Session: 34-03 Turbulence modeling, CFD models, and assessment

Paper Number: 151860

An Assessment of Algebraic and Differential Reynolds-Stress Models for a Highly-Bent Serpentine Aircraft Intake 

Serpentine ducts are highly bent diffusers commonly used as intakes of modern low-bypass aero-engines. These systems draw atmospheric air to the aero-engine installed in the aircraft fuselage. The performance of the compressor stages are coupled to the upstream flow features that develop in the intake, and this coupled intake-compressor interaction requires thorough analysis over the entire flight envelope, to account for changing aerodynamic features that develop with changing aircraft angles of attack, Reynolds number, Mach number, as well as engine mass flow rate. Serpentine intakes feature three-dimensional surface curvature and strong pressure gradients that cause unsteady vortex formation and flow separation, and these flow features have a misalignment between the eigenvectors of the Reynolds-stress and mean-strain-rate tensors. These large upstream flow features are ingested by the engine, significantly reducing the performance of the compression stages of the engine. Quantifying the performance of such a complex system requires a detailed analysis of the aforementioned parameters, and high-fidelity CFD approaches, such as large-eddy simulations, are too computationally costly to study such a large and complex parameter space. Steady Reynolds-averaged Navier-Stokes (RANS), on the other hand, is sufficiently affordable to conduct the required parameter sweeps. This highlights the requirement for robust turbulence modelling strategies for studying such a complex system. The current industrial eddy-viscosity turbulence models are inappropriate for the study of such a complex system (Gerolymos et al., 2010), and this partly because of an assumed alignment of the eigenvectors of the Reynolds-stress and mean-strain-rate tensors. Explicit algebraic Reynolds-stress models (EARSMs) overcome this assumption with an algebraic anisotropic correction to the Reynolds-stress tensor, which allows for a modelled non-linearity between the Reynolds-stress and mean-strain-rate tensors. Differential Reynolds-stress models (DRSM) utilise transport equations for the six components of the symmetric Reynolds-stress tensor, and because each term of the tensor is solved separately, this removes algebraic relationships between the two aforementioned tensors. Turbulence production and anisotropy are two quantities of importance when studying vortical structures and surface curvature, and both EARSM and DRSM can provide tangible improvements in the representation of these quantities in a RANS calculation. In the case of a DRSM, improvements in history effects should also be expected, and this is because of the additional transport equations used to describe the Reynolds-stresses in the domain. When studying a serpentine intake with a RANS calculation, deficiencies in the turbulence model will degrade the accuracy of the calculation in the developing flow, leading to results that rarely match the experimental measurements at the location where the intake interfaces with the engine (commonly known as the aerodynamic interface plane). This degradation occurs because of the assumptions made about the state of turbulence in model development, and these assumptions should be considered when selecting a turbulence model for complex industrial flows. In the present work, we consider the non-linear algebraic Reynolds-stress correction of Wallin and Johansson (2000) and the elliptic blending Reynolds-stress model (EBRSM) of Manceau and Hanjalić (2001). These models are used to calculate the flow through a highly-bent serpentine intake benchmark known as the military engine intake research duct (MEIRD) (Haug et al., 2018), in the operating conditions of 2.50 * 10^{5}< Re_{b,in} < 1.51* 10^{6} with low Mach numbers. This geometry is used as a turbulence modelling benchmark and we focus on the near-wall performance of the elliptic blending equation and the pressure-strain correlations, as well as wall-pressure values of the numerical and experimental studies. This provides an assessment of capabilities of EARSM and DRSM, and offers guidance on model selection when studying complex internal flows. Furthermore, the analysis in this work provides a fundamental understanding of the EBRSM performance and highlights future opportunities for model refinement.

Presenting Author: Sean Hanrahan The University of Melbourne

Presenting Author Biography: Sean is a final year PhD candidate at the University of Melbourne (Australia). His research specialises in physics-informed machine learning and turbulence modelling for industrial applications, with a focus in developing algorithms that are constrained to the known physics of wall-bounded turbulence. Sean currently studies advanced Reynolds-stress turbulence models in serpentine ducts, with an interest in the pressure-strain correlations of the near-wall region.

Authors:

Sean Hanrahan The University of Melbourne
Julian Scheibel Universität der Bundeswehr München
Andreas Grois Universität der Bundeswehr München
Marcel Stößel Universität der Bundeswehr München
Melissa Kožul The University of Melbourne
Suad Jakirlić Technische Universität Darmstadt
Dragan Kožulović Universität der Bundeswehr München
Richard Sandberg The University of Melbourne