Deliverable D.6.5: Final Public Report€¦ · 3 Water turbines Water (hydro) turbines transform...
Transcript of Deliverable D.6.5: Final Public Report€¦ · 3 Water turbines Water (hydro) turbines transform...
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 1 of 39
This project has received funding from the European
Union’s Seventh Framework Programme for research,
technological development and demonstration under
grant agreement no 612279
Accurate Simulations in Hydro-Machinery
and Marine Propellers (ACCUSIM)
Deliverable D.6.5: Final Public Report
Document identifier: ACCUSIM_Final_Public_Report_D.6.5_v01.doc Preparation date: 30/01/2018
Version: 01 State: Final
Distribution: Public Issued : Dragica Jošt
Verified : Aljaž Škerlavaj Approved : Enrico Nobile
Call Identifier: FP7-PEOPLE-2013-IAPP Call Theme: Industry-Academia Partnerships and Pathways (IAPP)
Contract start date: 01/02/2014 Duration: 48 months
Project coordinator: Kolektor Turboinštitut Participating partner: University of Trieste
Project website: www.accusim.eu
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 2 of 39
Document History Version Date Amended by Changes
01 30/01/2018 Dragica Jošt Document creation
Contributors
Both partners contributed and agreed with the content of this deliverable, which is the property of
the ACCUSIM consortium.
This document may be copied and reproduced. It may not be modified neither whole nor partially for
any purpose without written permission from the ACCUSIM project coordinator with acceptance of
the Project Consortium.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 3 of 39
Executive Summary
The report consists of two parts. Part 1 is a general purpose report, meant for general public interested in renewable energy sources and in particular in water turbines and numerical (computer) simulations. It comprises some basics about numerical simulations of flow in different types of water turbines. Part 2 is meant for a more professional public. It comprises a short overview of results achieved
during the ACCUSIM project. In this part the improvement of efficiency prediction with advanced
(scale resolving) turbulence models, the prediction of different forms of cavitation and finally design
optimization are presented in more detail. A comparison of results obtained with commercial code
ANSYS CFX and the open-source code OpenFOAM is also presented. At the end the references are
provided, where the reader can find more information about numerical simulations and accuracy of
results.
Note
In the email correspondence (13/03/2017) with Project Officer Olivier Delaunoy it was accepted that this deliverable can be created in month 48 instead of in month 40 of the project.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 4 of 39
Table of Contents
Part 1: Numerical Simulations of Flow in Different Types of Water Turbines
1 Introduction ..................................................................................................................................... 7
2 Numerical flow simulations ............................................................................................................. 8
3 Water turbines .............................................................................................................................. 10
3.1 Francis turbines ..................................................................................................................... 10
3.2 Axial Turbines ........................................................................................................................ 13
3.3 Pelton turbines ...................................................................................................................... 16
Part 2: Some of the Results Achieved during the ACCUSIM Project
1 Introduction ................................................................................................................................... 21
2 Scale resolving turbulence modelling ........................................................................................... 22
3 Cavitation ...................................................................................................................................... 26
3.1 Calibration of mass transfer models for the numerical prediction of sheet cavitation around
a hydrofoil ......................................................................................................................................... 26
3.2 Cavitation in water turbines .................................................................................................. 26
3.2.1. Cavitation in Francis turbines ........................................................................................ 27
3.2.2. Cavitation in Kaplan turbines ........................................................................................ 28
3.2.3. Cavitation in a prototype of a bulb turbine ................................................................... 29
3.2.4. Cavitation in Pelton turbines ......................................................................................... 29
3.3 Sheet and cloud cavitation prediction in double suction centrifugal pump ......................... 30
3.4 Numerical predictions of the cavitating flow around model scale propellers working in
uniform and non-uniform inflow ...................................................................................................... 32
4 Comparison of results obtained with OpenFOAM and ANSYS-CFX .............................................. 33
5 Optimization .................................................................................................................................. 34
5.1 Optimization of a double-sided centrifugal pump ................................................................ 34
5.2 Multi-objective optimization of the Francis turbine runner cone ........................................ 36
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 5 of 39
Abbreviations
CFD Computational Fluid Dynamics
mF modeFRONTIER© (optimization software)
KTI Kolektor Turboinštitut
UniTS University of Trieste
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 6 of 39
Part 1
Numerical Simulations of Flow in Different Types of Water Turbines
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 7 of 39
1 Introduction
A growing world population, striving for a better quality of life, is demanding access to reliable, low-
cost electricity supply. Global demand for electricity is increasing from 3% to 6% annually on average
in different parts of the world. Around 20% of the world’s electricity is produced by hydropower. In
Europe and around the world, the present development of intermittent renewable energy sources,
such as wind and solar energies, nowadays increases the need for energy storage. Pump Storage
Power plants are the cheapest solution for large scale energy storage, and have, at present, a global
efficiency much higher than other solutions for energy storage. Therefore development of hydro
machines with high efficiency and good cavitation and dynamic characteristics is of paramount
significance.
CFD (Computational Fluid Dynamics) has been a useful tool in design of all turbine parts for thirty
years. Numerical analysis of flow in water turbines and pumps is not only important because it
enables to reduce expensive and time consuming model tests, but also because it gives the insight
into the flow details in all turbine parts. On the basis of numerical results it is easier to find the
reasons for low efficiency or for cavitation and to improve the hydraulic shapes of all turbine parts.
Furthermore, for small projects the model tests are too expensive and CFD analysis is the only way to
foresee whether the required efficiency and cavitation characteristics on the prototype will be
obtained.
