Experiences with OpenFOAM for water turbine applications
Paper i proceeding, 2007
OpenFOAM has been used successfully at Chalmers for water turbine applications since the beginning of 2005. OpenFOAM has been validated for the flow in a Kaplan water turbine runner and draft tube, and for the swirling unsteady flow in a combustor. The flow in the combustor resembles the flow in a water turbine draft tube (which is a diffuser), with its adverse pressure gradient and unsteady flow features. The results compare well with the experimental results. For the flow in the draft tube the OpenFOAM results are almost identical to those of the CFX-5 CFD code. For the flow in the combustor the OpenFOAM results show the same trend as the results using the Fluent CFD code. OpenFOAM gives similar results as the CALC-PMB in-house CFD code that was developed specifically for water turbine applications (however, no comparisons are shown here).
Many interesting flow features in water turbines are unsteady. There is an interaction between steady and rotating parts of the machine, which should be properly resolved. There are also flow instabilities like the break-up of the vortex after the runner, yielding a highly unsteady flow (here referred to as the vortex rope). A CFD code used for water turbine applications should thus be able to include rotor-stator interaction and allow unsteady flow features to appear without numerical damping. OpenFOAM is not yet fully developed for full rotor-stator interaction in water turbines. In the present work a rotating inlet boundary condition is shown, which makes it possible to include guide vane wakes or non-axisymmetry from the spiral casing from a previous simulation of the distributor. The present work evaluates a newly developed filtering technique for the $k-\omega$ SST turbulence model. It is shown that the filter is necessary to get unsteady and accurate time averaged results of the flow in the combustor.
The Reynolds numbers in water turbines are high, the geometries are complicated, and the computational grids are usually coarse and skew. It is thus necessary to use models and methods that can give accurate results also under non-ideal conditions.