Prediction of the loss generation in anisothermal compressible flows applied to Nozzle Guide Vane with the Large Eddy Simulations
Monday January 20th – 2 p.m.
To improve the efficiency of aeronautic engines, the turbine entry temperature has strongly increased in recent years. Such high temperatures induces high thermal stresses for the turbine blades and vanes which reduces the blade lifetime. To overcome this thermal issue, efficient turbine cooling systems need to be designed. To do so, the accurate prediction of blade wall temperature and losses generated by these systems is required. Taking the opportunity of recent developments of high-fidelity predictions, this PhD thesis funded through the FUI project CASCADE with the support of Safran Helicopter Engines (SHE), aims to evaluate the prediction of blade wall temperature and losses for cooled high-pressure vanes with Large Eddy Simulations (LES). To do so, academic and complex anisothermal configurations featuring film cooling are investigated. Results obtained in the present work show that LES is able to predict the flow aerodynamics and blade wall temperature for all configurations studied. The prediction is clearly improved if the mesh is sufficiently refined in high dynamic regions and if turbulent fluctuations are taken into account at the inlet of the computational domain especially for cases presenting separation bubbles. To ease the use of LES in an industrial context and reduce the CPU effort associated to the resolution of the flow in the cooling system of turbine blades and vanes, a new coolant ejection model is introduced and evaluated. This model is shown to well reproduce the coolant jets aerodynamics and provides a good prediction of the wall temperature without meshing the cooling system. To accurately evaluate and investigate the losses in this context of turbine blade cooling, the approach Second Law Analysis (SLA) is adopted. Contrary to total temperature and total pressure balances, SLA directly gives access to 3D loss maps which are constructed from the entropy source terms resolved on mesh. As a result, the loss generation mechanism can be locally investigated and does not require any averaging procedures contrary to 1D loss models. These loss maps are split in an aerodynamic contribution and a mixing contribution which is linked to mixing process between hot and cold flows. The study of these loss maps shows that aerodynamics losses are generated in high sheared regions (boundary and mixing layers) while mixing losses are produced in the film cooling and in the wake of the vanes. Advanced analysis of loss maps indicate that turbulent fluctuations dominate the loss generation mechanism. This last finding evidences the benefits of SLA to predict losses from LES fields. Indeed and contrary to RANS, the turbulent contributions to losses are directly resolved on mesh with LES and does not require any modelling strategies. As a consequence of this PhD work, SLA coupled to LES is shown to be a very promising methodology to predict the flow aerodynamics and losses for the design of future geometries of industrial turbine vanes and blades.