In the last fifteen years a large progress in numerical simulation of flow in water turbines has been
achieved. Nowadays the possibility of simultaneous calculation of flow in rotating and non-rotating
turbine parts and development of more powerful computing platforms, enable coupled analysis of
the whole turbine in a reasonable time. By performing unsteady simulations, using more advanced
turbulent models, the accuracy of numerical results became similar to the experimental ones.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 8 of 39
2 Numerical flow simulations
Fluid flow is defined by flow equations (Navier–Stokes equations and continuity equation). Numerical
flow simulation means that we discretize, approximating a continuous mathematical problem with
an algebraic one, the region of interest where the equations will be solved. In the process of domain
discretization the region of interest is divided into small elements (cells). At the nodes of the
elements the values of velocity components and pressure are calculated during the numerical
simulations.
The main steps of each flow simulation are:
Defining a computational domain
Mesh generation
Setting up boundary conditions
Defining flow and solver parameters
Setting initial conditions
Solution of the problem
Post-processing.
The definition of the computational domain is very important for the accuracy of results and for
computational time (CPU). With reduction of computational domain to the region of interest, we can
significantly reduce CPU but in some cases we lose some information which might influence the flow
in the region of interest.
Quality and density of computational mesh significantly influence the results. The meshes should be
as orthogonal as possible. Density and mesh refinement near the walls depend on flow conditions
and turbulence models used in the simulations. Computational domain can be reduced by exploiting
symmetry or periodic boundary conditions.
Fluid properties are fluid density and dynamic viscosity. Solver parameters are number of iterations,
convergence criterion, etc. Discretization schemes have to be defined as well.
Good initial conditions can significantly improve the convergence and reduce CPU. If we numerically
analyze several similar cases, the results of a previous one can be used as initial condition for the
next case.
When all the conditions and parameters are defined, the solver starts to solve discretized equations.
The computing time of numerical simulations depends on many parameters, such as number of
elements in the mesh, steady or transient simulation, turbulence model, performances of the chosen
computing platform etc.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 9 of 39
In the post-processing phase, when flow simulation is completed, the flow can be visualized with
streamlines, velocity vectors, pressure contours etc. From velocity and pressure some additional
information can be calculated, such as flow energy losses in the flow, forces on surfaces, torque
acting on some parts, etc. For water turbines and pumps the most important results are flow energy
losses and torque on the shaft from which the efficiency of the machine can be predicted.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 10 of 39
3 Water turbines
Water (hydro) turbines transform energy of water into electrical energy via rotation of the turbine
runner, which in turn runs the electric generator through rotation of the turbine shaft. Such turbines
can be classified as high head (Pelton), medium head (Francis) or low head (propeller, Kaplan and
bulb) turbines. The specific speed nq , defined by the expression below, is a measure of the operating
condition (head and flow rate) of a turbine:
4/3
H
Qnn
q , (1)
where Q is discharge (m3/s), H is water head (m) and n is rotor speed (rpm).
Fig. 1 visually represents the choice of turbine type according to available head and flow rate.
Figure 1: Types of water turbines, according to head H and flow rate Q.
3.1 Francis turbines
Francis turbines are the most widely used water turbines. More than 60% of installed hydropower is
produced by Francis turbines. They operate at a wide range of head, from 6 to 600 m. Specific speed
nq is between 12 and 120.
Typical Francis turbine consists of a spiral casing with stay vanes, a guide vane cascade, a runner and
a draft tube. The amount of water which flows through the turbine is regulated by opening or closing
the guide vanes (see Fig. 2). Shape and number of runner blades depend on design conditions.
Turbines for higher head values have more blades than those which operate at moderate head. In
Q [-]
H [m]
Pelton
Francis
Kaplan Bulb
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 11 of 39
Fig. 3 a Francis turbine for extremely high head is presented. Its runner consists of 15 full length
blades and 15 splitters.
Figure 2: Definition of guide vane opening. Angles of guide vanes are adjusted to the flow conditions.
Figure 3: Francis turbine (Workshop Francis 99). Spiral casing with 14 stay vanes – blue, wicket gate cascade with 28 vanes – orange, runner – red, draft tube – green.
Results of numerical simulation are: velocity components, pressure and turbulence quantities in a
complete computational domain. In post processing we can visualize the flow with vectors or
streamlines and check pressure distribution on guide vanes and runner blades (see Fig. 4). From the
results of numerical simulation torque on the shaft, losses in all turbine parts and turbine efficiency
can be calculated.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 12 of 39
Figure 4: Results of numerical flow analysis at the BEP (Best Efficiency Point): pressure contours on guide vanes and runner blades and streamlines in distributor (spiral casing, stay and guide vane cascades) and in the draft
tube.
From measured or calculated efficiency values at several guide vane openings and several values of
head the efficiency hill chart diagram is obtained in such a way, that iso-lines with constant efficiency
values are drawn. From the hill chart diagram efficiency for all operating conditions can be seen. In
Fig. 5 an efficiency diagram for a middle head Francis turbine is presented. Iso-lines of efficiency are
in black while blue lines present constant guide vane openings. Efficiency values are divided by the
efficiency value at the Best Efficiency Point (BEP).
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 13 of 39
H/HBEP
Q/QBEP Figure 5: Efficiency hill chart diagram. Efficiency values are divided by the efficiency value at BEP. Isolines of
constant efficieny curves –black, iso-lines of constant guide vane opening – blue.
3.2 Axial Turbines
Low-head turbines are axial-type (with axial entry and axial discharge) reaction turbines, with 3 to 7
runner blades. Kaplan turbines are used for head H up to 80 m, whereas bulb turbines are used for
head up to 15 m. Specific speed nq of Kaplan turbines is between 95 and 250. For bulb turbines nq is
higher than 170. Different types of axial turbines are presented in Fig. 6.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 14 of 39
Kaplan turbine
Saxo turbine
Bulb turbine
Figure 6: Different types of axial, double regulated turbines.
