| Railway transportation is one of the important modes of transportation in our country,and "Fault—Safety" is the core principle of railway equipment.With the increase of railway operating mileage year by year,the running speed and traffic density gradually rise,and the frequency and distribution of turnout have multiplied.Turnout is one of the key equipment of the railway system,when the turnout fails,the need for timely repair and maintenance,and how to quickly and accurately identify the fault type is very important to ensure the safety of the train.At present,the maintenance mode of switch equipment is the traditional way that the field maintenance personnel can check the fault type according to the switch action current or power curve collected by the microcomputer monitoring system.On the one hand,this method relies on the fieldwork experience,operational skills,and professional knowledge of the maintenance personnel,but this manual diagnosis method is inefficient,leading to frequent false and missed alarms.On the other hand,the collected current or power curve is prone to translation and distortion of the time axis.The efficiency of traditional fault diagnosis can not meet the needs of the rapid development of railway field.Therefore,taking the S700 K switch machine as an example,this paper proposes the application of an improved DTW algorithm and GA-LSSVM algorithm to switch fault diagnosis based on the above problems.The main work of this paper is as follows:Firstly,analyze the turnout structure and the principle of collecting turnout action power curve,and analyze the common turnout fault types and fault causes.The turnout power curve collected by the microcomputer monitoring system is taken as the original data,and the serial number label is given to each curve according to the fault type.The traditional method of power curve segmentation extraction is abandoned,and the characteristic parameters are extracted without segmentation.The study is carried out from the integrity of the curve,and the Mallat algorithm is used to decompose the original data once to exclude noise interference and retain the maximum complete information of the curve.Secondly,the improved DTW algorithm and GA-LSSVM algorithm are respectively applied to the turnout fault diagnosis.(1)Fault diagnosis based on improved DTW: first use the Mallat algorithm to decompose and denoise the original data,then determine the optimal threshold value of improved DTW through experiments,and finally calculate the regularization distance between the reference template set and the test sample,determine the fault type using the minimum regularization distance,and output the maximum possible fault number.(2)Fault diagnosis based on GA-LSSVM: Based on the decomposition and denoising of data samples by the Mallat algorithm first,the feature parameters are extracted using timedomain statistics to obtain the feature vector set of each curve.Then the 8-class feature vector set training samples are trained on the model with GA-optimized LSSVM parameters,and finally the test samples and GA-LSSVM are used for fault classification.Finally,the power curves of switches collected by a station of Lanzhou Bureau are used as data samples,and the classification effect of the two diagnostic methods is verified by Matlab.The simulation results show that both diagnostic methods have high diagnostic efficiency.In terms of diagnostic accuracy,the improved DTW fault diagnosis accuracy reaches 97.78%and the GA-LSSVM fault diagnosis accuracy is 93.33%,and the accuracy is improved compared with that before optimization.In terms of diagnosis speed,the average diagnosis time of the improved DTW is 0.2118 s,while the diagnosis time(including training time)of the GALSSVM model is much larger than that of the improved DTW method.Compared with other algorithms,GA-LSSVM has a shorter testing time,higher accuracy,and consumes within the allowable time. |
