Simulations of the vortex in the Dellenback abrupt expansion, resembling a hydro turbine draft tube operating at part-load

This work presents an OpenFOAM case-study, based on the experimental studies of the swirling flow in the abrupt expansion by Dellenback et al.[1]. The case yields similar flow conditions as those of a helical vortex rope in a hydro turbine draft tube working at part-load. The case-study is set up similar to the ERCOFTAC Conical Diffuser and Centrifugal Pump OpenFOAM case-studies [2,3], making all the files available and the results fully reproducable using OpenSource software. The mesh generation is done using m4 scripting and the OpenFOAM built-in blockMesh mesh generator. The swirling inlet boundary condition is specified as an axi-symmetric profile. The outlet boundary condition uses the zeroGradient condition for all variables except for the pressure, which uses the fixed mean value boundary condition. The wall static pressure is probed at a number of locations during the simulations, and post-processing of the time-averaged solution is done using the OpenFOAM sample utility. Gnuplot scripts are provided for plotting the results. The computational results are compared to one of the operating conditions studied by Dellenback, and measurements for all the experimentally studied operating conditions are available in the case-study. Results from five cases are here presented, based on the kEpsilon model, the kOmegaSST model, and a filtered version of the same kOmegaSST model, named kOmegaSSTF [4,5]. Two different inlet boundary conditions are evaluated. It is shown that kEpsilon and kOmegaSST give steady solutions, while kOmegaSSTF gives a highly unsteady solution. The time-averaged solution of the kOmegaSSTF model is much more accurate than the other models. The kEpsilon and kOmegaSST models are thus unable to accurately model the effect of the large-scale unsteadiness, while kOmegaSSTF resolves those scales and models only the smaller scales. The use of two different boundary conditions shows that the boundary conditions are more important than the choice between kEpsilon and kOmegaSST, for the results just after the abrupt expansion.