Research On Compound Fault Diagnosis Of Gearbox Based On Deconvolution Method

Gearbox is one of the key components used to change the speed and transmit power in mechanical rotating equipment.It has been widely used in modern machinery and equipment such as aviation,power systems,agricultural machinery,industrial production,and metallurgy.As an indispensable transmission component of rotating machinery and equipment,gearbox is prone to gear wear,cracks,and broken teeth under long-term overload operation conditions.Once the gearbox fails or causes partial damage,it may affect the transmission chain system.Even the operating state of the entire machinery and equipment,potentially causing huge risks of safety accidents and economic losses.In order to improve the productivity of rotating machinery and equipment,prevent major accidents,and reduce maintenance costs,real-time monitoring and fault diagnosis of its operating conditions are of great significance.For this reason,this article takes gearbox as the research object,and conducts feature extraction and fault diagnosis based on the vibration signal time-frequency analysis method.The main research contents are as follows:(1)Taking the gearbox as the research object,introduce its failure form,failure cause,vibration mechanism,vibration characteristic frequency and corresponding vibration characteristics,etc.,to lay the foundation for subsequent research on feature extraction and fault diagnosis.(2)Aiming at the nonlinear and non-stationary characteristics of the gearbox vibration signal,an adaptive empirical wavelet transform(AEWT)and singular value decomposition(SVD)feature extraction method is proposed,and combined with the kernel extreme learning machine(KELM)to achieve a single failure of the gearbox Research on fault diagnosis.First,the scale space method is used to adaptively divide the frequency spectrum of the vibration signal to ensure the effectiveness of the EWT decomposition,and then the component with a larger correlation coefficient with the original signal is selected as the sensitive component to eliminate redundant noise components.After that,SVD is used to calculate the singular values of the sensitive components to construct the fault feature vector;finally,the feature vector is used as the input of KELM to establish the KELM gear state recognition model.This method can effectively identify gear faults under different fault types,and the overall identification is relatively high.However,this method requires prior knowledge when performing fault diagnosis,and the effect of feature extraction for compound faults of gearboxes is not good.(3)Aiming at the problem that the impact characteristics of gearbox compound faults are difficult to effectively separate,a gearbox compound fault diagnosis method based on infogram and multi-point kurtosis optimal minimum entropy deconvolution correction(MK-MOMEDA)is proposed.This method uses the information graph obtained by the average spectral negative entropy and selects the best bandwidth and center frequency to band-pass filter the composite fault signal to reduce the influence of noise components;then,calculate the multi-point kurtosis spectrum of the filtered signal to identify the spectrum The fault impulse cycle component contained in it,and the appropriate fault cycle interval is set,and the MOMEDA operation is performed to extract different fault characteristics;finally,the separated signal is subjected to envelope demodulation analysis to determine the fault type.Experimental results show that this method can effectively separate the compound faults of the gearbox.However,it is also found in the research that the noise reduction ability of the information graph is limited,which makes the diagnosis effect of some experiments poor.(4)In order to further solve the above problems,a new deconvolution method is introduced,namely maximum second-order cyclostationary blind deconvolution(CYCBD),and adaptive maximum second-order cyclostationary blind deconvolution(ACYCBD)and1.5-dimensional Gearbox compound fault diagnosis method enhanced by derivative spectrum.The method first uses cyclic spectrum analysis to detect the cyclic frequency components related to the fault characteristics in the composite fault vibration signal,and constructs the cyclic frequency set of different target types;then,according to the different types of cyclic frequency sets,the third-order cumulant sparsity(TCS)as an indicator,adaptively select the optimal filter length of CYCBD,so as to obtain the optimal filter signal of CYCBD containing different fault impact components;finally,introduce a new composite 1.5-dimensional derivative spectrum for feature enhancement,and analyze the s prominent fault characteristic frequencies to identify the fault types.The effectiveness of the proposed algorithm is verified by experiments,and the results show that the method can effectively separate each single fault signal from the gearbox compound fault signal collected by the sensor,and realize the gearbox compound fault diagnosis.