Summary of the first Francis-99 workshop

The first Francis-99 workshop was organized on 15 and 16 December 2014. Total 14 research papers were presented in the workshop. More than 50% participants were from the different hydropower industries. Over 50 researchers from different countries America, Austria, Canada, China, Czech republic, Germany, Italy, Korea, Russia, Sweden, Slovenia, Switzerland, and Spain were participated in the workshop. Extensive numerical studies on the model Francis turbine was performed by the researchers. The steady-state measurements were conducted on a model Francis turbine. Three operating points, part load, best efficiency point, and high load, were investigated. The complete geometry, meshing, and experimental data concerning the hydraulic efficiency, pressure, and velocity were provided to the academic and industrial research groups. Various researchers have conducted extensive numerical studies on the high-head Francis turbine, and the obtained results were presented during the workshop. The first workshop attempted to determine the state of the art in the simulation of high-head Francis turbines under steady operations, namely, at part load, best efficiency point and high load. The motivation resided in the continuous development of more powerful computers, thereby facilitating the use of more advance turbulence models and accurate numerical schemes. Efficiency, pressure, and velocity measurements were conducted on the Francis-99 test case at the Waterpower laboratory, Norwegian University of Science and Technology (NTNU), Norway. Many questions were raised during the workshop concerning the challenges related to accurate simulations of high-head Francis turbines. One of the main challenge was the optimization of the numerical models without compromising the numerical accuracy. It was also highlighted that the numerical model many times mislead and results in incorrect prediction of the flow field.

Outcome of the first Francis-99 workshop

Main objective of the workshop was to evaluate the numerical techniques applied to investigate the hydraulic turbines and prove open platform to the researchers for conducting numerical studies in high head turbines. For that geometry and mesh were proposed to the participants. Many participants also created their own meshes to improve the quality and investigate in detail specific issues such as near-wall modeling. The complexity of the turbine model geometry and high Reynolds number imply compromises by the mesh. A quality assessment of the mesh is necessary for a proper assessment of the simulations. A mesh sensitivity analysis is not always possible due to computational limitations. Systematic information about the mesh quality parameters, minimum angle, volume change, aspect ratio, etc. is a first step but not sufficient per se. In addition to the mesh, the numerical schemes and type of code should also be stipulated. The same mesh will not provide the same results for different codes; it is also a function of the code used, and each code operates differently.

The hydraulic efficiency obtained by the different participants is very close to the experimental ones. The overestimation at part load is attributed to the omission of the seal leakage losses. Often, numerical results are misleading because the torque and head are over-predicted with an inlet flow boundary condition; see Lenarcic et al. Because the torque generated by the turbine and the available head are related to each other, the hydraulic efficiency is fairly well predicted. Using the head as the inlet boundary condition provides a higher flow rate, decreasing the hydraulic efficiency. The reason is an under-prediction of the viscous losses. The use of a wall function assuming equilibrium between the production and dissipation of turbulence is widely used in the simulation of hydraulic turbines. The boundary layer of hydraulic turbines is never fully developed because of the continuously changing geometry and rapid change in pressure gradients. Resolving the boundary layer up to y+ = 1 is not feasible, even with RANS models. There is a need to develop wall functions that enable the estimation of viscous losses under boundary development if accurate simulations are to be developed. Improved simulations and results enable reliable estimation of the blade loading.

In a high-head Francis turbine, pressure amplitudes generated by the rotor-stator interactions are a major concern. A recently developed flow modeling technique was applied to investigate the rotor-stator interaction in the model Francis turbines for the first time. This technique obtained good agreement with the measured values. It has provided a compromise between numerical accuracy and the required computational power for simulating the complete turbine.

At the Francis-99 workshop, numerical simulations were conducted using three modeling approaches: (1) modeling of a complete turbine, (2) modeling of the components, and (3) passage modeling. The simulation of a complete turbine is more expensive than the other two approaches. No significant difference was found between the results obtained with the complete turbine and component modeling approaches. It was considered that the component modeling approach would provide the optimum solution when accurate boundary conditions are prescribed. Further, one can select a passage modeling approach and create a fine mesh near the boundaries to reduce the necessary computational power and time. This approach provides good results but does not consider the influence of the neighbouring passages. However, using recently developed techniques of passage modeling can provide reliable results, including dynamic pressure loading generated by rotor-stator interaction.

Input to the second Francis-99 workshop

To minimize the error due to the use of different mesh types and densities, it was noted that a mesh with a uniform density would be provided. The simulations would be conducted using the provided mesh only and validated with the provided experimental data. The mesh would be provided with different y+ values to accommodate different turbulence models. At the second Francis-99 workshop, both steady-state and transient operating conditions will be investigated. Numerical models will be validated with the steady-state experimental results, and the same model will be used for transient conditions. Load variations and start-stop conditions will be investigated under transient operations. The primary focus will be numerical modeling and investigation of the hydraulic turbine during transient conditions.