Best practices for computational fluid dynamics simulations of indoor droplet dispersion: Impact of grid, human models, thermal boundary conditions, droplet size, and interactions

Computational fluid dynamics (CFD) simulations of respiratory droplet dispersion are vital for implementing effective epidemic prevention strategies. However, the accuracy of CFD simulations depends on numerical and physical modeling parameters, whose selections vary significantly across previous studies. This study aims to unify diverse approaches in CFD simulations of indoor droplet dispersion by conducting a comprehensive sensitivity analysis based on the latest experimental data. We assessed the impacts of computational grids, human geometric models, thermal boundary conditions, droplet size distributions, and inter-droplet interactions on indoor droplet dispersion and infection risk. The findings reveal that polyhexcore grids offer accurate airflow and droplet concentration predictions with the shortest computational time. Simplified human models distorted the surrounding airflow and droplet deposition, underestimating infection risk by up to two times compared to three-dimensional (3D) scanned human models. The heat flux weakened thermal plumes, leading to an 8.5% absolute risk reduction (ARR) in infection relative to temperature boundaries. Simplified droplet size distributions preserved dispersion patterns but overestimated infection risk. Notably, the addition of droplet breakup and collision to CFD simulations decreased droplet concentrations in the breathing zones, achieving an ARR of 3.44% in infection risk. These insights provide guidance for enhancing the fidelity of indoor droplet transmission simulations and infection risk assessment.