초록/요약

As an effective strategy to facilitate delivering customized products within short lead time, hybrid manufacturing via a two-stage process has received attention from academia and industry. In this paper, we study a two-stage hybrid manufacturing system in which semifinished products are manufactured in a make-to-stock fashion in the first stage and end-products are produced from semifinished goods in a make-to-order (MTO) mode in the second stage. The rate of MTO production can be controlled within given limits, depending on the status of the system. The primary goal of this paper is to study a policy for coordinating order admission, MTO production rate, and inventory replenishment controls. Formulating the problem as a Markov decision process model, we characterize the structure of optimal control policies to maximize the long-run average profit. Using a numerical experiment, we study how the flexibility in MTO production rate affects the optimal policy and the optimal profit. We also examine the effect of the number of alternative MTO production rates on the optimal profit. We propose three heuristic policies implementable for general cases. The first heuristic describes two linear switching functions for admission and production controls and a selection rule for MTO production rate control. The second heuristic specifies fixed thresholds for the control decisions using the local information. The third heuristic presents linear switching functions that approximate the optimal threshold curves. Unlike second and third heuristics, the first heuristic does not require a grid search to determine the control parameters. We implement numerical studies to examine the marginal impact of system parameters and the effect of the number of alternative MTO production rates on the performance of the heuristics. Compared to the optimal policy, the average percentage performance of the first and third heuristics is less than 1% for both numerical studies. On the other hand, the average percentage performance of the second heuristic is larger than 3%, and it exceeds 10% for a set of particular problem examples.