Abstract
We solve a variation of a classic make‐to‐stock inventory problem introduced by Gavish and Graves. A machine is dedicated to a single product whose demand follows a stationary Poisson distribution. When the machine is on, items are produced one at a time at a fixed rate and placed into finished‐goods inventory until they are sold. In addition, there is an expense for setting up the machine to begin a production run. Our departure from Gavish and Graves involves the handling of unsatisfied demand. Gavish and Graves assumed it is backordered, while we assume it is lost, with a unit penalty for each lost sale. We obtain an optimal solution, which involves a produce‐up‐to policy, and prove that the expected time‐average cost function, which we derive explicitly, is quasi‐convex separately in both the produce‐up‐to inventory level Q and the trigger level R that signals a setup for production. Our search over the (Q, R) array begins by finding Q 0, the minimizing value of Q for R = 0. Total computation to solve the overall problem, measured in arithmetic operations, is quadratic in Q 0. At most 3 Q 0 cost function evaluations are required. In addition, we derive closed‐form expressions for the objective function of two related problems: one involving make‐to‐order production and another for control of an N‐policy M/D/1 finite queue. Finally, we explore the possibility of solving the lost sales problem by applying the Gavish and Graves algorithm for the backorder problem.
Get full access to this article
View all access options for this article.
