Research On Fault Diagnosis Of Tamping Vehicle Tamping Device Based On KPCA-RVM

With the development of the railway,the safe of it become more and more important today.As the tamping machine which is always used in maintain the height of the railway track always in the design standard,the normal operation of the parts of this vehicle is a basic element to keep the efficiency of it.So when the fault happened in one part of it,it is quite important to find and report the fault information.The tamping device as the working part of the tamp train,has a great influence to the tamp train's working.With the wildly use of the machine learning technique,it has been used in fault diagnosis area.Using these algorithms in dimension reduction and pattern recognising is very common.Base on the above,this paper had done as the followed:(1)The research status of the fault diagnosing of tamp train in domestic and overseas,and the common faults of the tamp device have been searched had been studied.(2)The dimension reduction algorithms have been searched,and the kernel function based Principle Component Analysis algorithm has been selected to help to process the signal of tamp device.Introduce that combine kernel functions with traditional PCA methods to achieve better processing of nonlinear signals to remove noise and reduce dimension for having more accurate fault characteristics(3)The recognising algorithms which is used to make classification had been studied,and Relevance Vector Machine method is selected to be studied.Relevance Vector Machine is a new classifier based on Support Vector Machine.Compared with SVM,RVM has the advantages of sparseness and better probability.And the improved One Against One method is used to realize the construction of RVM multi-classifier on the basis of traditional RVM.(4)Combine the above KPCA method with the multi-class RVM algorithm to obtain a new diagnostic model and apply it to the diagnostic tamping device.Compared with the traditional PCA dimensionality denoising algorithm and the method of constructing SVM classifier,the model proposed in this paper has better fault diagnosis ability.