Research On Reliability Modeling And Remaining Useful Lifetime Prediction Of Multi-state System

Reliability,security,availability and maintainability are critical to process fluid mechanical equipment,and timely state detection and maintenance have become the necessary capabilities of modern equipment.However,it is found that the multi-performance states that characterize the running behavior of the system are assumed to be a working state by binary reliability theory.Moreover,these transition states before faults can reflect the future development of the system and predict the remaining service life.Additionally,process fluid mechanical equipment often consists of dozens or even thousands of different multi-state function components.Due to the environment factors(load and aging),interaction between equipment and components,the resulting state space exponential growth and uncertain problem are the bottlenecks of multi-state system reliability theory and application.Therefore,it is of great theoretical and practical value to carry out reliability modeling and remaining useful lifetime prediction of multi-state system.Based on the theory of modern test technology and forecasting method,this paper presents a series of problems in operational process,such as degradation,space exponential growth and uncertainty of performance state,from the practical engineering problems of process fluid mechanical equipment and the inherent scientific problems of multi-state system.The main contents of this paper are as follows:(1)Research on reliability measurement method of multi-state systemIn order to effectively characterize the reliability of multi-state system under degraded conditions,to meet the actual performance requirements of the system and to quantitatively describe the different parts of the function in the system,this paper makes a detailed analysis of the performance state,and puts forward four new dynamic reliability metrics(i.e.reliability and performance state dwell time,performance status expectations and performance state average expectation),and four importance calculation methods under performance restrition((Bimbaum importance,Fussel1-Vesely importance,risk achievement worth,and Risk reduction worth).(2)Research on reliability modeling method of multi state system based on monitoring dataIn order to solve the multi-state reliability theory application problem and provide mathematical description of its operation behavior,the basic definition of multi-state system model structure based on monitoring data,feature space construction method,the optimal degeneration feature selection method based on genetic algorithm and the mathematic mapping relationship between observing process and reliability model based on WPHM model are established.(3)Research on automatic segmentation of multi state system degradation stateAiming at the problem of optimal model structure selection and the uncertainty,curse of dimensionality,an automatic segmentation algorithm of degradation state,based on manifold learning method and spectral clustering algorithm,is given: the lifetime data is automatically divided into degraded states corresponding to the continuous region of the data sequence.(4)Research on state identification and remaining useful lifetime prediction of multi-state systemAiming at the problem of state identification and remaining useful lifetime prediction,SVM method is used to construct SVM classifier,so that state identification is implemented.The future development of system operation is predicted by SVR,and the prediction of remaining useful lifetime is implemented by inputing pseudo operation information from SVR into prediction function of WPHM.(5)Experimental verification of chemical pumpA test bench for multi-conditions of chemical pumps is built to monitor the flow,vibration,pressure,rotation speed and shaft power,and the accuracy and effectiveness of manifold learning and SVM multi-conditions identification method are verified.Taking bearing failure as an example,the reliability modeling,state segmentation,state identification and remaining service life prediction method of multi state system are validated,and the reliability measurement index of bearing by reliability measurement method is obtained.