The aim of this thesis is to design and carry out bench-scale laboratory experiments specifically designed for the validation of fire models, and to use the experimental data for a validation study of the Fire Dynamics Simulator (FDS). The focus of the experiments is on one of the key components of fire models, the modeling of buoyancy-driven flows. The experimental setup is simplified by neglecting pyrolysis and combustion and its objective is to achieve high precision and reproducibility. Therefore, an electrically heated block of copper is used as a heat source and particle image velocimetry (PIV) is applied for measuring the flow velocities. Two different setups are investigated: an undisturbed open buoyant plume above the heat source, and a buoyant spill plume emerging from a compartment opening.
Depending on the setup, different characteristic values of the flow are evaluated. For the undisturbed buoyant plume setup, maximum velocities in the plume, plume widths and flow integrals are determined. Furthermore, a centerline analysis is carried out in order to localize the transition from laminar to turbulent flow as a function of the Grashof number. The obtained values lie in the range 4 x 10⁸ < Gr < 2 x 10⁹ and therefore agree well with previous studies. For the buoyant spill plume compartment setup, position and maximum velocity of the outflow as well as volume flows in front of the opening are compared. The heat flows out of the opening cannot be measured directly and are therefore estimated based on the available data.
The validation study with FDS leads to mixed results. For the open plume setup, generally a good agreement is achieved. Provided that a sufficiently fine resolution is used, the important flow characteristics are reproduced in the simulations. With the spill plume setup, significant differences between experiments and simulations are observed. In order to analyze them in detail, modifications of the experiments are required. Therefore, in conclusion of the study, areas of potential improvements for future experiments and the suitability of PIV for this kind of experiment are discussed.