Kaplan and bulb turbines are double-regulated turbines, which means that the inclination of guide
vanes, as well as of runner blades can be adjusted. The angle of guide vanes and runner blades for
small and large values of flow rate is presented in Fig. 7. A typical mesh for a Kaplan turbine is
illustrated in Fig. 8, and its corresponding flow path in Fig. 9.
a) b)
Figure 7: Adjustment of guide vane opening and angle of runner blades to flow conditions: (a) Small guide vane opening and small angle of runner blades (for a small flow rate value); (b) Wide guide vane opening and large angle of runner blades (for a large flow rate value)
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 15 of 39
Figure 8: Typical grid for a Kaplan turbine. Unstructured mesh in the semi-spiral casing with stay vanes and
structured mesh in the other turbine parts. Total number of nodes is around 8.3 million, total number of
elements is around 11.5 million.
Figure 9: Flow in a Kaplan turbine
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 16 of 39
In Fig. 10 an efficiency diagram at constant head for an axial turbine is presented. For each angle of
runner blades a partial efficiency curve (in blue) has to be obtained with measurements or
numerically. The envelope of partial efficiency curves presents the efficiency of the turbine at
constant head. From efficiency curves for several values of head a hill chart diagram can be obtained.
With double regulation, high efficiency for a wide range of operating conditions can be achieved.
Q
Figure 10: Efficiency diagram for a double regulated axial turbine at constant head – model test.
3.3 Pelton turbines
Pelton turbines operate at high head and small flow rate values. For Pelton turbines nq is less than
20.
Two-phase flow (water, air) in Pelton turbines is turbulent and unsteady. While reasonable results for
Francis and Kaplan turbines can be obtained by steady state analysis, runner flow simulations for
Pelton turbines have to be time dependent (transient). Besides, we have two fluids, water and air.
The shape and thickness of water jets and evacuating water sheets are not known in advance, and
they have to be calculated during the simulation. Numerical analysis of flow in a Pelton turbine is
therefore much more complex and time consuming.
Optimization of distributors with injectors is extremely important for the efficiency of Pelton
turbines. Discharge should be distributed equally between all injectors. Losses in the distributor have
to be as small as possible, but the quality of the jets is even more important. Jets should be compact
and velocity within the jet should be as uniform as possible. Secondary velocities caused by bends in
the distributor are most undesirable as they cause jet dispersion and deviation.
Numerical simulation is usually divided into two parts. At first, steady state simulation in distributor
is performed. Calculated thickness and velocity of the jets are then used as inlet conditions for
transient simulation of flow in the runner.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 17 of 39
Figure 11: Model of a two-jet Pelton turbine. Test rig – left, numerical model – right. In numerical simulations
the casing is not included.
For Pelton turbines efficiency prediction is a bit less accurate than for the Francis and axial turbines,
but the results are nevertheless very helpful. Numerical results help to improve the shape of the
distributor with injectors in order to get smaller losses and good quality of the jets. Numerically we
can verify whether the evacuating water sheets impact the previous buckets and if there is any
interaction between the evacuating sheets and the incoming jets.
Some results for a 2-jet Pelton turbine with horizontal axis and for a 6-jet turbine with vertical axis
can be seen in Fig. 12 and Fig. 13, respectively. As well, a comparison of calculated and measured
efficiency is presented.
a)
b)
c) Figure 12: Results for a 2-jet Pelton turbine.
a) Jets and evacuating water sheets b) Pressure distribution on runner buckets c) Measured and numerically obtained efficiency values
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 18 of 39
a)
b)
c)
d)
Figure 13: Numerical simulation of flow in distributor and runner of a 6-jet Pelton turbine.
a) Computational domain and grid for distributor with injectors. In the cones behind the injectors the jets are forming during the simulations.
b) Streamlines in distributor and jets c) Jets and evacuating water sheets and pressure on runner buckets. d) Comparison of measured and calculated values. Flow rate was input data, head, torque and efficiency
were calculated from the numerical results.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 19 of 39
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 20 of 39
Part 2
Some of the Results Achieved during the ACCUSIM Project
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 21 of 39
1 Introduction
The ACCUSIM project aims to identify, develop and apply accurate, high-fidelity CFD approaches for
the numerical predictions of the unsteady, turbulent and possibly cavitating flow in hydraulic
turbines, pumps and marine propellers, and to develop advanced shape and functional optimization
strategies for the best design of these systems.
In accordance with these aims, the main activities during the projects were related to the topics:
Advanced (scale resolving) turbulence modelling of flow in water turbines, pumps and
marine propellers
Use of the OpenFOAM CFD open source libraries
Cavitation in water turbines, pumps and marine propellers
Design optimization
In this 2nd part of the report some of the results are presented with particular reference to the papers
where more details about numerical simulations and accuracy of results can be found.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 22 of 39
2 Scale resolving turbulence modelling
Flow in water turbines, pumps and marine propellers is turbulent, with Reynolds number ranging
between 105 and 107. Direct numerical simulation (DNS) of the turbulent flow for complex
geometries and high Reynolds number is not possible nowadays, so one of the turbulence models
has to be used. Flow in water turbines, pumps and marine propellers is unsteady because of runner
rotation and also because of turbulent structures which are forming and disappearing in the flow.
Unfortunately, transient simulations are very time consuming. Therefore often steady state
simulations are used, especially in the design process, where a lot of different geometries have to be
analysed.
Francis turbines
For Francis turbines efficiency and cavitation are usually quite accurately predicted by steady state
simulations, while unsteady phenomena, like rotating vortex rope in a draft tube, can be predicted
only by transient simulations with more advanced turbulence models.
