Optimization of Semiconductor Monitoring via Residual-Based Computational and Statistical Methods

DOI

10.2991/978-94-6239-668-5_87How to use a DOI?

Keywords

Anomaly Detection; Kernel Density Estimation; Ridge regression; Principal Component Analysis; Semiconductor Manufacturing

Abstract

Automated detection of process deviations in semiconductor manufacturing is crucial for ensuring product quality and operational efficiency. Industrial datasets, such as the dataset, are challenging because they are relatively high-dimensional, contain highly correlated features, and include binary outcomes. This study develops a computational framework for residual-based monitoring, combining statistical modeling with algorithmic techniques to detect deviations.

For the monitoring procedure, the data set was split evenly: the first half was used to create baseline models and define control limits, while the remaining half served for testing. Principal Component Analysis and Ridge regression are employed to minimize the negative impacts of multicollinearity. Principal Component Analysis generates linear transformations of the features, reducing the dimensionality of the data while preserving most of its variation. Ridge regression, on the other hand, preserves all features and applies a penalty to large coefficients while maintaining the relationship between predictors and the response. In both scenarios, deviance residuals and randomized quantile residuals (RQRs) were computed. These residuals are proven to be approximately normally distributed, which enables accurate monitoring.

The actual monitoring was conducted using traditional control charts, median absolute deviation (MAD) limits, and kernel density estimation (KDE)–based limits. Among these techniques, traditional charts are straightforward and interpretable, MAD limits reduce the influence of extreme residuals, and KDE leverages non-parametric density estimation to capture subtle process deviations. These methods integrate algorithmic computation, statistical reasoning, and automated evaluation to establish a foundation for detecting process abnormalities.

The results indicate that KDE-based monitoring detects deviations more quickly and flags a larger number of observations than the other methods. Traditional and MAD-based approaches produce similar but slightly less sensitive results.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Ulduz Mammadova
PY  - 2026
DA  - 2026/05/14
TI  - Optimization of Semiconductor Monitoring via Residual-Based Computational and Statistical Methods
BT  - Proceedings of the International Conference on Current Problems in Engineering and Applied Sciences (ICCPEAS 2025)
PB  - Atlantis Press
SP  - 826
EP  - 836
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6239-668-5_87
DO  - 10.2991/978-94-6239-668-5_87
ID  - Mammadova2026
ER  -