Research On Fault Diagnosis Of Hydraulic System Of Tamping Truck Based On Multi-features And OAO-RVM

Tamping machine is a kind of large maintenance machinery,that widely used in all parts of the country's railway construction and maintenance.The tamping machine to ensure that the design standard for railway track construction mainly b y tamping operation,and the work is used to ensure the smoothness of the railway.Tamping machine has a great contribution to ensuring the safety,stability and speed of railway transportation throughout the country.The maintenance of the railway has its own characteristics,that the maintenance o f the railway is completed during the interval between the train operations for most of the time.Therefore,in order to complete the maintenance of the railway maintenance work on time,it is necessary to ensure the normal operation of the large-scale maintenance machinery such as the tamping machine during operation.It has important research significance for the fault diagnosis of tamping machine.In this paper,the processing method of mechanical vibration signals,the feature extraction method of signa l faults,and the method of pattern recognition and classification are studied.A fault diagnosis model of tamping machine based on multi-features and relevance vector machine(RVM)is presented.The purpose is to ensure the normal operation of the railway maintenance work by identifying the fault diagnosis of the tamping machine hydraulic system.First,two kinds of intelligent recognition algorithms,support vector machine(SVM)and RVM,are studied.RVM is a new intelligent identification method based on SVM and Bayesian theory.It has made some improvements to the defects of the support vector machine and has some new advantages.It has better applicability to fault diagnosis of tamping machine.Second,the method of processing vibration signal was studi ed.The mechanical vibration signal is non-linear and non-stationary.Aiming at this kind of characteristic,we mainly study the empirical mode decomposition(EMD)and its improved method,ensemble empirical mode decomposition(EEMD),and deal with two kinds of vibration signal processing methods.EEMD solves modal confusion that EMD has.Through simulation experiments,it is proved that the EEMD is more suitable for the processing of mechanical vibration signals than the EMD.Third,the feature extraction method of mechanical vibration signal was studied.In order to better extract the features of various faults in mechanical diagnosis,a plurality of features are selected from the time domain and frequency domain of the mechanical vi bration signal to construct a feature vector.The feature vector constructed in this way contains a lot of redundant information or information that is not highly correlated with fault identification.In order to solve this problem,a feature reduction method based on K-L divergence and local linear embedding is proposed.Through simulation experiments,it is proved that the proposed dimension reduction method has better dimension reduction effect.Finally,the multi-classification method was studied.Fault diagnosis generally requires diagnosis and identification of multiple faults.Through the research on multi-classification methods,a multi-classification model that improving OAO-RVM is proposed.Through simulation experiments,it is proved that the proposed improved OAO-RVM multi-classification model has better accuracy.