Research On Denoising Method Of Bearing Vibration Signal

Rolling bearing is the core component of rotating machinery,and its working state is the key factor to ensure the safe operation of mechanical system.Since the working environment of rolling bearings is harsh and changeable,resulting in rolling bearings abrased,it is necessary to detect possible defects in rolling bearings.Usually,signal analysis technology is used to analyze the state of rolling bearings.However,the fault characteristics of the actual collected bearing vibration signals are often mixed with random noise,which increases the difficulty of bearing fault diagnosis in the following part.In the thesis,the application of signal denoising technology in rolling bearing fault detection is studied.A denoising method is proposed to process the rolling bearing vibration signal with noise mixed and to improve the output signal-to-noise ratio.The main innovations of this thesis include the following aspects:1.Aiming at the problem that rolling bearing fault characteristics are easily submerged by noise under strong noise interference,the paper proposes a vibration signal de-noising method combining Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)and Wavelet Shrinkage.This method takes advantage of the multiresolution characteristics of empirical mode decomposition and wavelet analysis,and performs wavelet shrinkage on the principal IMFs obtained by CEEMDAN to eliminate the influence of noise to the signal.In addition,to solve the problem of difficult threshold selection in wavelet denoising,an adaptive wavelet threshold selection method based on control chart is proposed,which realizes the adaptive selection of thresholds of various scales of wavelet,and effectively solves the problem of difficult threshold selection caused by the non-stationary of the bearing fault signal.The simulation results show that the adaptive wavelet threshold selection method based on the control chart has stronger robustness and better denoising results than the traditional threshold selection methods such as BayesShrink and SureShrink.Compared with a single CEEMDAN or Wavelet Shrinkage,the combined method proposed in the paper can effectively remove the noise components in the rolling bearing vibration signal,and the signal-to-noise ratio is increased by 9.05dB in the case of low signal-to-noise ratio.2.In order to solve the problem of poor denoising effect of traditional singular spectrum analysis method in non-stationary environment,the paper proposes an improved denoising method of tensor spectrum analysis.Since there are some difficulties in group selection in tensor spectrum analysis,an adaptive feature tensor grouping method based on empirical mode decomposition is proposed,which can solve the problem of signal reconstruction in tensor spectrum analysis.The simulation results show that,compared with the singular spectrum analysis method,the tensor spectrum analysis can effectively remove the noise and increase the signal-to-noise ratio by 11.05dB in the case of low signal-to-noise ratio,which also means the vibration signal has more obvious fault characteristics.