The ACCUSIM group participated in the first Workshop Francis 99, organized by the Norwegian
University of Science and Technology (NTNU), Norway, and Luleå University of Technology (LTU),
Sweden. The goal of the first workshop was to determine the state-of-the-art of numerical
predictions for steady operating conditions. The ACCUSIM group performed several numerical
simulations by two CFD (Computational Fluid Dynamics) codes, ANSYS CFX and OpenFOAM. Steady-
state simulations were performed by k-ɛ and SST (Shear Stress Transport) turbulence models, while
for transient simulations the SAS (Scale Adaptive Simulation) and SST ZLES (Zonal Large Eddy
simulation) models were used. With proper grid refinement in distributor and runner, and 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 [1].
Figure 1: Flow in the draft tube at Part Load (bottom), BEP (middle) and High Load (top)
a) streamlines;
b) vortex structures. Results obtained with the ZLES turbulence model.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 23 of 39
Axial turbines
For axial turbines the efficiency prediction at large flow rate values is very poor mostly because of
overrated flow losses in the draft tube. During the ACCUSIM project a comparison between
numerical results and measurements for a six-blade Kaplan turbine for middle head and for a three-
blade bulb turbine for extremely low head was done in order to determine an appropriate numerical
setup for accurate and reliable simulations of flow in axial turbines. Values of discharge, torque and
losses obtained by different turbulence models were compared to each other and to the
measurements. Steady state simulations were performed with various turbulence models. The effect
of curvature correction (CC) and Kato-Launder (KL) limiter of turbulence production was tested.
Transient simulations were performed with shear-stress-transport (SST) turbulence model, the scale-
adaptive-simulation (SAS SST) model, and with zonal large-eddy-simulation (ZLES). Details about
turbulent structures in the draft tube were illustrated in order to explain the reasons for differences
in flow energy losses obtained by different turbulence models. Also the effects of different advection
schemes (high resolution scheme – HRS and bounded central differential scheme – BCDS) and mesh
refinement were tested.
On the basis of a detailed analysis of flow in a Kaplan turbine and in a bulb turbine with different
turbulence models and at different operating regimes it can be concluded:
• Results of steady state simulations in the Kaplan turbine were improved by using the Kato Launder
limiter of production term (KL) and the curvature correction (CC), therefore KL and CC were used also
in simulations of flow in the bulb turbine.
• For Kaplan and bulb turbines it was found out that steady state analysis is not suitable for all
operating regimes. While for the Kaplan turbine the prediction of efficiency by RANS two-equation
models and by SSG RSM was quite accurate for small and optimal runner blade angles, and
significantly failed only at full rate, for the bulb turbine steady-state simulations failed entirely. In
both cases the main reasons for discrepancy between measured and calculated efficiency were
underestimated torque on the shaft and overestimated flow energy losses in the draft tube.
• Transient simulations by the SST, SAS SST and SAS SST ZLES were performed at one operating point
for maximal runner blade angle. Compared to the steady-state simulations, results improved
significantly. The largest improvement was achieved by SAS SST ZLES.
• Comparing transient results of SST HRS, SAS HRS and SAS BCDS, it can be concluded that the
improvement due to the usage of BCDS instead of HRS was even larger than the improvement due to
the usage of SAS instead of pure SST. With BCDS, the agreement with measurements was improved
mostly because of smaller losses in the runner and better prediction of torque on the shaft.
• For bulb turbine, meshes of different density were used. The results of the SAS SST ZLES model
were better than the results of the SAS SST model on all meshes. Positive effect of mesh refinement
in the draft tube was clearly seen. While with mesh refinement only of the runner no improvement
was obtained, the best results were obtained when both meshes, the runner and the draft tube were
refined.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 24 of 39
Steady-state simulation, SST model, basic grid
Transient simulation, SAS SST with zonal LES in the draft tube, basic grid
Figure 2: Flow in the draft tube (stream lines and velocity distribution on several sections) and contours of viscosity ratio on mid cross-section for a low head bulb turbine.
• Finally, in both cases the simulations by SAS SST ZLES were performed at several operating points
for three runner blade angles. Comparing the results of the steady state analysis to the results of the
SAS SST ZLES the agreement of the former ones with measurements was improved at all operating
regimes. For Kaplan turbine the discrepancy was everywhere smaller than 0.8 %. For bulb turbine, in
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 25 of 39
spite of large improvement, the discrepancy with measurements was even on the finest mesh still
about 2.1 %.
• Too long CPU time is the main disadvantage of transient simulations and the reason for their
limited use in design process. It can be expected that with future development of hardware and
software the problem will be overcome.
More about an improvement of results for axial turbines can be found in [2], [3], [4] and [5].
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 26 of 39
3 Cavitation
Cavitation refers to the process by which vapour forms in low pressure regions of a liquid flow. In
water turbines, pumps and marine propellers the consequences of cavitation are flow instabilities,
excessive vibrations, damage of material surfaces, and reduced performance of the machines.
3.1 Calibration of mass transfer models for the numerical prediction of sheet
cavitation around a hydrofoil
Cavitating flows, which can occur in a variety of practical cases, can be modelled with a wide range of
methods/models [6]. Here, a so-called homogeneous model [7, 8] is applied to the numerical
predictions of sheet cavity flow around a hydrofoil. In the considered model the working fluid is
treated as a homogeneous mixture of two fluids, i.e. water and vapour, behaving as a single one, and
the mass transfer rate due to cavitation is modelled by the mass transfer model. Here, three
widespread mass transfer models are alternatively used. The considered mass transfer models share
the common feature of employing empirical coefficients to tune the condensation and evaporation
processes, whose values affect the accuracy and the stability of the numerical predictions. Thus, in
order to ensure stable and accurate predictions, the empirical coefficients of the considered mass
transfer models are properly and congruently tuned using a calibration strategy driven by the
modeFRONTIER [9] optimization platform. The numerical predictions based on the three different
well-tuned mass transfer models are very close to each other and in line with the available
experimental data.
