Research On Fault Diagnosis Method Of Rolling Element Bearings Based On Morphology And Deconvolution

As a typical representative of clean and renewable energy,wind power has a broad space for development.Because of complex structure and tough working conditions,wind turbines gradually occur failure after experiencing a long term service.If the failure can not be detected in time,the safe operation performance of the wind turbine may be seriously decreased.To ensure the stability and safety of wind turbines,prevent major accidents,and reduce maintenance costs,it is of great significance to carry out real-time monitoring and fault diagnosis of its operating conditions.In this study,the rolling element bearings of wind turbine with a high failure rate is taken as the diagnosis target,and the vibration-based bearing signal processing methods are explored.The main contents are as follows:(1)Based on the brief description of the structural features and working mechanism of wind turbines,the mechanical structure,mostly invalidation form,failure reason,vibration mechanism,fault characteristic frequency,and corresponding vibration characteristics of the rolling element bearings are introduced,which provides the theoretical basis for the realization of bearing fault diagnosis.(2)The weak fault-related features of the vibration signal,which originated from wind turbine rolling element bearings,are generally immersed in environmental noise and harmonic interference and difficult to extract.This issue is addressed by proposing a new enhanced morphological filtering scheme for fault diagnosis.Firstly,a new morphology analysis method,named morphological comprehensive filter-hat transform(MCFH),is constructed to extract fault-related impulses.Secondly,an adaptive scale selection strategy is developed to obtain an appropriate filter scale for MCFH.Thirdly,an improved envelope derivative energy operator is utilized to enhance the impulse characteristics of the signal after morphological filtering and to suppress the frequency of in-band noise.In both simulation and experimental studies for wind turbine bearing,the proposed method delivered better fault feature extraction and noise reduction performance than the traditional methods.However,this method still needs to be discriminated against manually in diagnosis,which is inevitably affected by people’s experience and can not achieve automatic fault diagnosis.(3)The working conditions of wind turbine rolling element bearings are always complex,therefore obtained vibration signals are non-stationary and non-linear.However,the traditional method based on the time-frequency domain has some problems,such as inaccuracy and poor adaptability when extracting fault features.To solve the problem,a novel feature extraction algorithm for fault diagnosis is proposed based on local mean decomposition(LMD)and morphological fractal dimension(MFD),and combined with extreme learning machine(ELM)to conduct wind turbine bearing fault diagnosis.Firstly,the raw vibration signal is adaptively decomposed by LMD and the sensitive product functions(PFs)are determined by correlation coefficients between PFs and the raw signal.Secondly,the fractal dimension of selected PFs is estimated by morphology to construct fault feature vector.It is taken as the input of ELM to develop a fault diagnosis model.Finally,the results of the experiment showed that the proposed method improves performances for detecting the bearing faults.However,it is also found in the research that the detection ability of this method is not outstanding when there are multiple faults in the bearing of wind turbines at the same time,which are denoted as compound faults.(4)The compound fault signal of bearings is coupled with each other among different faults types,which makes it difficult to extract the impulse characteristics of compound faults by traditional methods.Based on adaptive resonance-based signal sparse decomposition(ARSSD)and multipoint kurtosis optimal minimum entropy deconvolution adjustment(MK-MOMEDA),a fault diagnosis scheme of bearings is presented.In this scheme,ARSSD is firstly adopted to separate multi-fault features,and the cuckoo search algorithm(CSA)is used to optimize the high and low-quality factors to obtain the optimal low resonance component consist of the transient impactive component in RSSD.Secondly,multipoint kurtosis(MK)is utilized to extract the fault impulse periodic information in optimal low-resonance component.Thirdly,the MOMEDA method is employed to deconvolve different fault periodic impulse train with appropriate period range.In the end,the obtained filtered signal is conducted on the envelope spectrum to identify the fault type,respectively.The proposed algorithm is verified to be able to recognize the two and three faults of rolling element bearings in the experimental rig.