计量学
可解释性
预言
计算机科学
多元统计
半导体器件制造
过程(计算)
特征(语言学)
尺寸计量学
先进过程控制
人工智能
过程控制
机器学习
数据挖掘
工程类
数学
统计
电气工程
哲学
薄脆饼
操作系统
语言学
作者
Siho Han,Jihwan Min,Jui Ma,Gyuil Hwang,Taeyeong Heo,Youngeun Kim,Sung-Jin Kang,Hyojun Kim,Sang‐Jong Park,Kisuk Sung
出处
期刊:IEEE Conference on Prognostics and Health Management
[Institute of Electrical and Electronics Engineers]
日期:2023-06-05
卷期号:: 30-37
被引量:8
标识
DOI:10.1109/icphm57936.2023.10194015
摘要
In Prognostics and Health Management, virtual metrology is crucial for advanced process control, accounting for the condition of manufacturing machinery. Traditionally, virtual metrology has been tackled using statistical and machine learning approaches, which require extensive domain knowledge and feature engineering. Moreover, the high-dimensional nature of complex industrial systems renders the interpretation of metrology results increasingly difficult. In this work, we introduce PIE-VM, an attention-based multivariate time series regression model incorporating process information for virtual metrology in atomic layer etching. Experimenting on real-world data collected and provided by PSK Inc., a large semiconductor manufacturing equipment company based in South Korea, we empirically demonstrate that our method predicts etch depths more accurately than baseline approaches. Also, we show that our model provides useful information for advanced process control based on its inherent interpretability.
科研通智能强力驱动
Strongly Powered by AbleSci AI