Independent Component Analysis approaches for process variation monitoring and Mura defect inspection in TFT-LCD manufacturing

• Main Authors: Yen-Hsin Tseng, 曾彥馨
• Other Authors: Du-Ming Tsai
• Format: Others
• Language: en_US
• Online Access: http://ndltd.ncl.edu.tw/handle/86727915801177654540

博士 === 元智大學 === 工業工程與管理學系 === 101 === In this dissertation, two ICA-based approaches have been proposed for process monitoring of 1-D time-series data and mura detection of 2-D images in TFT-LCD manufacturing. For 1-D signal process monitoring and control, independent components (ICs) are used as source signals for statistical process control. To improve the yield of Liquid Crystal Display (LCD) panels, process control becomes a critical task in LCD manufacturing. In this study, a control chart based on Independent Component Analysis (ICA) is proposed to monitor TFT-LCD process variation. The proposed method can be effectively used in the monitoring of LCD critical process parameter, called Total Pitch (TP). TP is a parameter that is used to control alignment errors in TFT-LCD process. TP variations will cause serious defects like mura (brightness unevenness of a panel) and small bright points on the display area of LCD panels. Since the collected data could be a mixture of noise and different source signals, ICA is first applied to separate mixed data into independent components. The X-bar and R control charts are then used to monitor the separated source signals. Experimental results on real measured data of TP in the TFT-LCD process show that the proposed method can reliably detect process variations. For mura inspection in 2-D images, a machine vision approach is proposed for detecting local irregular brightness in low-contrast surface images and, especially, with focus on Mura defects in LCD panels. A Mura defect embedded in a low-contrast surface image shows no distinct intensity from its surrounding region, and the sensed image may also present uneven illumination on the surface. All these make the Mura defect detection in low-contrast surface images extremely difficult. A set of basis images derived from defect-free surface images are used to represent the general appearance of a clear surface. Each LCD image is then constructed as a linear combination of the basis images, and the coefficients of the combination form the feature vector for discriminating Mura defects from clear surfaces. In order to find minimum number of basis images for efficient and effective representation, the basis images are designed such that they are both statistically independent and spatially exclusive. An ICA-based model that finds both the maximum negentropy for statistical independence and minimum spatial correlation for spatial redundancy is proposed to extract the representative basis images. Experimental results have shown that the proposed method can effectively detect various Mura defects in low-contrast LCD panel images. It is also computationally very fast for real-time, on-line inspection.