Numerical investigation of the flow in a swirl generator, using OpenFOAM

Abstract: This work presents results from OpenFOAM simulations conducted on a swirl generatordesigned to give similar flow conditions to those of a Francis turbine operating at partialload. Francis turbines are one of the most commonly used water turbines. In these turbines,there is however a frequent problem occuring at part load. Due to a swirling flow inthe draft tube, a transient helical vortex rope builds up and creates severe pressure fluctuationsin the system that increase the risk for fatique. To predict and control such flowfeatures is therefore critical. A test rig was thus developed at the "Politehnica" Universityof Timisoara, Romania, to provide a detailed experimental database of such flow features.This test rig has four parts: leaning strout vanes, stay vanes, a rotating runner which isdesigned to have zero torque, and a convergent divergent draft tube.

In this work, numerical results are compared and validated against measurements realizedon the swirling flow test rig at the Polytechnica University of Timisoara in Romania.The computational mesh is created with ICEM-Hexa and the parts have been meshedseparately and then merged together, using General Grid Interfaces (GGI) to couple themnumerically. The finite volume method is used to solve both the unsteady and steady stateReynolds Averaged Navier Stokes equations and the standard k-epsilon model is used to closethe turbulence equations. Steady-state simulations is a preliminary method, which is lesstime-consuming and predicts the general behavior of the flow field. It also provides goodinitial conditions for the unsteady simulations. For the unsteady simulations, the mesh ofthe rotating part of the domain is rotating and the coupling between the stationary androtating parts is handled by a sliding GGI interface.

The simulation results shows a developing vortex rope in the draft tube which gives riseto oscillations of flow properties in the whole system. The size and shape of this vortexrope, as well as the frequency of the flow property oscillations it gives rise to, is highlydependent on the rotational speed of the free runner. The results show that a rotationalspeed of 920 rpm on the runner, corresponds best with the measurements out of the threerotational speeds 870 rpm, 890 rpm and 920 rpm. The rotational speed of 870 rpm gives apositive moment on the runner, an rpm of 890 of almost zero moment, and a speed of 920rpm gives a positive moment on the runner. This leads to the conclusion that the tourquein the test rig was not zero when measurements were carried out. The rotational speedof 920 rpm is however not the most corresponding rotational speed, for the results canprobably give a better velocity profile if the rotational speed would be increased to maybe930 rpm, and if LES or DES would be used for resolving the turbulence.