Research On Fault Diagnosis Of Hydraulic System Of Tamping Car Based On HMM-SVM

The tamping vehicle plays a very important role in the daily maintenance of railway.The normal maintenance of the tamping car ensures the safe operation of railway,so it is very necessary to ensure the tamping car in the normal working state.Among the components of the tamping vehicle,the hydraulic system,as a key link in the tamping vehicle operation,provides power source for the normal tamping operation of the tamping vehicle.All kinds of abnormal conditions will inevitably occur in the process of the hydraulic system.If the fault of the hydraulic system is not diagnosed and eliminated in time,it will lead to the abnormal operation of the hydraulic system,and even lead to the failure of the tamping car.Therefore,the fault diagnosis research on the hydraulic system of tamping vehicle is a very important subject,and also an important supplement to the natural science foundation research of the tutor.According to the characteristics of multiple and mixed vibration signals of hydraulic system fault,this paper analyzes the noise reduction processing that will greatly hinder the follow-up work,and researches and proposes an improved hydraulic system fault diagnosis Model based on Hidden Markov Model(HMM)to optimize the Support Vector Machine(SVM).This paper to improve the EMD threshold noise reduction method for vibration signal de-noising processing,on the basis of the noise reduction processing,using Principal Component Analysis(PCA),Principal Component Analysis,PCA)after processing of the vibration signal feature extraction form feature vector,to extract feature vector input to the HMM model for training,respectively,for different categories of the HMM model and SVM model.In view of kernel function in SVM model,v-fold cross verification method is adopted to optimize its related parameters,and a better SVM classifier is obtained.Since THE HMM model plays a strong role in the similarity reflection between different samples of the same kind,after using HMM to optimize SVM,more accurate fault diagnosis results can be obtained under the condition of small samples.This article adopted the viterbi algorithm for computing the probability of observation sequence,in order to made the result more accurate,is to minimize the interference by other factors,this article observed sequence,sorted from big to small and large probability in addition,when the rate is not less than the sum of 90%,will remove the residual vector sequence,vector sequence of the vector of main component.Based on the results of previous sequence processing,HMM model and SVM model were connected in series and parallel,respectively.After SVM diagnosis and recognition,simulation results of hydraulic system fault diagnosis were obtained.The simulation results show that the accuracy ratio of serial HMM-SVM model and parallel HMM-SVM model in fault diagnosis of hydraulic system is higher than that of single HMM model or SVM model.Compared with the parallel HMM-SVM model,the serial HMM-SVM model is characterized by simpler operation,faster operation speed,stronger anti-interference and lower complexity.Therefore,this paper proposes to use the serial HMM-SVM fault diagnosis model with the same efficiency.