| Summary: | 碩士 === 國立成功大學 === 製造工程研究所碩博士班 === 93 === With the trend of automation the manufacturing process of optical electrical industry is more complicated and the technology keeping progressing. In order to improve the product’s quality and to reduce the production’s cost, it is very important to have a control system. In the future, the TFT-LCD will replace the products including the TV, computer, monitor, etc. Under the huge scale of industry, it will be the key factor of surviving on the competitive market to make a profit.
In this research, applying Artificial Neural Network for a sputter process constructs a multiple-input single-output process quality control system. Due to the way the raw data are collected in the sputter process is bad, causing many constraints in this study. This research proposes a preprocessed method with the concept of pattern for this dynamic system. After preprocessing, the prediction model wasn’t still created, so this research uses the Control Chart, Mahalanobis Distance and Fisher Discriminant Analysis separately to diagnose the thick quality. These application studies demonstrated that, in comparison to conventional control chart, the multivariate analysis is more accurate and efficient. Although the prediction model with good performance isn’t constructed, the method and concept of data process in this research have fairly good contributions.
|
|---|
