Identification Of Rail Corrugation Based On Variational Mode Decomposition And Support Vector Machine

As the artery of urban rail transit,the subway has the advantages of large passenger capacity,safety and speed,which greatly facilitates people’s daily travel.However,with the long-term service and operation of subway trains,there will be wave-shaped wear on the rails(rail corrugation).Rail corrugation can cause abnormal vibration of the track and train,reduce the service life of structural parts,and affect ride comfort,and even endanger the safety of vehicles.Therefore,how to efficiently identify rail corrugation and evaluate the degree of corrugation has important guiding significance for the safe operation of urban rail transit.The vibration impact caused by rail corrugation is directly reflected in the vertical vibration acceleration signal of axle box.At present,the research on the detection of rail corrugation based on axle box signal is generally based on the premise of ideal wheel.But the wheels often have polygonal wear during the long-term service of the train,which causes that the axle box signal contains not only the information of rail corrugation,but also the polygonal wheel information.In this thesis,a rail corrugation component extraction method based on variational mode decomposition is proposed.Using this method,the rail corrugation component in axle box signal is extracted under different wheel wear conditions,and the corrugation depth is quantitatively evaluated.The research work is as follows:(1)This thesis summarizes the basic methods of rail corrugation detection,discusses the theoretical feasibility of detecting rail corrugation based on the vertical vibration acceleration signal of axle box,establishes a dynamic simulation model of a subway vehicle using SIMPACK,and gives the simulation methods of rail corrugation and wheel polygonal wear.(2)Based on the vehicle dynamics simulation model,the dynamic response of the vehicle under the combined action of wheel polygonal wear and rail corrugation is analyzed.The qualitative detection method of rail corrugation based on short-time Fourier transform is studied.The time-frequency characteristics of wheel polygonal wear and rail corrugation under typical working conditions are analyzed.The wavelength and position of the rail corrugation are qualitatively identified with the help of the time-frequency diagram,and the state of the wheel is judged,Finally,the method is verified based on the operational data.(3)According to the comparative analysis of two groups of typical analog signals,the paper points out that the variational mode decomposition can effectively solve the problem of mode mixing in empirical mode decomposition.According to the characteristics and complexity of dynamic simulation and the operational data,the method to determine the decomposition level of axle box signal is proposed,and the rail corrugation components are extracted to verify the feasibility.The results show that the variational mode decomposition method is effective in extracting the corrugation component under different wheel wear conditions.(4)Based on the time domain characteristics,wavelet packet energy feature and thirdorder cumulant,the feature vector of the rail corrugation component extracted by the variational mode decomposition is constructed.Firstly,the feature vector of the rail corrugation component under the ideal wheel is input into the support vector machine for training and testing.The classification of the wave depth is realized,and the accuracy is 92%.The combination of different types of parameters verifies the correctness.Finally,under the condition of the wheel polygon wear,the feature vector of the corrugation component is constructed and input into the support vector machine for testing.The accuracy rate is 88%.