Research On System Redundancy Optimization Problems Based On Different Coverage Models

Fault tolerance refers to the ability of a system to function properly even if some of its components fail.Adding redundant components to a system is a common fault tolerance method.However,due to the limitation of resources,there is an upper limit of redundancy in the system,which leads to the redundancy optimization problem.The study on redundancy optimization can be classified into two categories according to the assumption on fault coverage.The first category is based on the imperfect fault coverage model(IFCM),and system reliability may decrease as redundancy increases.Hence,the main purpose of this problem is to calculate the optimal redundancy of the system.The second category is based on the perfect fault coverage model(PFCM),and there is generally a monotonically increasing relationship between the reliability and redundancy of the system.Hence,the main concern is the trade-off between system reliability and resource constraints.This problem is also known as the reliability redundancy allocation problem(RRAP).In the IFCM,the coverage is limited to faulty components regardless of their irrelevancies.The recently proposed Irrelevancy Coverage Model(ICM),as an extension of IFCM,can cover both faulty and irrelevant components,but the solution for optimal redundancy in ICM is not provided.In addition,traditional RRAP research is limited to the perfect coverage model(i.e.,PFCM),and does not consider the problem in the imperfect coverage model(including IFCM and ICM).Also,the case where spare components can be shared between subsystems is ignored.In response to these problems,based on different coverage models,this thesis conducts system redundancy optimization research from the following three aspects:1)The closed-form solution based on the ICM is proposed in this thesis to calculate the optimal redundancy of the parallel-series system.The parallel-series system is chosen for two considerations.Firstly,it is a standard fault-tolerant system widely used in practice.Secondly,the parallel-series system contains components which are not always relevant in the system,which is exactly where the ICM can play a role.Three kinds of optimal redundancies for different objectives are investigated in this thesis,namely reliability maximization,cost minimization,and multi-objective optimization for both reliability and cost.2)New optimization objectives of reliability redundancy allocation problem are proposed based on the IFCM and ICM,and solved by using a genetic algorithm.To avoid the impact of the optimization algorithm as much as possible,the implemented GA is first compared with other studies under the same assumption.To accelerate the evaluation of system reliability,a two-stage GMDD-based method is also proposed in this thesis.The experimental results show that the time used to calculate the reliability of the system by the two-stage GMDD method can be reduced by more than 30% compared with the original GMDD,and as the number of components in the system increases,the effect of reducing the time consumption will be more obvious.3)To further improve the system reliability in RRAP,a redundancy strategy for sharing spares between subsystems is proposed in this thesis.When several subsystems use the same components,these subsystems can share their spares to improve the utilization of spares,thereby improving system reliability while the number of spares remains unchanged.The experimental results show that in the PFCM,the average improvement of this redundancy strategy is 39.10% compared with other strategies in the MPI(Maximum Possible Improvement)index,and in the IFCM the average improvement is 18.02%.