本論文旨在提供一個新的抽樣檢驗計劃,探討一產品包含兩個不相同組件 之情況下,這些組件該如何抽樣。檢驗模式是以貝氏決策理論為基礎,採 用貝氏方法可將成本效應及事前機率分配同時納入考慮。檢驗計劃之設計 採修正檢驗(Rectifying inspection) 及屬性抽樣(Attribute sampling) ,批量中相同組件為良品或不良品將假設為可交換性之隨機變 數。 我們探討三種不同檢驗模式:模式1 ─兩批量可同時進行抽樣檢 驗之模式,模式2 ─兩批量分段抽樣檢驗最後同時處置之模式,模式3 ─ 兩批量分段抽樣檢驗即時處置之模式。基於最佳期望成本及時效因素上的 考量,實際上我們將深入探究模式1 及模式2 二種情況。檢驗模式之建構 及推演乃藉由R.A. Howard 與 J.E. Matheson(1981)所發展之決策輔助分 析工具─因果關係圖(Influence diagram) 來完成,進而導出問題整體決 策函數及演算法則。
In the thesis,we propose two Bayesian rectifying inspection and attribute sampling models for a product containing two distinct importanat items. The functional states of items in a lot are assumed statistically exchangeable. Study on the case of single item has been often,but almost none is with regard to the case of two distinct items. In the first model samples of two lots are taken simultaneously while in the second model samples from one lot are taken first and after the observations, a size of samples of the other lot is determined. For both models, an action is taken with respect to the two lots after observing sampling outcomes,either to accept the remaninig items of the lots or to inspect all of them. The objective functions of both models are to minimize the expected total cost. In the research, influence diagrams are used as a tool to derive the total objective functions of the models. The first model is better in time-consuming while the second one is better in cost- consuming. The functions are also analyzed so as to make the problems feasible on computations.Several experimental results of the two models are presented and discussed.