利用製程能力指標選擇群體供應商之貝氏方法

Abstract

選擇供應商在供應鏈管理中是一個重要問題。在生產設置的初始階段，決策者通常面臨著從幾個可用的製造供應商中選擇最佳製造供應商（具有最大製程能力指標值）的問題。在選擇最佳的供應商中有許多因素需要被考慮的，如品質、成本、服務等等，其中，品質應當是最主要的考慮因素。關於選擇最佳製造供應商的問題，在很多文獻中已受到相當大的注意，但主要都是從頻率學派(frequentist)的觀點。在此論文中，我們考慮用貝氏(Bayesian)的方法來做群體供應商的選擇問題，意思就是選擇一組供應商，且這個群組有很高的信心水準包含具有最高的製程能力指標值的供應商。基於觀測的資料和事先的可用資訊，我們發展了一個符合實際的過程來做群體供應商的選擇，對於工程師在實際的應用上有相當的助益。為了探討所提出方法的有效性，我們做了一些模擬來研究我們所提出的貝氏方法用在群體供應商選擇上的績效，並且透過我們所定義的損失函數與Huang and Lee (1995)所提出的方法做比較。 目前大多數現有的方法用來評估製程能力都是透過傳統頻率學派的觀點來做推導。在此論文中，我們將對 Vännman 1995年所定義的製程能力指標Cp(u,v)從貝氏學派的觀點來考慮製程能力估計與檢定的問題。首先，我們推導製程是否有能力的後驗概率，p。為了使此貝氏過程在實例的應用中更實際化，我們用表格列出製程已達到預先指定的能力水準所需之最低 Cp(u,v)_hat 值，當後驗概率 已達到理想程度如0.95。為了說明如何應用我們所提出的方法到由工廠所收集的實際資料上，我們介紹了關於LCM 製造過程的應用示例。最後，我們應用此處所推導的後驗概率至之前所提出的群體供應商選擇的貝氏選擇規則中，對有關正確選擇率和所選子集的平均大小我們也做了一系列的模擬研究。

Supplier selection is an important part of supply chain management. In the initial stage of production setting, the decision-maker usually faces the problem of selecting the “best” one(s) from available manufacturing material suppliers. There are many factors, such as quality, cost, service and so on, that need to be considered in selecting the best supplier. Among them, quality should be given the major consideration. This problem of selecting the best supplier(s) has received considerable attention in the literature but mainly from the frequentist point of view. In this dissertation, we consider dealing with the group supplier selection problem by the Bayesian approach, meaning selecting a group of suppliers that would include the supplier with the largest value of process capability index with a high level of confidence. Based on the observed data and available prior information, we develop a practical procedure for the group supplier selection, which is useful to engineers for their in-plant applications. To investigate the effectiveness of the proposed method, a simulation study was carried out evaluate the performance of the proposed group supplier selection rule via Bayesian approach and to compare with the method proposed by Huang and Lee (1995) under the defined loss function. Most existing methods to assess the process capability were derived from the traditional frequentist point of view. In this dissertation, we consider the problem of estimating and testing the process capability based on the capability index Cp(u,v) defined by Vännman (1995) from the Bayesian point of view. We first derive the posterior probability, p, that the process under investigation is capable. To make this Bayesian procedure practical for in-plant applications, we tabulate the minimum values of Cp(u,v)_hat for some values of u and v, for which the posterior probability p reaches desirable levels with the pre-specified capability level, say, 0.95. To illustrate how we apply the proposed procedure to actual data collected from the factory, an application example of LCM manufacturing process was also presented. Finally, we use Cp(u,v) as the process capability index in our proposed Bayesian group supplier selection rule, and a series of simulation studies about the posterior probability of correct selection and the average size of selected subset were undertaken.