Development of a Novel Metric for Indoor Infection Risk and Advancing Automated Mitigation through Ventilation and Machine Learning

Abstract

Airborne infection risk (IR) in indoor environments, particularly in workplaces and healthcare settings, requires effective mitigation strategies. Accurate assessment relies on a probability of infection (POI) metric that incorporates particle number, accumulated viral load, and clinical data, capturing the complexity of airborne pathogen transport. However, practical implementation necessitates expressing IR in measurable terms such as CO₂ concentration for real-time monitoring and adaptive ventilation control. This thesis develops a CO₂-based IR prediction framework, integrating computational fluid dynamics (CFD) simulations and machine learning to enhance ventilation strategies. A key aspect of IR quantification is accumulated viral dosage, where continuous exposure to airborne pathogens over time influences infection likelihood. This highlights the significance of exposure duration and ventilation effectiveness (VE) in risk assessment. Despite the importance of outdoor air dilution, optimal ventilation rates remain uncertain due to the tradeoff between contaminant removal and the risk of air recirculation. To address this, the research systematically analyses airflow dynamics, the age of air, and VE, providing insights into their influence on IR. It further examines ventilation system configurations, infector positioning, and airflow patterns to assess their impact on both IR and CO₂ distribution. The role of VE in optimising airflow rates is explored, demonstrating strategies to minimise IR while avoiding excessive ventilation demand. This thesis introduces a novel POI metric using CFD and a transient Eulerian-Lagrangian approach to model the spatial-temporal dynamics of airborne infectious particles. Unlike traditional concentration-based assessments, this metric accounts for particle count, exposure duration, and clinical data, enabling the generation of detailed infection risk maps. To enable practical CO₂-based IR prediction, a machine learning-driven framework is developed for real-time ventilation control based on CO₂ sensor inputs. This facilitates automated infection risk mitigation, allowing intelligent building systems to dynamically adjust ventilation rates in response to changing conditions. The findings contribute to a comprehensive framework for adaptive ventilation control, bridging CFD modelling, CO₂- based prediction, and machine learning to support intelligent, data-driven building systems. By integrating real-time infection risk assessment with adaptive ventilation strategies, this research advances safer, energy-efficient infection control solutions for indoor environments.

Date of Award	9 Jun 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Yasser Mahmoudi Larimi (Supervisor), Amir Keshmiri (Supervisor) & Majid Sedighi (Supervisor)