Research On Fault Diagnosis Method Of Switch Machine Based On ANFIS

As a basic piece of equipment in the railway signalling system,the switch machine plays a very important role in the safe and efficient operation of the railway transport system,and the railway maintenance department currently uses "planned maintenance" to ensure the normal operation of the switch machine.This maintenance method is characterised by a mismatch between maintenance cycles and equipment requirements,i.e.rutters with signs of failure are not completely eliminated within the current maintenance cycle;it is difficult to avoid over-maintenance problems,resulting in maintenance having a negative impact on the working conditions of the rutters instead;the inability to accurately determine rutter failures in a timely manner also makes it difficult to improve the efficiency of the maintenance department and to control the safety risks of maintenance personnel during the maintenance process.In order to improve the safety and maintenance efficiency of railway transport,the implementation of intelligent fault diagnosis for switch machines has been a hot topic of concern in the industry.The core of the research on switch machine fault diagnosis methods has two main aspects: one is how to effectively acquire,analyse and process fault data,which must contain sufficient information to reflect the characteristics of the fault.As the field of switch machine fault diagnosis continues to be studied in depth,the extraction of characteristic information for switch machine faults is no longer limited to electrical signals,but the vibration signals generated by faults can also reflect the actual working conditions of the switch machine.The second issue is to determine how the feature information can correctly characterise the corresponding fault,as different faults may extract similar feature information,and the mapping relationship between switch machine faults and feature information in practice has fuzzy characteristics,which is not well described by traditional classification algorithms.Therefore,this paper applies the Adaptive Neural Fuzzy Inference Network System(ANFIS)to the fault diagnosis of high-speed railway rutters.The main work accomplished in this paper is as follows:(1)This paper takes the ZDJ9 switch machine as the research object and analyses the action mechanism and failure mode of the switch machine.According to the characteristics of the vibration signals of the rutters,the fault vibration signals are first decomposed into Intrinsic Mode Functions(IMFs)of different frequency modes using Ensemble Empirical Mode Decomposition(EEMD),and the IMFs are selected according to Pearson correlation coefficients and The IMFs are selected based on the Pearson correlation coefficient and the cliffness.The Improved Time-Domain Multiscale Dispersion Entropy(ITMDE)algorithm is obtained by adding the standard deviation of time-domain features to the IMFs,and the signal reconstruction of the IMFs is carried out to extract the entropy values of the fault feat ures.(2)An adaptive neuro-fuzzy network-based switch machine fault diagnosis model was developed.The characteristic entropy values extracted from IMFs by ITMDE algorithm are taken as the model input,and the type of working conditions corresponding to the entropy values are taken as the output,and the model is set to generate fuzzy rules autonomously using fuzzy C-mean clustering method,and the mapping relationship between the characteristic entropy values and working conditions is described in fuzzy language.Dynamic Particle Swarm Optimization Algorithm(DPSO)is used to learn the parameters of the affiliation function of the model and complete the training update of the fuzzy rule weights.(3)The Improved Adaptive Neural Fuzzy Inference Network Sys tem(IANFIS)is obtained by combining the Approximate Linear Dependence(ALD)algorithm.To address the problem that the fault diagnosis model based on offline training of historical data does not perform well in the test of samples with changing character istics,the model parameters are updated by calculating the minimum approximation error between the fault chara cteristics of different samples.If the approximation error of the new sample meets the requirements,the offline model is updated to adapt the model to the actual situation,and finally the adaptive fault diagnosis function of the model is realized.Finally,the model is simulated and validated using fault data test samples.The results prove that the diagnostic rate of the model can meet the field requirements and the diagnostic effect is satisfactory,which verifies the effectiveness,accuracy and adaptability of the method.