Computational Modelling of Cavity Arrays with Heat Transfer using Implicit Large Eddy Simulations

This PhD programme was sponsored by the United Kingdom Atomic Energy Authority (UKAEA). The aim of this study is to conduct advanced computational modelling of a cooling device used in the fusion process which recycles waste energy. The development of efficient, water cooled tiles, that can sustain heat loads of approximately 20 MW (in quasi-steady state conditions) is the motivation of the current work. The information presented here will contribute to thermal-mechanical analysis, to be conducted at the Joint European Torus (JET) in future years. The devices known as “Hypervapotrons” have been used successfully at JET to provide a ion dump that dissipates residual energy from the fusion process. A capability to model the flow structure and heat transfer, across a large number of geometric and material options is provided within. Differences in geometry, result in changes to the flow structure and heat transfer rates. The desire to optimise such designs relies upon the fundamental understanding of the flow field within the main section, where the geometry may be defined as a cavity array. The benchmark case of a lid driven cavity flow was used for the validation of the flow field solution. Solutions using high resolution methods in the formulation provided a good comparison with established experimental data. Therefore, validation of incompressible, Implicit Large Eddy Simulations (ILES) for a wall bounded, three dimensional, turbulent flow is provided within. The sensitivity of the high order reconstruction in conjunction with the characteristics based scheme (Drikakis & Rider, 2005), to resolve turbulent flow structure is provided here. The solution response to grid resolution and a regularised velocity profile at the upper lid surface is also detailed. The investigation provided insight and confidence in the turbulence modelling approach which is relatively recent. It was also demonstrated through the lid driven case (and later in the Hypervapotron cases) that high order reconstruction was a simulation prerequisite, based on grid resolutions used within. Additional validation was also provided against numerical and analytical solutions for the Conjugate Heat Transfer (CHT) and scalar temperature field. Where appropriate both unsteady and steady problems based on a composite, three layer medium are detailed to provide preliminary validation for the implementation of the temperature scalar and conjugate boundary conditions. Unfortunately, it was not feasible to solve the coupled problem with an explicit solver as used in this study. However, it is suggested that the initial stages of thermal boundary layer development may be observed leading to the locations of incipient boiling. Two different Reynolds numbers were considered for the Hypervapotron ”Standard” geometry, Re=12000 and Re=18000. The different flow structures show that the cavity aspect ratio of the Standard design promotes lower flow speeds at the cavity base, since two or three counter rotating vortices coexist inside the cavities depending on Reynolds number. A detailed analysis on the impact of the number of repeating units within the computational domain is also provided. Results are presented of ensemble averaged quantities based on the Reynolds decomposition. The temperature distribution present in the solid, fluid and its interface for the thermally developing case is achieved. In addition the total and decomposed heat fluxes are presented for the Hypervapotron (Standard design) which provides similar comparison with recent Reynolds Averaged Navier-Stokes (RANS) simulations.