Research On The Fault Diagnosis Methods For Gearbox Based On The Resonance-based Sparse Signal Decomposition

Gearbox is the key component of mechanical equipments. It plays a crucial rolein power and rotation transmitting of the mechanical system. To a large extent, thesafety and reliability of mechanical equipments was affected directly by the operationstate of gearbox. Therefore, condition monitoring and fault diagnosis of gearbox ispracticably valuable and realistically significant. Diagnosis technique based onvibration analysis is widely applied in engineering because of its practicability. Thistechnique firstly collects the vibration signal of gearbox, then extracts the faultfeature from the acquired signal and finally identifies the gearbox fault.Based on tunable Q-factor wavelet transform, Selesnick put forward a novelmethod named the resonance-based sparse signal decomposition recently. Differentfrom the traditional signal decomposition methods, the resonance-based sparse signaldecomposition method achieves signal decomposition according to the differentQ-factors of the harmonic component and transient impact component. Firstly, thesparse representation of components with high Q-factor and low Q-factor by tunableQ-factor wavelet transform were established, and then, the nonlinear separation ofthose components was achieved by the morphological component analysis method,finally, the high resonance component and low resonance component that contains theharmonic signal and the transient impact signal were obtained respectively. However,the decomposition result of the method is closely related to the decompositionparameters, and the decomposition parameters of the method was selected manually,which may result in the reduction of the decomposition accuracy because of themismatch between the components and the corresponding basic functions of tunableQ-factor wavelet transform.Supported by the project of Natural Science Foundation of China, which namedas “Research on the resonance-based sparse signal decomposition method and itsapplication in mechanical fault diagnosis (Project Approval Number:51275161)” andthe independent research project of the state key laboratory of advanced design andmanufacturing for vehicle body of Hunan University, which named as “Research onthe early fault diagnosis and residual life prediction techniques of key automotivecomponents and parts (Project Serial Number:71375004)”, aiming at these problemsof the resonance-based sparse signal decomposition method above, this thesis combined the resonance-based sparse signal decomposition method with the geneticalgorithm and the stepwise iterative optimal method respectively, and applied them tothe fault diagnosis of gearbox.The main researches and the acquired innovative achievements in the thesis areas follows(1) The decomposition result of the resonance-based sparse signal decompositionmethod is closely related to the decomposition parameters, and the optimaldecomposition result is difficult to be obtained with the decomposition parameterswhich were selected manually. Aiming at this problem, a novel method for the faultdiagnosis of rolling bearings based on the resonance-based sparse signaldecomposition with the optimal Q-factor is proposed in this thesis. In this method, theoptimal Q-factor is obtained firstly by the genetic algorithm, with the goal ofmaximizing the kurtosis of the low-resonance component of the resonance-basedsparse signal decomposition. Then, the vibration signal of a rolling bearing isdecomposed into the high-resonance component and the low-resonance component bythe resonance-based sparse signal decomposition method with the optimal Q-factor.Finally, the low-resonance component is analyzed by the Hilbert envelope method, thecycle of the periodic impulse component can be acquired and the faults of the rollingbearing can be diagnosed. The simulation and application examples show that theproposed method is effective in extracting the impulse signal from rolling bearings.(2) Aiming at the problem that there is a large calculating burden and lowcomputation efficiency in the genetic algorithm, a novel method for the faultdiagnosis of rolling bearings based on the resonance-based sparse signaldecomposition and the stepwise iterative optimal method is proposed in this thesis. Inthis method, the optimal Q-factor is obtained by the stepwise iterative optimal methodinstead of the genetic algorithm. The simulation and application examples show thatthe proposed method is effective in extracting the impulse signal from rollingbearings and in improving the computation efficiency.(3) Aiming at the limitation that the traditional demodulation analysis methodcan not diagnose the compound fault diagnosis of gearbox effectively, a novel methodfor the compound fault diagnosis of gearbox based on the resonance-based sparsesignal decomposition and the genetic algorithm is proposed in this thesis. In thismethod, the optimal decomposition parameters are obtained firstly by the geneticalgorithm, with the goal of maximizing the composite index consisting of the smoothness index of the high-resonance component and the kurtosis of thelow-resonance component of the resonance-based sparse signal decomposition. Then,the vibration signal of a gearbox is decomposed into the high-resonance componentand the low-resonance component by the resonance-based sparse signaldecomposition method with the optimal decomposition parameters. Finally, thehigh-resonance component and the low-resonance component are analyzed by theHilbert envelope method respectively, and the compound fault diagnosis is carried outaccording to the envelope spectra. The simulation and application examples show thatthe proposed method can separate the fault characteristics of gear and rolling bearingeffectively.In the thesis, the improved resonance-based sparse signal decompositionmethods are proposed combined with the genetic algorithm and the stepwise iterativeoptimal method respectively and applied to the fault diagnosis of gearbox. Thesimulation and application examples show that the proposed methods have a goodapplication prospect in the fault diagnosis of gearbox.