Table of contents

Francis-99 is a set of upcoming workshops jointly organized by the Norwegian University of Science and Technology (NTNU), Norway and Luleå University of Technology (LTU), Sweden in the same spirit as the previous Turbine-99 workshops. The Francis-99 workshops aim during the coming years to determine the state of the art of high head Francis turbine simulations (flow and structure) under steady and transient operating conditions as well as promote their development and knowledge dissemination openly. Three workshops are initially planned:

- Workshop 1: steady operation of Francis turbines (December 2014)

- Workshop 2: transient operation of Francis turbines (December 2016)

- Workshop 3: FSI of Francis turbines (December 2018)

A high head Francis turbine model, named the Tokke model, has been designed and experimentally investigated at the Water Power Laboratory, NTNU. The complete geometry of the model and mesh are now freely available on the site www.francis‑99.org together with a large set of experimental pressure and velocity measurements. The organisers expect this geometry to become with time a reference test case to the hydraulic community for research and development on high head Francis turbines and the workshops a meeting place to discuss developments, potentials, issues... on a common and open test case.

The present proceeding contains the papers presented at the first workshop at NTNU the 15th and 16th of December 2014. 50 participants were present at the workshop and a total of 14 papers were presented. A large variety of codes and models were used highlighting different issues in the simulation of high Francis turbines.

The editors:

Prof. Michel J. Cervantes (LTU, NTNU)

Dr. Chirag Trivedi (NTNU)

Prof O.G. Dahlhaug (NTNU)

Prof. T. Nielsen (NTNU)

All papers published in this volume of Journal of Physics: Conference Series have been peer reviewed through processes administered by the proceedings Editors. Reviews were conducted by expert referees to the professional and scientific standards expected of a proceedings journal published by IOP Publishing.

CFD simulations focusing on capturing dynamic fluctuations of the flow for three operating points were performed for a scale model of a high head Francis turbine. A mesh sensitivity study showed an influence of the near wall resolution, consequently a low Reynolds mesh with a SST turbulence model was used. Rotor/stator fluctuations are well reproduced with the full turbine simulation at all operating points. Velocity contours and average velocity profiles from LDV measurements in the draft tube confirm that the flow physics is generally well reproduced. Simplified approaches such as profile transform and Fourier transform underestimated the measured fluctuations. As full turbine simulations were time-consuming, a simulation with only the draft tube was performed at part load to predict the fluctuations in the draft tube cone. The SAS-SST turbulence model was able to capture the vortex rope behavior.

The simulations of high head Francis turbine model (Tokke) are performed for three operating conditions - Part Load, Best Efficiency Point (BEP) and Full Load using software Ansys Fluent R15 and alternatively OpenFOAM 2.2.2. For both solvers the simulations employ Realizable k-e turbulence model. The unsteady pressure pulsations of pressure signal from two monitoring points situated in the draft tube cone and one behind the guide vanes are evaluated for all three operating conditions in order to compare frequencies and amplitudes with the experimental results. The computed velocity fields are compared with the experimental ones using LDA measurements in two locations situated in the draft tube cone. The proper orthogonal decomposition (POD) is applied on a longitudinal slice through the draft tube cone. The unsteady static pressure fields are decomposed and a spatio-temporal behavior of modes is correlated with amplitude-frequency results obtained from the pressure signal in monitoring points. The main application of POD is to describe which modes are related to an interaction between rotor (turbine runner) and stator (spiral casing and guide vanes) and cause dynamic flow behavior in the draft tube. The numerically computed efficiency is correlated with the experimental one in order to verify the simulation accuracy.

For the Francis-99 project initiated by the Norwegian University of Science and Technology (NTNU, Norway) and the Luleå University of Technology (LTU, Sweden) numerical flow simulation has been performed and the results compared to experimentally obtained data. The full machine including spiral casing, stay vanes, guide vanes, runner and draft tube was simulated transient for three operating points defined by the Francis-99 organisers. Two sets of results were created with differing time steps. Additionally, a reduced domain was simulated in a stationary manner to create a complete cut along constant prototype head and constant prototype discharge. The efficiency values and shape of the curves have been investigated and compared to the experimental data. Special attention has been given to rotor stator interaction (RSI). Signals from several probes and their counterpart in the simulation have been processed to evaluate the pressure fluctuations occurring due to the RSI. The direct comparison of the hydraulic efficiency obtained by the full machine simulation compared to the experimental data showed no improvement when using a 1° time step compared to a coarser 2° time step. At the BEP the 2° time step even showed a slightly better result with an absolute deviation 1.08% compared with 1.24% for the 1° time step. At the other two operating points the simulation results were practically identical but fell short of predicting the measured values. The RSI evaluation was done using the results of the 2° time step simulation, which proved to be an adequate setting to reproduce pressure signals with peaks at the correct frequencies. The simulation results showed the highest amplitudes in the vaneless space at the BEP operating point at a location different from the probe measurements available. This implies that not only the radial distance, but the shape of the vaneless space influences the RSI.

