Research On Vibration Testing And Analysis Of Wheel Surface Simulated Damages

With the rapid development of high-speed and heavy-haul railway,the problem of wheel damages becomes more and more serious.It affects not only the comfort of passengers,but also seriously the train safety.Therefore,it is very necessary to carry out vibration testing and analysis of wheel surface damages to provide important theoretical support and technical guidance for ensuring the safety and reliability of railway operationIn this paper,the wheel surface rolling fatigue damage,early damage,spalling degree and abrasion degradation simulated respectively have been detected and analyzed to realize effective identification of wheel surface damages.At the same time,under field working conditions the vibration detection and analysis of wheel polygon damage were carried out to verify the validity of the simulation test,which provided certain reference value for online detection of wheel damages.The main achievements and conclusions of this paper are as follows1.Under different tangential forces the rolling fatigue damage and vibration characteristics of the wheel surface are analyzed.With tangential forceincreasing,wheel surface damages change from slightly fatigue crack to peeling transition and more serious spalling.The multifractal method of local mean decomposition is used to analyze the vibration signals of wheels.Under different tangential forces,the fractal dimension and variation fluctuation are different to realize the distinction of different wheel damages2.Aiming at the weak and difficult detection of wheel early damage signals,a method based on multifractal KPCA and LSSVM is used.Resonance sparse decomposition is reconstructed for noise reduction to highlighting impact component of the wheel signals Combinning the multifractal with nuclear principal component method,the feature is extracted.Based on LSSVM,the model of wheel early damage is constructed for realizing classification and diagnosis3.Because of the advantages of high precision and fast convergence speed,but disadvantage of artificially setting number of decomposition,thus the method of setting the number of decomposition of VMD is proposed.Based on the advantages of improved multi-scale permutation entropy,an improved multi-scale permutation entropy partial mean method is proposed to obtain feature information.Combined with PSO-ELM,wheel spalling degree is distinguished.Compared with the results of GA-ELM and ELM classifier,the identification accuracy and efficiency were improved.4.Three popular learning algorithms,ISOMAP,LLE and LLTSA are used to reduce the dimension of high-dimensional data and the optimal dimension reduction results are obtained by comparison.Combined with KELM method,the wheel abrasion degradation are identified.5.Combination of the principle of Euclidean distance and cross correlation was proposed for choosing the optimal component method by respectively EWT and VMD decomposed.The multi-scale fuzzy entropy of the optimal components is calculatted and by comparison the feature vectors constructed are putted into PSO-ELM,compared with PSO-LSSVM.The validity of the simulation experiment method is verified by analyzing the vibration of wheel damages in laboratory and wheel damages under field working conditions.