Study On Fault Feature Extraction Of High-speed Train Bearing By Combining EEMD And Sparse Method

With the rapid development of high-speed railway construction in recent years,the safety of high-speed trains is of great importance.Besides,the bearing is the key part of running department,the state of which must be monitored.In order to judge its state accurately,vibration signal can be used,which is the most direct carrier of bearing state information.When the bearing fails,collisions will happen on its irregular surface,resulting in different frequency impact components in the vibration signal that has been collected.However,due to the complexity of high-speed train system,the frequency components of bearing vibration signal are various,and the background noise interferences are so strong that fault characteristics are easy to be submerged.Therefore,how to extract the of impact characteristics effectively has been the core problem of bearing failure condition monitoring.However,each component in the running part of high-speed train is closely coupled.Once a component fails,the frequency and statistical characteristics of vibration signal of rolling bearing will change with time,resulting in non-stationary characteristics of vibration signal.The empirical mode decomposition method is very suitable for the decomposition of non-linear and non-stationary signals.However,due to the serious mode aliasing problem of this method,the ensemble empirical mode decomposition method which can improve the mode aliasing problem is used to decompose the bearing vibration signals.After the signal is decomposed into several components,in the process of analyzing the fault vibration signal of rolling bearing,it is necessary to select the components with impact characteristics.Therefore,an intrinsic mode function selection criterion is proposed to screen the target components.In addition,the background noise of high-speed trains is very strong in the process of operation.After the signal is decomposed into several components,the impact characteristics will still be submerged.In this case,both the noise signal and the target feature signal are in a mode component,and the denoising effect cannot be achieved by dividing the signal in frequency domain.The impact characteristics of the rolling bearing fault vibration signal are sparse in the time domain,so it can be denoised by sparse representation.In this paper,the non-convex sparse promotion function and the total variation term are introduced to form the regularization term,the objective function is established,and the sparse solution is obtained by using the majorization-minimization,algorithm.In order to evaluate the denoising quality qualitatively and objectively,a denoising quality evaluation method is proposed,based on which the optimal value of the regularization parameters is obtained,and the effectiveness of the full-variation sparse denoising algorithm is verified by using non-convex regularization terms.In order to extract bearing fault features of high-speed trains,the ensemble empirical mode decomposition method is combined with the sparse table method of total variation.Firstly,the signal is decomposed to obtain several components.Secondly,the components that meet the requirements are filtered.Thirdly,sparse denoising is carried out.Finally,the components after denoising are reconstructed to obtain the impact characteristics.By applying the method in this paper to the simulation signal and the measured signal of high-speed train bearing fault diagnosis,it is verified that the method can realize the sparse representation of the signal and extract the fault features in the signal.