The paper presents numerical simulation of flow in Francis-99 water turbine under three operation modes: part load, best efficiency point and high load. Calculations were performed by means of Reynolds stress model and detached eddy simulation based on k-omega SST model. The paper focuses on the flows in the draft tube. The calculated mean velocity components in the draft tube are in close agreement with experimental results. Calculated r.m.s velocity components agree with experimental pulsations qualitatively.

Numerical investigations of hydraulic turbo machines under steady-state conditions are state of the art in current product development processes. Nevertheless allow increasing computational resources refined discretization methods, more sophisticated turbulence models and therefore better predictions of results as well as the quantification of existing uncertainties. Single stage investigations are done using in-house tools for meshing and set-up procedure. Beside different model domains and a mesh study to reduce mesh dependencies, the variation of several eddy viscosity and Reynolds stress turbulence models are investigated. All obtained results are compared with available model test data. In addition to global values, measured magnitudes in the vaneless space, at runner blade and draft tube positions in term of pressure and velocity are considered. From there it is possible to estimate the influence and relevance of various model domains depending on different operating points and numerical variations. Good agreement can be found for pressure and velocity measurements with all model configurations and, except the BSL-RSM model, all turbulence models. At part load, deviations in hydraulic efficiency are at a large magnitude, whereas at best efficiency and high load operating point efficiencies are close to the measurement. A consideration of the runner side gap geometry as well as a refined mesh is able to improve the results either in relation to hydraulic efficiency or velocity distribution with the drawbacks of less stable numerics and increasing computational time.

The paper presents numerical simulations of flow in a model of a high head Francis turbine and comparison of results to the measurements. Numerical simulations were done by two CFD (Computational Fluid Dynamics) codes, Ansys CFX and OpenFOAM. Steady-state simulations were performed by k- and SST model, while for transient simulations the SAS SST ZLES model was used. With proper grid refinement in distributor and runner and with taking into account losses in labyrinth seals very accurate prediction of torque on the shaft, head and efficiency was obtained. Calculated axial and circumferential velocity components on two planes in the draft tube matched well with experimental results.

CFD is a very powerful tool, which significantly reduces the need for model testing of Francis turbines. Numerical models are constructed so as to cover all the essential geometric and physical properties of the physical models. However, for the sake of simplicity, some geometric details are often neglected. In case of Francis turbines, labyrinths are usually not included in the numerical model and consequently the disc friction losses and labyrinth losses are ignored. This may lead to inaccurate prediction of turbine efficiency. In our study we have investigated the importance of considering all the geometrical details of labyrinth for the accurate prediction of efficiency of high head Francis turbine. The research was performed by using commercial software Numeca FINE^TM/Turbo.

In this work the incompressible turbulent flow in a high head Francis turbine under steady operating conditions is investigated using the open source CFD software package FOAM-extend- 3.1. By varying computational domains (cyclic model, full model), coupling methods between stationary and rotating frames (mixing-plane, frozen-rotor) and turbulence models (kω-SST, κε), numerical flow simulations are performed at the best efficiency point as well as at operating points in part load and high load. The discretization is adjusted according the y⁺-criterion with y⁺_mean > 30. A grid independence study quantifies the discretization error and the corresponding computational costs for the appropriate simulations, reaching a GCI < 1% for the chosen grid. Specific quantities such as efficiency, head, runner shaft torque as well as static pressure and velocity components are computed and compared with experimental data and commercial code. Focusing on the computed results of integral quantities and static pressures, the highest level of accuracy is obtained using FOAM in combination with the full model discretization, the mixing-plane coupling method and the κω-SST turbulence model. The corresponding relative deviations regarding the efficiency reach values of Δη_rel ~ 7% at part load, Δη_rel ~ 0.5% at best efficiency point and Δη_rel ~ 5.6% at high load. The computed static pressures deviate from the measurements by a maximum of Δp_rel = 9.3% at part load, Δp_rel = 4.3% at best efficiency point and Δp_rel = 6.7% at high load. Commercial code in turn yields slightly better predictions for the velocity components in the draft tube cone, reaching a good accordance with the measurements at part load. Although FOAM also shows an adequate correspondence to the experimental data at part load, local effects near the runner hub are captured less accurate at best efficiency point and high load. Nevertheless, FOAM is a reasonable alternative to commercial code that makes it possible to predict integral quantities and local parameters under steady operating conditions adequately.

In the present paper numerical investigations of a complete high head Francis turbine comprehensive of a spiral casing, stay and guide vanes and draft tube have been performed at three operating conditions, namely at part load (PL), best efficiency point (BEP), and high load (HL). The main target of the investigations is to assess the prediction accuracy of a reduced domain of the complete turbine using a novel mixing-plane formulation.

