Research On Dynamic Imbanlance Detection Method Of The Cardan Shaft In CRH5

With the rapid development of high-speed railways in China,the running safety has become an important research content of high-speed trains.CRH5 transmission system is composed of traction motor,gear box,cardan shaft and safety device,and transfers the driving force and torque of the motor to the gear box through the cardan shaft.The service environment of the cardan shaft is complex,it is easy to cause eccentricity in the running process,which will cause the dynamic imbalance,the vibration of transmission system intensified,and then affect the train running performance.It is of great significance to ensure the safe operation of the train by conducting in-depth research on the detection of the state of the cardan shaft and fault diagnosis of the dynamic imbalance.In this paper,dual-tree complex wavelet packet transform(DT-CWPT)and least squares support vector machine(LS-SVM)are applied to dynamic imbalance detection,and a dynamic imbalanced fault diagnosis model is proposed.The effectiveness of the proposed method is verified by bench test vibration data and measured line data.First of all,this paper analyzes the cardan shaft motion form and the mechanism of dynamic imbalance,studies the feature of dynamic imbalance.In order to solve the problem that the cardan shaft vibration signal is nonstationary and fault feature extraction is difficult,dual-tree complex wavelet packet analysis method is proposed.It is proved by simulation signals that the dual-tree complex wavelet packet transform method is compared with the discrete wavelet packet transform method can better guarantee shift-invariance,anti-aliasing and reduce energy leakage,and provides an effective analysis method for cardan shaft fault features.For the problem that the dual-tree complex wavelet packet filter bank cannot be completely cut off,an improved method is proposed.In the process of decomposition and reconstruction,two operators are added to eliminate frequency components which are outside the theoretical band.The effectiveness of the improved method is demonstrated by the filter characteristic curve.Next,taking the bench test data as example,the improved dual-tree complex wavelet packet method is used to determine the time-domain and frequency-domain feature parameters that can reflect the cardan shaft dynamic imbalance,construct the feature vector,and use principal component analysis to reduce data dimension.Finally,using the least squares support vector machine to classify data and taking the bench test vibration data and line data as examples,comparison of support vector machine and least squares support vector machine for classification performance.The result shows that the least squares support vector machine classification method can effectively reduce the classification time based on the accuracy.The classification accuracy of the least squares support vector machine to the bench test data is 100%,to the measured line data is 97.5%.