B6

111
3
rd
AHR nternational Meeting of the Workgroup on Cavitation and Dynamic Problems in
Hydraulic Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
*
Svlvain Tridon, LEGI-INP Grenoble-CNRS, BP53, Grenoble, 38041, France,
phone: 334 76 82 70 27, Iax: 334 76 82 50 01, email: sylvain.tridonlegi.grenoble-inp.Ir

EXPERIMENTAL DESCRIPTION OF SWIRLING FLOW
DOWNSTREAM A FRANCIS TURBINE RUNNER AND
INITIALIZATION OF NUMERICAL SIMULATIONS
SyIvain Tridon* and Stphane Barre
LEG-NP Grenoble-CNRS, BP53, Grenoble, 38041, France
GabrieI Dan Ciocan and Laurent Tomas
ALSTOM HYDRO France, 82 av. Lon Blum, Grenoble, 38000, France
ABSTRACT
The draIt tube oI a hydraulic turbine is the component where the Ilow exiting the runner is decelerated,
thereby converting the excess oI kinetic energy into static pressure. In the case oI machine
reIurbishment oI an existing power plant, most oI the time only the runner and the guide vanes are
currently modiIied. For Iinancial and saIety reasons, the spiral casing and the draIt tube are seldom
redesigned, even iI these components present some undesirable behaviour. In some cases, the
installation oI an upgraded runner leads to a peculiar and undesirable eIIiciency drop as the discharge
is increased above the best eIIiciency point value. It is Iound to be related to a corresponding sudden
variation in the draIt tube pressure recovery coeIIicient at the same discharge.
The swirling Ilow exiting the runner is complex and highly turbulent. The radial velocity is rarely
measured because a quite complicated measurement setup is needed. However, this velocity
component is greatly needed in order to initialize properly the numerical simulations and its inIluence
is important in spite oI its small magnitude. Velocity measurements downstream the runner include
radial component made at CREMHyG (Grenoble) by LDV and PIV techniques are presented. An
analytical Iormulation Ior this velocity component based on the Iormulation Ior the conical diIIuser
and on the three vortices structure is proposed and compared with measurements.

KEYWORDS
DraIt-tube, PIV, LDV.


1. INTRODUCTION
The draIt tube oI a hydraulic turbine is the component where the Ilow exiting the runner is
decelerated, thereby converting the excess oI kinetic energy into static pressure and then, as a
consequence, increases the global power oI the turbine. The swirling Ilow exiting the runner
is complex and highly turbulent and required LES calculations to be well perIormed |1|.
Numerical simulation oI swirling Ilow in draIt tube need boundary conditions at the inlet to
be perIormed. Several experimental studies have been devoted to the description oI the Ilow
structure under the runner. Most oI the time, only the axial and tangential velocity component
112
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
are measured by LDV technique. The radial velocity is rarely measured because a quite
complicated measurement setup is needed see Ciocan and al. |2|. However, this velocity
component is greatly needed in order to initialize properly the numerical simulations and its
inIluence is important in spite oI its small magnitude. OIten the radial component is neglected
or barely approached with the analytical Iormulation Ior a conical diIIuser,

see Mauri and al. |3|. The use oI the conical Ilow approximation may be inadequate iI Ior
example the Ilow is non-axisymmetric or iI the overall vorticity structure emanating Irom the
runner is perturbed and distorded. In most practical cases, some distortion eIIects initiated by
downstream perturbations like some Ilow rate asymmetry as experienced in the elbow and the
draIt tube may change the global structure oI the Ilow in the conical zone just downstream oI
the runner. These eIIects may be ampliIied by the presence oI a pressure recovery break-oII in
the draIt tube as experienced in some particular Ilow conIigurations see Tridon and al. |4|.
In these conditions the conical axisymmetric Ilow approximation do not hold and then, the
radial velocity component must be measured or estimated in a diIIerent way in order to obtain
a suIIiciently accurate Ilow description Ior the initialization and validation oI numerical
RANS or LES simulations.
In this paper, radial velocity component measurements made at CREMHyG (Grenoble) by
PIV technique with two diIIerent setups are presented. And an analytical Iormulation Ior this
component based on the three vortices structure presented by Resiga and al. |5| is proposed.
The present paper is organized in Iour parts. In the Iirst one the investigated case and velocity
measurement instrumentation are presented. The second part is devoted to the analytical
representation oI axial and tangential velocity proIiles Irom LDV measurements. In the third
part, the particular case oI the radial component is treated. Conclusion and discussion Iollows.

