Numerical and experimental studies on room airflow

In this study, the air flow in a small scale industrial hall is simulated numerically using two different computing codes, the commercial flow solver with a high-Reyholds number turbulence model and a university code with a low-Reynolds number turbulence model. The results are compared with measurements. Two different air supply arrangements with grille or nozzle types of air terminal devices are studied, both with isothermal and non-isothermal boundary conditions.
The air flow patterns are very complicated since heat and momentum sources are distributed. The supply air devices cannot be described in detail without the computational grid becoming impractically large. Therefore, an integral part of a CFD simulation is to simplify the actual flow situation withous losing essential features of the flow.
The air velocities with the high Reynolds number code were consistently higher than the measured ones, and therefore, one could say that these results were on the safe side considering the draught risk. The low-Reynolds number simulations predicted too low velocities in some cases. In some parts of the room een the qualitative air flow pattern was different in the two simulations, most notably in the regions where the buoyancy forces and the inertia forces of the supply air jet are of the same magnitude and the supply air jet is deflecting near the solid boundaries.
It turned out to be difficult to make accurate comparisons with the measurements because the comfort oriented, omnidirectional air speed measurements were not directly comparable with the air velocity computed with the turbulence models. One possibility is to use an artificial, modified velocity for comparison instead of the actual velocity. This method has been used in this study and it seems that especially in low speed areas the correspondence between the measured and calculated speeds is improved by using the modified velocity for comparison.