A simulation of readiness-based sparing policies

Abstract

We develop a simulation to complement a new optimization tool that establishes inventory levels for aviation weapon systems (WS) in the U.S. Navy. The optimization seeks cost minimization while achieving required readiness rates for hundreds of WS, each comprising thousands of indentured parts. Based on work in similar realms, the optimization employs the Vari-Metric model and a variant of a greedy heuristic algorithm to set stock levels and estimate overall WS availability. Our discrete event simulation is then used to test the assumptions of the new optimization tool, compare its performance to other optimization tools available, and provide additional metrics for decision makers. In testing the new optimization tool, we find that (a) there is no systemic bias in estimated readiness; and (b) 53 of 64 WS simulated yield results within 5% difference, with a worst-case difference of 8%. We also test two legacy optimization tools currently in use by the Navy and find they have a larger difference in expected readiness. Finally, we demonstrate additional insights and metrics from the simulation that are not available in the optimization tools.