Experimental verification of the Fire dynamics simulator (FDS) hydrodynamic model.

Abstract

The objective of this research has been to verify the hydrodynamic model that is contained within the Fire Dynamics Simulator (FDS). In the first part of the research, a series of buoyant salt water experiments have been conducted, with the purpose of generating experimental data for comparison with computational fluid dynamics (CFD) models. Two types of buoyant flows have been generated in the experiments; a natural transitional flow, and flows that resemble fire induced smoke flow within a residential building. Laser Induced dye Fluorescence (LIF) has been used to measure the fluid density in a single vertical plane of the flow. Measurements have also been made of eddy frequencies on the perimeter of the transitional flows, and of the temporal development of the fire similar flow fields. The uncertainty of the experimental measurements has been quantified. In the second part of the research, the salt water experiments have been simulated with the FDS, to assess the accuracy of the hydrodynamic model. The simulations of the transitional flows are found to be highly dependent upon the resolution of the computational grid. The findings highlight the fact that the numerical methods employed in the FDS can generate fluid behaviour in the computational flow field that does not occur in the real salt water flows. This "numerical fluid behaviour" is clearly seen in the transitional flow computations, because at the source of the flow, the buoyancy and the momentum of the fluid are orientated in perpendicular directions to each other. The comparison of the computational and experimental results for the transitional flows show that the trajectory of the computed buoyant plume is steeper than the trajectory of the real salt water plume. It is speculated that the disagreement in the plume trajectory may be due to the spatial distribution of pressure within the computational domain. Due to limited computational facilities, this research has been unable to determine if the FDS hydrodynamic model can accurately compute the natural transition to turbulence. Further simulations of the transitional flows are required with grid cell dimensions that are less than the compartment height divided by 100, to determine if the transition can be correctly computed. The simulations of the fire similar flows have shown, that the FDS performs well in modelling fully turbulent flow fields, as found in residential building fires. From the fire similar flow simulations a maximum grid cell dimension, of the compartment height divided by 50, has been recommended for the simulation fire induced smoke flows within multicompartment residential scale buildings. At this recommended resolution, and resolutions coarser than this, the Smagorinsky sub-grid scale (SGS) has been found to give more accurate results than the constant viscosity SGS model. A relationship has been determined, for the minimum fluid viscosity that is required for stable computations in simulations that use the constant viscosity SGS model.