top of page
Tirah Wu
TRICORNTECH CORPORATION
Business Development
Vice President
Education
Taipei Tech, B.A. in English
Experience
﹒Tricorntech Corporation Business Development
﹒Taiwan High-tech Facility Association Working Group
﹒SEMI Taiwan Membership Representative
Speech Title and Abstract
Physics-Constrained AI Digital Twin for Environmental Risk Prediction in High-Tech Facilities
Maintaining environmental stability is critical for semiconductor fabrication facilities. However, interactions between process exhaust emissions and rooftop airflow may create transient contamination risks that are difficult to predict using conventional monitoring or standalone CFD simulations. This study presents an AI-driven digital twin framework that integrates CFD modeling, real-time environmental sensing, and Physics-Informed Neural Networks (PINN) to enable predictive environmental risk management for high-tech facilities.
By embedding governing fluid dynamics and pollutant transport equations into the neural network training process, the proposed approach ensures physically consistent predictions even under limited data conditions. A real semiconductor facility environment was used as a case study to construct a digital twin model linking rooftop exhaust dynamics, airflow patterns, and contamination monitoring data. The framework enables prediction of potential contamination recirculation events and provides early risk insights for facility operation. The results demonstrate how physics-constrained AI can transform facility environmental management from reactive monitoring to predictive intelligence for next-generation smart fabs.
bottom of page