Session: 35-03 Compressors & ducts integration

Submission Number: 176996

The Impact of Modelling Fidelity on the Predicted Flow in Low-Pressure Compressor Transition Duct 

Modern aero gas turbines employ annular S-shaped transition ducts between compressor spools and the ability to design more integrated, shorter ducts is beneficial from a performance and weight saving perspective. However, the presence of blade wakes from the upstream compression system and the effect of strong duct curvature generates a complex aerodynamic flow which is challenging to predict using computational fluid dynamics (CFD). Understanding the interactions between the upstream compressor and the complex flow in the duct is crucial for minimising length, reducing loss and improving the quality of the flow presented to the downstream compressor. Consequently, accurate yet practical CFD predictions are essential for duct design and optimisation. However, as CFD fidelity increases there exists a trade-off between accuracy and time. This paper examines the benefits of increasing CFD fidelity versus the additional computational cost on the modelling compressor transition ducts downstream of a low-pressure compressor.

STAR-CCM+ was used to perform three levels of CFD fidelity (i) steady Reynolds Averaged Navier Stokes (RANS) predictions with different turbulence closures, and a mixing plane between the rotating and stationary components, (ii) unsteady RANS and (ii) detached eddy simulation (DES) both with a sliding mesh between the rotating and stationary components. Historically, RANS mixing plane models have been popular but their circumferential averaging limits the ability to capture the influence of upstream flow structures. Unsteady RANS with a sliding mesh offers the ability to do this in conjunction to adding some limited resolution of unsteady phenomena. DES is potentially more accurate and provides a high level of detail by directly resolving large-scale turbulent structures. However, unsteady RANS and DES require additional computation resources which potentially make these methods impractical in an industrial environment.

The predictions were compared to experimental data from a previous investigation at Loughborough University. These included time-averaged pressure and velocity data from five-hole probe area traverses and turbulence data from hot wire anemometry with the latter phase locked to the rotor to give information in the rotating frame. Four test cases were considered, combining two duct designs (a datum and an aggressive duct) and two upstream compressor configurations (a datum and a configuration with increased secondary flows). These provided varied test cases exhibiting relevant flow phenomena such evidence of limited flow separation and large compressor generated secondary flows.

The results demonstrate that RANS models employing a Reynolds Stress turbulence closure can provide adequate predictions of the flow topology, outlet guide wakes and the onset of separation in the duct. Unsteady RANS offers minimal improvement in the mean flow but is able to capture some more of the unsteady phenomena. DES does not provide a notable benefit in predicting the mean flow but offer a better insight into time-resolved secondary flow structures and the impact of upstream turbomachinery conditions. However, whilst RANS predictions can be run on a typical desktop workstation the unsteady RANS and the DES both required access to high-performance computing facilities. Overall, this study highlights the importance of selecting the appropriate CFD methodology based on the specific requirements of the application and the desired level of accuracy versus cost.

Presenting Author: Chris Entwisle Loughborough University

Presenting Author Biography: Presently studying for a PhD in "advanced compressor transition duct design" under the supervision of Professor Duncan Walker at the Rolls-Royce University Technology Centre, Loughborough University, Loughborough, United Kingdom
2023, MRes, Future Propulsion and Power, University of Cambridge, Cambridge, United Kingdom
2021, BEng, Mechanical Engineering, Loughborough University, Loughborough, United Kingdom

Authors:

Chris Entwisle Loughborough University
Dimitra Tsakmakidou Rolls-Royce
A Duncan Walker Loughborough University