Figure 3: Distribution of water and water vapour along a hydrofoil NACA66 obtained with Zwart model with standard (a) and calibrated (b) evaporation and condensation constants.
3.2 Cavitation in water turbines
Accurate prediction of cavitation in water turbines is more and more important because nowadays
due to commercial reasons turbines often operate at regimes far from their best efficiency point.
Besides, to reduce manufacturing costs, the dimensions of turbines are reduced; to get the same
power, rotating speed has to increase. The consequence of higher speed is smaller cavitation
(Thoma) coefficient.
During the ACCUSIM project different forms of cavitation in different types of water turbines were
simulated. Cavitation predictions were done for models of Francis and Kaplan turbines and for
prototypes of bulb and Pelton turbines. For all cases the homogeneous multiphase model was used.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 27 of 39
Mass transfer due to cavitation was modelled by the Zwart model with standard constants. In cases
of Kaplan and Francis turbines, cavitation was also modelled by Zwart model with evaporation and
condensation constants that were previously calibrated on a hydrofoil. For Kaplan and Francis
turbines the extent and shape of cavitation were compared to the experimental observation on the
test rig. The effect of cavitation on turbine efficiency was also investigated.
3.2.1. Cavitation in Francis turbines
For a Francis turbine we simulated leading edge cavitation, travelling bubble cavitation on suction
side near the trailing edge and vortex rope at part and high load.
Overload operating regime
Steady-state simulations of flow in a medium-head Francis turbine were performed at overload
operating regime with high flow rate. SST turbulence model with curvature correction and with the
Kato-Launder limiter of production term in equation for turbulence kinetic energy was used.
Simulations were done for different values of cavitation coefficient, which means from non-cavitating
to strongly cavitating regimes.
a) b)
Figure 4: Cavitation in the Francis turbine at full load. a) vapour at suction side of runner blades, left - standard, right - calibrated coefficients. b) Cavitation on suction side of runner blades near the trailing edges and rotating
vortex rope, left - test rig, right - CFD.
Numerical results showed three types of cavitation: inlet cavitation, cavitation at the suction side
near the blade trailing edge and vortex core cavitation behind the hub. Blades' inlets are not visible
on photos, because they are hidden behind the runner band, therefore the extent of inlet cavitation
cannot be compared to the observation on the test rig. Shape and extent of numerically predicted
cavitation at the suction side of the blades near the trailing edges, and shape and size of the
cavitating vortex core behind the hub are in good agreement with observation on the test rig. Results
were presented in [10].
Part load operating regime
Transient simulations of flow in a Francis turbine were performed with a goal to predict pressure
pulsation frequencies and amplitudes caused by rotating vortex rope at part load operating regime.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 28 of 39
Simulations were done with the SAS SST turbulence model with curvature correction on basic and
refined computational meshes. Without cavitation modelling too small values of frequency and
amplitudes were obtained. With mesh refinement the calculated amplitudes were a bit closer to the
measured values, while the accuracy of predicted frequency did not improve at all. Agreement
between measured and numerical values was significantly improved when cavitation was included in
simulations. In addition, the predicted value of the dominant frequency was slightly more accurate
when, in the Zwart et al. cavitation model, the default condensation and evaporation model
constants were replaced with previously calibrated ones. Results were presented in [10] and [11].
a)
b)
c)
d)
Figure 5: Vortex rope a) on the test rig; b) without cavitation modelling, iso-surface of evaporation pressure,
basic mesh; c) default cavitation constants, iso-surface of Vapour Volume Fraction = 0.1 d) calibrated cavitation constants, iso-surface of Vapour Volume Fraction = 0.1.
Figure 6: Comparison of numerical and experimental results at part load operating regime
Exp. - experimental values, 1 - no cavitation modelling, basic mesh, 2 - no cavitation modelling, fine mesh, 3 -
cavitation modelling, default parameters, basic mesh, 4 - cavitation modelling, calibrated parameters, basic
mesh
3.2.2. Cavitation in Kaplan turbines
For a Kaplan turbine numerical simulations were done at one operating point for maximal runner
blade angle and nominal head. Steady-state results obtained with the SST (Shear Stress Transport)
turbulence model were improved by transient simulations, where the SAS (Scale Adaptive
Simulation) SST model was used. Cavitating flow was simulated using the homogeneous model. Mass
transfer rate due to cavitation was regulated by the Zwart et al. model with default model constants
used in ANSYS CFX commercial code and also with the evaporation and condensation parameters
previously calibrated considering the sheet cavity flow around a hydrofoil. The numerical results
were compared with the observation of cavity size on the test rig and with the measured sigma break
curve. Steady-state simulations predicted significantly too small efficiency level and too small extent
of cavitation on the runner blades. With transient simulations, the shape and size of the predicted
sheet cavitation agreed well with the cavitation observed on the test rig. In addition, also the
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 29 of 39
predicted efficiency was more accurate, although the value of σ (cavitation or Thoma number) where
the efficiency dropped for 1% was a bit too large. The difference between the results obtained with
standard and calibrated model parameters of the Zwart mass transfer model was small. Details about
the simulations and the accuracy of results can be found in [12].
Figure 7: Comparison of shape and size of cavity, a) experiment, b) steady state simulation, calibrated
coefficients, c) time dependent simulation, calibrated coefficients.
