Cycle Time Forecasting Models for Defect Inspection Process in TFT-LCD Module Assembly

—Because most of the procedures in defect inspection process of TFT-LCD module assembly are examined manually through human vision, cycle time estimation for this particular process is complicated and usually deviated from actual observations considerably in practice. Hence, this study would like to apply the approaches of Bayesian network, linear discriminant analysis, and logistic regression to develop reliable prediction models for defect inspection cycle time. Potential explanatory variables like work-in-process, throughput, yield, and number of product mixes are considered for model construction. Applicability of these approaches is validated through an empirical study of TFT-LCD factory. From the perspective of prediction accuracy and flexibility, findings of this study suggest that logistic regression is a better choice for cycle time estimation than Bayesian network and discriminant analysis.