Multiscale High Fidelity Turbulence Modelling of Cavity Flows

Lead Research Organisation: University of Oxford
Department Name: Engineering Science

Abstract

This project falls within the EPSRC fluid dynamics and aerodynamics research area.

Resolving (instead of modelling) turbulence has been a very active and challenging research area. Capabilities of conventional methods are limited to relatively low Reynolds number conditions, and thus limit the range of applications of academic interest as well as practical relevance. Some recent efforts at the Southwell Laboratory in Oxford have indicated that the application of novel multiscale approaches like block spectral mapping to large eddy simulation (LES) may lead to much enhanced solution capabilities.
This DPhil project aims to extend the block spectral mapping methodology to the LES of buoyancy induced rotating cavity flows.
Buoyancy induced rotating cavity flows are strongly three dimensional, and inherently unstable/unsteady. They contain a full range of interacting flow structures from the smallest scales of turbulence to large pairs of contra-rotating vortices that take up most of the cavity. This lack of a clear gap in the scales, and the strongly three dimensional nature of the flow, means that less computationally expensive techniques (e.g. unsteady Reynolds-averaged Navier Stokes (URANS) modelling) do not accurately reproduce experimental measurements. Wall modelling is often carried out to reduce the computational cost of LES, however it adds a great deal of uncertainty due to its empirical nature, often leading to difficulty in accurately predicting the heat transfer. Hence, it is desirable to resolve the smallest scales possible of the near wall flow for accurate heat transfer predictions.

Practical wall resolved LES of rotating cavity flows is currently limited to rotational Reynolds numbers of around 105, two orders of magnitude smaller than in engine realistic configurations. As the computational requirements of a fully resolved LES have a superlinear dependence on the Reynolds number, it is clear that a fully resolved LES of a rotating cavity at engine realistic conditions will remain unfeasible for a while yet, even in academic settings.

In the proposed approach, fine mesh domains of an appropriate resolution for wall resolved LES are discretely embedded in a coarse mesh global domain. The fine domains are solved and used to derive unsteady stress terms; these can then be mapped around the rest of the computational domain using a block spectral mapping, this will allow the coarse global mesh to account for the action of the fine scales of turbulence whilst significantly reducing the computational cost of the simulation in comparison to a global fully resolved LES, allowing more industrially relevant flows to be investigated.

In unsteady flows, the block spectral mapping methodology has previously only been applied to cases with small scale and/or predictable unsteadiness. Neither of these are appropriate for cavity flows due to their unstable nature, so the project will involve developing a time accurate version of the methodology. It is recognized that implementing new methodologies in academic flow solvers is less useful to industry, so a secondary aim of this project is to be able to implement the new approach in ANSYS Fluent, a popular industrial flow solver.

The applications of this project mainly concern the internal air systems of turbomachines, where rotating cavities can be used to model the rotor/rotor and rotor/stator cavities. As gas turbines move from providing base load generation to flexible generation in a grid with renewables, the prediction of heat transfer in these cavities is becoming more important - if the rotor and the casing have very different thermal response times as is currently common, flexible operation of the gas turbine will lead to tip leakage flows caused by uneven thermal expansion of the rotor and casing, leading to reduced efficiency and greater emissions.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509711/1 01/10/2016 30/09/2021
2117004 Studentship EP/N509711/1 01/10/2018 30/09/2021 Tom Michael Hickling