Rolling Element Bearing Fault Feature Extraction Based On Cyclostationarity Enhancement Theory

Rolling element bearings are widely used in rotating equipment because of their characteristics of low friction coefficient and high rotation accuracy.However,as an essential supporting component,the large dispersion of the bearing’s lifespan makes it one of the least reliable parts in rotating machinery,directly affecting the equipment’s working accuracy and operating life.Therefore,researching the state detection and fault diagnosis of rolling element bearings can effectively monitor and evaluate their health status,provide strong support for the safe and stable operation of the equipment,and reduce the economic losses,personal injuries,and social impacts caused by rolling element bearing failures.Against this background,how to extract the fault features of rolling element bearings from complex signals has always been one of the hot and challenging research topics in the field of fault diagnosis.This dissertation is based on the cyclostationary theory,with rolling element bearings as the research object,and vibration signals and encoder signals as the source signal,focusing on the engineering application of the cyclostationary method theory in the fault feature extraction of rolling element bearing under different working conditions based on the cyclostationary method is carried out in depth.In the research,combining theory research and engineering practice,the shortcomings of the cyclostationary algorithm and its limitations in engineering application are summarized.To address these issues,a series of schemes are proposed,the main contents include:(1)The conventional cyclostationary method is described in detail,and its shortcomings and limitations in engineering application are analyzed:(1)The selfadaptive determination of optimization parameters of the original cyclostationary method,specifically the determined of the optimal window width related to the shorttime Fourier transform.(2)Adaptive determination problem of optimal demodulation frequency band for cyclostationary analysis.(3)Suppression of cyclostationary interference components under time-varying transmission path,multi-source coupling and variable speed conditions.(2)To address the problem that the full frequency band envelope spectrum of the conventional cyclostationary algorithm can not effectively reveal the fault characteristics related to the rolling element bearings under low signal-to-noise ratio and low spectral frequency resolution conditions,the optimization demodulation analysis of cyclostationary method is studied.Firstly,to solve the problem that the decomposition ability of the conventional frequency band division structure depends on experience and can not adaptive obtain finer demodulation band group under low spectral frequency resolution conditions,a proportional band division structure is proposed,which has the advantages of low computational cost and high precision.Secondly,aiming at the failure of the traditional diagnostic feature index when the theoretical and practical characteristic frequencies of rolling element bearings are different,an improved diagnostic characteristic index is proposed to enhance the effectiveness and robustness of the traditional diagnostic index.Finally,based on the traditional cyclostationary analysis,the fault feature extraction of the rolling element bearings under low signal-to-noise ratio and low spectral frequency resolution conditions is realized by combining the proportional frequency division structure and the improved diagnostic feature indicator.The proposed scheme not only expands the application scope of cyclostationary algorithm in the field of fault diagnosis but also provides a valuable reference for solving similar demodulation analysis problems.(3)Two schemes are proposed to address the issue that the conventional cyclostationary method is difficult to effectively reveal the fault features of planet bearings under time-varying transmission paths and multi-source coupling conditions.(1)First of all,aiming at the problem that window width parameters of traditional cyclostationary analysis methods depend on experience,an adaptive parameter determination scheme based on improved diagnostic indexes was proposed.Then,to address the issue that spectral kurtosis algorithm is susceptible to the interference of multi-source coupling components,the preprocessing of cepstrum pre-whitening is studied.Finally,an enhanced cyclostationary scheme based on cepstrum pre-whitening and spectral kurtosis is proposed,which realizes fault feature detection of planetbearing inner race.(2)Firstly,to optimize the length of the maximum correlation kurtosis deconvolution filter,a variable scale parametric structure is proposed,which solves the defects of the traditional grid method.Secondly,aiming at the problem that the cyclostationary optimal demodulation analysis method fails in the feature extraction of planet bearings,a combined method of optimal maximum correlation kurtosis deconvolution and cyclostationary optimization demodulation analysis is proposed.Two schemes combine the advantages of the pretreatment method and the cyclostationary method,the advantages are complementary.Their complementary advantages enable the fault feature extraction of planet bearings under time-varying transmission paths and multi-source coupling conditions,which provides a valuable exploration experience for the application of the cyclostationary algorithm in the planetary gearbox,and the study provides important technical support for fault diagnosis and maintenance of planet bearings.(4)There are some problems in vibration signal,such as the lower frequency limit of vibration sensor,the limitation of the installation environment and the ambiguity of frequency spectrum under variable speed conditions,an exploratory research was carried out on fault feature extraction of rolling bearing using encoder signal(torsional vibration signal)as signal source.Firstly,aiming at the problem of wide parameter variation range and high precision in the Savitzky-Golay filter,a parameter decomposition structure with lower calculation cost and higher parameter decomposition accuracy is proposed for the feature extraction related to the rolling element bearing.Secondly,to expand the evaluation range of background noise in the original diagnosis index,an improved diagnosis index considering sideband modulation and random slip related to the rolling element bearing is proposed.Finally,aiming at the issue that the failure of the conventional cyclostationary method under the condition of strong amplitude modulation and frequency modulation,a combination of adaptive Savitzky-Golay and optimized cyclostationary is proposed for fault feature extraction of the rolling element bearings.Additionally,the IAS signal-based enhanced preprocessing method of rolling element bearing fault features is studied,and an adaptive down-sampling multi-period differential mean scheme is proposed to achieve the enhanced pre-processing of rolling element bearing fault features under low signalto-noise ratio conditions.