Abstract
In recent years, the demand of thin-film transistor liquid crystal display (TFT-LCD) increases rapidly. Taiwan became the second major country in 2004 with the expanded global market for LCD industrial. In general, the mura defect which is a Japanese term meaning imperfective blemish may occur during LCD panels manufacturing process. At present, the image quality of LCD panels has been determined subjectively by means of human visual inspection. In fact, the inspectors need to remember a large number of instructions because mura defects are usually difficult to detect. The inspection tasks include a series of complicated procedures that increase the workload for the inspectors.This research focuses on the improvement of LCD inspection process in order to reduce the inspectors’ workload. The inspection information was gathered through questionnaire survey at first. Then the knowledge extracting was to analyze association rules through interviewing experienced inspectors. The nested IF-THEN inspection rules between test patterns were analyzed by means of two-dimensional matrix and group technology (GT). In terms of inspection test patterns, the occurring sequence of pattern was rearranged in order to make inspection tasks more efficiently. Furthermore, this study aimed to construct an expert system for inspecting LCD defects according to inspection rules and strategies. According to the results of the experiment and expert evaluation, the expert system was proposed as a training support system to aid the trainees to learn inspective skills more effectively. Therefore, the performance of the inspection training could be increased. The cost of manufacture could be also reduced significantly.