單邊規格下多條生產線製程之群組選擇

Abstract

本研究主要探討供應商群組選擇問題，由於現今生產技術先進，諸多製程的不良率極低，且不同製程間的良率僅有些微差異，造成決策者難以選擇最佳的供應商。製程能力指標建立品質規格與實際製程之間的連結，並在極低不良率的製程下，被廣泛地用來評估製程產出產品是否符合規格以及提供品質保證。本文藉由單邊規格下多條生產線的製程能力指標 CpuM來選擇出擁有較佳製程能力的一組供應商，然而一般比較母體平均值大小的多重比較技術並不適用於評選多組製程良率指標，因此本篇以供應商的CpuM_hat相減作為檢定統計量，並採用Bonferroni方法，達到有效衡量最佳品質製程的目標，且提供假設檢定所對應的臨界值、正確選擇供應商之信心水準保證以及在要求之檢定力下所需的樣本數，作為供應商選擇的輔助資訊。最後，我們應用Bonferroni方法與其他多重比較方法Multiple Comparisons with the Best (MCB) 作為範例，並設定不同參數和指標估計值，說明兩種方法在供應商群組選擇問題所做的決策不同。

In this thesis, we deal with the group supplier selection problem. Due to advanced production technology, the decision maker is hard to select the processes with extremely high yields and small differences between each other. Process capability indices (PCIs) establish the relationships between the actual process performance and the manufacturing specification. Therefore, based on the processes with extremely low fraction of nonconformities, PCIs have been extensively used to evaluate and measure whether the process meets the specifications and they also provide quality assurance at the same time. This research considers process capability index for one-sided processes with multiple independent manufacturing lines, CpuM and selects a group containing the best suppliers. However, the conventional multiple comparisons techniques for evaluating different means are inadequate to evaluate yields indices. Thus, we adopt the Bonferroni method and subtraction test statistic to tackle the group selection problem. Critical values for testing procedure, correct selection assurance, and sample sizes required for a designated selection power are also investigated. Finally, we compare our method by Multiple Comparisons with the Best (MCB) method under different parameters and estimations in examples. It is noted that the decisions which Bonferroni and MCB methods make are not quite similar.