以偏最小平方法建構光阻膜厚之虛擬量測模式及其實證研究

Constructing a Virtual Metrology Framework for Halftone thickness based on Partial Least Squares and an Empirical Study

葉雅綸(Yeh, Ya Lun)

指導教授 ： 簡禎富

國立清華大學/工學院/工業工程與工程管理學系/碩士(2016年)

摘要

面板產業為資本密集且高度競爭的高科技產業，為了維持競爭優勢，開始使用四道光罩技術於製程當中，透過使用半色調網點光罩，能使製程道數減少，然而此類光罩容易導致光阻厚度不均的情況，因此需透過檢測以確保產品品質。廠商在考慮時間成本與資本設備成本的情況下，會使用抽檢來監督產品品質，但抽檢並無法保證全面品質管理，因此本研究建立光阻膜厚之虛擬量測模式，透過蒐集過去機台參數資料，使用偏最小平方法（partial least squares, PLS），建構光阻膜厚預測模型。此虛擬量測模式建立後，不僅能使機台的量測頻率減少，更能協助監測整個生產設備，即時反應異常製程，減少產品作業週期，達到高效能高產能的目標。並與台灣某知名光電面板公司合作，進行實證研究，透過平均絕對百分比誤差（mean absolute percent error, MAPE）檢驗方法效度。驗證資料組之平均絕對百分比誤差為3.96%，代表此模型的預測能力良好。

關鍵字

半色調網點光罩 ； 偏最小平方法 ； 虛擬量測

並列摘要

In a highly competitive and capital-intensive industry, such as panel industry. To keep its competitive advantage, panel industry try to use the halftone mask into the process. Using halftone mask reduce the process flow into four. However, halftone mask easily lead to resist non-uniformity, these make panel industry need to control the resist uniformity to make sure that product quality is fine. Considering the cost of time and equipmetn, industry always use sampling to monitor the product quality, but sampling does not guarantee total quality management. In this study, we collect history data, using partial least squares to construct a virtual metrology framework to predict the halftone thickness. After the prediction model is built, not only reduce the frequency of measuring but help panel industry to inspect the whole production equipment, react deviant problem and reduce product cycle time then achieve high capacity goals. By cooperating with a well-known Taiwanese panel company to test the method validity and the MAPE of the validation data set was 3.96%, means a good representative of the prediction model.

並列關鍵字

Halftone Mask ； Partial Least Squares ； Virtual Metrology

參考文獻

Chang, Y.-J., Kang, Y., Hsu, C.-L., Chang, C.-T., and Chan, T. Y. (2006), "Virtual metrology technique for semiconductor manufacturing," Proceedings of IEEE International Joint Conference on Neural Network, pp. 5289-5293.

Devarakonda, N., Subhani, S., and Basha, S. A. H. (2014), "Outliers detection in regression analysis using partial least square approach," Advances in Intelligent Systems and Computing, Vol. 249, pp. 125-135.

Garthwaite, P. H. (1994), "An interpretation of partial least squares," Journal of the American Statistical Association, Vol. 89, No. 425, pp. 122-127.

Geladi, P. and Kowalski, B. R. (1986), "Partial least-squares regression: a tutorial," Analytica chimica acta, Vol. 185, pp. 1-17.

Hung, M.-H., Lin, T.-H., Cheng, F.-T., and Lin, R.-C. (2007), "A novel virtual metrology scheme for predicting CVD thickness in semiconductor manufacturing," IEEE/ASME Transactions on Mechatronics, Vol. 12, No. 3, pp. 308-316.