Research On Some Key Problems Of Reliability Design For Complex System

Reliability design of complex system is an analysis and evaluation technique for ensuring the high reliability,which not only affects the performance and economic utility of the product,but also the operation reliability and safety of user.In this thesis,several key problems associated with evaluation,prediction,optimization and allocation in complex system design are deeply researched,some theory methods and models are developed for supporting the optimization and design of the highly complicated system.The main content and results are lised as follows:(1).Storage reliability under periodic inspection and maintenance.For certain types of systems which spend most of their life in storage,the non-parametric estimates of failure probabilities at each testing time points are introduced to the parameter models which describe the reliability degradation,and the integrated technique about parameter and non-parameter is proposed to evaluate the inherent storage and initial failure probability.Based on the data of periodic inspection and maintenance,two kinds of models,improved as bad as old and improved as good as new,where the failure ratios in each storage periods keep unchanged or increasing,were proposed to evaluate the storage reliability.Because of the small sample of test and insufficient data of storage failure,the estimation of real-time reliability and Monte Carlo(MC)method are investigated in the case of failure time series of all and part of the components.The illustrative examples indicate that,the integrated techniques are superior to the traditional method for estimating the storage reliability,the models of periodic inspection and maintenance are reasonable and effective,and the MC method is reasonable and practicable under the case of insufficient data.(2).Reliability of single cold standby system consists of dependent components.Aiming at dependence among the component lives,the multidimensional copula is introduced to model the joint distribution function and describe the dependence of components.Using the conditional probability theory,the models for one cold standby and multiple active dependent series and parallel units are developed.Based on the FGM copula,the explicit expression of system reliability and MTTF are obtained.In the case of exponential life,the reliability function and MTTF of system are expressed as the linear combination form of exponential reliability under different failure ratios.By an illustrative example,the effects on the system reliability because of the different number of components and dependent degree are investigated.The results indicate that the reliability and MTTF of series system increase with the dependence degree,but decrease for the parallel system.(3).Evaluation technique for reliability of multi-state coherent system.Consider the monotone and coherent of multi-state system,the multiple discrete function theory is introduced to describe the structure function,and the logic approaches for lower bound points of upper equivalence class and upper bound points of lower equivalence class which control the state vector of system are proposed,and the expressions of state structure function,reliability and expected states are deduced for the multi-state coherent system.For avoiding the complexity of computation caused by the number of state,the Demogen law and the new Block diagram algorithm are developed to simplify the expression of reliability.Demonstrated by illustrative example about certain type of aero engine,the effectiveness of the logic vector measure which controlling the state equivalence class and the Block diagram algorithm is verified.(4).Reliability analysis for repairable multi-state phased-mission system(MS-PMS).For analyzing the component states which are not comparable in the MS-PMS design,the grouped stated and reverse order method are proposed for constructing the multi-state decision diagram(MDD)of MS-PMS.Using MDD to describe the state of system level and Markov measure to model the dependence and dynamic behavior,the hierarchical module method used to estimate the repairable MS-PMS including unordered state components are developed.Based on the analysis method for components with non-comparable states,the MS-PMS with comparable state components is also analyzed.An example for civil aircraft which carries out the flying missions through three different phases,i.e.,climbing phase,cruising phase and descending phase,is used to illustrate the effectiveness and practicability.(5).Optimum choice for the reliability of system-component using the discrete data of technics.First,the concept of importance weight associated with customer preference is introduced based on the discrete data of importance-reliability-cost.According to the equivalence structure about series,parallel,series-parallel and parallel-series system,the integer programming of reliability allocation for system-component,under the balance consideration for the importance of component and cost,are formalized.The developed optimum nonlinear programming is converted into linear forms using the relaxation technique,and the relaxation-genetic algorithm for searching the optimum solution of nonlinear programming is proposed based on the discrete data from different technics.Finally,the effectiveness and rationality are demonstrated through the illustrative example.