2. EXPERIMENTAL SETUP
a. FRANCIS TURBINE MODEL PRESENTATION
The investigated case corresponds to the scale model oI a Francis turbine oI high speciIic
speed, 0.55 (n
q
86). The scale model supplied by ALSTOM is installed on the test rig oI
the CREMHyG. The turbine model has a spiral casing oI double curvature type with a stay
ring oI 10 stay vanes, a distributor made oI 20 guide vanes, a 19 blades runner oI a 365mm
outlet diameter, and a symmetric elbow draIt tube with one pier. The draIt tube is composed
oI three diIIerent parts, the cone in Plexiglas, the elbow in Iiberglass and the straight diIIuser
in sheet steel.


Fig.1 Efficiencv break-off obtained bv increasing and decreasing the discharge
B6
113
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno

The global 'eIIiciency oI the draIt tube is quantiIied using his static pressure recovery
coeIIicient, deIined with the discharge and the diIIerential pressure between inlet and outlet
sections oI the draIt tube,

By increasing the discharge (see Fig.1) the eIIiciency presents an important drop oI more than
one percent very close to the best operating point. This break-oII presents a hysteresis eIIect;
by decreasing the discharge the drop is less important. It is Iound to be produced by a
corresponding drop in the draIt tube pressure recovery. The perIect similarity between the two
proIiles on Fig.1 conIirms the idea that the accident occurred in the draIt tube. The Iour
operating points where Iull velocity measurements are perIormed on the survey section are
selected around the eIIiciency drop. They are detailed in the Tab. 1 and marked by a square on
Fig.1. The operating point n2 is the closest Irom the best operating point and OP3 just
Iollows the eIIiciency drop.

PI Accident Discharge coeI. M Energy coeI. \
1 No 0.368 1.18
2 No 0.374 1.18
3 Yes 0.380 1.18
4 Yes 0.390 1.18

Tab.1 Four operating points near efficiencv drop selected for full velocitv measurements

b. LASER DOPPLER VELOCIMETRY INSTRUMENTATION


Fig.2 Sketch of the Francis turbine model for the flow survev section in the cone

The 2D velocity proIile survey is perIormed by the LDV measurement method on three
diameters at 120 in the cone at 115 mm under the runner and one diameter at 48 mm under
the runner which correspond respectively to 0.63 and 0.26 runner outlet radius R
ref
. The
experimental data used in this paper were obtained with a two-component probe Laser
Doppler Anemometer (LDA), operating in backward-scattered light on-axis-collection mode
and transmission by optical Iiber, with a 5 W argon-ion laser source. Bragg-cell shiIting at
40MHz is used to resolve directional ambiguity oI the velocity.
114
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
Spherical silver-coated glass particles are introduced in the test rig Ilow. These particles are
hollowed in order to match the water density and are able to Iollow Ilow Iluctuations
Irequency up to 5 kHz. The mean diameter oI these particles is 10 m. Three cylindrical glass
optical windows with plane and parallel Iaces are placed at 60 with their axes 115 mm under
the runner which correspond to 0.63 UXQQHU RXWOHW UDGLXV 5
UHI
(Fig.2). They are used as
interIaces and exempt rotate the cone between each azimuthal direction. The geometrical
reIerence position oI the measurements is obtained by positioning the laser beams on the
windows Iaces with accuracy better than 0.05 mm. Both axial and circumIerential
components oI the velocity are measured. Accuracy in the velocity measurements is estimated
to be 2 oI the measured value; see Ciocan and al. |6|.

