2025년 7월 8일
Ontology-based Process Data Preprocessing
10-2025-0091382 · KP25083
발명자
Taekwon Jee
출원인
SemiAI, Co.,Ltd.
정식 명칭
METHOD AND APPARATUS FOR ONTOLOGY-BASED PROCESS DATA PREPROCESSING FOR HYBRID INDEX GENERATION
요약
This invention relates to an ontology-based preprocessing method and apparatus for semiconductor manufacturing data that integrates heterogeneous data by collecting process data from a digital twin fab through a queue-based system and semantically normalizing and classifying it using an ontology hierarchy. The data is then transformed into a hybrid index by combining dimensionality reduction techniques such as PCA and ICA with physics-based models, and mapped to process performance metrics such as yield and defect rate for machine learning input. Multiple machine learning models are ensembled to improve prediction accuracy, and the processed data, prediction results, and model metadata are stored in a database, enhancing analytical efficiency and predictive performance while supporting process optimization.