Convolutional neural network based multi-input multi-output model for multi-sensor multivariate virtual metrology in semiconductor manufacturing
Annals of Operations Research
Abstract
Virtual metrology (VM) in semiconductor manufacturing is to predict product physical quality measurements using the processing information at a wafer fabrication process. The process state parameters from various sensors on production equipment are used for the accurate and reliable prediction of process outcomes, and several deep learning methods have been considered, including convolutional neural network (CNN). Recent studies in the literature have demonstrated the successful VM modeling with CNN for univariate-response prediction. Multivariate process outputs, however, are overlooked in deep learning-based VM modeling although the joint information among the process outputs can be used to improve the prediction performance. In this work, we propose a CNN-based multivariate VM model using multi-sensors process sensor data. We evaluate the proposed model on a real-life case for VM modeling at an etching process in wafer fabrication.
Links
Published paper
Citation
@article{choi2024,
author = {Choi, Jeongsub and Zhu, Mengmeng and Kang, Jihoon and K.
Jeong, Myong},
title = {Convolutional Neural Network Based Multi-Input Multi-Output
Model for Multi-Sensor Multivariate Virtual Metrology in
Semiconductor Manufacturing},
journal = {Annals of Operations Research},
volume = {339},
pages = {185-201},
date = {2024-03-16},
url = {https://link.springer.com/article/10.1007/s10479-024-05902-z},
doi = {10.1007/s10479-024-05902-z},
langid = {en}
}