在半導體和TFT-LCD產業中，為了能自動地將虛擬量測系統導入全廠，本研究提出一個適用於全廠導入之虛擬量測自動換模機制，係將初期所建立之一組的推估模型作為基模，自動移植或複製至其他尚未有推估模型之同型機台中。此外，換模程序必須能恢復並維持虛擬量測應有精度，因此本研究亦提出換模成功的判定條件，經由訂定類神經網路、複迴歸之虛擬量測推估精度以及製程參數整體相似度指標等三項指標之換模門檻值，當此三項指標連續三點降至換模門檻值時即表示換模完成，進而可即時地進行與虛擬量測系統相關之所有服務。以TFT-LCD廠薄膜與黃光製程為例，實驗結果證明該機制能迅速地換模及保持推估準確性，並符合有效達到全廠導入之虛擬量測之目標。

This research devises a virtual metrology automatic model refreshing scheme for fab-wide deployment of the virtual metrology (VM) systems in the semiconductor and TFT-LCD industries. The functionality of the proposed scheme is to create a conjecture model as the base model at the initial stage, and then automatically fan out or port the base model to the same type of equipment that wants conjecture models. Besides, a successful model refreshing process must be able to recover and maintain the conjecture accuracy of the VM systems. Therefore, this research also defines the refresh thresholds for the neural network (NN) index, the multiple regression (MR) index, and the global similarity index (GSI) to differentiate whether a model refreshing process completes. If the values of the three abovementioned indexes are lower than the refresh threshold (RT) for three continuous check points at the same time, it means the model refreshing process is successful, and the VM system is ready to provide services on time. The chemical vapor deposition (CVD) process and the photo process in the TFT-LCD industry are taken as the examples to verify the proposed scheme. Results show that the proposed scheme can perform rapid model refreshing, maintain high conjecture accuracy, and achieve the goal of fab-wide VM system deployment.

參考文獻
[1] F.-T. Cheng, H.-C Huang, and C.-A. Kao, “Developing a Dual-Phase Virtual Metrology Scheme ,” to appear in IEEE Transactions on Semiconductor Manufacturing., November 2007
[2] Y.-T. Huang, H.-C. Huang, F.-T. Cheng, and T.-S. Liao, and F.-C. Chang, “Automatic Virtual Metrology System Design and Implementation,” in Proc. IEEE International Conference on Automation Science and Engineering, Washington, D.C., U.S.A., August 2008.
[3] P.-H. Chen, S. Wu, J.-S. Lin, F. Ko, H. Lo, J. Wang, C.-H. Yu, and M.-S. Liang, “Virtual Metrology: a Solution for Wafer to Wafer Advanced Process Control,” in Proc. 2005 IEEE International Symposium on Semiconductor Manufacturing, San Jose, CA, Sept. 2005, pp.155-157.
[4] Y.-J. Chang, Y. Kang, C.-L. Hsu, C.-T. Chang, and T. Y. Chan, “Virtual Metrology Technique for Semiconductor Manufacturing,” in Proc. 2006 International Joint Conference on Neural Networks, July 2006, pp. 5289-5293.
[5] J. L. Elman, “Finding Structure in Time,” Cognitive Science, Volume 14, pp. 179-211, 1990.
[6] Donald F. Specht, “A General Regression Neural Network,” IEEE Transactions on Neural Network, Vol. 2, No. 6, pp. 568-576, Nov. 1991.
[7] D. R. Anderson, D. J. Sweeney, T. A. Williams, and J.-C. Chen, Statistics for Business and Economics- A Practical Approach, Singapore: Thomson, 2006.
[8] 曾登琳, “Evaluating Reliance Level of a Virtual Metrology System” ,國立成功大學研究所碩士論文, 2007.
[9] I. H. Witten and E. Frank, Data Mining: Practical Machine Learning Tools and Techniques, San Francisco, CA: Morgan Kaufman, 2005.
[10] S. Sharma, Data Mining: Practical Machine Learning Tools and Techniques, San Francisco, CA: Morgan Kaufman, 2005.