c. PARTICLE IMAGE VELOCIMETRY INSTRUMENTATION
i. MEASUREMENTS ON THREE VERTICAL PLANES
During the Iirst measurement campaign, the 3D instantaneous vertical velocity Iield in the
cone is investigated with a Dantec M.T. 3D PIV system, which consists oI a double-pulsed
laser, two double-Irame cameras, and a processor unit Ior the acquisition synchronization and
the vectors detection by cross correlation. The illuminating system is composed oI two Nd-
YAG lasers, each delivering a short impulse oI 10 ns and 15 mJ energy at 1 kHz Irequency.
Thus the time interval between two successive impulses is Iixed to 100s according to have
approximately a 10 pixels displacement in the axial direction. The output laser beam oI 532
nm is transIormed into a sheet oI 4 mm width and 25 divergence.
Two Nanosense cameras with a resolution oI 12801024 pixels operated in dual-Irame mode
are used Ior 200200mm
2
investigation area. The cameras are placed in a stereoscopic
conIiguration, Iocused on the laser-sheet, synchronized with the two pulses. They capture the
position oI seeding particles oI ~10m diameter by detecting their scattered light. PIV
measurements also use hollowed spherical silver-coated glass particles oI 10m.
For the optical access, the cone is manuIactured in Polymethyl methacrylate (PPMA) with a
reIractive index oI 1.493, equipped with a narrow window Ior the laser`s access and two large
symmetric windows Ior the cameras access, having a Ilat external surIace Ior minimizing the
optical distortions. The corresponding two-dimensional vector maps, obtained Irom each
camera by a Iast Fourier transIorm-based algorithm, are combined in order to obtain the out-
oI-plane component, characterizing the displacement in the laser-sheet width.
The correlation between the local image coordinates and real space coordinates is realized
through a third order optical transIer matrix, which includes the correction oI distortions due
to diIIerent reIractive indices in the optical path and to the oblique position oI the cameras.
The calibration relation is obtained acquiring images oI a calibration target with equally
spaced markers, moved in Iive transversal positions in order to have volume inIormation. The
target displacement in the measurement area, with accuracy within the narrow limits oI 0.01
mm in translation and 0.1 deg in rotation, insured a good calibration quality; Iliescu and al. |7,
8|. The overall uncertainty oI the PIV 3D velocity Iields is 3 oI the mean velocity value.
The measurements are perIormed Ior three vertical planes at 120 which correspond with the
diameters oI the LDV measurements survey.
ii. MEASUREMENTS ON FOUR HORIZONTAL PLANES
During the second measurement campaign, the horizontal velocity Iield under the
runner is investigated with a LA VISION PIV system, which consists oI a double-pulsed
laser, one double-Irame camera ProX with a resolution oI 16001200 pixels, and a processor
unit Ior the acquisition synchronization and the vectors detection by cross correlation. The
B6
115
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
illuminating system is composed oI two Nd-YAG, each delivering a short impulse oI 10 ns
and 20 mJ energy at 20 Hz Irequency.



Fig.3 PIJ measurements under the runner setup during calibration step

For the optical access, the above part oI the elbow is equipped with a window directing to the
runner see Fig.3. This window is manuIactured in Polymethyl methacrylate (PPMA) with a
reIractive index oI 1.493. A special circular calibration plate oI diameter 355mm is designed
Ior the measurements - see Fig.4 - and held under the runner with a magnet. The elbow is
dismantled Ior calibration plate positioning. Then the elbow is reassembled and the test rig
Iilled with water. The camera is installed under the elbow and Iocused on the laser-sheet.
AIter the acquisition oI the calibration images the calibration plate is Ialling down by an LDV
window in the cone and takes oII the draIt tube by the large window near the draIt tube outlet.
The measurements are perIormed Ior Iour horizontal planes at respectively 24 mm, 48 mm, 91
mm, and 141 mm under the runner which corresponds to respectively 0.13, 0.26, 0.50 and
0.77 runner outlet radius R
ref
.



Fig.4 Circular calibration plate under the runner

3. ANALYTICAL REPRESENTATION FOR AXIAL AND TANGENTIAL VELOCITY
Throughout this paper the velocity is made dimensionless by the runner outlet velocity
(runner angular speed multiplied bv runner outlet radius, R
ref
), and lengths are made
dimensionless with respect to the runner outlet radius R
ref
(Fig.2).

