Improvement Of System Reliability Through Redundancy Allocation

Redundancy allocation is a widely used method for reliability improvement.The optimal redundancy allocation scheme is generally obtained by solving a typical op-timization problem called the redundancy allocation optimization problem.However,two concerns somewhat restrict the applications of most redundancy allocation meth-ods to real world applications:lack of reliability data and the simplified assumptions when reliability data are available.When reliability data is missing or cannot be collected for real engineering prob-lems,allocating redundancy through quantitative methods(such as solving redundan-cy allocation optimization problems)becomes infeasible.In this case,the only availa-ble information of a system includes system functions,system structures,comprising components,and possible failures and the causes.In order to fully utilize the available qualitative information of the system,a qualitative redundancy allocation method should be proposed for redundancy allocation when reliability data are not available.When reliability data are available,more accurate redundancy allocation methods,like the quantitative redundancy allocation methods,should be applied.However,most of quantitative redundancy allocation methods are applied with two simplified assumptions:component independency and binary states.Specifically,the first as-sumption assumes that the components in a system are all independent and they do not affect each other.The second assumption assumes only two states for each com-ponent and the entire system,the fully operational state and the completely failed state.Although these two assumptions can simplify the redundancy allocation process,they sometimes lead to quite inaccurate reliability estimations and consequently result in doubtful or misleading redundancy allocation schemes,which can cause serious problems in reliability critical applications.Thus new methods should be proposed to relax those simplified assumptions when solving the redundancy allocation problems.This dissertation focuses on developing new redundancy allocation methods to solve these two concerns in redundancy allocation.This dissertation includes four parts of the work(corresponding to Chapters 3 to 6).In Chapter 3,a qualitative re-dundancy allocation method is proposed to provide optimal redundancy allocation schemes when reliability data are not available.The next three chapters proposed quantitative methods to relax these two simplified assumptions for redundancy alloca-tion when reliability data are available.Specifically,In Chapter 4,the first simplified assumption is relaxed and a method is proposed to solve the redundancy allocation problem with a special type of component dependency,the failure dependency.In Chapter 5,both simplified assumptions are relaxed and a method is proposed to solve the redundancy allocation problem with failure dependency and multiple states.Chapter 6 extends the work in Chapter 5 by proposing a method to solve the redun-dancy allocation problem with component dependency and multiple states.The com-ponent dependency discussed in Chapter 6 is a more general case which is not limited to the failure dependency discussed in Chapters 4 and 5.In all those four chapters,illustrative examples and case studies from real applications are presented to demon-strate the applicability of the proposed methods.