Numerical and experimental investigation of shaft-like compartment fires

The present study focuses on under-ventilated compartment fires, by analysing the results of an experimental and numerical study about fires issuing from heptane pools in a compartment with an open door. This compartment has an aspect ratio (height/width) of the order of 1.8 and a doorway with an aspect ratio up to 4, which is very different from conventional compartments. This type of geometry is therefore expected to yield new results. The key parameters in this case are the ventilation factor (based on the door height and its area) and the pool fire diameter. To analyse the combustion regimes, the fire Heat Release Rate (HRR) is split into two parts, namely, the HRR released inside and outside the compartment. To quantify these quantities, two experimental methods have been combined. The first one is based on temperature measurements inside the compartment. The second one combines measurements from radiative heat fluxes outside the compartment and images taken with visible cameras that provide the flame shape in order to evaluate the corresponding heat released rate. The experimental data have been compared to dedicated numerical simulations carried out with the CFD code Fire Dynamics Simulator. These comparisons have revealed that in a well-ventilated or moderately under-ventilated regime, integral quantities such as mean temperature, fire heat release rate inside and outside the room appear to be well predicted by the simulations. However, under the same conditions, the temperature profiles show differences that can be locally significant. This observation might be explained by a difficulty in correctly reproducing the flame dynamics and the air flow entering through the door. Furthermore, in the severely under-ventilated case, the temporal evolution of the average quantities (HRR and mean temperatures) are very poorly estimated by the simulations, as well as the local distributions. In this case the simulations predict that the combustion of all the combustible vapours occurs outside the door, which does not correspond to the physical observations. This phenomenon have been observed at ${\dot{Q}}^{\ast }\approx 1.6$, where ${\dot{Q}}^{\ast }$ is the dimensionless heat released rate. Therefore, improvements of the CFD code are required to improve the simulations, probably also involving dedicated efforts on the ignition and extinction sub-models.