ArcelorMittal award for best thesis at Ghent University: Operations Research Rules!

ArcelorMittal award for best thesis at Ghent University.

The student Louis-Philippe Kerkhove has received the ArcelorMittal prize for the best thesis during the graduation ceremony for the Commercial Engineers (Handelsingenieurs) on 22/09/2012. The thesis was supervised by Prof. dr. Mario Vanhoucke from the Faculty of Economics and Business Administration (www.feb.ugent.be) from the Ghent University.

In his thesis "A study of exact and meta-heuristic planning techniques for unrelated parallel machines with common servers in the textile industry", Louis-Philippe has used various Operations Research techniques to solve unrelated parallel machine problems at a Belgian company. Through the use of simulation experiments, Louis-Philippe has shown how Operations Research can be helpful in detecting problems in planning and in suggesting improvements.

Every year, ArcelorMittal gives this prize to the best thesis at the Faculty of Economics and Business Administation. It is not the first time that this prize goes to a student who uses an Operations Research methodology to solve the problem in the thesis. Operations Research certainly rules, and not only at OR-AS!

Louis-Philippe will continue to work on this topic at the OR&S research group and will therefore contribute to the future development of OR-AS' software tool ProTrack.

Abstract of his thesis: This paper develops and applies exact and meta-heuristic solution techniques on a parallel machine scheduling problem at a Belgian knitted textile manufacturer. The production system of this manufacturer consists of multiple production locations with unrelated parallel machines. First a MIP model and both local and population based meta-heuristics are applied to the PMS problem without taking into account the limited number servers available for changeover. The objective of these optimization methods is a tailored goal function designed to maximize the economic value for the production company. Computational tests using real production information from the manufacturer show that both local and population based meta-heuristics significantly outperform the current scheduling methods when tested on a wide range of generated problems. The performance of the local search meta-heuristic (simulated annealing) appears to be superior to that of the population based meta-heuristic, both in terms of speed and in terms of solution quality. The second part of this paper introduces the limited availability of servers to the problem. The potential impact of this factor is examined and several heuristic methods to anticipate and avoid potential delays due to server availability are tested.