An Open Queueing Network Representation of the Reparable Item Problem.
HOUSTON UNIV TX
Pagination or Media Count:
This research examined various approaches to the reparable item problem and demonstrated significant shortcomings in those approaches which hamper their effectiveness. The METRIC-based approaches had problems handling the variability of empirical data and state dependent behavior. Queueing approaches ran into significant state space problems when they attempted to solve realistically sized problems or address complex issues. We develop a new paradigm for the reparable item problem which abandons stock levels as decision variables in favor of depot allocations of repair funding. It also assumes that the depot repair process is not constrained by the availability of unserviceable assets or by workshop capacity. We use this new paradigm to propose an open queueing network representation of the repair process. This generates an item availability probability distribution function, thus opening up a broad range of different objective functions. We demonstrate a set of specific techniques for creating these representations from empirical data with a U. S. Air Force data set. In a comparison of the fitting and forecasting performance of our model out performed a METRIC based model 38 out of 40 comparisons. Based upon the queueing network representations generated in this demonstration, we developed a global marginal allocation model to determine the best allocation of the depots repair funding between competing bases and competing items at each of the bases.
- Operations Research
- Logistics, Military Facilities and Supplies