Session: 34-09 High-fidelity CFD – General

Submission Number: 178962

Benefits of Scale-Resolving Simulations for Blade Row Configurations 

As the aerospace industry advances toward more efficient and sustainable aero-engine technologies, accurate and computationally efficient turbulence modelling becomes increasingly critical for achieving future performance and emission-reduction targets. Modern high-bypass ratio designs rely on smaller core compressors that operate at lower Reynolds numbers, where transitional and separated flows challenge the reliability of conventional RANS models. Scale-resolving simulations offer higher fidelity but have traditionally been too expensive for iterative design processes. Recent advancements in GPU acceleration, however, open the possibility of performing such simulations with turnaround times that bring high-fidelity analysis closer to design practice.

This study demonstrates the application readiness of the Ansys GPU-accelerated CFD solver for turbomachinery applications. Two compressor cascades are examined: The Low Reynolds Number Outlet Guide Vane (LRNOGV) with Re = 150,000 and Ma = 0.6, and the Transonic Cascade TeamAERO (TCTA) with Re = 1,350,000 and Ma = 1.21. The corresponding experimental investigations were conducted at the Transonic Cascade Wind Tunnel of DLR.

Particular attention is given to the numerical setup required to accurately reproduce the experimental results, with emphasis on the effects of the axial-velocity density ratio (AVDR) and the spanwise extent of the computational domain. The impact of different turbulence modelling strategies, i.e. RANS and LES, is analysed in detail. Wall-modelled and wall-resolved LES are incorporated. Furthermore, the transient shock-boundary layer interaction characteristic for the TCTA case is recognized. An in-situ signal-processing approach is employed to reduce the computed cycle time units (CTU) to the essential minimum.

Performance is evaluated across mesh types and precision settings to establish robustness for industrial use. A Python-based workflow is utilized to enhance standardization, by minimizing user error and ensuring automation and a hardware-agnostic process.

Complementary investigations address Machine-Learning-driven tuning of RANS turbulence model. Data-driven parameterization informed by flow features reduces discrepancies between RANS predictions and LES references, while transition modelling further improves predictive fidelity. Crucially, the trained model demonstrates generalization capability, maintaining accuracy across off-design inflow angles.

This work, conducted within the Horizon Europe Sci-Fi Turbo project, illustrates a pathway toward integrating, GPU-accelerated solvers, scale-resolving simulations and data-driven turbulence modelling, into next-generation aero-engine design workflows, reducing predictive uncertainties while advancing both performance and sustainability objectives.

Presenting Author: George Klavaris Synopsys Northern Europe Ltd.

Presenting Author Biography: George Klavaris is a Senior R&D CFD Engineer at Ansys (now Synopsys), specializing in turbomachinery aerodynamics, Verification and Validation (V&V), and next-generation CFD methodologies. His work focuses on GPU-accelerated CFD, automation and data-driven turbulence modelling using Adjoint Methods and Maxhine Learning. He represents Ansys UK (now Synopsys) in the Horizon Europe Sci-Fi Turbo project, contributing to the integration of high-fidelity simulations and data-driven modelling into industrial aero-engine design workflows. George holds an MSc in Computational Fluid Dynamics from Cranfield University (2019) and an MSc in Computational Engineering and Design from the University of Southampton (2017).

Authors:

George Klavaris Synopsys Northern Europe Ltd.
Andreas Gantner Synopsys GmbH
Tobias Danninger Synopsys GmbH
Wolfgang Bauer Synopsys GmbH