Research On Diagnosis Methods Of Incipient Fault For Rolling Bearing Based On Processing Of Noisy Vibration Signal

As one of the key components of mechanical equipment to ensure the normal operation,the safety monitoring of rolling bearings has been paid more and more attention from industry and academia.However,when the running state of bearings is abnormal,such as early pitting,peeling,cracking,mild wear and other incipient faults,the vibration acceleration signal of bearings always has the characteristics of low amplitude and fault characteristics which are easily to be embedded by interferences.However,in actual working conditions,the generation process of fault vibration signal is inevitably accompanied by background noise,which easily conceals the characteristics of incipient faults and brings great difficulties to diagnosis of incipient fault for bearing.Therefore,a reasonable de-noising model,fault characteristic extraction mechanism and diagnosis algorithm can effectively prevent the occurrence of major faults to ensure the safe operation of bearings.Hence,the research of incipient fault diagnosis under background noise has being a widely concerned problem by scholars at home and abroad.In this paper,the rolling bearing is taken as the research object.Since the vibration signal is the most direct reflection of the bearing state,from the angle of signal analysis and processing,several key problems namely vibration signal de-noising,characteristic extraction and characteristic recognition for incipient fault of bearing are studied.The major work and achievements of this paper are shown as follows:(1)Aiming at the disadvantage that the characteristic of incipient fault for rolling bearing with strong background noise is easy to be concealed,two improved de-nosing algorithms are proposed as follows.On the one hand,considering the disadvantage of traditional adaptive equalization algorithm which is difficult to ensure the convergence and stability at the same time,an improved least square adaptive equalization(RLS)combined with least mean square(LMS)de-noising model is proposed.Based on the traditional RLS algorithm,an improved RLS de-noising model is constructed by introducing momentum factor and variable forgetting factor to de-noise the bearing incipient fault signal with strong background noise.On the other hand,based on the improved adaptive equalization algorithm and considering the limitation of the adaptive equalization algorithm in dealing with wideband signals,a new interactive with dynamic adjusted function method for de-noising of incipient fault vibration signal for rolling bearing is proposed on the basis of LMS and empirical mode decomposition(EMD).The simulation results of bearing incipient fault experimental signal show that the proposed de-noising method can achieve effective de-noising of bearing micro-fault vibration signal,and has more advantages than the traditional de-noising method.(2)Incipient fault characteristics are not obvious and easily concealed by background noise,so two feature extraction algorithms are proposed to overcome the disadvantage of extracting fault symptoms from incipient fault vibration signal with rudimental noise.On the one hand,aiming at the low diagnostic accuracy caused by the difficulty of characteristic extracting of incipient fault effectively and completely,a characteristic extraction method based on optimized LMD component and projection energy is proposed.Firstly,aiming at the difficulty of selecting PF component after the LMD decomposition,a weighted fusion model of PF components is constructed by introducing genetic algorithm based on traditional LMD algorithm.In order to maximize the correlation of the weighted fusion components with its incipient fault original signal,a new rule for calculating the weights is designed.Secondly,the equal interval energy projection method is introduced to construct the characteristic extraction model of incipient faults from the perspective of energy distribution,and a new rule of characteristic selection is designed to extract the characteristic of incipient faults and effectively eliminate redundant information.On the other hand,considering the limitation of LMD decomposition,a new method for characteristic extraction of incipient fault based on peak-to-peak sample entropy of VMD decomposition with parameter optimization is proposed.Firstly,to solve the problem that the decomposition parameter K of VMD is difficult to determine,a K estimation method based on LMD decomposition is designed.Secondly,aiming at the weak impact of incipient faults,the peak-to-peak values of vibration signals are constructed,and the sample entropy peak-to-peak values is extracted as fault characteristics.(3)Aiming at the diagnosis accuracy is not high caused by small characteristic samples,the intelligent classifier support vector machine(SVM)is used to recognize the characteristics of incipient fault.The simulation results show that the proposed diagnosis methods can effectively diagnose incipient faults under background noise,and the diagnosis accuracy is higher than that of traditional diagnosis methods.