Using Multiple Criteria Optimization and Two-Stage Genetic Algorithms to Select a Population Management Strategy with Optimized Reliability

Abstract

An important aspect of good management of inventory for many single-use populations or stockpiles is to develop an informed consumption strategy to use a collection of single-use units, with varied reliability as a function of age, during scheduled operations. We present a two-phase approach to balance multiple objectives for a consumption strategy to ensure good performance on the average reliability, consistency of unit reliability over time, and least uncertainty of the reliability estimates. In the first phase, a representative subset of units is selected to explore the impact of using units at different time points on reliability performance and to identify beneficial consumption patterns using a nondominated sorting genetic algorithm based on multiple objectives. In the second phase, the results from the first phase are projected back to the full stockpile as a starting point for determining best consumption strategies that emphasize the priorities of the manager. The method can be generalized to other criteria of interest and management optimization strategies. The method is illustrated with an example that shares characteristics with some munition stockpiles and demonstrates the substantial advantages of the two-phase approach on the quality of solutions and efficiency of finding them.