Simulation of Atrium Smoke Filling by Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) is evidently relevant to the study of fires, yet the intermediate chemistry has yet to be factored successfully into combustion models. Consequently, predicted airflow patterns, together with pressure and temperature contours, are mostly used in evaluating the performance of smoke control systems. But even using these assumptions, very few studies exist comparing predicted results from CFD with experimental findings. This leaves research with a paucity of data on how smoke is likely to spread, fill and be controlled in large halls.

While hot smoke tests in a hall in Japan have yielded good-quality experimental data for smoke layer interface heights, the dominant approach to hazard assessment on big projects in East Asia is to use a software Fire Dynamics Simulator (FDS) developed at the National Institute of Standards and Technology in the USA. In this paper, FDS predicted results for smoke filling and exhaust will be examined side-by-side with experimental data. Our technique is to compare recorded transient smoke layer interface heights with the results predicted by FDS.

This work also responds to questions on how to determine smoke layer interface height. In addition to the default FDS method (FDS menu), two methods are proposed to determine heights from sharp changes in, respectively, vertical temperature profiles and particle tracking. Functional analysis is applied to justify predictions, with results suggesting that CFD offers fairly good predictions on smoke layer interface height.