Rapid Aerosol Infection Risk Prediction for a Classroom Using Computational Fluid Dynamics Simulation Coupled with Data-driven Machine Learning

HYEONJUN LEE, Donghyun Rim, Pennsylvania State University

     Abstract Number: 561
     Working Group: Indoor Aerosols

Abstract
This study offers an exclusive analysis of airborne infectious disease transmission risks in classroom environments, employing Computational Fluid Dynamics (CFD) simulations to explore critical factors such as ventilation strategies, air change rates, occupant arrangements, source locations, and particle sizes. Through 224 generated simulation cases, this research also incorporates data-driven supervised learning methods—specifically, Long Short-Term Memory (LSTM) and Artificial Neural Networks (ANN)—for rapid airborne infection risk prediction. Key findings reveal that displacement ventilation decreases airborne infection risk by 49%-77% in comparison to mixing ventilation. Moreover, the conventional Wells-Riley model was identified as lacking in its ability to accurately predict spatial infection risks. The study further challenges the universally beneficial role of higher air change rates, revealing an increased risk of up to 166% for occupants seated in the back rows of a classroom when air change rates were elevated. The research also suggests that relying solely on physical distancing is insufficient in specific settings and emphasizes the need to consider other factors like airflow patterns. Lastly, while the ANN-based model struggled capturing the intricate relationship of inhalable particle number and infection risk, there is still a chance of improving the level of prediction. The results of this study can be used to inform HVAC system design and operation strategies to reduce the risk of infection transmission in indoor environments.