Comparing turbulence models for buoyant plume and displacement ventilation simulation.

Computational fluid dynamics may be used to predict the details of airflow in rooms served by displacement ventilation systems, provided a suitable turbulence model can be found. Since buoyant plumes are central to the displacement ventilation strategy, four turbulence models - three eddy-viscosity models (the 'standard' k-s model, a modified k-s model, and an RNG k-s model) and the Reynolds stress model - were applied to simulate airflow in a turbulent buoyant plume. Corresponding experimental data from the literature were used for validation, although for a plume stronger than expected in rooms as no reliable plume data for room air flow were found. The Reynolds stress model predicted velocity, temperature, and turbulence quantities satisfactorily while the eddy-viscosity models performed poorly. The eddy-viscosity models were then applied to predict airflow in a furnished room with displacement ventilation. The computed airflow patterns, mean velocities, temperatures, and contaminant concentrations agree reasonably well with the experimental data obtained from a full-scale test chamber but the discrepancies in some locations were large.