Extraction of turbomachinery loss mechanisms from high-fidelity simulations

With recent advances in high-power computing, Direct Numerical Simulations (DNS), high-fidelity fluids simulations which are able to capture the full turbulence spectrum, are becoming a more common-place tool for the analysis of turbomachinery. As a result, new challenges arise in the management and post-processing of the vast amounts of data which are produced by such simulations, especially at representative Reynolds numbers. Currently, many post-processing strategies which analyse high-fidelity simulations using averaged flow fields are anchored in a “steady” description of the flow, which is usually based on Reynolds Averaged Navier-Stokes (RANS) averaging. This can mask the true unsteady behaviour of the flow, where the evolution of multiple flow structures can give rise to the mean behaviour. To make better use of DNS to inform design, this work aims to investigate unsteady flow structure in realistic turbomachinery flows using high-fidelity simulations, and develop tools which can be applied to extract and analyse flow structure obtained from high-fidelity simulations of various turbomachinery applications. In this work, methods were first developed for the extraction, tracking and analysis of turbulent structures via the kinetic energy of the unsteady part of their velocity components. Turbulent structures were then extracted from two different three-dimensional compressor test cases of varying complexity: one axial geometry and one shrouded mixed-flow impeller. Exploring two test cases of different size and complexity showed the possibility and necessity of adapting methods to individual test cases. For the two test cases, the structure analysis methods developed were able to identify structures which were significant to the energy exchange in the flow. This enabled the identification of several turbulence producing mechanisms. For the axial test case, turbulent transition was driven by separation bubbles on both pressure and suction surfaces and structures associated with the separation bubbles were identified to contribute to raising the entropy of the flow at higher rates than the non-structure flow in the same regions. For the case of the impeller, two key turbulence production mechanisms were observed - transition through leading-edge suction surface separation and a new mechanism driven by an interaction of leading-edge separation and endwall. The structure analysis was able to quantify the importance of instantaneous Reynolds stresses, turbulence production, pressure fluctuations, body forces (due to rotation), diffusion and dissipation to these processes. In general, turbulence production dominated energy transfers, however pressure and dissipation terms were also found to be significant. Overall, the methods enable a quantification of the key drivers for energy exchange within the flow even for complex geometries. This work has demonstrated a modular pipeline (structure identification, segmentation, tracking/interactions, property analysis and transport equation evaluation) for use in future work, in which blade geometry modifications can be evaluated in terms of their impact on the structure tracking graph and properties of interest.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2023

Author dc:creator

• Martin, Hanna

Advisor dc:contributor.advisor

• Wheeler, Andrew

• high-fidelity CFD
• turbomachinery
• turbulence

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/29a9f52d-d636-45a8-8fdd-e0b4e74519dc/download
• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.112589

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/374561