The computational domain is simplified simulating one single passage of the runner, thus assuming rotational periodicity and steady state conditions. The results were compared with experimental data published by the workshop organization.

All CFD simulations were performed at model scale with an in-house adapted, 3D, unstructured, object-oriented finite volume code based on the OpenFOAM-V2.2 framework and designed to solve steady-state incompressible RANS-Equations. The pressure-based solver uses a SIMPLE-C like algorithm and is capable of handling multiple references of frame (MRF). The influence of the turbulence has been considered applying the shear-stress transport model (SST). Full second order upwind scheme for advection discretization has been used for all computations.

In the present paper, fully 360 degrees transient and steady-state simulations of a Francis turbine were performed at three operating conditions, namely at part load (PL), best efficiency point (BEP), and high load (HL), using different numerical approaches for the pressure-velocity coupling. The simulation domain includes the spiral casing with stay and guide vanes, the runner and the draft tube. The main target of the investigations is the numerical prediction of the overall performance of the high head Francis turbine model as well as local and integral quantities of the complete machine in different operating conditions. All results were compared with experimental data published by the workshop organization. All CFD simulations were performed at model scale with a new in-house, 3D, unstructured, object-oriented finite volume code within the framework of the open source OpenFOAM library. The novel fully coupled pressure-based solver is designed to solve the incompressible RANS- Equations and is capable of handling multiple references of frame (MRF). The obtained results show that the overall performance is well captured by the simulations. Regarding the local flow distributions within the inlet section of the draft-tube, the axial velocity is better estimated than the circumferential component.

This work investigates the flow in the scale model of the high-head Tokke Francis turbine at part load, best efficiency point and high load, as a contribution to the first Francis-99 workshop. The work is based on the FOAM-extend CFD software, which is a recent fork of the OpenFOAM CFD software that contains new features for simulations in rotating machinery. Steady-state mixing plane RANS simulations are conducted, with an inlet before the guide vanes and an outlet after the draft tube. Different variants of the k- and k-ω turbulence models are used and a linear explicit algebraic Reynolds stress model is implemented. Sliding grid URANS simulations, using a general grid interface coupling, are performed including the entire turbine geometry, from the inlet to the spiral casing to the outlet of the draft tube. For the unsteady simulations, the k-ω SSTF model is implemented and used in addition to the standard k- model. Both the steady and unsteady simulations give good predictions of the pressure distribution in the turbine compared to the experimental results. The velocity profiles at the runner outlet are well predicted at off-design conditions. A strong swirl is however obtained at best efficiency point, which is not observed in the experiments. While the steady-state simulations strongly overestimate the efficiency, the unsteady simulations give good predictions at best efficiency point (error of 1.16%) with larger errors at part load (10.67%) and high load (2.72%). Through the use of Fourier decomposition, the pressure fluctuations in the turbine are analysed, and the main rotor-stator interaction frequencies are predicted correctly at all operating conditions.

The three-dimensional numerical investigation for turbine-99 at the best efficiency operation point, part load operation point and full load operation point was conducted by using the different turbulence models. By comparing the results of numerical simulation and experimental results, it was found that: there is a certain deviation between the numerical simulation results obtained by different turbulence models and experimental values, and the deviation increase with the reduction of output. Compared to other turbulence model, the result obtained by the standard k-e turbulent model has a relatively small difference with the experimental results. The main causes for the big difference between the numerical simulation and model test include two aspects: (1) the mesh generation and boundary conditions setting lead to differences between the research object and the actual model, (2) it is difficult to accurately simulate the unstable flow such as impact, flow separation and vortex in the turbine. Therefore, in the future actual flow pattern simulation, besides the reasonable choice of turbulence model, based on the actual flow characteristics, the boundary conditions and the simulation results will be amended to reduce the deviation between the numerical simulation and experimental results as much as possible.

Steady state and non-linear harmonic (NLH) flow simulations were performed within the framework of the Francis-99 project in order to assess the capacity of the NLH method to capture the main pressure fluctuations associated with the rotor-stator interactions between the distributor and the runner of the turbine. This paper focusses on the methodology developed to obtain harmonic solutions and presents preliminary results from the simulations using the flow solver NUMECA FineTURBO on intermediate grid level meshes. Comparisons of the first simulations to experimental data reveal good agreement concerning the predicted pressure amplitudes notably at high load operating condition.

Two-dimensional particle image velocimetry (PIV) measurements in the draft tube cone of the Francis-99 model have been performed to complete the actual experimental data set with radial velocity data. The velocity profiles obtained presented some variation, which reason has not yet been identified. The presented results are therefore presented as preliminary until the reason is assessed. The axial velocity profiles corroborate well with the ones previously measured with laser Doppler velocimetry (LDV) for all operating points investigated. The radial velocity measured is small in magnitude for all operating points compared to the axial velocity. A gyroscopic effect induced by the swirl leaving the runner and the draft tube bend seems to induce an asymmetry in the draft tube cone.