Research On Fault Diagnosis Method Of Rolling Bearing Of Tamping Car Based On RVM

Rolling bearing is a kind of mechanical parts which belong to rolling friction.It is also an easy to damage device for large mechanical equipment such as tamper and subway.Due to the influence of the complex factors of the outside world and itself,the damage degree of the bearing is greatly improved.Rolling bearing is widely used,but the failure of bearing equipment will cause significant security risks and social and economic losses.The usual methods can’t make effective diagnosis.Therefore,in order to minimize failure rate and save maintenance costs,the use of efficient fault diagnosis method to solve the problem of failure has become a very important research topic at this stage.This paper studies the vibration mechanism of rolling bearing in the different fault conditions and characteristics,introduces the basic principle of relevance vector machine,analysis and verification methods of feature extraction of rolling bearings,expounds the kernel function in the relevance vector machine(RVM)plays an important role in the design of the bearing fault diagnosis method of vector machine.A rolling bearing is normal,inner fault,outer ring fault and roller fault in four different states made specific analysis.First,signal feature extraction is the premise and foundation of fault diagnosis research,and a signal feature extraction method based on harmonic wavelet packet is proposed.The method of test data by harmonic wavelet packet decomposition,the wavelet coefficients is obtained;then the wavelet coefficients of different scales of the energy value of the energy normalized fault feature vector;secondly,the kernel function is an important factor in the relevance vector machine and optimization effect the method is not ideal problem,proposed a new algorithm based on adaptive genetic algorithm(IAGA)optimization method of RVM kernel parameter.The algorithm utilizes the advantages of IAGA algorithm,multiple population,fewer iterations and strong global search ability,adaptively selects the optimal kernel parameters,and applies it to the intelligent fault diagnosis of rolling bearings.The simulation results show that the parameters of the RVM kernel optimized by IAGA are better than the adaptive genetic algorithm(AGA)and the genetic algorithm(GA)to optimize the RVM kernel parameters.Again,for the rolling bearing fault causes complex belongs to the multi classification problem,and classification method of single difficult to effectively identify problems,introduces the voting mechanism and classification theory,RVM and improvement of "one to one"(OAO)method combined with the improved OAO-RVM model is constructed,the effective implementation of the fault identification and classification.The verification results show that with the traditional support vector machine(SVM)method,the number of RVM classification method of relevance vector required less fault diagnosis accuracy is higher;and the "general" relevance vector machine(OAO-RVM),"one to many" relevance vector machine(OAR-RVM)method.The improved OAO-RVM method is proposed in this paper can reduce the fault classification error rate,but also saves the time of fault diagnosis.