RANS/LES Hybrid Turbulence Model For Fire Simulation

Fire embraces nearly all the effects found in subsonic chemically reacting flow. It is a sophisticated problem involving many physical and chemical processes. Of great importance to the spread and control of fire is the smoke flow in fire, which, due to the complex characteristic of fire, is often turbulence. Therefore, the success of fire simulation will depend directly on the accurate modeling of turbulence.Currently, two approaches are often employed to model turbulence: Reynolds-averaged Navier-Stokes Equations method (RANS) and Large Eddy Simulation method (LES). The predictions of many important problems made by RANS are not satisfactory, while LES is becoming more and more important in fire simulation because of its budget between accuracy and cost. LES, however, has a strict demand for grids, especially near rigid walls. RANS/LES hybrid turbulence model can combine the good performance of RANS within the attached layer and of LES in separated region. Starting from the general definition of Detached-eddy Simulation (DES) given by Travin, this thesis constructs two new hybrid turbulence models based on the k — e model with the buoyancy modification by introducing the energy-containing turbulence length scale related with the grid size. When the grids are coarse, these newly-constructed hybrid turbulence model will degrade to the k—ε model with the buoyancy modification, while when the grids are fine enough, one model can be seen as the extension of Smagorin-sky sub-grid model of LES with the effect of non-equilibrium, and the other become a two-equation sub-grid model.The new models are applied to a natural convection problem and a forced convection problem, respectively to test their performances. The calculations are carried out using various grid conditions. When compared with the calculation results of LES and experimental data, the results obtained with the hybrid models agree quite well with the experimental data and are at least not worse than those of LES, while using less grids than LES. Furthermore, the new hybrid models are employed to simulate two typical fire scenarios: Steckler room fire and shopping mall fire. The results give correctly the profile of temperature and velocity in the fires, and successfully predict the strongentrainment and mixing effects near the exit between the smoke flow and outdoor environment. All these show that the hybrid models developed in this thesis are suitable for the simulation of such complicated problem as fire, which is characterized by high Reynolds number and commonly-found walls.However, it will not suffice to improve only the turbulence model. The turbulence combustion model play an important role in the fire simulation.