派工法則對加班模式影響之分析

Abstract

時至今日，企業面臨快速變化的市場需求，無不致力於改善作業流程，以更低的價格提供更好的產品與服務，其中尤以現場的生產排程問題較為複雜而難以管理。現行的軟體系統如企業資源規劃（ERP）即希望能夠藉由資訊科技解決此類生產排程問題。然而，這些系統皆並未考慮到企業的產能是有限的，往往造成規劃的結果不可行。 目前大多數的現場排程問題為零工式生產排程（job shop scheduling）問題，派工法則（dispatching rule）因具有簡單易懂及求解快速的特性，為求解零工式生產排程問題常用的方法之一。但使用派工法則得到的排程結果並不一定能滿足所有生產排程的目標，例如無法滿足產品的交期。在這種情況下，企業常使用加班使工件能在交期之前完成，這與過去研究僅使用派工法則求解一次即結束不同。本研究探討不同的派工法則產生的排程起始解對加班模式的影響，同時針對過去加班模式的缺失提出修正的方法。實驗模擬考慮機器數、工件數、最大操作時間、交期鬆緊因子（due date allowance factor）、每日最大加班限制等5個實驗因子設置，利用總延遲時間、延遲工件數、平均流程時間3項績效評估指標，對55個派工法則進行加班前後的績效評估。 實驗結果顯示，大多數加班前表現較佳的派工法則在加班後的表現亦不差，少數加班前表現較差的派工法則可透過較大的加班量達到與加班前表現較佳的派工法則接近的表現，但加班帶來的成本不可忽視。在交期設定較緊的情況下，我們發現SPT.TWK及SPT.TWKR等派工法則可在多項績效評估指標上皆有十分理想的表現。未來的研究者可將本研究延伸至不同的生產系統，如流程式生產（flow shop）等。此外，加班的成本及可能與正常工作時間不同的工作效率亦可加入考慮。

Due to the rapidly changing demand, corporations now put more emphasis on improving their workflow in order to provide better products and services with lower prices. The production scheduling problem is especially complicated and difficult to manage. Although current software and information systems, such as ERP, are designed to solve this kind of problem, they neglect the limitations of capacity, and usually produce infeasible planning results. At present, most scheduling problems are job shop related. Dispatching rules, which are easy to understand and efficient, are one of the most common approaches to solving job shop scheduling problems. Nevertheless, the results generated by dispatching rules do not necessarily meet the goals of production scheduling. For instance, the due date of products may not be satisfied. In this case, having employees work overtime has been commonly used as a way to meet the delivery schedules, which is different from previous research where the scheduling problem was only solved by dispatching rules. In this study, the effects of different initial solutions resulting from different dispatching rules on overtime models are explored. The shortcomings of previous overtime models are also identified and a modified overtime model is proposed in this work. The experimental setting takes five experimental factors into account: the number of machines, the number of jobs, the maximum processing time, the due date allowance factor, and the maximum daily overtime. Fifty-five dispatching rules are evaluated both before and after overtime by three performance metrics: total tardiness, number of tardy jobs, and mean flow time. The experimental results show that most dispatching rules that outperform others before overtime usually also perform well after overtime. On the other hand, by using more overtime, some of the dispatching rules that perform poorly before overtime can catch up with those performing well before overtime. However, the cost of overtime cannot be neglected. In the case of tighter due date settings, we observe that SPT.TWK and SPT.TWKR outperform other dispatching rules both before and after overtime in terms of the three performance metrics. In future implementations, our model could be extended to different production systems, such as flow shops. In addition, the cost and efficiency of overtime compared to normal working hours could be considered.