116
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
a. PHASE-AVERAGED VELOCITY
The two-component LDA system take 100 000 acquisition per measuring position in
coincidence. Each velocity measurement over 100 000 is stored with his corresponding
encoder position. The velocity samples are sorted in 180 phase bins giving a circumIerential
resolution oI 2 degrees.
Phase-averaged contour oI axial and tangential velocity Ior OP4 and axe 1 (see Fig.2) are
shown in Fig.5. The passage oI the 19 runner`s blades is easily observable. The extent oI the
blade-vortex in term oI radius is in the range between 0.6 R
reI
and 0.9 R
reI
. From these results
it appears clearly that the blades wakes must be taken into account when modeling this Ilow.


Fig.5 Axial and tangential velocitv phase averaged - OP1

At the beginning oI this study |3|, we evoke three diIIerent scenarii to explain the physical
origin oI the eIIiciency break-oII. One oI these is a possible apparition, Ior operating
conditions corresponding to the break-oII, oI a global 3D large-scale instability implicating
the entire Ilow in the diIIuser. A simpliIied analysis based on a parallel and symmetrical Ilow
have been done by Resiga and al. |4| in the context oI the FLINDT project. We start Irom this
analysis and adapt the analytical representation oI the velocity components to our case. Then,
a simpliIied stability analysis can be done, which is presented here. This result will pave the
way Ior a global 3D stability calculation which can be perIormed as a Iuture extension oI the
present work.

b. BLADES-VORTEX FORMULATION
We introduce in this paper the analytical representation oI the velocity components previously
used by Resiga and al |4|. These equations correspond to the superposition oI three Batchelor
vortices. One Batchelor vortex with a characteristic radius inIinite which correspond to a solid
body rotation, and , and two Batchelor vortices, one co-rotating and the other
counter-rotating with respect to , and co-Ilowing/counter-Ilowing with respect to
, respectively. Faller and Leibovich |9| have previously used these axial velocity
Iunctional Iorms with only two vortices to Iit their experimental data Ior a radial guide-vane
swirl generator.
The blades-vortex are added in the both equations by two anti-symetrical Batchelor vortices,
it means Iour parameters more: the radius oI the blades-vortex R
3
, the center oI these vortex
along the blade R
4
and the axial and circumIerential amplitude oI these vortices,

and

respectively. We obtain the Iollowing equations:

B6
117
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno




Axial and tangential velocity proIiles oI equations 3 and 4 deIine a twelve parameters set
to be determined by Iitting the experimental data.
For each operating point, with a set oI experimental data ,the vector error is
deIined as,


A special least-square procedure has been developed in the Matlab environment in order to Iit
the equations to the experimental data inimizing the term . It used an initial set oI
eight parameter given by the user and Iurnished aIter minimization oI a converged value oI
. This analyze maybe perIormed in a quasi axisymetric reIerence with respect to the vortex
itselI and then, when done, translated into the draIt tube reIerence.
II R
0
is the dimensionless survey section radius, then the discharge coeIIicient can be obtained
by integrating the axial velocity proIile, but this integration is taken error Iunction erf and
exponential integral Ei, so it is not detailed here. By observing the phase-averaged velocity
contours in Fig.5, R
3
can be estimate to 0.2 R
reI
and R
4
to 0.65 R
reI
.
Fig.6 shows the analytical proIiles obtained with these equations. The quality oI the Iit can be
assessed by observing that most oI the time the curves approach the experimental points.


