Modelling strategies for the prediction of hot streak generation in lean burn aeroengine combustors

The accurate prediction of the fluid dynamic conditions at the exit of gas turbine combustors are of paramount importance in the aero-thermal design of the aero-engine. In fact, both the heat loads and the aerodynamic performance of the high pressure turbine (HPT) are substantially affected by the entry conditions, such as velocity components, temperature and turbulence intensity. The problem is particularly serious in new generation devices based on a lean burn concept. Compared to standard Rich-Quench-Lean (RQL) scheme, the absence of dilution jets and the use of highly swirled flows for flame stabilization make the control of combustor exit temperature distribution a complex task. Therefore, the high-fidelity prediction of the hot streak formation within the combustor, as well as its propagation through the HPT, are becoming key aspects. In this work, different strategies for turbulence modelling are tested, mainly focusing on scale-resolving approaches such as Large-Eddy Simulation (LES) and Scale Adaptive Simulations (SAS), which are becoming increasingly popular with the availability of more powerful computational resources. At the same time, classical eddy-viscosity models based on RANS approach are considered, as they still represent the standard simulation strategy in the industrial framework, when a fast response is required in a preliminary design phase. The different methodologies are benchmarked on an experimental test article representative of an aeronautical lean burn combustor cooled by means of effusion. The benchmark performed at engine relevant conditions allowed to draw interesting conclusions for the purposes of the aero-thermal simulation. SAS proved to be a valid alternative to LES, returning on the whole the same level of accuracy. As expected, the disagreement obtained with RANS was significant. The sensitivity to the turbulent Prandtl number was investigated to provide an insight on its reliability in compensating the underestimation in turbulence mixing. (C) 2018 Elsevier Masson SAS. All rights reserved.