Research On Fault Recognition And Prediction Method Of High-speed Train Shock Absorber Based On Machine Learning

As a key component of the train,the performance of the shock absorber directly affects the safe operation of the train.At present,the maintenance method adopted by the hydraulic shock absorber of high-speed trains in our country is regular maintenance.This maintenance method will not only increase the maintenance cost,but also may cause safety hazards due to untimely maintenance.In view of this,this article starts from the vibration data generated during the operation of the high-speed train,through data processing and statistical analysis,the artificial neural network is used to establish the fault diagnosis model of the shock absorber,and according to the characteristic parameters of the shock absorber in the performance degradation process Change trends give early warning to it.Since the research requires a large amount of vibration data corresponding to the known performance of the shock absorber,these data can be obtained through a simulation model.In view of the above situation,this article mainly carried out the following work:(1)According to the working principle and structural parameters of the single-cycle hydraulic shock absorber,the mathematical model is established in combination with the knowledge of fluid mechanics,and then the simulation model is established based on Simulink according to the mathematical model.The characteristic curve is obtained by changing the structural parameters of the shock absorber model,and the change law of the characteristic curve is analyzed to obtain the conclusion that the damping coefficient of the shock absorber will decrease during long-term use,which provides a theoretical basis for the setting of simulation model parameters.(2)Establish a high-speed train co-simulation model based on Simpack and Simulink and verify it.The verified model changes its running speed and the damping coefficient of the shock absorber and then simulates,records the vibration data of the shock absorber under various working conditions,calculates the performance indexes of the following vehicles under different working conditions,and explains the performance indexes through analysis The potential threat to the safe operation of trains caused by the degradation of the damping coefficient of the shock absorber is proved,and it is proved that it is necessary to take real-time detection of the performance state of the shock absorber,which further illustrates the necessity of this research.(3)Perform feature extraction on the collected simulation data under different working conditions.Feature extraction mainly includes time-domain,frequency-domain,and time-frequency domain features.The time-domain features can be obtained by directly calculating the collected signal;the time-domain signal is subjected to fast Fourier transform to calculate its frequency-domain features;in order to obtain the time Frequency domain features need to convert the data.Wavelet packet transform and aggregate empirical mode decomposition can convert time domain data into time-frequency domain data.According to the advantages of wavelet packet multi-level decomposition in frequency bands,the vibration energy of different frequency bands is calculated and used as The characteristic parameters in the time-frequency domain are used to decompose the vibration data into mode functions in different frequency bands using aggregate empirical mode decomposition,and the correlation coefficients of the mode functions under different working conditions are calculated and used as features in the time-frequency domain for subsequent faults Provide data support for diagnosis and early warning.(4)Reduce the dimensionality of the feature set and establish fault diagnosis and early warning models.The features before and after dimensionality reduction were used for model training and verification.The results showed the necessity of feature dimensionality reduction,and the parameters of the model were adjusted by the grid search method to further improve the accuracy of the model.Finally,the data is used to test the model.The test results prove the effectiveness of the method and model,and provide a research idea for data-driven shock absorber fault identification and early warning.