Reliability Assessment Method For Space Rolling Bearing Based On Condition Vibration Feature

The space rolling bearing is the key component in the space motion mechanism,and its operating state directly affects the performance of the entire space motionmechanism. Practice show that the failures in space motion mechanism mostly comefrom the bearing failure. However, the bearing cannot ensure the reliability with thebackup because the application of bearing in the space occasions is limited. The spacebearing bears alternating temperature, the combined effects of the atomic oxygenerosion and other extreme environments, this condition can easily result in the failure ofaccuracy of the space rolling bearing and accelerate the space rolling bearing damage.In order to ensure the operating safety and reliability of space rolling bearing, and avoidmajor accidents, reliability evaluation for space rolling bearing are necessitated to studyin the vacuum.The traditional reliability assessment method, which used the probability theoryand the mathematical statistics theory as the main mathematical tools, determined thefailure distribution of equipment and got average reliability of a number of equipmentwith the use of large amounts of repetitive failure samples. However, each of spacerolling bearings typically operates under different conditions and environments. Thedamage degree, the failure degree and the fault degree is varied. Therefore, operationalreliability is also bound to be different. Operational reliability assessment of thespecified rolling bearing is an individual problem. The average reliability, obtained byusing probability statistics method with large sample data, cannot fulfill therequirements of the operating reliability evaluation for single rolling bearing.Due to the state characteristics of bearing can provide important information forreliability assessment, based on state characteristics, the reliability modeling andanalysis techniques is an important way to solve the reliability evaluation need for asingle space rolling bearing. At present, the characteristics that reflect the state ofbearing are friction torque, vibration and temperature. However, the friction torque andtemperature can not effectively reflect the changes in the state of space rolling bearinglifetime，therefore the paper chosen the vibration signal which indulgences the rich lifeevolution information to act as the state characteristics to assess the operating reliabilityfor rolling bearing. Proportional hazard models is one of the most commonly reliabilityassessment method which is based on condition vibration feature. The key of reliability analysis and evaluation based on condition vibration feature is to extract thecharacteristic indicators which can reflect the operating status of space rolling bearingand determine model between the characteristic indicators and reliability. Based on thecharacteristic indicators extracted by the real-time vibration data, the reliability ofrolling bearing can be assessed by the model. Meanwhile, combined with performancedegradation trend prediction theory, the reliability trend is predicted on the basis of theestablished proportional hazard model. The main works of this paper are as follows:①The primary problem for reliability modeling and analysis based on conditionvibration feature is the problem of characteristic indicator construction; therefore, thepaper studied the indicator construction methods based on the integration ofmulti-domain feature set. Time domain features, frequency domain features,time-frequency domain features and weibull feature is extracted to form themulti-domain feature set, and the manifold learning method is used to merge theoriginal features and reduce the dimension, so it can solve the conflict and redundancyproblem between the feature set.②Aiming to the operating reliability evaluation requirement of single rollingbearing, this paper proposes the weibull proportional hazard model reliabilityassessment method based on condition vibration feature. This method can overcome theproblem that traditional reliability method can not evaluate the reliability of singleequipment. The extracted feature act as the weibull proportional hazard modelcovariates and the model parameter can estimate with the use of maximum likelihoodestimation. The operating reliability can be assessed by the established model.③Aiming to predict the reliability trend for space rolling bearing, this paperpropose the reliability trend prediction method based on performance degradation.Based on the extracted feature, the LS-SVM model is constructed and trained foraccurately predicting the performance degradation trend. Then, the degradation trendresult is substituted into the established PH model so as to predict the reliability trend.④On the basis of the above theory, trend prediction and operating reliabilityevaluation module is developed with C#as the development platform and verified theeffectiveness and engineering application of the system by examples.