本論文應用六標準差DMAIC手法來提升封裝廠之晶片切割良率。在定義階段由高階主管參與而組成的六標準差專案團隊,首先了解客戶的聲音及針對客戶退貨品進行分析,之後再進行異常流程分析。在衡量階段,則針對量測機台的重複性及再現性探討並了解現行之製程能力,以作為改善的依據。在分析階段,利用特性要因圖列出可能造成異常的原因,之後並利用迴歸顯著性檢定來驗證造成不良的因子;並針對這些真因提出矯正措施。在改善階段,則是利用實驗設計法來求得因子的最佳設計,並驗證改善效果。最後在控制階段,則是由專案團隊制訂管制計畫表交由產線執行及監控製程,以完成整個專案。在經專案團隊將近半年的努力後,生產線的異常終於解除,也達到客戶期望的目標;並因此為公司獲取近兩佰萬美元的效果。
In this thesis, a workflow to improve the wafer saw yield of assembly house with the six sigma “DMAIC” is proposed. In “Define”, top managers attend and build up a “Six Sigma” project team to understand the customer’s complaint and analyze the rejected products in the abnormal process flow at the first stage. In “Measure”, the project team aims at the reproducibility and repeatability of measuring machine to come out a good Gage R&R; in addition, they also analyze the current ability of process and set up the target of improvement. In “Analyze”, they use fishbone diagram to list down all the possible factors that could cause the abnormality and use regression analysis to verify the real factors, meanwhile, they also provide the preventive actions to prevent the problem happening again. In “Improve”, project team uses DOE to get the optimal solutions and verify the effectiveness. Finally, in “Control”, project team generates a control plan and provide it to production line for implementing and monitoring the process flow. 6 months past, the production line issue was solved and achieved what customer wished then got about 2 million US dollars effectiveness.