Research On Degradation State Identification And Prediction Of Switch Machine Based On KFCM Clustering

The switch machine is one of the basic equipment of railway signal.Because it is installed outdoors,the working environment is greatly affected by the outside world,and its complex structure,high conversion frequency,large usage and wide application range lead to frequent failures in the process of use.However,the fault diagnosis method for the switch machine can only be repaired after the equipment fails,which cannot predict the occurrence of faults,nor can it realize the real-time monitoring of the running state of the switch machine in the whole life cycle.In the working process,the switch machine does not only have two states of normal and fault,but will go through different degradation stages from normal to complete failure as the working time increases,which is a changing process.Once the state degradation of the switch machine occurs,its operating reliability will be reduced.In serious cases,the turnout may not switch.Therefore,it is of great significance to study the degradation state of the switch machine and predict its degradation trend for reducing the maintenance work of railway site.The main research contents of the thesis are as follows:Firstly,the S700 K switch machine is taken as the research object to analyze the relationship between its power curve and the conversion force.In view of the small change in the degradation characteristics,a method of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise was proposed,which combines with energy entropy and correlation coefficient.The degradation features with high correlation coefficient are selected to construct the degradation feature vector set that could reflect the trend of curve change.Secondly,Fuzzy C-Means clustering method is used to optimize the clustering center of the Kernel-based Fuzzy C-Means clustering method,which realizes automatic partition of degradation feature vectors,and recognize the degeneration state of switch machine.The clustering results are evaluated according to the classification coefficient and average fuzzy entropy,and the clustering effect of the proposed method was compared with that of various clustering methods.Finally,the Kernel Principal Component Analysis is used to achieve the feature dimension reduction of the degradation feature vector set,and the degradation performance index was obtained.The weight and threshold of Elman neural network were optimized by Particle Swarm Optimization algorithm,and the prediction model of degradation trend is established to improve the accuracy of the prediction model.Compared with the prediction results of BP neural network and Elman neural network,the accuracy of the proposed method was proved to be higher than that of BP and Elman neural network.Based on the whole process of the switch machine from normal to fault,the thesis established the degradation state recognition model and degradation trend prediction model,divided the running state of the switch machine into four stages scientifically and reasonably,accurately predicted the degradation trend of the switch machine based on Elman neural network,realized the health management of the switch machine,and contributed to the timely maintenance of switch equipment in railway field.In this way,the maintenance cost can be reduced to ensure the safe operation of trains.