Multiple Enhanced Sparse Decomposition Method For Gearbox Compound Fault Diagnosis

As an important power transmission component in the rail transit system,the gearbox's operating state is directly related to the service performance of high-speed trains.Once the train gearbox fails and cannot be detected and warned in time,it will seriously endanger the operation safety of high-speed trains and even cause major safety accidents.In recent years,with the rapid development of high-speed railways in China,the monitoring and diagnosis of gearboxes has become a key technology that needs to be developed.In the actual operation,the train gearbox has poor working conditions,many vibration components,and large background noise,which leads to the compound fault of the gearbox with many vibration sources and large noise interference.Therefore,research on the accurate separation method of multi-source signals of gearboxes is of great significance for the state detection and life prediction of rotating machinery.This research is financially supported by the Youth Program of National Natural Science Foundation of China "Research on Optical Image Monitoring Method for Train Bogies"(No.51405320)and the Suzhou science and technology plan project "Research on Key Technologies of Rail Vehicle Wheelset Bearing Health Diagnosis and Life Prediction"(No.SYG201511).Aiming at the problem of multi-source vibration signal decomposition of gearbox,we propose a multiple enhanced sparse decomposition method,which solves the problem of low accuracy of complex fault signal reconstruction by traditional methods,and significantly improves the effectiveness of gearbox fault diagnosis and status detection.The main research contents are as follows:(1)Gearbox vibration mechanism analysis.The study of gearbox failure mechanism can provide overall guidance for the accurate separation of multi-source vibration signals.Therefore,this paper first analyzes the characteristics of gearbox vibration signals under different states.Then,the parametric wavelet basis models matching the internal characteristic structure of the gearbox multi-source vibration signal are constructed.Based on the constructed wavelet basis,the sparse representation over-complete dictionaries are constructed through parameter expansion.(2)Research on multi-source sparse decomposition algorithm and its application in fault diagnosis.First,the multi-source sparse decomposition model is constructed for the gearbox compound fault diagnosis.Secondly,based on the analysis of the vibration mechanism of the gearbox,we have established a novel data fidelity project capable of comprehensively solving multiple gearbox compound failure problems;then,the non-convex penalty function with good amplitude retention ability is constructed.Finally,the convexity-preserving condition of the objective function is further deduced,and the corresponding sparse solution method is used to optimize the solution.Simulation and engineering tests prove that the algorithm can accurately realize gearbox compound fault diagnosis.(3)Considering that the sparsity within and across groups property of the bearing signals can be used to eliminate the noise between each adjacent pulse,we propose a novel method of group sparsity-enhanced signal decomposition to promote this sparsity.The proposed non-convex group sparse penalty function can not only effectively solve the problem of amplitude underestimation introduced by the traditional convex penalty function,but also can effectively filter out irrelevant noise components between fault pulses based on the sparse characteristics between groups within the group.At the same time,in order to solve the problem of sparse decomposition parameter selection,this paper proposes an adaptive regularization parameter selection strategy.Simulation and engineering tests verify that this sparse decomposition algorithm can effectively implement fault diagnosis of various gearbox bearing fault signals.In summary,the sparse decomposition method proposed in this paper can effectively implement signal reconstruction and fault diagnosis for multiple fault types of gearboxes.It plays an important role in improving the decomposition accuracy of gearbox compound fault signals under the background of strong noise and improving the reliability of fault diagnosis of rotating machinery represented by gearboxes.