On a Random Yield Problem with Non-zero Lead Times

Project: Research

View graph of relations


The proposed project has arisen from recent experiences with a pharmaceutical firm at which arriving materials must be sample-tested for microbial contamination control. If a sample test fails, the entire shipment of a material is disposed of or returned. That is, each order shipment is subject to so-called all-or-nothing random yield. A sample test usually takes a significant amount of time to run, in the range of multiple weeks, which may be much longer than the order delivery time. The firm needs to implement a decision-support system to guide its inventory replenishment of such materials.We will consider a single-item multi-period periodic-review inventory system in which the firm conducts a sample-test immediately upon the arrival of each order shipment. The total time taken by the supplier to fulfill an order and by the firm's sample test is significant, and we refer to it as the (order) leadtime. If the arriving order fails the sample test, the entire shipment is defective and is disposed of or returned. Demand occurs randomly in each period. The firm's objective is to minimize its expected total discounted cost, which includes purchasing, inventory-holding and shortage costs, by optimizing its inventory replenishment policy.In the extensive literature on inventory management in the presence of random yield and/or supply disruption, there are only a few reported results for problems with positive leadtimes. The structure of an optimal inventory control policy is an open area of research due to the challenge intrinsic to problems with random yield. Furthermore, an optimal policy structure will not be easy to implement, even if it can be fully characterized, and hence developing easily implementable heuristics and bounds is inevitable. However, characterizing the optimal policy structure will provide not only a benchmark for evaluating heuristics and bounds, but also insights into heuristic development and bound construction. Although a number of heuristics have been proposed and tested under a general form of yield randomness, little is known for the problem with an all-or-nothing yield pattern. The proposed project will aim to fill these two gaps.Preliminary analytical results are presented in this proposal, and a set of tasks for achieving the research goals are described. Our recent work on related problems and the techniques learned during those experiences have encouraged us to take up the challenge described here.


Project number9042129
Grant typeGRF
Effective start/end date1/01/1527/05/19

    Research areas

  • Inventory Model ,Random Yield ,Supply Unicertainty ,Stochastic Process,