Evaluation of Turbulence Models and Boundary Conditions for Hybrid Ventilation in Reduced-scale Classroom Model

TL;DR

The paper addresses how turbulence models and boundary conditions influence CFD predictions of hybrid ventilation in a reduced-scale classroom. It employs multiple RANS models in ANSYS Fluent and validates them against detailed PIV measurements on a 1/12-scale model, exploring both standard inlet definitions and an extended domain that includes the exterior. Key findings show that geometry and inlet boundary conditions at the door dominate prediction accuracy, with extended-domain simulations yielding the closest agreement to experiments, while the choice of turbulence model has a comparatively minor effect. This work provides a validated CFD framework for evaluating mechanically assisted natural ventilation designs and highlights the need to accurately represent exterior geometry when predicting indoor airflows relevant to infection risk mitigation.

Abstract

In this paper, we study the ventilation airflow in a model classroom, where exhaust fans throw out the used air, to replace it with outdoor air through open door. Hybrid ventilation, or mechanically assisted natural ventilation, of this kind is used as a retrofit design to reduce infection risk from airborne transmission. The air stream entering the door forms a jet-like flow, driven by the suction effect of exhaust fans. We compute the jet velocity using Reynolds averaged Navier Stokes (RANS) method and compare with velocity field measured using particle image velocimetry. Different turbulence models are found to match experimental data near the door, but they over-predict the peak jet velocity further downstream. There is minimal variation between the results obtained using different turbulence models. The computational results are found to be sensitive to inlet boundary conditions, whether the door entry is specified as a pressure inlet or velocity inlet. The geometry of the space outside the door also has a significant effect on the jet velocity. Changing the boundary condition takes the computational results closer to the experimental data; the velocity profiles computed with the extended domain being the closest to the measured peak velocity. Interestingly, the centerline velocity decay computed with the extended domain aligns well with the experimental data. The other cases, irrespective of turbulence model, show much lower decay rate that seem to align with wall jet scaling. This suggests that geometry and boundary conditions at the door is critical to predict the airflow in hybrid ventilation.

Figures (12)

• Figure 1: Validation of CFD methodology with experimental data for forced convection case from Nielsen et al.Nielsen2010Annex
• Figure 2: Geometry and dimensions of (a) realistic classroom, (b) reduced-scale model, showing PIV set up, and (c) computational domain and grid used in CFD. A brief schematic of the PIV set up is also included in part (b).
• Figure 3: Flow caracteristics in the scaled breathing level at z$=0.11$ m: (a) velocity magnitude, (b) turbulent kinetic energy and (c) turbulent intensity, along with the profiles of (d) $y$-component and (e) $x$-component of velocity, and (f) turbulent kinetic energy at different $y$-locations.
• Figure 4: Velocity profile comparison with PIV and anemometer results for different turbulence models at four streamwise $y$-locations.
• Figure 5: Effect of varying turbulence Intensity at the inlet boundary on (a) TKE, (b) velocity