Synchroextracting Transform And Its Application In Mechanical Fault Diagnosis

Vibration signals obtained from rotating machinery contain a lot of information related to equipment operating conditions.How to effectively deal with complex non-stationary signals is of great significance to the health management of mechanical equipment.In engineering practice,affected by the complex structure and operating conditions of the mechanical equipment,when a fault occurs,the defect signal often shows complex non-stationary characteristics and is affected by noise.Time-frequency analysis technology characterizes the time-varying features of non-stationary signals through the joint distribution in the time domain and frequency,and has been widely used in many fields.However,due to the complexity of defect signals,there are still many problems in the application of traditional time-frequency analysis technologies in mechanical equipment fault diagnosis.Such as the emphasis amplitude and frequency modulation components in the defect signal are difficult to be effectively characterized.The fault impact component is difficult to be effectively extracted.The weak fault feature component is greatly affected by noise and the adjacent fault feature components are difficult to be effectively separated.In order to solve these typical problems as much as possible,so that the time-frequency analysis technology can be better applied in the fault diagnosis of mechanical equipment.Based on short-time Fourier transform and Synchroextracting transform,this dissertation proposes a series of improved time-frequency analysis methods by using the idea of parameterization and post-processing and transient assumptions,which can be applied to the addressing of complex fault signals under different working conditions.The main work done in this dissertation includes the following five aspects:(1)To effectively characterize the chirp component in the fault signal of mechanical equipment.In this dissertation,the principle of synchrosqueezing transform and reassignment method is studied,and a longitudinal synchrosqueezing transform method is proposed through their mathematical relationship.Using the time-frequency distribution generated by the reassignment method,the frequency estimation operator in the original synchrosqueezing transform is optimized in the frequency direction,that is,in the longitudinal direction,so that it can more accurately estimate the instantaneous frequency of the chirp signal.To further improve the time-frequency resolution of the longitudinal synchrosqueezing transform,the synchroextracting operator is also optimized by the reassignment method,and the longitudinal synchroextracting transform is proposed to generate a time-frequency representation with high energy concentration.Moreover,it is effectively applied to characterize the time-varying law of the chirp component in the defect signal.(2)For efficient processing of fault-related strong AM-FM signals,this dissertation proposes a high-order synchrosqueezing transform.By performing high-order Taylor expansion on the amplitude and phase of the signal,it is substituted into the original short-time Fourier transform.A high-order frequency estimation operator is constructed.Compared with the original frequency estimation operator,it can more accurately capture the dynamic features of AM-FM signals.At the same time,using the idea of synchrosqueezing transform,the components of the time-frequency representation that are most relevant to the signal dynamic features are retained and other components are discarded,and high-order synchrosqueezing transform can be constructed to generate time-frequency distributions with high time-frequency resolution.The high-order synchrosqueezing transform can more accurately characterize the time-varying law of the strong AM-FM signal,and with the increase of the order,the extracted features will be more accurate and the computational burden will be heavier.The feature extraction of gravitational wave signal and the fault diagnosis of bearing outer ring can verify the effectiveness of the proposed method in the processing of signal with strong amplitude and frequency modulation.(3)To solve the problem that weak fault feature components are difficult to be characterized effectively and are greatly affected by noise,the idea of parameterization is introduced to propose a high-order synchroextracting chirplet transform.On the basis of chirplet transform,a series of time-varying demodulation parameters are first constructed by using the idea of general linear chirplet transform,and these time-varying modulation parameters are introduced on the basis of short-time Fourier transform,so as to obtain a parameter-adaptive chirplet transform,which can generate a compact time-frequency representation when address a multicomponent signal.Moreover,a high-order frequency estimation operator is constructed based on the result of the parameter-adaptive chirplet transform,and then the high-order synchroextracting chirplet transform is obtained by using the idea of synchronous extraction operation.This method can accurately characterize the time-varying law of weak fault features,at the same time,it has good noise robustness.(4)To solve the problem that the adjacent fault feature components are difficult to be effectively separated,a multi-scale chirplet synchroextracting transform is proposed in this dissertation.Through the second-order optimization of the phase kernel function in the chirplet transform,the window is rotated twice within a window length,so that the window function fit the varying instantaneous frequency ridge better over the entire time-frequency domain.Finally,a multi-scale chirplet synchroextracting transform is obtained by retaining the features most relevant to the instantaneous frequency of the signal.The proposed method is applied in the simulation analysis of the sun gear fault signal and the fault diagnosis of the rotor-rub impact in the industrial field.The results show that the proposed method can well characterize the dynamic features of the adjacent fault feature components,and at the same time,it is suitable for the processing of measured vibration signals.(5)It is often difficult to effectively extract the transient impact features of rotating machinery such as bearing.In this dissertation,a local maximum time-reassigned synchroextracting transform is proposed to accurately capture transient fault characteristics.The transient shock caused by the fault often occurs in a very short time,thus it has a wide frequency band,which means that the time-frequency model in the traditional time-frequency analysis method cannot well extract the transient shock components caused by bearing faults.Based on the time-reassigned synchrosqueezing transform,this dissertation constructs a new group delay estimation operator by searching for the local maximum of the time-frequency distribution in the time-frequency domain,and then performs the synchronous compression operation in the time direction.The energy concentrated time-frequency representation can be obtained,and the moment when the transient shock occurs can be accurately captured.It provides a more physical explanation for the fault diagnosis of mechanical equipment.