The Application of Artificial Neural Network in Virtual Metrology System
碩士 === 國立中興大學 === 資訊科學與工程學系 === 96 === In the semiconductor or TFT-LCD panel industry, the shorter the production cycle the faster the rate on returns (ROR). The production processes are divided into manufacturing and metrology processes. Manufacture process cycle times are unable to do be reduced....
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Online Access: | http://ndltd.ncl.edu.tw/handle/76392448853209693837 |
| Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系 === 96 === In the semiconductor or TFT-LCD panel industry, the shorter the production cycle the faster the rate on returns (ROR). The production processes are divided into manufacturing and metrology processes. Manufacture process cycle times are unable to do be reduced. The Virtual Metrology System (VMS) can reduce some steps or time of required in the metrology process without the lowering of yield rates, thus not only to make increases on ROR but can also reduce the number of purchases of the metrology machine to further reduce the manufacturing costs. This paper applies Neural Network technology to virtual metrology and focuses on the prediction accuracy and training rate. This paper also tries to find the proper neural network model for virtual metrology dataset by discussing the characteristics and restrictions between different neural networks models to obtain the maximum prediction accuracy. This paper applies the virtual metrology system to a semiconductor manufacturing factory ETCH data to prove that it get better accuracy by applying proper neural network model.
|
|---|