Doochan Ko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Doochan Ko has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Daniel Jin Hun Kim
Department of Computer Science, Hanyang University, Seoul, Republic of Korea
1 shared publication - Inwhee Joe
Department of Computer Software, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 04763, Republic of Korea
1 shared publication - Richard Wong
Department of Computer Science, Hanyang University, Seoul, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 17 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 30476-30490
Arc Detection and Severity Analysis for Plasma Etching Processes
With the rising demand for higher quantities and quality of semiconductors, refining production processes is essential. Plasma etching, a key step in semiconductor manufacturing, is valued for its precision and speed but is also vulnerable to plasma arcing - an electrical discharge that can cause significant damage to wafers and equipment. Rapid and reliable detection of arcing incidents is critical for maintaining yield and minimizing downtime. Traditional detection methods, such as static thresholds applied to Radio Frequency signals, often struggle with non-linear signal characteristics and ambiguous edge cases, resulting in false positives or missed events. To address these challenges, this paper introduces a novel, real-time, configurable statistical anomaly detection framework specifically designed for arcing incidents in plasma etching. The system combines dynamic feature monitoring with a multi-feature validation approach, reducing false positives and improving detection accuracy. In addition, a proof-of-concept severity scoring system is introduced to provide tunable, heuristic-based scoring of detected events, enabling engineers to assess potential impact. Evaluated under log-level labeling constraints using proprietary etching logs, the proposed framework exhibits promising detection performance relative to naive threshold-based approaches and representative machine learning baselines evaluated for comparison, including LSTM, one-class SVM, and Isolation Forest, while maintaining compatibility with real-time process constraints. The system’s modular design further allows for potential adaptation across different etching configurations, offering a practical and efficient proof-of-concept framework for real-time process monitoring in semiconductor manufacturing.
- 17 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 30476-30490
Open AccessArc Detection and Severity Analysis for Plasma Etching Processes
With the rising demand for higher quantities and quality of semiconductors, refining production processes is essential. Plasma etching, a key step in semiconductor manufacturing, is valued for its precision and speed but is also vulnerable to plasma arcing - an electrical discharge that can cause significant damage to wafers and equipment. Rapid and reliable detection of arcing incidents is critical for maintaining yield and minimizing downtime. Traditional detection methods, such as static thresholds applied to Radio Frequency signals, often struggle with non-linear signal characteristics and ambiguous edge cases, resulting in false positives or missed events. To address these challenges, this paper introduces a novel, real-time, configurable statistical anomaly detection framework specifically designed for arcing incidents in plasma etching. The system combines dynamic feature monitoring with a multi-feature validation approach, reducing false positives and improving detection accuracy. In addition, a proof-of-concept severity scoring system is introduced to provide tunable, heuristic-based scoring of detected events, enabling engineers to assess potential impact. Evaluated under log-level labeling constraints using proprietary etching logs, the proposed framework exhibits promising detection performance relative to naive threshold-based approaches and representative machine learning baselines evaluated for comparison, including LSTM, one-class SVM, and Isolation Forest, while maintaining compatibility with real-time process constraints. The system’s modular design further allows for potential adaptation across different etching configurations, offering a practical and efficient proof-of-concept framework for real-time process monitoring in semiconductor manufacturing.