Numerical investigation of Francis turbine draft tubes with respect to geometry modification and turbulence treatment

Abstract

The design of draft tubes plays a significant role in the design process of modern hydro turbines as the possible pressure recovery realized within this component strongly influences the overall performance of the turbine. The draft tube flow field is characterized by swirling inflow, adverse pressure gradient due to a stream wise increasing cross sectional area and a strong curvature of the mean flow caused by the geometrical design. Depending on the operational conditions the flow field can be dominated by unsteady flow phenomena and recirculation zones. All together these complex flow features represent a challenging task when designing a draft tube for a specific turbine. But the same is valid for the use of current numerical methods when predicting draft tube flows. A description of draft tube design based on numerical steady state solutions with additional aspects from recently conducted turbine model testing is given. The influence and effect of the hydraulic section shaping on steady-state and dynamic operational behaviour in various operation conditions is investigated by state of the art measuring technique. For the laboratory investigations modern measurement technique and special modelling is applied to come to investigate the complex flow behaviour in selected operation conditions. Additionally the obtained measurement results are taken as basis for a study on numerical methods for draft tube calculations which focuses on transient methods and turbulence treatment. When solving the Reynolds-averaged Navier-Stokes equations the underlying modelling of turbulence is playing a substantial role. The assumptions made in the model development define the limitations of the turbulence model and therefore the capability of resolving physical flow phenomena. In the case unsteady phenomena dominate the flow as observed in part load conditions classical steady state solutions get more and more unreliable. For analysing part load conditions unsteady methods are preferable. A detailed comparison between model test, steady-state and transient calculations with suitable turbulence modelling approaches will be presented. 1.