| In recent years,with the rapid development of science and technology,the performance of mechanical systems,mechanical and electrical systems and electrochemical systems is increasing,the systems become more and more complex.The performance of most systems will gradually degenerate in the complex conditions,and the failure state of the components is no longer be either one or the other,rather than fuzzy characteristics.In many cases,they present a variety of complex characteristics such as nonmonotone,fuzziness,dynamic and multi-state.At the same time,with the improvement of system performance,the service life will be longer and longer,and in the course of performance degradation,the characteristics of nonlinearity are presented,which are the fundamental causes of system complexity.Reliability refers to the ability of a product to perform its specified functions without fault during a certain period of time and under certain conditions.Residual life is the length of time from the current moment to the final failure moment in the work or storage condition.Reliability and residual life have been paid more and more attention as important indexes to measure the quality and performance of the systems and components.In engineering practice,there are many limitations in the analysis and diagnosis of deterioration systems using traditional reliability analysis methods and residual life prediction techniques.Traditional reliability analysis methodsusually consider the fault state and failure probability of the system and components as a static fixed value,which ignore the dynamic change of reliability index with time,so they can't accurately evaluate the reliability of the deterioration system.The traditional life prediction methods usually use the failure time data as the statistical analysis object and use the statistical criteria to determine the distribution model.Then the residual life expectancy of the product or system is predicted according to the established life distribution model.At present,the traditional life prediction method is facing some problems,which can be summarize in the following two aspects:(1)The degradation process is complicated.The performance and working environment of modern products are complex and changeable,and the lifedata of long-life products are difficult to obtain.State variables are not easy to observe during degradation,and the observed variables present a nonlinear relationship with state variables;(2)Accelerated stress is multiple.In a certain period of time,it is difficult to obtain sufficient failure data through the accelerated life testing from long life products;It is also difficult to accurately describe the effect of factors such as work environment on the performancedegradation of products by single acceleration factor.For meeting the requirements in engineering practice,this thesis aims to solve the issue of the reliability analysis based on Fuzzy Dynamic Bayesian network,the residual life prediction of the nonlinear state space model and the multi stress accelerated degradation model degenerate system in deterioration system.The main research contents and innovation are listed as follows:(1)Multi-state system reliability analysis method based on fuzzy dynamic Bayesian network.In this paper,a method of reliability analysis of fuzzy dynamic bayesian network multi-state system is proposed to solve the problems of fuzzy,dynamic and non-monotonicity of deterioration polymorphic systems.The method introduces the fuzzy set theory,combining the traditional fault tree analysis method with the Bayesian network,and taking the linear function into the construction of the fuzzy subset of the root node,establishes a dynamic fuzzy subset to describethe law of the failure probability of the root node over time.The methodmakes a comprehensive analysis of the fuzziness and dynamics in the fault information.The variation regularity of the fuzzy importance of each node is calculated,and the fuzzy multi-state CPT is used to describe the logical relationship between the components in a multi-state system.This method reflects the changes in the system's fault status with the increase of the running time of various components in the system.It can reflect the integrity of the information to a greater extent,and solves the relationship between components and system failure polymorphism,fuzzy dynamics and component failure.Finally,a case study ofthe lifting system of the lifting beam is presented to validate the proposed method,which further validates the effectiveness of the method and provides a new idea for the reliability analysis of the deterioration system.(2)Rapid Prediction of Residual Life of Degradation System Based on Nonlinear Statespace Model.In view of the nonlinear problems of the performance degradation in deterioration systems,this paper studies the law of system degradation using state space model.The paper establishes the relationship between observable variables and unobservable variables through state-space model,and studies its model parameter estimation method based on the degradation of system performance.Then this paper takes lead-acid battery as an example,analyzing its failure mechanismthoroughly,the battery capacity is quickly predicted by the variation of the discharge voltage,and the capacity degradation model is established by using the nonlinear state space model,which realizes the rapid prediction of the residual life.The main contribution of the method is presented when the key performance parameters are not easy to obtain or takes longer in a particular degradation system,the fast and accurate prediction of the key performance parameters is realized through the regular changes of its functional parameters,and then the prediction of residual life is realized by establishing the degradation model of key performance parameters.It presents a very good application value in practical engineering.(3)Residual Life Prediction of Degradation System Based on Multi-stress Acceleration Model.Aiming at the simultaneous influence of multiple stresses which are not simply superimposedin the process of engineering system degradation,this paperintroduces the Eyring model,polynomial acceleration model,generalized logarithmic linear model,Cox model and other typical stress acceleration models,and explicit the significance of multi-stress accelerated model for predicting residual life of deteriorating system.The modeling method of Wiener process degradation based on the multi-stressconstant acceleration is used to model the performance degradation under multi-stress acceleration,and to predict the residual lifeof the degraded system under multi-stress acceleration.The proposed method is applied on Lithium-ion batteries to demonstrate the effectiveness and validity of the proposed approach.In this case,the Wiener process is used to establish the capacity degradation model with the simultaneous influence of the temperature and discharge ratio,and the life distribution model and the residual life prediction method are given.Through this research work of this paper,it provides a better solution for the residual lifeprediction of the degraded system working in the multi-stress environment and has an important practical application value. |
PDF Full Text Request