本論文主旨在提出一個以最佳經濟成本為考量決策模式,計算當TFT-LCD組裝完成後,成本包裝出貨至客戶手中前,該如何行動以使期望總成本為最小。模式考量產品檢驗成本、抽樣資訊以及產品失效成本,可選擇行動方案分兩階段:(一)決定抽樣數,(二)根據抽樣結果對批量剩餘組件之最佳處置。研究方法以決策樹為工具,來建立問題決策模式,並以決策樹理論推導出整體決策目標函數,並據此分析模式性質,再以C++語言撰寫模式程式。 文中採用案例公司所提供的實際數據,透過Arena裡的Input Analyzer的轉換分析求得機率分配與其參數並搭配自行撰寫的程式來計算期望最小支出成本。由計算出的結果發現,當產品檢驗成本/產品失效成本的比值與產品期望良率相差不大時,100%檢驗與0%檢驗時,經常並非最佳選擇。因此,本文旨在建立一個適用於TFT-LCD產業的最佳經濟型抽樣檢驗計畫,並且針對該產業的期望良率與產品檢驗成本/產品失效成本的比值作敏感度分析,並彙整所得到的結果,建立幾組符合TFT-LCD產業製程良率的最佳經濟型抽樣檢驗計畫表,以供決策者使用。
This thesis proposes a decision tree model to minimize expected total cost for handling the TFT-LCD products after assembly. The factors that influence the total cost include inspection cost, product failure cost, product quality, and lot size. The model consists of a two-stage decision: At stage 1, determining the sample size, and at stage 2, deciding what to do about the remaining products of the lot after observing the sampling outcome. The objective function of the model can be derived according to the decision tree theory. Some properties of this model using Beta priors to represent the product quality are presented, and such findings will help facilitate the computation of the model objective function. A computer program coded in Visual C++ is provided for computing the optimal decisions and the optimal objective value of this model. We studied a real life case in the TFT-LCD industry. A set of data regarding the product quality of a company was collected and used to construct the prior distribution of the product quality. The inspection cost and product failure cost are also estimated using the relevant information from that company. An optimal decision policy on the TFT-LCD products for that company is provided and discussed. Sensitivity analysis as to the model parameters is also performed and the results will provide a good reference for the management to improve their operational procedure. Finally, the research provides several inspection tableaus that can be easily used by the work persons in the assembly line.