A Variable Neighborhood Descent Heuristic to the Redundancy Allocation Problems of Series-Parallel System

• Main Authors: Chia-Chuan Wu, 吳佳娟
• Other Authors: Yun-Chia Liang
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/67063840473436894454

碩士 === 元智大學 === 工業工程與管理學系 === 92 === Series-parallel system has been frequently used to maintain the reliability and security of a product or system. In order to increase the system reliability, the inclusion of the redundant component is a common approach. Unfortunately, such inclusion will often lead to increasing weight and cost of a system. Thus, when the system cost and weight constraints are given, how to effectively select the optimal combination of components and redundancy level for each type of components so that the system reliability becomes the maximum has been considered an NP-hard redundant allocation problem (RAP). The paper applies the algorithm of Variable Neighborhood Descent (VND) to solve RAP in the series-parallel system. That is, using the systematic neighborhood local search method to obtain the optimal redundancy allocation under both cost and weight constraints. This research also develops a hybrid algorithm called VND_SA in which an acceptance mechanism borrowed from simulated annealing is employed. A set of benchmark problem is tested to verify the effectiveness and efficiency of both algorithms. Our result indicate that percentage deviation from best-known solutions is the lower than 0.3% using VND algorithm. In addition, VND_SA can not help improve solution quality as expected though it did expand the search area. Our research concludes that VND is a simple and fast algorithm to obtain the near-optimal solutions for RAP and does not need complex parameters design.