Research On Fault Diagnosis Of Hydraulic System Of Tamping Truck Based On Fuzzy Support Vector Machine

Due to the hydraulic system of tamping machine operating in high strength and high load condition, some faults arising cannot be solved in time during repairing railway track, which will affect the smooth operation of other trains. Therefore, it is necessary to research the fault diagnosis methods of tamping machine hydraulic system. This paper studies the fault vibration signal de-noising methods and fault feature extraction method, and studies a fault classification method of tamping machine hydraulic system based on fuzzy support vector machine, aimed at achieving the fault classification accurately and rapidly when the faults of tamping machine hydraulic system occur, while use improved Particle Swarm Optimization algorithm to optimize the parameters of FSVM to improve the accuracy of classification.This article introduces the improved EMD threshold de-noising method, de-noising method based on hidden Markov tree model in dual tree complex wavelet domain and de-noising method of improved EMD threshold combined with hidden Markov tree model in Q-Shift dual tree complex wavelet domain, and on the basis of empirical research, which study a fault feature extraction method of tamping machine hydraulic system based on the Complete Ensemble Empirical Mode Decomposition combined with fuzzy entropy. Firstly, intrinsic mode functions of fault vibration signals were obtained by the empirical mode decomposition method. Secondly, the improved threshold method is used to de-noise from selected the first few major intrinsic mode functions in accordance with the continuous mean square error criterion. Once more, hidden Markov tree model in Q-Shift dual tree complex wavelet domain is used to second de-noise from the de-noised and remaining IMF components. Then, subsequent de-noised IMF components are reconstructed to produce the final de-noised fault signal. Finally, the IMF components of the final de-noised fault signal are obtained by the Complete Ensemble Empirical Mode Decomposition, and Fuzzy entropy algorithm is used to calculate the fuzzy entropy of the first few major IMF component which constitutes the fault feature vector. Experimental results show that the de-noising method of improved EMD threshold combined with hidden Markov tree model in Q-Shift dual tree complex wavelet domain can achieve better de-noising effect and extracted features by using the mentioned fault feature extraction method can well reflect faulty information.The tamping machine hydraulic system fault classification model and method based on fuzzy support vector machine are studied, while use improved Particle Swarm Optimization algorithm to optimize the model parameters of fuzzy support vector machine, aimed at improving the accuracy of fault diagnosis model. Firstly, calculate the Fuzzy membership degree of each sample point by fuzzy C-means membership function and so change the penalty parameters of each sample point to construct FSVM classification model. Then, the improved Particle Swarm Optimization algorithm is applied to select the best penalty coefficient C and kernel width σ for FSVM model, so as to construct the optimum fault diagnosis model based on FSVM, and use the model to classify the faults. Simulation results demonstrate that the method has a high recognition rate for the fault type of tamping machine hydraulic system.