Research On Complicated Multicomponent Signal Analysis And Its Application In Rotating Machinery Fault Diagnosis

Rotating machinery is a vital part of mechanical equipment,and its running status directly affects the service performance of the equipment.Once faults occur in gears,bearings,rotors,or other key parts,they will reduce the efficiency of entire machines,or even lead to economic losses or casualties.Vibration signals from rotating machinery are rich in information about the running status of the key parts.Therefore,it is crucial to accurately extract fault characteristics from such signals to ensure the normal and safe operation of mechanical equipment.However,the vibration signals of rotating machinery contain complicated frequency components,such as integral or fractional harmonics of rotational frequency,resonance frequency components,etc.Meanwhile,as rotating machinery often works under time-varying speed conditions,the constituent components exhibit intricate characteristics,such as being close to each other,spectral overlapping,or even crossing on the time-frequency plane.Thus,it is difficult to extract fault characteristics from the multicomponent vibration signals of rotating machinery with complicated characteristics.In this dissertation,aiming at the difficulties of accurately extracting constituent components and depicting time-varying features,we summarize the characteristics of the vibration signals from the key parts of rotating machinery and then carry out the research on the analysis methods for the complicated multicomponent signals by taking the extraction of sensitive components rich in fault information and the depiction of time-varying characteristics as the grip.The main research contents are as follows:Focused on the difficulties of accurately extracting fault signatures and deterioration features due to the complicated frequency components,an adaptive spectral segmentation and analysis method for complicated multicomponent signals is studied.On the one hand,to simplify spectral structures and extract fault signatures,an adaptive spectral segmentation and modulation feature analysis method is proposed,which solves the core parameter setting problem in the spectral segmentation decomposition methods such as the variational mode decomposition,and develops a sensitive component screening and analysis strategy.On the other hand,to obtain the deterioration features,a mapping relationship between the constituent components and the parts is established,and a method of feature component extraction and feature signal construction is proposed,thereby achieving the goal of characterizing part running status via statistical indicators of feature signals.Aiming at the difficulty of the constituent component extraction from the multicomponent signals with spectral overlaps,an adaptive decomposition method is proposed based on spectral segmentation.The spectral segmentation decomposition methods such as the empirical wavelet transform have fine anti-noise performance,but they are subject to mode mixing and integrity issues when extracting constituent components from multicomponent signals with spectral overlaps.Combined with the advantages of spectral segmentation decomposition methods and angular domain resampling,the mode mixing and integrity issues are addressed,the constituent components are accurately extracted,and the time-varying characteristics of multicomponent signals with spectral overlaps are accurately depicted.The effectiveness of the proposed method in the extraction of overlapping frequency components and the characterization of time-varying features is validated through analyses of four data sets(including a simulated multicomponent signal with spectral overlaps,two experimental signals of a wind turbine planetary gearbox,a ground test signal of a civil aircraft engine,and a field measurement signal of a hydraulic turbine).For the multicomponent signals consisting of crossing frequency components,focused on the time-frequency blurs caused by the close distribution of constituent components in linear time-frequency post-processing transforms,the time-frequency post-processing methods get rid of the restriction of sufficient component spacing by exploiting the capability of Vold-Kalman filter to extract constituent components.Furthermore,we propose the general parameterized synchrosqueezing transform under the signal decomposition framework to accurately reveal the time-frequency signatures of the multicomponent signals containing both crossing frequency components and frequency-modulated components.The effectiveness of the proposed method in suppressing noise artifacts and alleviating time-frequency blurs is validated through analyses of three data sets(including a simulated signal,an experimental signal of an induction motor,and a field measurement signal of a hydraulic turbine).The proposed method is able to reveal time-varying features of the multicomponent signal consisting of crossing frequency components.For the nonstationary signals with close instantaneous frequencies,the proportional-extracting Chirplet transform(PECT)with high time-frequency resolution and high order approximation property is proposed to address the issue that the conventional time-frequency analysis methods cannot accurately reveal their time-frequency structures due to the limited resolution.By adopting proportional kernel functions,the proposed PECT greatly enhances the time-frequency resolution and the precision of time-frequency coefficients,and suppresses the time-frequency blurs.To further improve the discriminability of the constituent components with close instantaneous frequencies,we extend the PECT to the synchrosqueezing transform framework and propose the proportional-extracting synchrosqueezing Chirplet transform(PESCT).The proposed PECT and PESCT are validated through the vibration data from a fix-shaft gearbox,a planetary gearbox,and an in-situ hydraulic turbine.The analysis results show that the proposed methods can achieve concentrated time-frequency representations with high precision,and clearly depict time-frequency structures of the multicomponent signals containing the components with close instantaneous frequencies.