期刊名称：IOP Conference Series: Earth and Environmental Science
印刷版ISSN：1755-1307
电子版ISSN：1755-1315
出版年度：2019
卷号：240
期号：6
页码：1-9
DOI：10.1088/1755-1315/240/6/062011
出版社：IOP Publishing
摘要：Rotor-stator interaction is the main cause of harmonic pressure pulsations and resulting dynamic stresses in pump-turbine runners in normal operation. Considering the guide vane passing excitation frequency and the channel length of pump-turbine runners, it shows that compressible effects in the runner channels need to be taken into account for pump-turbines in the medium to high head range. This affects the choice of blade number combinations, the assessment of pressure pulsations and resulting dynamic stresses in components. The compressible effects are primarily linked to the propagation of pressure waves at a finite speed of sound in water. A lower effective wave propagation velocity is sometimes considered to reduce complex fluid-structure interaction effects into simpler pure acoustic problems. In state of the art computational fluid dynamics (CFD) software, using compressible fluid properties allows to reproduce compressibility effects and to combine them with physically accurate transient rotor-stator modelling by sliding interface techniques. The relative motion of the guide vane cascade and the runner is resulting in a periodic variation of the outer channel ends geometry and is also the source of excitation; both effects are thus included in the model which could not be achieved by modal analysis using finite elements. This paper presents the assessment of the prediction capabilities of CFD simulations using commercial software against an experiment carried out on a pump-turbine reduced scale model using air at ambient conditions as a working medium. Using air allows separating the acoustic problem from fluid-structure interaction effects that are not covered in this comparison, and that are intrinsically hard to transpose from model to prototype. The analysis of results focusses on the predicted resonance frequency; its dependency on time step size and number of inner coefficient loops is shown as well as the influence of different model extent and spatial discretization.
