| Abstract: | An evaluation of some numerical methods for turbulent reacting flows in furnace-like geometries is carried out. The Reynolds averaged Navier–Stokes equations and the two-equation k–? model together with either finite-rate or infinite-rate reaction models are solved numerically. Either single- or multiple-step reactions together with the ‘eddy dissipation concept’ (EDC) are used to model reacting flows with finite reaction rates. The numerical scheme is finite difference based, together with a multi-grid method and a local grid refinement technique. These methods have been used to calculate the combustion of propane in a single- and multiple-burner configurations. In the former case, the sensitivity of the solution to variations in some model parameters (determining the reaction rate) and numerical parameters (mesh spacing) has been studied. It is noted that different dependent variables exhibit different levels of sensitivity to the variation in model parameters. Thus, calibration and validation of models for reacting flows require that one compares the most sensitive variables. For engineering purposes, on the other hand, one may calibrate and validate models with respect to the most relevant variables. Our conclusion is that since sensitivity of the temperature distribution is relatively mild, one can still use EDC-like methods in engineering applications where details of the temperature field are of minor importance. |
