Research On Rotating Machinery Fault Diagnosis Based On Variational Mode Decomposition

Rolling bearings and gears are the most widely used key components in rotating machinery.Since some defects frequently exist in those key parts,it is very important to detect them as early as possible.However,the collected vibration signals are obviously nonstationary,meanwhile those weak signatures are always immersed in heavy background noise.Thus,the available fault diagnosis methods have some difficulties in identifying incipient faults.Based on the theory of vibration analysis and the newly proposed signal processing techniques,some novel signal de-noising and weak fault diagnosis techniques have been developed in this thesis.Firstly,the theory of variational mode decomposition(VMD)as well as traditional signal processing methods are simply introduced,then the characteristics of VMD are deeply analyzed using simulated signals.It is found that VMD has some advantages in refraining from modes aliasing and detecting impacts compared with EMD.Secondly,a novel hybrid fault diagnosis approach is developed for denoising and nonstationary feature extraction in this work,which well combines the VMD and majoriation –minization based total variation denoising(TV-MM).TV-MM approach is utilized to remove stochastic noise in the raw signal and to enhance the corresponding characteristics.Since the parameter ? is very important in TV-MM,weighted kurtosis index is also proposed in this work to determine an appropriate ? used in TV-MM.The performance of the proposed hybrid approach is conducted through the analysis of the simulated and practical bearing vibration signals.Results demonstrate that the proposed approach has superior capability to detect roller bearing faults from vibration signals.Thirdly,modulation intensity distribution(MID)combined with VMD is applied to detect second-order cyclostationary components in gear fault diagnosis.Considering the shortcomings of MID in the analysis of multi-harmonic modulation signals,VMD is used as the signal preprocessing before MID analysis.Results of simulation and the experimental analysis have demonstrated the effectiveness of the method.Finally,a fast fault diagnosis method based on nonlocal mean algorithm(NLM)is proposed for the purpose of online fault detection.Coefficients of wavelet transform is used to estimate the noise variance of the original signal,while the appropriate bandwidth parameter is determined via Stein unbiased estimation.Moreover,mirror extension is adopted to process the signal denoising on the boundaries.The effectiveness of NLM algorithm in noise reduction and signal enhancement are verified through the analysis of the simulated signals.Results of bearing and gear fault diagnosis on different test-rigs further demonstrate its usefulness.