Article ID: | iaor19941801 |
Country: | Singapore |
Volume: | 8 |
Issue: | 1 |
Start Page Number: | 90 |
End Page Number: | 103 |
Publication Date: | May 1991 |
Journal: | Asia-Pacific Journal of Operational Research |
Authors: | Fukumura Satoshi, Yamakawa Eiki |
Keywords: | planning, artificial intelligence: expert systems |
This paper describes an expert system for berth planning developed to assist shipment planning at a steel works. The problem in berth planning is to compile daily shipment schedules, deciding ‘what cargo, when, by which crane, at which berth.’ Many factors must be considered, including the progress of the cargo in the production process, ship arrival schedules, weather, and agreed delivery dates, with lost crane and ship time minimized. This complex task has traditionally been assigned to experts, but standardization was desired to improve planning performance and allow for job rotation and the eventual retirement of experts. Observations and interviews are conducted, and a branch-and-bound model is chosen to structure the expert’s knowledge. In assigning each job on the schedule, a tree structure is supposed and expert heuristics are used in the new branching and bounding processes. With this paradigm, production rules are easily understood and refined. The system was put into practical use in April 1987, with satisfactory results in reducing demurrage and lost crane time.