3.2.3. Cavitation in a prototype of a bulb turbine
For a prototype of a bulb turbine numerical flow
simulation was performed with a purpose to obtain
pressure distribution for the stress analysis. To get
accurate pressure on runner blades the cavitation
modelling has to be included. The prototype diameter of
the turbine was very large (Dp = 7.5 m.) therefore
hydrostatic pressure has to be included in the simulation.
Due to hydrostatic pressure a cavity at upper blades is
much larger than at the bottom ones. During transient
simulation a size of a cavity at each blade increases when
the blade rotates up and decreases when it rotates down.
Figure 8: Pressure distribution and
attached cavity at suction side of bulb turbine runner blades
3.2.4. Cavitation in Pelton turbines
Cavitating flow in Pelton turbines consists of three components: water, air and water vapour.
Transient simulations include modelling of free surfaces and mass transfer between water and
vapour due to cavitation and condensation processes.
The presence of water vapour does not necessary cause material erosion. The conditions for
cavitation pitting on Pelton buckets are [13]:
• Vapour cavity is sticking to the bucket surface.
• Water vapour is condensed in a very short time.
• The condensation of water vapour is developed in absence of air.
In Fig. 9 regions with water vapour on the inner and on the back side of the bucket are presented. At
the inner side the vapour is condensing very slowly. At the back side the condensation is also slow
and the vapour is in contact with air. So no cavitation damages are expected in this case.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 30 of 39
Figure 9: Detail of computational grid (left), water vapour on the inner side of the bucket (middle), water
vapour at the back side of the bucket (right). Water vapour is coloured by wall distance.
3.3 Sheet and cloud cavitation prediction in double suction centrifugal pump
Cavitation at the leading edges of the impeller blades in a pump is quite a usual phenomenon, due to
the limitations in the available suction head (or, in simple words, due to too low "submergence" of
the pump). Usually the sheet-type cavitation (also called blade cavitation) is of small size and occurs
on the suction side of the blade, so the pump performance is only negligibly affected. At specific
conditions, the so-called cloud cavitation can appear (Fig. 10), when large clouds of vapour are torn-
off of the sheet cavity, and are carried along with the stream.
Two types of numerical approaches for the prediction of cloud cavitation in a centrifugal pump were
used: a single-blade simulation and a (symmetrical) half-geometry simulation. For both approaches,
simulations were performed with the default and with the modified [8, 14] parameters of the mass
transfer model. Simulations, performed with the modified parameters, were found to be less stable
than the ones with the default parameters.
The single-blade simulation with the modified parameters of the mass transfer model produced
cloud cavities of dubious origin (Fig. 12). The results obtained on a single-blade mesh are of limited
use due to the negligence of the suction chamber effect on the inlet velocity distribution.
It is possible to conclude that simulations performed on the geometry of a half-pump, using the
default and the optimized parameters, can predict the pinch-off of vapour clouds (Fig. 13).
Figure 10: Cloud cavitation, observed on the double-suction pump at HPP Fuhren [15].
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 31 of 39
Figure 11: Geometry of the double-suction pump. Left: numerical model (in green: inlet pipe with suction
chambers; in blue: impeller; in red: volute with outlet pipe). Right: experimental setup with visible observation window.
Figure 12: Prediction of cavity (isosurface of 10% water vapour) with single-blade passage simulation
Figure 13: Half-geometry simulation with default constants (isosurface of 10% water vapour).
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 32 of 39
3.4 Numerical predictions of the cavitating flow around model scale propellers
working in uniform and non-uniform inflow
The numerical predictions of the turbulent cavitating flow around two model scale propellers,
recognized as international benchmarks, were performed. In particular, cavitating flow around PPTC
propeller working in uniform inflow [16] and cavitating flow around the E779A propeller working in
uniform as well as non-uniform inflow [17] were simulated.
The simulations were performed using commercial and open source CFD (Computational Fluid
Dynamics) codes. The cavitating flow was modelled using the homogeneous model along with three
different widespread mass transfer models, previously calibrated considering the sheet cavity flow
around a two-dimensional hydrofoil [8]. The turbulence effect was modelled using the RANS
(Reynolds Averaged Navier Stokes) approach.
The numerical results were compared with the available experimental data. The simulations
performed with the three different calibrated mass transfer models were very similar to each other
and in line with the experimental data, even though the numerical cavitation patterns were generally
slightly overestimated [14, 18].
Ex
per
imen
t
Zw
art
Figure 14: PPTC propeller; RANS simulation performed with Zwart mass transfer model;
cavitation patterns depicted using isosurfaces of vapour volume fraction equal to 0.2
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 33 of 39
4 Comparison of results obtained with OpenFOAM and ANSYS-CFX
Comparison of results obtained with OpenFOAM and ANSYS-CFX was done for 6 cases:
• High head Francis turbine (Tokke model from Workshop Francis99)
• Middle head Francis turbine
• Kaplan turbine
• Bulb turbine
• Cavitation prediction: Attached cavity flow around a hydrofoil
• Cavitation prediction: Attached cavity flow around E779A model scale marine propeler
Approximately the same level of accuracy of numerical results was achieved with both codes. CPU
time was significantly longer in case of simulations with OpenFOAM.
Figure 15: High-head Francis turbine, computational domain and mesh with a detail of mesh in stay and guide vane cascade (left), and comparison of predicted efficiency to the measured values (right).
Figure 16: Prediction of the attached cavity for a hydrofoil, results of CFX and OpenFOAM, both with the Kunz
mass transfer model.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 34 of 39
5 Optimization
The results of two design optimization cases are presented: an optimization of a runner of a double-
sided centrifugal pump and an optimization of a Francis turbine runner cone.
5.1 Optimization of a double-sided centrifugal pump
Centrifugal pumps are widely used in industrial applications. Compared to single-entry centrifugal
pumps, double-sided pumps allow transportation of greater flow rates due to smaller proneness to
cavitation, and offer counter-balancing of axial hydraulic forces due to double-entry design [19].
