Reliability-redundancy-location allocation with maximum reliability and minimum cost using search techniques

详细信息    查看全文

• 作者：Xiang Qiu^a ; ^d ; ^{qiuxiang008@gmail.com} ; Shaukat Ali^b ; ^{shaukat@simula.no} ; Tao Yue^b ; ^c ; ^{tao@simula.no} ; Li Zhang^d ; ^{lily@buaa.edu.cn}
• 关键词：Reliability block diagram ; Redundancy structure ; Search algorithms ; UML ; Optimization ; Empirical studies
• 刊名：Information and Software Technology
• 出版年：2017
• 出版时间：February 2017
• 年：2017
• 卷：82
• 期：Complete
• 页码：36-54
• 全文大小：3950 K
• 卷排序：82

文摘

A safety critical system requires an automated and optimal allocation of redundant component instances to its existing components, including: 1) the selection of components (locations) on which the redundancy must be applied, 2) how many redundant component instances of varying reliability and cost should be allocated to each selected location.ObjectiveOur work aims to searching for the near optimal allocation solutions achieving the higher reliability of the system within the allowed cost. Such allocation must be made earlier, for example, while designing the architecture of the system to avoid unnecessary complexity of addressing unsafe situations discovered in the system development and deployment phases.MethodWith the above objective in mind, we propose a search-based allocation approach based on the overall objectives of maximizing the overall system reliability and minimizing the cost of introducing and allocating redundancy structures to the system. The architecture of a system modeled using the Unified Modeling Language (UML) along with redundancy structures is encoded as an optimization problem. To guide a search algorithm to solve the problem, we propose a fitness function based on the two optimization objectives: high reliability and low cost.ResultsWe empirically evaluated the performance of four search algorithms (Genetic Algorithm, (1 + 1) Evolutionary Algorithm, Alternating Variable Method (AVM) and Random Search) together with the proposed fitness function on 10 real-world Subsea Oil&Gas Production Systems of varying complexity. Results show that the AVM algorithm significantly outperforms the rest.ConclusionBased on the results of empirical evaluation, we found that AVM can provide the best allocation of redundancy structures as compared to the rest of the algorithms. On average, AVM provided 0.008% of more reliability while saving 26.78% on allocation cost as compared to RS. Our novel solution based on the results of empirical evaluation is implemented as a software tool.