Beschreibung
Pipeless plants are an innovative batch production system in which the chemical substances are transported in mobile vessels between fixed processing stations. They are characterized by a reduced cleaning effort and by a high degree of flexibility. The parallel handling of several batches leads to shorter production times. The optimal operation of a pipeless plant is a highly complex combinatorial planning and scheduling problem. Decisions have to be made on the sequencing and timing of the batches, on the assignment of the equipment, and on the collision-free routing of the mobile vessels. The processing duration of an operation may depend on the chosen equipment and on the duration of the routing of the vessels. A method that finds near-optimal but yet highly accurate and feasible schedules taking into account the specific characteristics of the pipeless plant was missing in previous research. This thesis proposes a simulation-optimization framework that unites a problem-specific evolutionary algorithm (EA) with a discrete event simulator. It is the first optimization approach for pipeless plants that is based on a detailed plant model, which considers the schedule-dependent routing of the mobile vessels, and the variable durations of chemical processing steps. The framework was successfully applied to a wide range of problems, including an industrial case study. Compared to a heuristic as-soon-as-possible schedule, the makespan could be reduced by up to 17 %. The best constraint handling was a combination of repair algorithms that eliminate most infeasibilities of the candidate solutions generated by the EA, and a decoder that postpones infeasible recipe steps during schedule building. Feasible, high-quality schedules were obtained within few minutes of computation time, which makes the approach suitable for online applications. The algorithmic framework can be quickly applied to new problem instances without requiring an in-depth parameter tuning.
Herstellerkennzeichnung:
Shaker Verlag GmbH
Am Langen Graben 15a
52353 Düren
DE
E-Mail: info@shaker.de
