Research On Nonlinear Mode Decomposition Strategy For Time-varying Speed Bearing Fault Diagnosis

Rolling bearings are one of the key parts of rotating machinery equipment that can bear loads,which often run in harsh working environments.Its operating state directly determines the stability and reliability of the entire system.In the industrial production process,bearings and other components often operate under varying speed conditions.They bear more complex stresses and are prone to failure,which may cause equipment damage or even casualties.Therefore,it is particularly important to carry out bearing condition monitoring and fault diagnosis under varying speed conditions to ensure the safe operation of machinery equipment.However,under varying speed conditions,bearing fault signals are characterized as non-stationary,thereby resulting in spectral smearing phenomenon.The fault characteristics cannot be directly identified in the frequency spectrum,making traditional fault diagnosis methods designed for constant speed no longer applicable.In this dissertation,the rolling bearings in the rotating machinery are taken as the research object,and aiming at the bearing fault diagnosis under varying speed conditions.Considering the problem that the traditional order analysis method based on tachometer will lead to the increase of cost and the limitation of application,and that the tacholess order analysis method based on time-frequency representation exists in the fuzzy time-frequency ridges,the application of variational nonlinear chirp mode decomposition(VNCMD)based on the variational mode decomposition method(VMD)in the bearing fault diagnosis under varying speed conditions is carried out.The specific work of the paper is as follows:(1)Firstly,the form of bearing failure and the characteristics of its vibration signal are described.On this basis,the differences between bearing failure characteristics under varying speed conditions and constant speed conditions are pointed out.Considering the superiority of the nonlinear mode decomposition methods in the analysis of non-stationary signals,the VMD and the VNCMD methods are studied in depth.Combined with the signal modulation model,the differences between the VMD and VNCMD methods are described.Through simulation signal analysis,the validity of the VMD and the VNCMD methods in signal decomposition and instantaneous frequency estimation is verified.Furthermore,for the wideband characteristics of bearing fault signal under varying speed conditions,the limitations of the VNCMD method in the initial parameter setting are demonstrated.(2)Aiming at the problem that the constant initial instantaneous frequency of the original VNCMD method cannot effectively track the faulty components of strong speed fluctuations,which causes the algorithm not to converge,a multi-target ridges guided VNCMD method is proposed.In terms of multi-target ridges estimation,a multi-band signal separation method is first used to enhance the shaft rotation frequency and fault characteristic frequency information.Then,a local ridge detection algorithm based on time-frequency representation is used to extract the shaft rotation frequency and fault characteristic frequency respectively.Finally,the estimated multi-target ridges are used as the time-varying initial instantaneous frequency of the VNCMD algorithm,which promotes the convergence of the algorithm and optimizes the instantaneous frequency.In order to avoid resampling errors in order analysis,a characteristic frequency ratio(CFR)criterion is proposed to accurately determine the bearing fault type.Simulation studies and experimental verifications confirm that the proposed method can improve the estimation accuracy of the shaft rotation frequency and the fault characteristic frequency,and successfully identify the fault bearing type under varying speed conditions.(3)In the actual industrial application,the background noise is often large,thereby resulting in poor resolution of the time-frequency representation of the original vibration signal.Hence,it is difficult to accurately estimate both the shaft rotation frequency and fault characteristic frequency,which affects the decomposition effect of the multi-target ridges guided VNCMD method.Aiming at this problem,an optimization tendency guiding mode decomposition(OTGMD)method based on the dominant synchronous ridge is proposed through analyzing the optimization trend of the VNCMD method.In terms of dominant synchronous ridge extraction,the traditional fast spectral kurtosis method is very easy to fail due to the interference of strong background noise.Therefore,a feature separation method based on low-pass filtering and iterative envelope analysis is used to enhance the shaft rotation frequency component and fault characteristic frequency component respectively.Then,the dominant component ridge is estimated in the shaft rotation frequency component by the local ridge detection algorithm.In terms of the initial instantaneous frequency setting of the VNCMD method,the estimated ridge of the dominant component is optimized and used as the reference frequency,and the initial parameters of the remaining components are automatically set based on the optimization tendency guidance algorithm.Based on this,the enhanced fault component is extracted synchronously,the target component is decomposed and its instantaneous frequency is optimized adaptively.In order to avoid the influence of noises and harmonic components,a characteristic frequency ratios library(CFRL)stopping criterion is proposed,which can accurately control the iterative process of the algorithm and determine the bearing fault type.All analysis results clearly show that the proposed method can significantly improve the accuracy of the original VNCMD method and present good results in bearing fault diagnosis under varying speed conditions.In summary,this dissertation focuses on the needs of bearing fault diagnosis under varying speed conditions and theoretical flaws of VNCMD method,and proposes two new VNCMD decomposition strategies from the perspectives of multi-target component ridges selection and dominant synchronous ridge optimization tendency guidance.Combined with the proposed feature preprocessing technology and CFRL stopping criterion,an accurate estimation of the instantaneous frequency is achieved,which has theoretical and practical significance for bearing fault diagnosis under varying speed conditions.