面板產業本是高成本產業之一,近年來更隨著大尺寸面板廠的興起,玻璃基板尺寸的放大,使得設備成本不斷攀升,建廠成本更是達到數百億的規模。製造成本的逐步墊高,迫使面板業者無不撙節成本以求利潤最大化,降成本活動更成為面板業者必做的功課項目。 在製造過程中,在製品存在之目的是為了確保生產線不會因為缺料而造成產線中斷,然而在製品對於製造業者本身而言即是一項投入成本,除其本身所包含之物料成本外,製造過成中所衍生之人事成本、時間成本以及儲存成本等,無不加重製造業者的負擔。因此,在製品的存在對於製造業者而言可說是一項必要之惡,如何有效管控製造過程中的在製品以保護生產線不受缺料的風險,同時又有效降低成本支出將是一項重要的工作。 本研究以DMAIC為架構,首先定義問題以確認問題所在,並於資料收集完成後加以整理並轉換為可用之型式,在分析階段則應用資料探勘手法中之迴歸模型、決策樹模型以及相關係數進行比較,設定出合理之在製品水準。之後再將限制理論中所發展出之DBR排程概念導入改善作為,設立在製品水位之警戒與回饋機制。最後以產出數量之達成狀況為監控目標,持續驗證在製品水準之適用性,如此不斷循環以降低在製品數量,消除不必要之成本支出,使企業能有效追求利潤。 TFT-LCD industry is one of the capital-intensive industries. In recent years, the expenditure of equipment gets higher and higher owing to the larger-size glass substrates and the investment of new plant, somehow even excess ten billion New Taiwan Dollars. The manufacturing cost increases gradually what enforces the manufacturers maximize their profit through cost down. Therefore, cost-down activities become one of the most important things to TFT-LCD manufacturers. In the manufacturing process, the WIP could protect the production line from starving, but it becomes a serious loading as well to the manufacturers. As a result, the WIP is the necessary evil. How to control the WIP level to keep the production line running and lower the cost of manufacturing at the same time is a significant job. In this research, the DMAIC provides a rigorous and pragmatic framework for set a WIP level. First, find out the problem and define it clearly. After data collection, it should be converted to fit the software attributively. At the analysis stage, set the WIP level by using three methods, regression, decision tree and correlation coefficient. Next step, create a warning and feedback mechanism from the concept which TOC developed. Finally, monitor the target achievement of output and keep rolling this process to lower the WIP level. This process provides a possible mechanism for the manufacturers, who are eager to eliminate the unnecessary expenditure, to reduce the production cycle time and to improve the production efficiency.