経験知識と最適化手法を組み合わせた連続焼鈍設備のスケジューリング

抄録

The scheduling problem in Continuous Annealing Line (CAL) can be described as a combinatorial optimization problem. Because of the vastness of the search area and the complicated problem structure, it is difficult to solve the problem only using the optimization algorithms.
In this paper, a solving algorithm that combines the heuristic knowledge and optimization algorithms is presented and the new methodology how to use the heuristic knowledge in solving the scheduling problem is also discussed. The conventional expert system development methodology, which usually aims to built the expert problem solving model in detail and realize the model in the computer system, is nearly impossible to treat the numerous problem patterns and trial-error process that is necessary to solve the scheduling problem. The methodology described in this paper makes much account of practical use of the optimization algorithm and classify the heuristic knowledge considering its relation with the optimization algorithm in order to determine how to use the knowledge in constructing the solving algorithm.
The scheduling algorithm developed through the methodology can offer a useful solution ten times faster than the human expert and shows good maintainability.