Research On The Method Of Gearbox Fault Feature Extraction Based On Sparse Representation

Gearbox plays an important role in mechanical equipment,once a failure occurs,it will cause huge economic losses and even cause casualties.Therefore,it is of great significance to the condition monitoring and fault diagnosis of the gearbox.The gearbox vibration signal usually contains a lot of noise,so it is difficult to accurately extract its fault feature.Therefore,this paper studies the method of gearbox fault feature extraction based on sparse representation.Firstly,to solve the problem that the existing single dictionary is not effective in the reconstruction accuracy of gearbox fault vibration signals,a compound dictionary based on the mathematical model of gearbox distributed fault and impact fault signal is proposed.Combining the orthogonal matching pursuit algorithm to reconstruct the gearbox fault signal.Through comparison and analysis with DCT,Fourier and Ilmf dictionaries,the proposed compound dictionary is better than the other three dictionaries in terms of mean square error,signal-to-noise ratio and similarity.The noise immunity of the compound dictionary is further studied,and the results show that the compound dictionary has better robustness to noise.Secondly,after the vibration signal is reconstructed by the compound dictionary combined with the orthogonal matching pursuit algorithm,the frequency band component with rich fault information is still needed to be selected for analysis,while the common signal decomposition methods such as band-pass filtering and EMD cannot achieve adaptive decomposition or exists modal aliasing phenomenon.A method of gearbox fault feature extraction based on compound dictionary and envelope frequency band optimization Fourier decomposition algorithm is proposed.The method divides the envelope of the reconstructed signal frequency band and effectively overcomes the phenomenon of modal aliasing in Fourier decomposition algorithm.Compared with EMD and fast spectral kurtosis,the proposed method has better performance in extracting the fault features of the gearbox.Thirdly,there is the problem of subjective selection for atomic parameters of compound dictionary.A sparse representation parameter optimization method based on wingsuit flying search algorithm is proposed.The optimal sparse representation algorithm is used to solve the parameters of the optimal dictionary atoms in each iteration of the compound dictionary,which improved the reconstruction accuracy of the sparse representation and highlighted the main fault information features.Compared with other commonly used parameter optimization algorithms,the proposed method can significantly improve the accuracy of gearbox fault feature extraction.Then,in order to solve the problem that the orthogonal matching pursuit algorithm has to match the vibration signal with low signal-to-noise ratio for several times before it can match to the effective dictionary atom,which increases the calculation amount and may introduce noise.It is improved from two aspects.One uses SVD to construct an over-complete dictionary,and the other uses the minimum variance and inner product fusion index as the index for selecting the optimal dictionary atom for each iteration of the orthogonal matching pursuit algorithm.A signal decomposition method based on scale-space and improved sparse representation is proposed.Compared with EMD and VMD decomposition algorithms,the proposed method has better time-frequency characteristics,no mode aliasing phenomenon,and better feature extraction effect.Finally,a gearbox fault experimental platform is built to verify the effectiveness of the gearbox fault feature extraction method proposed in this paper.