Computational modelling of the HyperVapotron cooling technique for nuclear fusion applications

Efficient heat transfer technologies are essential for magnetically confined fusion reactors; this applies to both the current generation of experimental reactors as well as future power plants. A number of High Heat Flux devices have therefore been developed specifically for this application. One of the most promising candidates is the HyperVapotron, a water cooled device which relies on internal fins and boiling heat transfer to maximise the heat transfer capability. Over the past 30 years, numerous variations of the HyperVapotron have been built and tested at fusion research centres around the globe resulting in devices that can now sustain heat fluxes in the region of 20 – 30MW/m2 in steady state. Unfortunately, there have been few attempts to model or understand the internal heat transfer mechanisms responsible for this exceptional performance with the result that design improvements are traditionally sought experimentally which is both inefficient and costly. This thesis seeks to develop an engineering model of the HyperVapotron device using commercial Computational Fluid Dynamics software. To establish the most appropriate modelling choices, in-depth studies were performed examining the turbulence models (within the Reynolds Averaged Navier Stokes framework), near wall methods, grid resolution and boiling submodels. Validation of the models is accomplished via comparison with experimental results as well as high order Implicit Large Eddy Simulation methods. It is shown that single phase cavity flows and their related heat transfer characteristics (time-averaged) can be accurately captured if the SST k-omega turbulence model is employed using a fine nearwall grid throughout the cavity (e.g. y+ < 1 throughout). Separately, multiphase solutions with tuned wall boiling models also showed reasonable agreement with experimental data for vertical boiling tubes. As more complex multiphase HyperVapotron models were constructed, it became clear that there is an intrinsic incompatibility between the fine grids required for the single phase heat transfer predictions and the coarser grids plus wall functions required by the boiling model. Ultimately, the full 3D solution was based on the coarser grids as the fall off in accuracy in single phase heat transfer only becomes significant for HyperVapotron designs with deeper cavities. Since it is also shown here that deeper cavities are generally less efficient, these grid induced errors become less relevant if the primary objective is to find optimised performance.Comparing the CFD solutions with HyperVapotron experimental data suggests that a RANS-based, multiphase model is indeed capable of predicting performance over a wide range of geometries and boundary conditions. Whilst a definitive set of design improvements is not defined here, it is expected that the methodologies and tools developed will enable designers of future High Heat Flux devices to perform significant virtual prototyping before embarking on the more costly build and test programmes.