Lifetime State Indentification Research For Space Rolling Bearings Based On Vibration Spectrum

The space rolling bearing is the key components in the space motion mechanism,the spacecraft such as satellite, space shuttle and spaceship, can run normally toachieve a predetermined function and achieve the life expectancy,which largelydepends on the performance,lifetime and reliability of rolling bearing in themechanical components of spacecraft. The rolling bearing service in space is subjectedto comprehensively effect by extreme environments such as low temperature andalternating temperature, high-energy particle radiation, atomic oxygen erosion, dustand erosion. So, there are great differences in failure behavior and mechanism withconventional environment. Therefore, in order to meet development needs of the highreliability and long-life of spacecraft, life evaluation and prediction for space rollingbearing are necessitated to study in the vacuum, and the research foundation is themethods of description and identification of the lifetime state for space rollingbearings,namely the lifetime state indentification for sapace rolling bearing. At present,the friction torque and temperature cann’t effectively reflect the changes in the state ofspace rolling bearing lifetime, and a new thought is obtained,which study the methodsof characterize lifetime state for space rolling bearings based on vibration. While thespace rolling bearing lifetime state changed, the amplitude of the bearing vibrationsignal in the time domain and the probability distribution would be changed, and thefrequency component, the energy of different frequency components, and the mainenergy peak position of the spectrum also would be changed. According to theproblems in characterization and identification for lifetime state of space rollingbearing, the main research work and conclusions are as follows:①For the issue of the characterization of the space rolling bearing lifetime state, themechanism of characterizing lifetime state for space rolling bearing is proposed. Thedifferent lifetime state and the degree of wear of the space rolling bearing is reflectedby the change of vibration signal characteristics. It reflects the space rolling bearingoperation state through the vibration signal features extracted in the time waveformand frequency spectrum. Therefore, the difference among different lifetime state can beindicated clearly. Finally, the accurately characterization and identification for thelifetime status of space rolling bearing is obtained. In order to automatic identify andfully reflect the lifetime state of space rolling bearings, comprehensive utilization of time domain and frequency domain characteristic parameters,16characteristicparameters of the time-domain and14frequency domain characteristic parameters areselected to construct the high-dimensional mixed-domain feature vector ascharacteristics of vibration spectrum for the space rolling bearing. Then, the lifetimestate eigenvectors of space rolling bearing is obtained. At the same time, the effect ofspeed and load for the space rolling bearing lifetime is analysed, and the vibration ofthe space rolling bearing is also analysed in this researched. Then, the methods andstrategies of complete life test for space rolling bearing is proposed;②According to the problems of de-noising and background noise filtering for thespace rolling bearing, the de-noising and filtering methods based on the ensembleempirical mode decomposition(EEMD) for the space rolling bearing are proposed. Thevibration signals of rolling bearing are obtained in the simulation vacuum environment.However they are disturbed by the noise and background noise, amonge which fromtest system and the background noise is from the equipment which maintain analogvacuum environment.In order to increase signal to noise ratio, accurately extract thelifetime state eigenvectors of space rolling bearing, two types of noise should bede-noised and filtered. For vibration signal de-noising of space rolling bearing, theensemble empirical mode decomposition (EEMD) can effectively suppress thephenomenon of mode mixing, and according to the product of the energy density ofintrinsic mode functions (IMFs) from the white noise by EMD and the correspondingaveraged period of IMFs is a constant, an automatic algorithm of choose IMFcomponents to reconstruct signal is designed, and an adaptive de-noising method basedon EEMD for vibration signal is proposed. For background noise filtered, according tothe filtering characteristics of EEMD, calculating the correlation coefficient betweenthe IMF component of the background noise and the IMF component of the vibrationsignal which includes background noise, then, the IMF component is selected based onthe correlation coefficient.At last, the background noise which come from the vacuumenvironment is effectively filtered out. Aim to the selection for two importantparameters, the amplitude coefficient k of the white noise and the number of ensembletrials, M, the adaptive selection methods for EEMD parameters is proposed, based onthe influence law of distribution uniformity on the signal extreme point because ofdifferent amplitude coefficient of wihite noise.In the end,the decomposition accuracyand computational efficiencys is ensured.③Aim to the problems of identification of lifetime state for space rolling bearing, the methods of lifetime state identification based on manifold learing is proposed. Thefeatures of imformation couping and time-varying are presented in the vibration signalof bearing, then, the features are presented in the high-dimensional lifetime stateeigenvectors, which go against classification and identification. Thus, thehigh-dimensional lifetime state eigenvectors is compressed based on the excellentperformance characteristics of compression and classification from manifold learning,the real manifold is extracted in the original observation spatial, and the goodclassification characteristics,high sensitivity and low-dimensional life stateeigenvectors is obtained.Then,the low-dimensional feature vector of training samplesand testing samples is inputed into the K-nearest neighbors classifier(KNNC), and thetesting samples is made classification decision of KNNC based on the neighborhoodand the class label information of training samples, at last,the lifetime state for spacerolling bearing is identified.④On the basis of research of this paper, the hardware platform of test system isbuilded based on the vibration characteristics and lifetime testing requirements forspace rolling bearing.The lifetime state identification system of space rolling bearing isimplemented, including signal acquisition signal, processing, lifetime stateidentification and others functions. This system can be used not only for managementof large data set, but also can be management of lifetime state sample database basedon database technology, and the integrity and security of lifetime state sample databaseis ensured. The system mainly includes: vibration signal acquisition module, vibrationsignal analysis module (including waveform analysis, probability analysis, correlationanalysis and spectrum analysis, vibration spectrum), waterfall analysis module, HHTanalysis module and life state identification module. At last, the lifetime state of spacerolling bearing is exact identified using this system and the result shows that thesystem is feasible and valid.