Rans/Les Analysis for the Turbulent Mixing in a Gas Turbine Injector

Abstract

[EN] CFD plays a very important role in the design of gas turbine combustors. Accuracy in mixing and combustion modelling can have a great impact in the final performance of these components. Combustor flows mean a huge challenge in this subject, as they present complex unsteady structures and the range of scales to be simulated or modelled is very wide. Moreover, swirl flows, present in most gas turbine injectors to increase their performance, make the simulation process more challenging. Steady simulations, like RANS, are unable to offer accurate results for mixing and combustion problems involving swirl flows. While DNS and standard LES models can provide high-fidelity results, they require a grid resolution and computational cost that most industries cannot afford. The objective of this thesis is evaluating the performance of hybrid turbulence models for mixing problems in order to establish an intermediate solution with good accuracy at a reasonable cost. U-RANS, DES, SAS, D-LES and ILES have been applied in an injector test-case for a better understanding of the flow features and validation of the different turbulence approaches. For the first time, unstructured meshes were employed for the specific test-case demanded by the collaborative partner, PERM engines. Hybrid RANS/LES models were used with Fluent’s pressure based solver and, independently, these same approaches were implemented within an in-house unstructured HLLC Godunov-type solver. In the context of high-resolution methods, this thesis takes an initial step implementing k-ω SST, DES, SAS models and species mixing in this solver, and the required time-analysis tools. After a preliminary validation of the models, steady and unsteady cases were run in the injector, with promising results. In terms of implementation, this work leaves a powerful tool with strong potential in turbulence analysis, for further study of combustion and many other different flows