The VT is a numerical model of a real turbine representing the behaviour of a real turbine as accurately as possible. As in a real turbine interactions with the VT have to be possible to change operational parameters and geometry. The flow conditions resulting from interactions should immediately be available and be visualized.

Realization of a VT therefore requires an online visualization environment integrating automatic mesh generators for turbine components and a simulation code able to provide results within short latencies suitable for working interactively. To realize a VT the virtual reality environment COVISE [1] and the open source computational fluid dynamics (CFD) toolbox OpenFOAM are applied. OpenFOAM has to be integrated as a module into COVISE which in turn provides already implemented modules for automatic mesh generation for the wicked gate, radial and axial runner.

A key factor to enable interactive work is the very fast execution of underlying simulations. Therefore simulations imperatively have to be accelerated up to a point suitable to allow interactive working. This requires the use of adequate computational resources along with different simplifications applied to the simulation model. Hereby it is crucial to find an appropriate trade off between execution time and the quality of results as simplifications of the simulation model always entail a loss of accuracy of the results.

To investigate possible acceleration through simplification and the associated loss of accuracy, possible simplifications are identified and investigated. As a test case a 540k nodes wicked gate mesh was used. On this mesh turbulent and laminar Navier-Stokes simulations, Euler simulations and potential flow simulations are performed. The realizable speedup is determined and the resulting quality is examined.

The fastest execution of a turbulent Navier-Stokes simulation with a runtime of 142sec wall time was performed using 16 domains. Laminar simulations show a speedup of 1.3 against turbulent simulations. Taking into account the better convergence behaviour even a speedup of 2.4 can be observed. The results qualitatively can be considered viable whereas quantitatively some distinct differences can be observed. An inviscid (Euler) simulation, besides velocity on walls due to slip wall boundary conditions, shows almost identical results to a laminar simulations with a speedup of 1.9 against turbulent simulations. Again taking into account better convergence a speedup of 2.7 against turbulent simulations is possible. The last case considered is a potential flow simulation with an execution time of 4sec for 540k nodes on 16 domains. The resulting quality, however, can be considered of limited or restricted use.

1

http://www.hlrs.de/organization/av/vis/covise

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £140 +P&P)