Determining Optimal Machine Replacement Events with Periodic Inspection Intervals.

摘要

This research will examine the optimal maintenance and replacement policies for a generic machine with periodic inspection intervals. The considered reliability models consist of a single machine that can fail during operation or else may be found to be inoperative during regularly-scheduled maintenance inspections. A distinction will be made between spontaneously- occurring failures during operation and those that are discovered during inspections. Since the elapsed time between inspections is constant, the resulting stochastic reliability process becomes non-Markovian, and thus a Semi- Markov Decision Process (SMDP) framework must be employed in order to determine the cost-optimal stationary policy consisting of repair and replace decisions and inspection intervals. Using the methodology developed here, a system controller will be able to readily develop an inspection-based strategy to optimize the overall costs of maintaining systems with a variety of failure characteristics over a finite time horizon.