標題: 以線性規劃模式於預防保養排程探討半導體機台產出極大化
Developing Linear Programming Models To Maximize Throughput For Cluster Tools By Preventive Maintenance Scheduling in Semiconductor Manufacturing
作者: 陳權勇
Chuan-Yung Chen
姜齊
Chi Chiang
管理學院管理科學學程
關鍵字: 預防保養排程;集束型製程機群;線性規劃;Preventive maintenance scheduling;Cluster tools;Linear programming
公開日期: 2006
摘要: 在半導體產業,技術的領先主宰一切,而技術的背後是高資本的機台。因此,在不失良率而使產能利用率極大化,對半導體廠的毛利率影響甚鉅。在半導體廠,由於數百台機台的緊密配合度,製造執行系統的複雜性,預防保養工作之間相互依賴性及因應產量需求所要求的產線平衡,預防保養排程是一項極富挑戰性的任務。 在本研究中,吾人提出一個兩階層級式的模式,在上層為一長期預防保養計劃模型,底層為短期預防保養排程模式,而本研究探討的焦點是解決底層問題。本研究應用線性規劃(LP)模式,推算出所有必須在某段時間(時窗)內完成預防保養工作的最佳排程。此模式可將整合相互依賴集束型製程機群的不同預防保養項目、生產規劃資料譬如人力限制、在製品、和預防保養的時窗政策及費用並歸納求最佳解-預防保養排程,而本研究中,集束型製程機群主要探討的為化學機械研磨機(chemical mechanical polishing , CMP)。管理者透過此模式可以有效地控制成本支出並成功地提昇機台利用率,進而機台產出極大化。 此外,本研究將採用數學規劃軟體LINGO為求解之方法,並以實例操作證明模式之有效性,期能提供相關預防保養安排規劃者或決策者一完整且具有量化分析之排程。利用此規劃參考模式,作為日後進行預防保養安排及規劃時之重要參考依據。
As semiconductor technology continues its rapid advancement, semiconductor fabs costs skyrocket, and their effective lifetime decreases. In turns, to maximize availability of equipment would be a tough challenge for modules, and utilization to manufacturers. Hence, the stability and quality are critical to reflect her fab yield. Preventive Maintenance (PM) scheduling is indeed a very challenging task in semiconductor manufacturing. In this paper, a two-level hierarchical modeling framework is proposed. At the higher level is a model for long-term planning and at the lower level is a model for short-term PM scheduling. Solving the lower level problem is the major focus of this paper. By developing Linear programming (LP) models for scheduling all due PM tasks for a group of cluster tools, over a planning horizon, interdependence among different PM tasks, production planning data such as, manpower constraints, projected WIP levels and associated PM time windows and costs, all of which are incorporated in the model. The cluster tool implemented is Chemical Mechanical Polishing, CMP, equipment. Results of a preliminary simulation study with LINDO program are from the performance of the model-based PM schedule, which has shown very encouraging improvement on maximizing tools’ throughput.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009362528
http://hdl.handle.net/11536/79954
顯示於類別:畢業論文