This project concerns a fire modelling validation exercise involving a 3 MW compartment test fire conducted by the Loss Prevention Council. The simulations were performed using SMARTFIRE V2.1 (build 370) and involved two representations of the fire source, a simplistic volumetric heat release rate model and a gaseous combustion model. For comparison purposes the fire was also simulated using the CFX V2.2.2 software. In SMARTFIRE, the fire can be represented as a transient volumetric heat and fuel mass source through the use of diffusion or eddy-dissipation combustion models. The code uses the SIMPLE and SIMPLEC algorithm and can solve turbulent (two equation K-Epsilon closure with buoyancy modification) or laminar flow problems under transient or steady state conditions. Radiation is represented through the use of a six-flux radiation model.
The test comprises of a burning wood crib within an enclosure with a single opening. The test compartment had a space volume of 6m x 4m x 3.3 and contained a door (vent) measuring 1.0m x 1.8m centrally located on the rear 6m x 3.3m wall. The door was open throughout the experiment. After approximate 500 seconds, the mass loss of the wood crib reaches a quasi-steady condition and the peak value is 0.1978 kg/s. If complete combustion is assumed, the peak heat release rate is estimated as 3.5 MW by using 17.9 MJ/kg of the heat of combustion of pine wood. Two vertical trees of thermocouples were installed within the fire compartment. One was situated directly above the fuel source to measure the fire plume temperature. The other was located in the left rear corner.
The results generated using the SMARTFIRE and CFX heat release rate models are similar and the results generated using the SMARTFIRE and CFX combustion models are similar. As may be expected, results generated using the gaseous combustion model are in better agreement with the experimental data than those generated using the simple volumetric heat release rate model.
The plume temperatures predicted by the simple volumetric heat source model are much higher than those predicted by the combustion models and the corresponding measurements. More importantly, the heat source models failed to reproduce the trend of temperature variation above the middle of the crib.
The temperatures along the vertical line above the middle of the crib at the heights of 1.5m and 3.0m predicted by SMARTFIRE and CFX (along with the experimental results) are plotted in Figure 1. The predictions are generated using the combustion models. The numerical predictions appear somewhat higher than the measured values however, they are reasonably close to the measured values. Once again, SMARTFIRE and CFX produce very similar results. The trend of temperature variation is captured by both codes. After the peak temperature has been attained, SMARTFIRE predicts a steady decrease in temperature, while some oscillations occur in the CFX predictions.
The SMARTFIRE simulation demonstrates that the peak trend in the plume
temperature variation is mainly caused by the changes of the fire plume shape. A
possible explanation for the observed and predicted peaks in temperature is due
to the plume leaning back away from the centre line. This is supported by the
fact that the maximum plume temperature is reached well before the maximum in
heat release rate. After the initial phase of fire growth, the hot combustion
products accumulate beneath the ceiling creating a gradually deepening hot
layer. In conjunction with the fresh air being entrained into the compartment by
the fire, the downward movement of the hot upper layer pushes the fire plume
back so that it tilts away from the opening towards the rear wall. Since the
line of the thermocouple tree is not in the centre of the fire plume, the
temperatures along it predicted and measured are reducing after they reach the
For information relating to the SMARTFIRE fire field model visit our
SMARTFIRE web pages. You will
find a range of publications relating to SMARTFIRE and our other combustion
research on the FSEG Publications pages both CMS press and