2000 LP-based heuristics for scheduling chemical batch processes.pdfVIP

LP-based heuristics for scheduling chemical batch processes International Journal of Production Research, 38 2000 , No. 5, 1029-1051 Ferdinand Bl?mer and Hans-Otto Günther Dept. of Industrial Management, Technical University of Berlin, Wilmersdorfer Str. 148, D-10585 Berlin, Germany Corresponding author: Prof. Dr. H.O. Günther, Dept. of Industrial Management, TU Berlin, Sekr. WIL-B-1-1, Wilmersdorfer Str. 148, D-10585 Berlin, Germany Tel.: +49-30-314-22669; fax: +49-30-314-21672; E-mail: HO.guenther@ww.tu-berlin.de Abstract: A mixed-integer linear programming MILP model for scheduling chemical batch processes is presented. Since computational times are prohibitive for most problems of realistic size, a two-stage solution procedure is suggested. In the first stage an initial solution is derived by use of a LP-based heuristic. The proposed heuristic defines a time grid which includes only a limited number of feasible periods in which a processing task is allowed to start. Thus, the size of the original multi-period MILP model is reduced in a controlled manner and optimal solutions to the relaxed model are obtained within reasonable computational time. The second stage consists of an improvement step which aims to compress the initial schedule by left-shifting operations over the time-axis. In order to evaluate the applicability of the heuristics a number of numerical experiments were performed. It is shown that near-optimal solutions are obtained for large size problems with only modest computational effort. Zusammenfassung : Auf der Grundlage des ?State-Task-Netzwerk“ Konzeptes wird ein gemischt- ganzzahliges Optimierungsmodell für die Ablaufplanung von Batchprozessen in der chemischen In- dustrie entwickelt. Da die Rechenzeiten für realistische Problem dieser Art zumeist nicht vertretbar sind, wird ein zwei