A Simulation-Based Method for Capacity Adjustments in Job Shop Production Scheduling

生産スケジューリング業務においてスケジュールに納期遅れジョブが発生した場合, 計画担当者は現有の設備・工数内で遅れジョブを出さないようスケジュールの改善を試みるが, 遅れジョブの発生が不可避と判断した場合には, 工場の保有能力を上位の計画部署より認められた許容範囲内で追加し遅れジョブの削減に努める.しかしながら, 最小の追加能力で納期遅れジョブ数を最小にするには, どの期間, どの設備に, どの程度の能力を追加すればよいかを決定することは難しい.ここではパラメータ空間探索改善法を能力追加を決定する問題に適用し, シミュレーション法に基づく系統的な能力調整法を提案する.さらに, 現実の計画条件データを用いてその方法の有用性を示す.\nWhen tardy jobs are involved in a production schedule, an improved schedule is sought for to eliminate tardy jobs under the given available capacities. When the occurrence of a tardy job is judged to be inevitable, the reduction of tardy job is attempted by increasing resource capacities within the allowances granted by the upper-level planning department. However, it is generally difficult to decide the capacities when, to which machine and what amount, to add so as to minimize of the number of tardy jobs after suppressing the added amount of the capacity. In this paper, a systematic simulation-based method for capacity adjustments is proposed which applies the Parameter Space Search Improvement method to the determination of additional capacities. Two parameters are introduced to span the parameter space. One parameter is used to control the total amount of adding capacities, and another parameter to control the distribution of capacities among resources. The best capacity adjustment plan is sought for on the parameter space. In the proposed procedure, the resource utilization under the given (initial) planning condition is first investigated in order to setup the relationship between the parameters and the variables which decide the amount of capacity addition to each resource. As many production schedules as the possible combination of capacity adjustment plans are then generated by the simulation-based scheduling method, and incorporating the cutting off of redundant capacities, the capacity adjustment plan which gives the schedule with the minimum number of tardy jobs is selected as the best capacity adjustment plan. The practical usefulness of the proposed method is demonstrated by using scheduling condition data on a practical large-scale system