In the modern world of rapidly-improving technologies it is important to design excellent products in
short time. The optimization techniques bring many benefits over the traditional "trial-and-error"
design process: shorter design phase, exploration of design space in a more systematic way,
development with less hard-to-spot human-based errors, etc. In turbomachines, usually multiple
objectives have to be optimized. One of the first multi-objective optimization studies was performed
by Lipej and Poloni [20].
The optimization study of a double-sided centrifugal pump with specific speed nq=62 (per impeller
side), presented in Fig. 17, was performed within modeFRONTIER® optimization platform [21]. During
the optimization only the impeller geometry was allowed to be modified, while the rest of the
geometry was fixed. The objective of the study was twofold:
To compare different Response Surface Methods (RSMs), also called metamodels or surrogates,
for their ability to predict pump efficiency [22],
To propose and validate an economical decoupled simulation method for the optimization of a
pump [23].
Figure 17: Geometry of the double-sided centrifugal pump.
In the first phase, during the study of RSMs, a simplified pump geometry comprised of (symmetrical)
half of volute and of one impeller channel (Fig. 18). Since the impeller channel and volute were
simulated simultaneously, we call this model coupled model. An example of successful modification
is presented in Fig. 19.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 35 of 39
In the second phase of the study, based on flow simulations of the volute, flow losses in the volute
were estimated with a simple function ΔHvolute=(f(α2,hub)+ f(α2,midspan)+ f(α2,shroud))/3, where α2 is angle
between circumferential and absolute velocity at the impeller outlet. Afterwards, geometry
optimization was performed only on one impeller channel (Fig. 20). We call this model decoupled
model. Two optimization objectives were used, namely pump efficiency and reluctance to cavitation.
For the latter, a simple procedure for its estimation was proposed. Results for forty most promising
candidates for high pump efficiency were validated with full-geometry CFD numerical simulations.
The following conclusions were drawn:
The presented method with estimation of flow losses in the volute represents a good
approach for quick optimization;
Angle α2 at hub and at shroud needs small adjustments;
Low value of angle α2, when compared to the average α2 value, acted beneficially on
reducing the volute flow losses.
Figure 18: Geometry and mesh for the coupled model.
Figure 19: Blade geometry before (left) and after (right) the optimization. Relative efficiency increased
for 0.9 %. Green surface: inlet surface. Yellow: outlet surface.
Figure 20: Geometry and mesh for the decoupled model. Blue arrows represent inlet and outlet boundary
conditions.
Volute
mesh
Single-channel impeller
mesh
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 36 of 39
5.2 Multi-objective optimization of the Francis turbine runner cone
A multi-objective optimization with the genetic algorithm MOGA-II in modeFRONTIER was used to
optimize the shape of a simplified runner-cone extension (RCE) for a Francis turbine of higher specific
speed nq. The objective of the optimization was the minimization of the draft tube pressure losses for
two operating regimes. To evaluate the progress, a baseline case without the extension was
calculated first. Six geometric design variables, defining the shape of RCE, were being modified
during the optimization. Finite-volume based commercial software ANSYS CFX was employed to
calculate the flow field on a structured mesh. Shapes of RCE on Pareto front were long and thin. At
the best-efficiency point (BEP) the draft tube pressure losses were larger than without the RCE. At
the maximal discharge point (MAX) the draft tube pressure losses were smaller than without the
RCE; turbine efficiency increases for 0.15% at full size computation grid.
Figure 21: Case 0, total pressure contours at meridional section, QBEP at left, Qmax at right
Figure 22: Design ID_54, total pressure contours at meridional section, QBEP at left, Qmax at right
Details about the optimization of the runner cone and the results of optimization can be found in
[24].
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 37 of 39
Conclusions
Scale resolved SAS SST model in combination with zonal LES was successfully used for flow simulation
in Francis, Kaplan and bulb turbines. In comparison to the steady-state results with the SST
turbulence model, significant improvement was achieved, especially for a Kaplan and a low head
bulb turbine.
Cavitation prediction with standard or/and calibrated cavitation models was done for several cases:
initial and travelling bubble cavitation as well as cavitating vortex rope at overload operating
regime for Francis turbine,
cavitating vortex rope at part load in a Francis turbine with analysis of pressure pulsations,
cavitation in a Kaplan turbine at high load and its effect on efficiency,
cavitation on a prototype size of a bulb turbine,
three-phase flow in a Pelton turbine with detailed analysis of results to verify occurrence of
pitting cavitation,
sheet and cloud cavitation in a double-suction centrifugal pump,
cavitating flow around model-scale propellers working in uniform and non-uniform inflow.
For all cases, except for the prototypes of bulb and Pelton turbines, numerical results were validated
with observations on test rigs and with measured values, and good agreement was obtained.
For design optimization two cases are presented: optimization of a double-sided centrifugal pump
and multi-objective optimization of a Francis turbine runner cone.
A comparison of results obtained with the commercial code ANSYS CFX and with the open-source
code OpenFOAM shows that approximately the same level of accuracy was achieved, although CPU
time was significantly higher in case of simulations with OpenFOAM.
Overall the results achieved during the ACCUSIM project are very good, as demonstrated by the large
number of technical papers and conference presentations, or even better by the exceptional results,
in comparison with other teams and research centers, obtained by ACCUSIM researchers in blind and
non-blind benchmarks.
