Abstract
Redundancy allocation problem (RAP) has been an active research area for the past decades. Generalized redundancy allocation problem (GRAP) extends it to a more realistic situation where the system can have a complex network structure, for example, its components are connected with each other neither in series nor in parallel but in some logical relationship. Because of that, solving GRAP has presented major challenges in practice. In this paper, we propose a simulation optimization method, called Nested Partitions for Reliability Optimization (NPRO), to solve GRAP efficiently. Due to a newly-developed partitioning strategy, NPRO can locate the optimal solution in an efficient manner. The incorporation of many useful techniques, including the proposed encoding approach, importance sampling (IS) and Latin hypercube sampling (LHS), further enables NPRO to reduce the number of simulation observations needed in the optimization process, facilitating quick generation of the optimal solution. An extensive numerical experiment is conducted to verify the efficacy and efficiency of NPRO. Results show that NPRO can find the optimal or nearly optimal solution of GRAP within limited computational budget and moreover, it significantly outperforms the other two existing algorithms.