加权
可靠性工程
半导体器件制造
索引(排版)
过程(计算)
数据挖掘
计算机科学
理论(学习稳定性)
薄脆饼
工程类
钥匙(锁)
状态监测
过程能力指数
半导体器件
维修工程
健康指标
系统工程
实时计算
半导体
利用
控制工程
工业工程
健康评估
电子工程
计量学
晶圆制造
作者
Jeongsun Ahn,Sang-Hyun Cho,Hong-Yeon Kim,Hongyeon Kim,Hyeonjeong Choi,Dain Ham,Yongjo Kim,Hyun-Jung Kim
标识
DOI:10.1109/tsm.2026.3665568
摘要
In semiconductor manufacturing, monitoring equipment status is vital in ensuring process stability and efficient equipment operation. Thus, there is an increasing need for a health index that can efficiently reflect the overall condition of the equipment as a unified metric. Previous studies have derived a health index based on real-time time-series data collected from single wafer processing, overlooking other significant perspectives essential for practical equipment monitoring. Parameters such as temperature and pressure with similar trends in time-series data of multiple consecutive wafers indicate that they are sensitive to the cumulative use of equipment and are similarly affected by equipment aging. Hence, it is necessary to consider the similarity of trends for efficient diagnosis and control. The type of physical components such as chambers, pumps, and pipes observed by each parameter is also an important criterion for accurately understanding the condition of the equipment and taking necessary actions. These can be considered as monitoring views. Therefore, this study introduces a new health index framework for monitoring semiconductor equipment, utilizing multi-view data. We propose three key views for monitoring and define multi-view data by suggesting a quantitative method to represent it in graphical form. Based on it, we propose a method for deriving the health index using an autoencoder-based approach and dynamic weighting techniques. We have validated the effectiveness of our approach with real datasets, demonstrating its potential as a valuable tool for monitoring the condition of semiconductor equipment.
科研通智能强力驱动
Strongly Powered by AbleSci AI