Research On Fault Diagnosis Of Hydraulic System Of Tamping Vehicle Based On Correlation Vector Machine

As the tamping machine plays more and more important role in the railway track maintenance,the demand for diagnosis and maintenance of tamping machine's hydraulic system is becoming more and more urgent.The hydraulic system of tamping machine is the key part of the power transmission,which is always under high load during tamping machine's work.In this case,the hydraulic system cracked frequently which will have negative impact on the normal operation of trains if we can not fix the malfunction in time.Based on the above consideration,troubleshooting in time is a very important research direction to reduce breakdown ratio.In this paper,we utilize the EEMD and multi-scale entropy to extract the malfunction' s features with which relative vector machine(RVM)is adopt for the subsequent multi-class classification problems.To further improve the performance of the model,we use IAGA to optimize the kernel parameters of the relative vector.Since the hydraulic system's fault vibration signal always shows the cycle non-stationary characteristics and its complexity within feature frequency band varies,how to extract the features that can maintain the signal's characteristics is very challenging.To solve this problem,in this paper we use EEMD and MSE to extract the malfunction ' s features.The EEMD not only maintains the advantage of EMD but also eliminates the modal aliasing problem,which can improve the accuracy of signal analysis.On the other hand,the multi-scale entropy introduces the scale factor to reflect the information from different scales of the vibration signal,which can more clearly and intuitively to distinguish the different fault states of hydraulic system.The kernel function is a very important element in the relative vector machine and it directly determines the final diagnostic accuracy.To improve the performance and generalization ability of RVM,we adopt the IAGA to optimize the kernel function,which makes full use of IAGA's good global search capability and fast convergence speed to adaptively select the optimal kernel function's parameters.In this paper,we propose a binary tree based RVM to classify the different fault states of tamping machine ' s hydraulic system.In our model,we utilize the IAGA to optimize the parameters of kernel functions to select the optimal parameters for the RVM model.To verify the performance of our method,simulation experiments are carried out and the results suggest that our method always identifies the fault categories accurately.