Comparison of metaheuristics for solving multi-product multi-stage multi-machine scheduling problems

  1. Lookup NU author(s)
  2. Dr Pupong Pongcharoen
  3. Professor Christian Hicks
Author(s)Pongcharoen P, Chainual A, Khadwilard A, Kaweesirikon C, Hicks C
Editor(s)Grubbstrom, RW; Hinterhuber, H
Publication type Conference Proceedings (inc. Abstract)
Conference NameFifteenth International Working Seminar on Production Economics
Conference LocationInnsbruck, Austria
Year of Conference2008
Legacy Date2-7 March 2008
Volume
Pages
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
Capital goods companies manufacture complex products with deep and complex product structures that give rise to many stages of assembly relationships. They schedule multiple products in a multi-stage, multi-machine system (MMMS) environment. Feasible schedules must correctly sequence the operations required for manufacturing components and also satisfy assembly precedence relationships. This paper presents a comparative study of the application of Genetic Algorithms, Particle Swarm Optimisation and the Ant Colony System for solving MMMS production scheduling problems in multi-product, multi-stage, multi-machine environments. The algorithms were designedaimed to minimise the combination of earliness and tardiness penalties. Four different sized scheduling problems were obtained from a collaborating company that manufactures complex capital goods. Simulation experiments were designed to test and evaluate the performance of the metaheuristics for solving these problems. The performance of the algorithms was compared in terms of the minimumean penalty costs obtained and the computational time required. The experimental results indicated that the efficiency and effectiveness of the algorithms was close particularly for small, medium and large problems. However, for the extra large problem, the Ant Colony System produced the lowest penalty cost but it required three times more computational time than the other algorithms.