Noise Reduction And Feature Extraction Of Gearbox Vibration Signal

Because the structure of large machinery is complex and most of them work in harsh environment,monitor the operation of machinery equipment and diagnose fault characteristics in a timely and accurate manner,which is an essential part to ensure the stability and safe operation of machinery equipment.Fault feature extraction is the key to fault diagnosis technology.However,most mechanical devices are operated under strong background noise conditions.The extracted vibration signals are often nonlinear and non-stationary.Therefore,effectively signal noise reduction methods are the precondition and key to extract fault features of mechanical equipment.In this paper,starting from the theory of dual-tree complex wavelet transform,higher-order cumulants,and variational modal decomposition theory,study the problem of denoising and feature extraction of vibration signals under Gaussian noise and natural background noise with gearbox as the research object.And the particle swarm optimization algorithm is introduced to optimize the parameters.The main research methods and conclusions are as follows:(1)Because the conventional threshold noise reduction method performssoft and hard thresholds respectively on the wavelet coefficients,the noise reduction effect of the gearbox vibration signal under strong background noise is not ideal,and the method of processing the real and imaginary parts with the threshold value will cause local phase distortion.Based on the translation invariance of the dual-tree complex wavelet transform,proposed a dual-tree complex wavelet transform and high-order cumulant gearbox vibration signal denoising method.The algorithm uses the fourth-order cumulant processing method for the wavelet coefficients of each layer,and carries on the signal-to-noise separation according to the statistical characteristic of the signal and the noise.Because the number of wavelet decomposition layers will directly affect the denoising effect of the signal,the particle swarm optimization algorithm is used to select the number of decomposed layer number of wavelet.Simulation and experimental signal processing results show that this method can effectively suppress noise interference,improve signal-to-noise ratio,and can meet gear fatigue under different signal and noise levels compared with dual-tree complex wavelet transform soft and hard thresholding methods.The need for follow-up processing of the collected vibration signals in wear tests.(2)Aiming at the problem of unsatisfactory fault characteristics of gearbox extracted under strong background noise,this paper proposes a fault diagnosis method of gearbox with parameter optimized modal decomposition and high-order cumulant.First,the particle swarm optimization algorithm is used to search for the number of optimal variational model decompositioncomponents.The fault signal is divided into several intrinsic mode function components(IMF)by variational modes,and each component are subjected to the fourth-order cumulant noise reduction processing to find the component containing the fault characteristic frequency.Then the selected IMF component is subjected to envelope demodulation operation to find its envelope spectrum,so as to effectively extract the feature frequency of the fault signal.The results of simulation and experimental signal processing show that this method can extract the fault characteristic frequency information more effectively than the EMD method.And it can realize the effective extraction of the fault features in the vibration signals of the gearbox and restrain the noise interference well,which has a certain feasibility and application value.(3)Aiming at the defects of frequency aliasing and parameter customization in dual-tree complex wavelet transform,this paper proposes a self-adaptive dual-tree complex wavelet transform gearbox fault diagnosis method.This method utilizes the dual-tree complex wavelet transform and variational modal decomposition techniques(DTCWT-VMD).Firstly,the signal is decomposed and reconstructed by dual-tree complex wavelet transform.Particle swarm optimization(PSO)is used to determine the component kurtosis value as a fitness function to select the optimal decomposition level of dual-tree complex wavelet.Secondly,the reconstructed low frequency signal is subjected to spectrum analysis to extract the fault characteristic signal.Single-branch reconstructed high-frequency components are subjected to variational modaldecomposition,and the main frequency component signals of each high-frequency component subjected to variational modal decomposition are obtained by the kurtosis value.Finally,analyze the spectrum of each main frequency component signal to identify the fault characteristics of the gearbox.The simulation results of the simulation signal and the gear box vibration experimental signal show that compared with the dual-tree complex wavelet transform,variational modal decomposition,and empirical mode decomposition,this method not only eliminates the frequency aliasing phenomenon,but also improves the signal-to-noise ratio.And select the autonomy of the main frequency components,and improve the ability to extract fault features from a strong noise environment.(4)The kurtosis is used as an index to determine the presence or absence of faults in the signal.It is a dimensionless parameter,has nothing to do with other external factors,and is sensitive to strong impact faults.and is very suitable for diagnosing surface damage.The larger the absolute value of the kurtosis index is,the more components contain the fault impact.According to the maximum kurtosis principle,the relevant frequency band signal is selected for analysis,and the transient signal can be effectively detected.Therefore,it is used as the judgment basis of the fitness function,and it is applied to the variational modal decomposition component number and the double-tree complex wavelet decomposition layer number through the particle swarm optimization and is used to obtain the high frequency components after the variational modaldecomposition.The main frequency component signal has a very good feature extraction effect.