Fig.6 Axial and tangential velocitv profiles OP2 and OP3

By Iitting the diIIerent parameters on a large band oI discharge we Iound a suitable parametric
representation Ior the swirling Ilow at the Francis runner outlet. Fig.7 shows the variation oI
vortex characteristic angular velocities with respect to . Linear least squares Iits accurately
represent , and . is the characteristic angular velocity oI the centered vortex which
increase in intensity with the discharge. Moreover, one should note that and are almost
constant over the investigated operating range. The variation oI vortex characteristic axial
velocities with respect to is shown in Fig.7, together with the corresponding linear Iits.
Fig.8 displays the dependence oI the vortex core radii on . Finally, Fig.8 presents the
variation oI the diIIerent blades-vortex parameter. A Iirst conclusion Irom Fig.7 and 8 is that
swirl parameters have a smooth, generally linear, variation in M over the investigated range.
118
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
As a result, one obtains the velocity components as C

Iunctionals w(r ,) and u (r ,), Iurther

employed in a parametric study oI the Ilow stability or other properties.


Fig.7 Characteristic angular and axial velocities versus discharge coefficient



Fig.8 Jortex core radii versus discharge coefficient and blades-vortex parameters
4. FOCUS ON RADIAL VELOCITY MEASUREMENTS
In this part we will compare the conical Ilow Iormulation Ior radial velocity with azimuthally
averaged data. Then a new analytical Iormulation based on the Iormulation Ior the conical
diIIuser and on the three Batchelor vortices superposition will be presented and compared
with phase averaged data and non-phase averaged data on a single diameter.

a. ANALYTICAL FORMULATION FOR RADIAL VELOCITY
Even iI we are conscious that the radial velocity component is obviously non-axisymetric we
propose an axisymetric analytical Iormulation Ior this component based on the three
Batchelor vortices superposition. As Ior axial and tangential velocity component in the
paragraph 3, this analyze maybe perIormed in a quasi axisymetric reIerence with respect to
the vortex itselI and then, when done, translated into the draIt tube reIerence. To manage it we
use the conical Ilow Iormulation Ior the solid body rotation and add the two counter rotating
Batchelor vortices with Iour new magnitude parameters , and ,


Blade-wakes can be neglected Ior the radial velocity component because oI their small
magnitude on the radial component. We use the special least-square procedure developed in
the Matlab environment in order to Iit this equations to the experimental data. The radial
velocity is obtained experimentally only by mean oI PIV technique so we are constraint to use
it. Fig.9 shows the variation oI vortex characteristic radial velocities with respect to .
B6
119
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno


Fig.9 Characteristic radial velocities versus discharge coefficient

Linear least squares Iits accurately represent and while Ior a parabolic Iit seems to be
quite satisIactory. is the characteristic angular velocity oI the centered vortex which
increase in intensity with the discharge.

b. COMPARISON OF ANALYTICAL FORMULATION WITH RADIAL
VELOCITY PROFILE MEASURED BY PIV TECHNIQUE
The 3D time-resolved PIV velocity measurements in a vertical plane under the runner provide
the radial, tangential and axial velocity components. For all radial, tangential and axial
components the mean velocity proIile is calculated on a complete diameter and compared
with the analytical velocity proIiles Ior OP1 to OP4 on the Fig.10. These mean velocity
proIiles are obtained by averaging 3000 instantaneous velocity Iields. In this case, it could be
noticed that analytical Iormulation Ior radial velocity component not Iit well the experimental
data on a single diameter contrary to spatially averaged data. Indeed, a strong radial velocity
component peak increasing with the discharge is Iound on the vortex center. In the case oI
spatially averaged mean values oI the previous paragraph this peak was Ilattened by the
averaging process because it is not centered on the geometry.


