesis systematically builds effective feature parameters which reflect the bearing’s condition information of urban rail train. Time domain parameters are closely related to the fault degree of rolling bearing, frequency-domain parameters can be used to diagnosis the fault type of rolling bearing, the size and distribution of wavelet packet energy spectrum parameters and ISC energy moments can be used as a judgment of rolling bearing’s fault diagnosis. To avoid information redundancy between the feature parameters and improve the speed of fault recognition, the thesis uses Kernel Principal Component Analysis (KPCA) method to optimize these parameters’dimension and gets the feature vector of the signal.3. Considering the accuracy of the fault diagnosis, the thesis puts forward a fault diagnosis method based on Least Squares Support Vector Machine (LSSVM) which is optimized by Particle Swarm Optimization (PSO). The LSSVM method has powerful learning ability and generalization ability, so it can effectively deal with small sample and nonlinear problem. Taking the feature vectors of vibration signal in normal bearing, bearing with outer race defect, bearing with inner race defect and bearing with rolling defect as the input of LSSVM model and using PSO method to optimize the parameters of this model instead of selecting the parameters blindly, so the fault diagnosis accuracy of rolling bearing based on optimized LSSVM model can reach97.5%. The study results show that the fault diagnosis algorithm based on PSO-LSSVM has higher diagnosis accuracy and higher computing speed, so it can effectively diagnose the fault of rolling bearing. On the basis of the above research results, an on-line fault diagnosis system of urban rail train running gear is developed, and it has been successfully used in Guangzhou metro trains, realizing the online fault diagnosis of rolling bearing.