Evaluation of RNG and LES Non-Isothermal Models for Indoor Airflow Using PIV Measurement Data

Abstract

Airflow significantly affects indoor environment and air quality. Computational fluid dynamics (CFD) modeling provides a useful and inexpensive tool for airflow analysis, but its application is limited by the ability to validate the model with accurate measurements. In this study, both a renormalization group (RNG) κ-e model and a larger eddy simulation (LES) model were evaluated using intensive air velocity data from a full-scale test room measured by particle image velocimetry (PIV) techniques. The results showed that the RNG κ-e model predicted airflow patterns similar to the PIV measurements throughout the room space. In approximately 60% to 70% of the ventilation space, the normalized mean square error (NMSE) between the predicted and measured velocity magnitude was less than 0.25, which is an indicator of adequate model performance. Furthermore, under winter conditions, approximately 85% to 90% of NMSE values in animal and human zones were less than 0.25. These factors indicate that the RNG model is an effective tool for predicting non-isothermal airflow distributions in a full-scale room with simple geometry. The low grid LES model (<300,000 grids), which is feasible using Fluent software and normal computer resources, did not predict the PIV measurement as closely as did the RNG model in this study.