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 38 of 39
References:
1. Jošt D, Škerlavaj A, Morgut M, Mežnar P and Nobile E 2015 Numerical simulation of flow in a
high head Francis turbine with prediction of efficiency, rotor stator interaction and vortex
structures in the draft tube Journal of physics: Conference series 579(1) p 012006 doi:
http://dx.doi.org/10.1088/1742-6596/579/1/012006
2. Jošt D, Škerlavaj A and Lipej A 2014 Improvement of efficiency prediction for a Kaplan turbine
with advanced turbulence models Journal of Mechanical Engineering - Strojniški vestnik 60(2)
pp 124-134 doi: 10.5545/sv-jme.2013.1222
3. Jošt D and Škerlavaj A 2014 Efficiency prediction for a low head bulb turbine with SAS SST
and zonal LES turbulence models IOP conference series. Earth and environmental science
22(1) p 022007 doi: http://dx.doi.org/10.1088/1755-1315/22/2/022007
4. Jošt D, Škerlavaj A, Mežnar P and Morgut M 2014 Napoved izkoristka nizkotlačne cevne
turbine z numerično analizo toka = Efficiency prediction for a low head bulb turbine with
numerical flow analysis (in Slovene) Kuhljevi dnevi 2014 Maribor 24-25 September 2014
(Ljubljana: Slovensko društvo za mehaniko) pp 79-86
http://www.drustvozamehaniko.si/zbornik/ZbornikKD2014.pdf
5. Jošt D, Škerlavaj A, Morgut M and Nobile E 2015 Effects of turbulence model and mesh
resolution on the performance prediction of a bulb turbine. NAFEMS World Congress 2015
incorporating the 2nd International SPDM Conference: proceedings San Diego
6. Koop AH 2008 Numerical Simulation of Unsteady Three-Dimensional Sheet Cavitation. PhD
Thesis University of Twente
7. Asnaghi A, Feymark A and Bensow RE 2017 Improvement of cavitation mass transfer
modeling based on local flow properties Int. J. Multiphase Flow 93(2017) pp 142-157 doi:
http://dx.doi.org/10.1016/j.ijmultiphaseflow.2017.04.005
8. Morgut M, Nobile E and Biluš I 2011 Comparison of mass transfer models for the numerical
prediction of sheet cavitation around a hydrofoil Int. J. Multiphase Flow 37(6) pp 620-626
doi: http://dx.doi.org/10.1016/j.ijmultiphaseflow.2011.03.005
9. www.esteco.com
10. Jošt D, Škerlavaj A, Morgut M, Stopar RR and Nobile E 2016 Numerical simulation of different
forms of cavitation in Francis turbines ANSYS Convergence Conference Ljubljana 25 May 2016
11. Jošt D, Škerlavaj A, Morgut M and Nobile E 2017 Numerical Prediction of Cavitating Vortex
Rope in a Draft Tube of a Francis Turbine with Standard and Calibrated Cavitation Model
Journal of Physics: Conference Series 813(1) pp 012045 doi: http://dx.doi.org/10.1088/1742-
6596/813/1/012045
12. Jošt D, Morgut M, Škerlavaj A and Nobile E 2015 Cavitation prediction in a Kaplan turbine
using standard and optimized model parameters Cavitation and dynamic problems; 6th IAHR
Meeting of the Working Group, IAHRWG (eds) Lipej A and Muhič S (Novo mesto: Faculty of
Technologies and Systems)
13. Rossetti A, Pavesi G, Ardizzon G and Santolin A 2014 Numerical Analyses of Cavitating Flow
in a Pelton Turbine J. Fluids Eng. 136(8) p 081304
IAPP 2013 – 612279
ACCUSIM_Final_Public_Report_D.6.5_v01.doc Page 39 of 39
14. Morgut M and Nobile E 2012 Numerical Predictions of Cavitating Flow Around Model Scale
Propellers by CFD and Advanced Model Calibration International Journal of Rotating
Machinery 2012 p 618180
15. Škerlavaj A, Titzschkau M, Pavlin R, Vehar F, Mežnar P and Lipej A 2012 Cavitation
improvement of double suction centrifugal pump HPP Fuhren IOP Conference Series: Earth
and Environmental Science 15(2) p 022009
16. https://www.sva-potsdam.de/pptc-smp11-workshop/
17. Salvatore F, Streckwall H and van Terwisga T 2009 Propeller Cavitation Modelling by CFD-
Results from the VIRTUE 2008 Rome Workshop The First International Symposium on Marine
Propulsors smp'09 Trondheim Norway.
18. Morgut M, Jošt D, Nobile E, Škerlavaj A 2015 Numerical investigations of cavitating propeller
in non-uniform inflow Fourth International Symposium on Marine Propulsors smp’15 Austin
Texas
19. Gülich JF 2008 Centrifugal Pumps (Berlin: Springer-Verlag) p 926
20. Lipej A and Poloni C 2000 Design of Kaplan runner using multiobjective genetic algorithm
optimization J. Hydraul. Res. 38(4) pp 73–9
21. ESTECO 1999-2016 modeFRONTIER User Manual (ESTECO: Trieste) p 504
22. Škerlavaj A, Morgut M, Jošt D and Nobile E 2017 Optimization of a single-stage double-
suction centrifugal pump Journal of Physics: Conference Series 796(1) p 012007 doi:
http://dx.doi.org/10.1088/1742-6596/796/1/012007
23. Škerlavaj A, Morgut M, Jošt D and Nobile E 2017 Decoupled CFD-based optimization of
efficiency and cavitation performance of a double-suction pump Journal of Physics:
Conference Series 813(1) p 012048 doi: http://dx.doi.org/10.1088/1742-6596/813/1/012048
24. Stopar RR, Škerlavaj A, Nobile E 2017 Multiobjective optimization of a hydraulic shape of
runner cone extension in Francis turbine (in Slovene language) Kuhljevi dnevi 2017
http://www.drustvozamehaniko.si/zbornik/ZbornikKD2017.pdf