Applying the AVM System to the TFT-LCD Industry

博士 === 國立成功大學 === 製造資訊與系統研究所碩博士班 === 100 === Processing quality of thin film transistor-liquid crystal display (TFT-LCD) manufacturing is a key factor for production yield. The processing quality of tool will affect the production yield. The conventional sampling approach cannot receive processing q...

Full description

Bibliographic Details
Main Authors: Wen-HuangTsai, 蔡文皇
Other Authors: Fan-Tien Cheng
Format: Others
Language:en_US
Published: 2012
Online Access:http://ndltd.ncl.edu.tw/handle/92417317215792161345
Description
Summary:博士 === 國立成功大學 === 製造資訊與系統研究所碩博士班 === 100 === Processing quality of thin film transistor-liquid crystal display (TFT-LCD) manufacturing is a key factor for production yield. The processing quality of tool will affect the production yield. The conventional sampling approach cannot receive processing quality of each product. Therefore, when the equipment is out of control, the defects can not be found in time, which will result in a great yield loss. Virtual metrology (VM) is a good resolution to the problem of conjecturing manufacturing quality of a process tool based on data sensed from the process tool without physical metrology operations. When considering fab-wide VM implementation, the number of conjecture models will increase rapidly with the growing number of tools. Under such condition, if we still take the traditional method to create models one by one with a lot of historical data, then huge cost in human resources and capital will make the fab-wide VM implementation impossible. The AVM systems can have the function of automatic fanning-out and model-refreshing to solve the above-mentioned problem and copy the conjecture model of a certain tool and apply it to other tools that have no conjecture models. As a result, AVM can be automatically deployed fab-wide without waste in tremendous man-power and time for establishing conjecture models. In TFT-LCD factory, all process tools are sorted to be three types, namely single-stage, dual-stage and mixed equipment. One of each type is chosen to demonstrate this AVM architecture for TFT-LCD manufacturing. Three experimental results are predicted well by AVM, so the AVM can also do that in the other tools. In addition, AVMS needs to handle a large volume of VM-related data, a novel AVMS architecture is proposed based on Main Memory Database (MMDB) technology. The novel AVMS architecture adopts a free commercial MMDB to significantly reduce total system cost and have better data-storage efficiency, superior data-query performance, and lower database cost. Finally, The AVM system has been successfully deployed in a fifth generation and sixth TFT-LCD factory and a benefit equation is established to evaluate the profitability of applying AVM fab-wide. The evaluation result shows that AVM fab-wide implementation is expected to gain extra production volume output due to cycle-time reduction.