Research On The Method Of Performance Degradation Evaluation And Fault Prediction For The Gearbox Of The High-Speed Train

With the increasing mileage of high-speed trains,the condition monitoring of trains has become a top priority to ensure the safe and stable operation of trains.As the core component of bogies of high-speed trains,the research on the method of performance degradation evaluation and fault prediction for the gearbox is of great significance to ensure the safe operation of trains.This paper focuses on the extraction method of gearbox degradation characteristics,the method of gearbox performance degradation evaluation and the related method of gearbox fault prediction.Firstly,The paper simulates the vibration signals of the high-speed train gearbox.According to the data related to the bearings and gears of the high-speed train gearbox,the vibration signals of the gearbox at normal condition and at different degradation stages are simulated.The simulations of Vehicle model dynamics are performed in the dynamics software to analyze the effect of external excitation on the gearbox.The simulation results are combined to generate the vibration signals of the high-speed train gearbox under external excitation.The simulation data are provided for degradation feature extraction,performance degradation evaluation and fault prediction in later sections.A degradation feature extraction method based on improved MCKD with Morlet complex wavelet filtering is proposed for the extraction of degradation features of high-speed train gearboxes.The gray wolf optimization algorithm is used to optimize the key parameters of MCKD and realize the adaptive selection of MCKD parameters.The filtered signal obtained by inputting the simulation signal and the vibration signal of experimental data into the MCKD with optimized parameters is filtered by Morlet complex wavelet for secondary filtering,and the filtering results show the feasibility and superiority of the degradation feature extraction method based on the improved MCKD and Morlet complex wavelet filtering.A method based on wavelet packet singular spectrum entropy and learning vector quantization(LVQ)neural network is proposed for the performance degradation evaluation research of the high-speed train gearboxes.The method decomposes the gearbox vibration signal by the wavelet packet and calculates the singular spectral entropy to form the feature vectors.The training samples’ feature vectors are input into the LVQ neural network clustering model to establish a performance degradation evaluation model,and then test samples’ feature vectors are input into the established model to evaluate the performance degradation.The simulation data and experimental data show that the proposed method can distinguish the different degradation states of gearboxes and detect the degradation state earlier.For the study of fault prediction after performance degradation evaluation,three deep learning methods,BP neural network optimized by genetic algorithm,radial basis neural network and convolutional neural network,are used to analyze the vibration trend of the high-speed train gearbox and the performance degradation evaluation model of the bearing.The results show that convolutional neural network achieves better results than other deep learning methods in predicting the vibration trend and performance degradation evaluation model,which provides a reference basis for the high-speed train gearbox fault prediction.