設施佈置中的模組化設計與高效率流動問題(I)

Abstract

基於個別加工需求，各項產品生產時將以不同路徑通過生產設施。當各項產品流動 路徑相互錯雜跨越時，易衍生出額外運送成本及品質控制問題。因此在設施規劃中，避 免流動路線的交錯至為重要。傳統上此問題被歸類為團塊設施規畫(Block Facility Planning)問題。傳統解法是刪減次要生產流動路線，並保留必要設施。但對於一般的 設施規畫，此傳統方法可能無法產生最佳解或無法施行。 我們提出一個新的想法。生產設施中的大多數機器其實可以被增購，並安裝於異 地。藉由增購機器，便能對同一產品增添不同的流動路線，以此避免原流動路線的交錯。 但本提案有兩大關鍵議題：(1)我們必須知道增購的機器數量 (2)相關的生產流動路線 如何重新規劃以避免交錯與跨越。 將設施規畫視為”多重圖(multi-graph)”，其機器之配置位置可視為”結點 (vertex)” ，相關生產流動則為”有向邊(directed arc)” 。增購機器的想法其實是 在多重圖中，以增加結點的方法使設施規劃圖”平面化( planar)”。此一構想為原創， 未被提出過。我們預計從理論與實驗兩個方向來檢驗新構想。 理論模型中將考慮：(1)機器增購成本 (2)各機器最大增購量 (3)總機器最大增購 量，並包含流線交錯成本。模型中各項成本之計算將真實化，以其增強模型理論解的適 用範圍。此外，當流線交錯無法避免時，將分析保留那些交錯能使成本降低。並考慮以 投產排程上的時間差來降低交錯成本。最後，將以電腦模擬測試並改進提案。 本研究之最終目的是以突破性方法改進台灣半導體業界生產流線的交錯問題，並於 重要期刊中發表。本計畫之流線的交錯問題，是自竹科的半導體製程啟發，並期待新方 法能於竹科內驗證成效。基於產學合作，我們並將於課堂中教授使此一新概念，並期待 能大幅提升生產設施之效能。

Each product flows through a manufacturing facility based on the operations that are performed on that product. When product flows cross, it means large additional transportation costs and quality control issues. Consequently, it is important to design a plant layout with few flow crossings. Traditionally, this issue is considered as a “Block Facility Planning” problem. This traditional method creates layouts by ignoring portions of the product flow. Unfortunately, the solution may not be feasible or optimal for many facility design problems. We would like to study a new approach. Since most facilities have or can purchase duplicate machines, these duplicates are used to eliminate the crossing of product flows. There are two interrelated issues. We must specify not only the number of duplications of each machine but also the product flow routings. That is, we must determine which machine of a given type will perform a specific operation on a given product. We model the facility as a multi-graph where the arcs represent the product flows and vertices represent the locations of the machines. The concept of adding vertices to a graph to make it planar has NOT been explored previously. We plan to provide both theoretical and experimental results for the heuristic and optimal methods we develop. Some features of the model will include costs of duplicating specific machines, the maximum number of duplications of a given machine, and the maximum total number of duplications permitted. Models will also be developed that permit the crossing of product flows at a cost. More realistic flow crossing costs than are currently used will be incorporated in the analysis. Then, when flow crossings cannot be avoided, we will analyze which crossings should be retained. Also, we will consider more general production settings where flow crossing may be avoided by proper product scheduling. Computer packages will be built to implement our methods and provide for technology transfer. Finally, we intend to include our methodology in courses in our department and the final goal is to produce high quality research published in select journals. This problem was motivated by and will be tested and validated at real, a local company that manufactures wafers and precision instruments. As part of our research, we will improve the flow pattern in their plants, which will impact quality and production costs. Also, our research will be integrated in our classroom teaching to educate students about concepts relating to effective facility design.