具多功能機台與考慮途程規劃之混合流線型生產排程

Abstract

近年來受到新冠疫情的影響，製造業面臨著前所未有的難題，許多企業透過使用具更多功能的工具、導入新製程等方式面對疫情下應將不確定性視為新常態的市場。於生產排程層面，在一般的混合流線型生產排程問題 (Hybrid Flow Shop Scheduling Problem) 之上，建構出能求解具多功能機台、途程規劃特性問題的方法，能夠在產能與生產彈性提升的過程中，使新投入的技術與資源產出最大效益。將設置時間、機台效率等因素納入考量則可以縮小排程計畫與實際生產現場之偏差，避免排程計畫不可行或有效性較低的情況。除此之外，面對充滿不確定性的市場，對於求得排程結果所需時間的限制也相對嚴苛，必須能夠適時根據現場狀況變化更新生產參數，快速求得當下最適排程計畫，避免因等待求解而導致生產中斷。 本研究針對具有多功能機台與途程規劃之混合流線型生產排程問題，將獨立設置時間與機台、作業組合效率納入考量，提出混合整數線性規劃模型與基於模型的多階段啟發式演算法，以最小化總延遲時間為目標求解。於前者，本研究調整與改良過往學者所提出的模型，使用作業先後順序建立決策變數，建構出能夠求解上述問題的混合整數線性規劃模型。然而此作法所建立之模型規模非常大，面對實務問題無法於合理的時間內求解以協助生管人員進行排程決策。本研究進而提出基於模型的多階段求解法，利用許多生產現場存在之工件少樣多量的特性，將多階段加工流程拆分成數個求解階段，並使用工件類別取代作業先後順序建立決策變數以降低模型規模，達到縮短求得最適排程計畫所需時間的目標。本研究以數值分析與實務案例驗證證實使用此方法可以在求解時間限制較短的情況下求得較好的解，即使放寬求解時間限制，在實務規模的情境下其表現依然優於使用混合整數線性規劃模型與既有的基因演算法求解。

Affected by Covid-19 in recent years, the manufacturing industry is facing unprecedented difficulties. To adapt to the changing market environment, many companies develop multifunctional machines and import new processes. Scheduling methods that can solve the problem with multifunctional machines and routing characteristics can improve productivity and production flexibility. By using the methods described above, the use of investment resources can be optimized. Taking factors such as setup time and machine efficiency into consideration can reduce the deviation between schedule and the actual production environment, which can avoid infeasibility and ineffectiveness. In addition, facing an uncertain market, it is necessary to obtain the optimal schedule in a very short period of time. It enables manufacturing system to update parameters according to the on-site conditions in a timely manner and quickly generate current optimal schedule while avoiding production interruptions due to solving process. This research proposes Mixed Integer Linear Programming (MILP) and Model-based Multi-phase Scheduling Heuristic (MMSH) to solve the above-mentioned hybrid flow-shop scheduling problem with multifunctional machine and routing which take independent setup time and machine-operation based efficiency into consideration, with the objective function of minimizing total tardiness. In the MILP, this study adjusts and improves the model proposed by previous scholars, uses the sequence of operations to establish decision variables to constructs a mixed integer linear programming model that can solve problems with the above characteristics. However, the scale of the model established by this method is very large, and it cannot be solved within a reasonable time when facing practical problems. In contrast, MMSH takes advantage of the characteristics of the small variety of workpieces in many actual situations, divides the multi-stage machining process into several phases for solving, and uses workpiece categories instead of operation sequences to establish decision variables to reduce model size and the elapsed time. In this study, numerical analysis and practical case verification demonstrate that this method is better when the elapsed time limit is tighter. Even if the elapsed time limit is relaxed, its performance is still better than using MILP or existing genetic algorithm in practical scale scenarios.