| Solid lubricated rolling bearing is the key component in space moving parts,whether the spacecraft like satellite and space station,can normally work to achieve a predetermined function as well as reach the life expectancy depend largely on the performance,life as well as the reliability of rolling bearings in the mechanical parts of spacecraft.As it is a challenge to realize the on-line monitoring of rolling bearing state information,qualification test,including failure test and censored test,must be conducted for the rolling bearing in simulated space conditions in order to assess and predict the bearing lifetime state,and to establish the database for the high-reliability long-lifetime bearings,so as to achieve the selection of the ’best’ bearing within the candidate rolling bearings.Now,the friction torque and temperature are the commonly used ways of monitoring the sate of solid lubricated rolling bearing.However,both the friction torque and the temperature bear less characteristic information about bearing lifetime,and are insensitive to the bearing degradation when the bearing is still running in its normal stage,so they show difficulties in timely reflecting the bearing performance degradation.Vibration monitoring,which is a way of monitoring the sate of solid lubricated rolling bearing,has received increasing attentions,and shows potential for indentifying and predicting bearing lifetime state.On the one hand,the vibration characteristics in the useful life stage of bearing is much feebler than that of the signals collected when typical or weak defect has appeared on the bearing.On the other hand,bearing vibration signal collected in simulated space conditions are contaminated by strong environmental noise and interference from the assets that used for maintain simulated space conditions such as vacuum pump,and from the other bearings operating in a same vacuum tank.The weak bearing feature is difficult to be extracted from the signal that contaminated by heavy noise.When the solid lubricated bearing in simulated space conditions is preload with axial position preload,as the bearing wears with time,the clearance between bearing parts increases and the preload level changes;Also,due to the volume restriction,the loop control of the bearing speed can hardly be achieved;Bearing vibration features are sensitive to bearing operating condition,degradation feature with trend for bearing subjected to variable conditions is hard to be extracted.Moreover,bearing test in simulated space conditions needs harsh conditions,and a great deal of human and material resources are often necessary to be invested,thus only limited samples in partial conditions can be obtained,it is a problem that how to construct high-precision prediction model under the condition of small-samples.Because of these characteristics of the bearing life test in simulated space conditions,it is a highly challenging task to achieve the lifetime state identification and life prediction based on the vibration analysis.Aiming at the problems that exist in the lifetime state identification and life prediction of bearing in simulated space conditions based on the vibration analysis,this paper researches on the weak feature extraction method based on adaptive signal component separation,the method of degradation trend feature extraction for bearing under variable conditions,and the construction of high-precision prediction model under the condition of small-samples,the specific contents are as follows:1)Aiming at the problem of extracting the weak bearing feature from the signal with heavy noise and interference,a feature extraction method based on adaptive signal component separation is proposed.The vibration signals of bearing in simulated space conditions are analyzed,and the source of the background noise is believed to contain the test system noise and the interference that arise from the assets used for maintain simulated space conditions as well as from other bearing platforms.The original signal is filtered using a set of band-pass filters with overlapping passband,a maximal-spectral kurtosis-minimal-redundancy criterion is proposed for the selection of the energy-dominated components.According to the empirical mode decomposition(EMD)characteristics of adaptive decomposition and filtering,the selected filtered signal are further decomposed to realize the adaptive signal component separation.By calculating the correlation coefficients between the IMFs of the noise bearing vibration signal and the noise signal itself,the correlation coefficients between the IMFs of the noise bearing vibration signal and that of the background noise,partial IMFs are selected and used to reconstruct the signal,the noise and interference can be effectively restrained,and the bearing weak feature can be extracted from the signal with heavy noise and interference.2)Aiming at the problem of extracting trend degradation feature for bearing under variable conditions,a method of quantification feature extraction based on signal amplitude normalization technique and feature similarity measuring is proposed.The relationships between bearing vibration and speed,as well as load are investigated,the amplitude normalization technique is proposed to reduce the effects of the variation of speed and load on the amplitude of the vibration.By taking the advantages of the time-frequency analysis in characterize the local characteristics of signal,two feature extraction methods are proposed: feature extraction based on subband analysis and feature extraction with reduced time-frequency representation.Quantification feature is extracted for rolling bearing based on the similarity measuring,accordingly to achieve the same scale evaluation of rolling bearing degradation under variable conditions and to realize the extraction of the degradation trend feature.3)Aiming at the problem of constructing the high-precision prediction model under the condition of small-samples,a method which can fuses censored histories for rolling bearing life prediction is proposed based on the optimized multiple kernel least square support vector machine.In simulated space conditions,due to the restrictions of test cycles,life test are conducted on only few bearings,and censored test is performed for relatively more bearing.For the limited life test samples,a method for rolling bearing life prediction is proposed based on multi-scale mutation particle swarm optimization and multi-kernel least square support vector machine.For the relatively large number of censored histories,in order to utilize the valuable degradation information within censored histories,a model based on functional principal component analysis is constructed to depict feature trends,and a method for the lifetime estimation of the censored sample is proposed based on similarity analysis.Finally,a prediction model which can fuse the censored histories is constructed and used for bearing life prediction,and to improve the prediction accuracy of the bearing in simulated space conditions.4)According to the characteristics of the bearing life test in simulated space conditions,a prognostic system for solid lubricated rolling bearing in simulated space conditions is developed.The developed system realizes vibration data acquisition,signal analysis,feature extraction,life prediction and system management functions,and can be used to realize the bearing lifetime state identification and life prediction,accordingly to achieve the selection of the ’best’ bearing.The developed system has been used for the lifetime state identification and life prediction of the solid lubricated rolling bearing in simulated space conditions.At the end of the thesis,the work of this paper is summarized,and expectation of the relative technology development is presented. |
PDF Full Text Request