基於主成份分析之DPC基板良率分析暨最佳化

Abstract

根據 McKinsey&Company指出，由於商業照明市場從傳統照明技術轉換至發光二極體(LED)，LED照明的市場需求預計將比去年中、長期的上升更快，其市場規模看起來與2011年的預期相似，由於其價格的快速侵蝕。預計2016年 LED照明市場規模約為370億歐元，2020年為640億歐元。由於市場的競爭激烈，所以有效的提高產品良率及減少成本消耗，對於陶瓷基板工廠為重要指標。產量分析是最重要的項目於陶瓷基板工廠。在新產品開發階段有可能有很多因子會影響產品產量。傳統上，產品良率提升需要幾個學習週期經過來解決產品損失問題。本研究提供一種在較少的學習週期經過中識別產量損失問題的新方法。首先，由主成份分析將自動光學檢驗地圖之缺點分類進行轉換成新的基礎。第二 由計算主成份分析之缺點率及良率損失地圖之集群分析進行分類事件。第三 失效分析樣品之選擇來解決產品良率損失問題。此外，新的良率損失基礎可以用來監控產品改善過程中，所選擇之失效模式之對策有效性。(自動光學檢驗地圖失效模式)According to McKinsey&Company , Since the commercial lighting market is on a clear transition path from traditional lighting technologies to the Light Emitting Diode (LED), the share of LED lighting is expected to rise faster than forecast last year in the mid to long term, its market size appears similar to that expected in 2011 due to its accelerated price erosion. The LED lighting market size overall is anticipated to be around EUR 37 billion in 2016 and EUR 64 billion in 2020. Yield analysis is one of the most important topics in ceramic material company. In the early stage of the new process development, several factors will have a greater impact on the yield. Traditionally, several learning cycle iterations are needed to solve the loss of revenue. This study describes a new method for diagnosing loss of earnings in less iterations. Firstly, the fault classification of Automatic optical Inspection (AOI) mapping data is transferred to a new foundation by using principal component analysis. Secondly, the defect rate is calculated, and the original AOI mapping data is reconstructed on the main foundation, which causes the loss of revenue space by Cluster Analysis. Third, we can choose the physical failure analysis sample to solve the output loss problem. In addition, the new loss of revenue base can be used to monitor progress and increase yield measures to reduce the effectiveness of failure modes (AOI mapping failures).