Surface Roughness Impact on Francis Turbine Performances and Prediction of Efficiency Step Up

In the process of turbine modernizations, the investigation of the influences of water passage roughness on radial flow machine performance is crucial and validates the efficiency step up between reduced scale model and prototype. This study presents the specific losses per component of a Francis turbine, which are estimated by CFD simulation. Simulations are performed for different water passage surface roughness heights, which represents the equivalent sand grain roughness height. As a result, the boundary layer logarithmic velocity profile still exists for rough walls, but moves closer to the wall. Consequently, the wall friction depends not only on roughness height but also on its shape and distribution. The specific losses are determined by CFD numerical simulations for every component of the prototype, taking into account its own specific sand grain roughness height. The model efficiency step up between reduced scale model and prototype value is finally computed by the assessment of specific losses on prototype and by evaluating specific losses for a reduced scale model with smooth walls. Furthermore, surveys of rough walls of each component were performed during the geometry recovery on the prototype and comparisons are made with experimental data from the EPFL Laboratory for Hydraulic Machines reduced scale model measurements. This study underlines that if rough walls are considered, the CFD approach estimates well the local friction loss coefficient. It is clear that by considering sand grain roughness heights in CFD simulations, it forms a significant part of the global performance estimation. The availability of the efficiency field measurements provides a unique opportunity to assess the CFD method in view of a systematic approach for turbine modernization step up evaluation. Moreover, this paper states that CFD is a very promising tool for future evaluation of turbine performance transposition from the model scale to the prototype.