Deep autoencoder with clipping fusion regularization on multistep process signals for virtual metrology

Jeongsub Choi, Myong Kee Jeong

Research output: Contribution to journalArticlepeer-review

4 Scopus citations

Abstract

We propose a new feature extraction method for virtual metrology with multisensor data in semiconductor manufacturing based on deep autoencoder. The multistep process-equipment signals at wafer fabrication process consist of multiple subprocesses and signals thereof. Existing extracted features based on signals reconstructed by autoencoder turns into fluctuating shapes coarsely reflecting the original values. We propose the clipping fusion regularization on the signals reconstructed by deep autoencoder in order to extract features from raw signals preserving data characteristics. The predictive performance with the features extracted by the proposed model is evaluated in a case of an etching process for wafer fabrication. The experimental result shows that the proposed method improves the accuracy of prediction models, compared with those with the existing feature extraction approaches.

Original languageEnglish (US)
Article number7782634
JournalIEEE Sensors Letters
Volume3
Issue number1
DOIs
StatePublished - Jan 2019
Externally publishedYes

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Instrumentation

Keywords

  • Deep neural network
  • feature extraction
  • manufacturing
  • semiconductor
  • sensor data
  • virtual metrology (VM)

Fingerprint

Dive into the research topics of 'Deep autoencoder with clipping fusion regularization on multistep process signals for virtual metrology'. Together they form a unique fingerprint.

Cite this