以Excel數據驗證田口方法改善封裝 製程切割品質研究

Abstract

中文摘要 本研究旨在驗證既有文獻中以田口方法改善晶圓切割崩裂的數據，探討其最佳化參數是否確實能提升製程能力。由於晶圓切割過程中的正面與背面崩裂會直接影響產品良率，因此如何降低崩裂一直是製程改善的重要課題。 在研究過程與方法上，本研究並未重新設計實驗，而是以Excel將文獻中提供的數據進行再分析與驗證，所引用之數據來自國立雲林科技大學工業工程組碩士論文—高志銘（2015）所撰寫的《以田口方法改善封裝製程切割品質之研究》。藉由建立運算模型，模擬田口方法所提出的最佳參數組合，以確認其改善效果是否具有再現性。 驗證結果顯示，正面崩裂的Cpk值由1.43提升至1.97，背面崩裂的Cpk值由1.49提升至1.87，分別提升約38 % 與25 %，同時平均崩裂值也顯著下降。此結果證實，田口方法確實能有效改善製程品質，而本研究所採用之 Excel 再驗證模式，亦提供一種快速且低成本的分析手法，不僅能重現文獻中的改善趨勢，也可應用於其他製程數據之可靠性驗證，對提升製造效率與品質控管具有實務應用價值。

Abstract The present study aims to verify the data reported in existing literature on improving wafer dicing cracks using the Taguchi method, and to examine whether the proposed optimized parameters can indeed enhance process capability. Since both front-side and back-side chipping during the wafer dicing process directly affect product yield, reducing chipping has always been a critical issue in process improvement. In terms of research process and methodology, this study did not redesign new experiments. Instead, Excel was employed to re-analyze and validate the data provided in the literature. The referenced data were drawn from the master’s thesis by Chih-Ming Kao (2015), A Study on Improving the Dicing Quality of Packaging Process by Taguchi Method, Department of Industrial Engineering, National Yunlin University of Science and Technology. By establishing a computational model, this research simulated the optimal parameter combinations proposed in the thesis to confirm whether the improvement effects are reproducible. The verification results show that the Cpk value for front-side chipping improved from 1.43 to 1.97, while that for back-side chipping increased from 1.49 to 1.87, representing improvements of approximately 38% and 25%, respectively. At the same time, the average chipping value was significantly reduced. These results confirm that the Taguchi method can effectively improve process quality. Moreover, the Excel-based revalidation model adopted in this study provides a fast and low-cost analytical approach. It not only reproduces the improvement trends reported in Kao (2015), but can also be applied to the reliability verification of other process data, offering practical value for enhancing manufacturing efficiency and quality control.