The first part of this work presents a comparison of predictions obtained with several two-equation type RANS turbulence models commonly used in industry against experimental data obtained by Whitelaw et al. All examined models yield a relatively poor match in the flow region very close to the wall; agreement with the measurements improves significantly when moving further away from the wall. This concerns both the internal normal stress profiles and the average velocity profiles, the latter show improved prediction of the recirculation zone area when moving further into the main stream. Downstream behaviour for both models shows an excellent match more than 6 diameters away from the jet inlet, defined as the region after which the flow essentially resumes its normal duct behaviour. Expanding upon these RANS results, another series of simulations using LES modelling with the standard Smagorinsky SGS model was conducted using the same grid and compared to the RANS-based results. Although performance in the most complex flow areas was slightly improved over RANS, this was at the cost of an increase of computation time by almost a factor of 6. The next stage involved developing a code based on the model for two-phase flow described in to predict the atomisation pattern for a non-vaporising (or "cold") flow based on the parameters of the previous simulations. This model implements transport equations for the liquid mass fraction and the average surface area per unit mass along with an equation for average density; resulting in an entirely Eulerian model which can be used to predict atomisation from first principles. Current work consists in development of additional source terms describing vaporisation in a strongly turbulent environment and further coupling with a combustion model applicable to the combustion chamber of an industrial gas turbine.
展开▼
