Abstract
The flat panel display industry has invested considerable resources in constructing large-size panels and rendered the process more complex, resulting in various defects and low yield. Engineers rely on their domain knowledge or rules of thumb for troubleshooting; however, limited domain knowledge, insufficient experience, faulty generalization, and bounded rationality lead to ineffective judgment. The objective of this study was to develop a framework for data mining and knowledge discovery from a database; the Kruskal-Wallis test and a decision tree were used to investigate a large amount of thin film transistor-liquid crystal display (TFT-LCD) manufacturing data and determine the possible causes of faults and manufacturing process variations. An empirical study was conducted at a TFT-LCD company in Taiwan, and the results demonstrated the practical viability of the framework.