Fig.10 J
r
, J

and J
:
velocitv profiles Measurements vs. analvtical profiles OP2 and OP3

It is now clear that the spatially averaging process destroy the Ilow asymmetry making
possible a global agreement between spatially averaged experimental data and a Batchelor
vortices based analytical modeling. This kind oI approach can and was used to initialize
numerical simulations oI the whole draIt tube Ilow. However, the radial velocity Iield is
clearly non axisymmetric and it also evolves strongly Irom operating points OP1 to OP4. It is
120
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
important to note that the eIIect oI the accident itselI is then clearly identiIiable on the radial
velocity proIile even at the runner exit which is located as Iar as possible Irom the elbow and
the diIIuser.
5. CONCLUSION
Two experimental investigation including LDV and PIV measurements have been perIormed
in the conical diIIuser and the draIt tube oI a Francis turbine. The measurements were
obtained at six operating points near the 'best eIIiciency point. Particularly, 3D time-
resolved velocity Iields were obtained by mean oI PIV technique.
The LDV measurements permit to obtain analytical representation oI axial and tangential
velocity component oI the swirling Ilow under the runner by the superposition oI one solid
body rotation and two counter-rotating Batchelor vortices see Resiga and al. |4|. The blade
wakes were added at the Resiga`s Iormulation and give better agreement with experimental
data. Obviously this representation is axisymetric so it is based on velocity proIiles on a single
diameter. In Iact the Ilow under the runner is Iar to be axisymetric as was shown in this paper.
The 3D time-resolved PIV in vertical plane under the runner and the classic PIV in horizontal
plane under the runner permit us to accede to measurements oI the radial velocity component.
Like the two others velocity components, the radial velocity is non-axisymetric. Anyway we
propose an analytical representation based on the Iormulation Ior the conical diIIuser and the
three Batchelor vortices superposition. The new Iormulation Ior radial velocity component Iit
well the experimental data Ior radial velocity proIile on a single diameter.
Our measurements and analysis leads to the conclusion that we made strong approximation by
initializing calculations with axisymetrical velocity proIiles.
6. ACKNOWLEDGEMENTS
The authors would like to acknowledge ADEME, ALSTOM and DGR Ior their Iinancial
support.

7. REFERENCES
|1| Duprat C., Balarac G., Metais O., and Laverne T., Simulation des Grandes Echelles d`un
Aspirateur de Centrale Hydraulique. Mecanique & Industries, 2009
|2| Ciocan G.D., and Avellan F., 2004, Flow Investigation in a Francis DraIt Tube: Advanced
Experimental Methods, Proceedings oI the 3rd ConIerence oI Romanian Hydropower
Engineers, Bucharest
|3| Mauri S., Kueny J-L and Avellan F., November 2004, Werle-Legendre Separation in a
Hydraulic Machine DraIt Tube, Journal oI Fluids Engineering Vol. 126, pp. 976-980
|4| Tridon S., Barre S., Tomas L., and Ciocan G. D., 3D Time-Resolved PIV Measurement in
a Francis Turbine DraIt Tube, October 2008, Proceedings oI the 24th Symposium on
Hydraulic Machinery and Systems
B6
121
AHR WG Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, Brno
|5| Susan-Resiga R., Ciocan G.D., Anton I. and Avellan F., January 2006, Analysis oI the
Swirling Flow Downstream a Francis Turbine Runner, Journal oI Fluids Engineering, Vol.
128, pp. 177-189
|6| Ciocan G.D., Avellan F., and Kueny J.-L., 2000, Optical Measurement Techniques Ior
Experimental Analysis oI Hydraulic Turbines Rotor-Stator Interaction, Proceedings oI the
Asme 2000 Fluids Engineering Division Summer Meeting, Boston, MA.
|7| Iliescu M.S., Ciocan G.D., and Avellan F., 2002, 3d PIV and LDV Measurements at the
Outlet oI a Francis Turbine DraIt Tube, Joint U.S. Asme - European Fluids Engineering
Summer ConIerence, Montreal, Quebec, Canada
|8| Ciocan G.D., Iliescu M.S., Vu T., Nennemann B. and Avellan F., February 2007,
Experimental Study and Unsteady Simulation oI the Flindt DraIt Tube Rotating Vortex Rope,
Journal oI Fluid Engineering, Vol. 129, pp. 146-158
|9| Faler, J. H., and Leibovich, S., 1977, Disrupted States oI Vortex Flow and Vortex
Breakdown, Physics oI Fluids, 20(9), pp. 13851400.
8. NOMENCLATURE
5
UHI
(m) runner radius
5
VHF
(m) measuring section radius

(rad/s) runner angular speed
4 (m3/s) turbine discharge
U (kg/m3) Iluid density

() eIIiciency
9
U
(m/s) radial velocity component
9

(m/s) tangential velocity component

9
]
(m/s) axial velocity component
() pressure recovery coeIIicient
$
UHI
(m) static pressure reIerence section area
122