ECCO 2024
Abstract Submission

8. The stochastic procurement and production lot-sizing problem: Models and a sample average approximation approach

Invited abstract in session TD-2: Heuristic scheduling, stream Heuristic scheduling.

Thursday, 14:30 - 16:00
Room: M228

Authors (first author is the speaker)

1. Caio Tomazella
Erasmus University Rotterdam
2. Raf Jans
Department of Logistics and Operations Management , HEC Montreal
3. Maristela Santos
Department of Applied Mathematics and Statistics, University of Sao Paulo
4. Douglas Alem
Business School, University of Edinburgh

Abstract

In this presentation, we approach the procurement and production lot-sizing problem with uncertain demand. We present three variants of the stochastic model: a static variant, in which all production and procurement decisions are made before demand realization; and two static-dynamic variants, where production is flexible enough to be delayed until demand is fully known. These models are complex and hard to be solved by a commercial solver, thus, in order to find high quality stochastic solutions, we propose an adjustable Sample Average Approximation (SAA) heuristic, which is divided into two phases. In Phase 1, the models are solved with a manageable number of scenarios several times in order to find the most prevalent decisions. In the models, these prevalent decisions are binary variables which have the same value in almost, if not all, solutions found. In Phase 2, these variables are then fixed, simplifying the stochastic model so it can be solved with more scenarios. The number of binary variables fixed in Phase 2 varies between 65% and 90%, depending on the parameters used, which explains the efficiency of the proposed method. Using the solutions obtained with the adjustable SAA, we evaluate the value of production flexibility, showing how much it affects the expected costs of the solution, and how these costs are distributed according to the possible demand realizations.

Keywords

Status: accepted


Back to the list of papers