Towards Full Predictions of the Unsteady Incompressible Flow in Rotating Machines, Using OpenFOAM

Doctoral thesis, 2012

The main objective of the present work is to validate methodologies for accurate numerical predictions of the incompressible flow of water in the U9 Kaplan turbine model. The term “prediction” implies that the use of detailed experimental data for boundary conditions should be avoided, and that all important features of the studied case should be included. That includes specifying boundary conditions at locations where the flow conditions can be easily estimated, and where reasonable variations in those estimations do not significantly affect the flow prediction. As an example, the U9 Kaplan turbine model has a sharply bent inlet pipe, and it is here argued that the secondary flow from that should be taken into account by including the pipe in the simulation. In the case of rotating machines, such as the U9 Kaplan turbine model, the interaction between rotating and stationary components (rotor-stator interaction) is a feature that must be included in the simulations. Three highly relevant well-documented cases have been used in the present work, the ERCOFTAC Centrifugal Pump, the Timisoara Swirl Generator and the U9 Kaplan turbine model. All three cases include rotor-stator interaction. The latter, being the main goal of the studies, has just recently been studied experimentally and is a computationally demanding case. Thus, the former two cases were used while validating the new implementations and evaluating the numerical settings, until the results were reliable and efficient. Two rotor-stator interaction methods were investigated, the steady-state frozen-rotor approach, and the unsteady sliding grid approach, and the results from four turbulence models were compared. The results show that both approaches can be used to couple the rotating and stationary parts of the domain. However, the frozen rotor yields an unphysical advection of the runner wakes, and such results should only be used for a first estimation or as initial conditions for full unsteady sliding grid simulations. The predictions compare very well with the experimental results, and the main differences can be explained by the geometrical simplifications that were made. The four turbulence models behave similarly, with a minor preference for different models in the different cases. The present work is done using the OpenFOAM OpenSource CFD toolbox. The code is chosen to facilitate an OpenSource distribution of the developments, to be shared in the scientific community, and to be directly useful in industry. It was not possible to achieve the results presented here with OpenFOAM before the start of the present work, which has significantly contributed to the validation of, and trust in, the new implementations.