3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
Experimental Investigation of Draft Tube Flow of an Axial Turbine
by Laser Doppler Velocimetry
Philippe GOUIN
Laboratory of Hydraulic Machinery, Laval University, Quebec, Canada
Claire DESCHENES
Laboratory of Hydraulic Machinery, Laval University, Quebec, Canada
Monica ILIESCU
Laboratory of Hydraulic Machinery, Laval University, Quebec, Canada
Gabriel Dan CIOCAN
Laboratory of Hydraulic Machinery, Laval University, Quebec, Canada
ABSTRACT
This paper presents the current results of an experimental study of the draft tube flow in a propeller turbine by Laser
Doppler Velocimetry. Aiming to assess the impact of radial velocity on flow behaviours, measurement campaigns
were performed in the draft tube to acquire two velocity components at the outlet and three components at the inlet.
A characteristic setup for measuring the three components of the velocity field is presented. Details regarding the
positioning system and rotating optical access design are given and the uncertainty issues are discussed. Time
averaged velocity profiles are presented for both inlet and outlet sections of the draft tube and correlations are
established between them.
KEYWORDS
Axial hydraulic turbine, draft tube flow, LDV, radial velocity, velocity profiles
1. INTRODUCTION
Draft tube flow is studied to a great extent in hydraulic turbines as this part plays a crucial role in the
energy recovery of the machines. Incoming flow to the draft tube induces unstable operation and
diminution of its energy recovery, which has a considerable effect on the efficiency of low head turbines.
Many numerical studies are thus dealing with these complex flows while fewer studies aim to address the
lack of good validation data, both facing interesting challenges. While the radial velocity is often put aside,
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
this paper intends to present briefly a method to acquire three velocity components at draft tube inlet and to
focus on measurement results in terms of the impact of velocity distribution at the draft tube inlet on the
outlet velocity profiles. Other experimental studies involving 3D-LDV measurements were performed
notably by Ciocan et al. [4].
In the scope of this paper, the experimental study of draft tube flow involves measurements on an axial
turbine model. The propeller turbine model is characterised by 24 stay vanes and 24 guide vanes, a runner
with six fixed blades, an intake and an asymmetric draft tube both with a central pier. It is also defined by a
specific speed of 0.424. The runner blades being fixed, the runner does not guarantee optimal flow
downstream except for best efficiency point. Moreover, the asymmetry triggered in the semi-spiral casing
is conveyed downstream in the draft tube whose optimisation gets laborious. The draft tube investigated is
characterised by two channels decentred from runner axis which are subject to flow rate unbalance.
2. EXPERIMENTAL TOOLS
Three measurement campaigns were performed to acquire the data. Laser Doppler Velocimetry, which
measures the velocity of a fluid by detecting the frequency shift between a laser source and scattered light
from tracer particles, was used for its non-intrusiveness and reliability [3]. The LDV system used in the
study (Dantec Dynamics) allows simultaneous measurement of two velocity components. Silver coated
hollow glass spheres of 10 µm were used as seeding particles throughout the study. Tab.1. shows the
measured components and section for the three campaigns with some LDV parameters.
Laser
Wavelength
Beams diameter
Beams separation
Focal length
Meas. volume diam.
Meas. volume length
5.8W Argon-Krypton
488 / 514.5 nm
1 mm ± 0.02
39 mm ± 0.02
400 or 1000 mm
189 or 238 µm
3.97 or 6.33 mm
Campaign
# - Section
Measured
components
1 - Inlet
2 - Outlet
3 - Inlet
Cz, Cθ
Cx, Cz
Cz, Cr*
LDV
focal length
(mm)
400
400 or 1000
400
Samples
per spatial
point
60000
10000
60000
Tab.1 LDV characteristics and measurement campaigns
3. MEASUREMENT SECTIONS
For the outlet section, two series of measures were necessary to obtain the velocity field near the exit plane
in the two channels. On each of them, measurements were made on a grid of 10 x 23 points covering 95%
of the cross-section. The transversal velocity component was put aside as the model structure did not allow
acquiring it without major modifications. Fig.1 illustrates the measurement plane, located at 0.02·RRef
upstream the outlet, also pointing out the runner rotation direction and the global coordinate system. The
reference radius (RRef) corresponds to the runner radius. In order to acquire velocities on these grids, the
LDV probe is mounted on a traverse system with degrees of freedom in y and z directions. The laser
propagates through an optical access whose quality is crucial; both the parallelism and finish of the
window surfaces must meet optical quality standards and the alteration of the hydraulic profile minimized.
For the inlet section, the first measurement campaign gathered information on axial and tangential
velocities at the inlet section, on three diameters over three cross-sections of the conical diffuser, where
asymmetric profiles were obtained [2]. This campaign used a traverse system in translation only, aligned
with the runner axis. The radial velocity, more challenging to acquire, is subject to restrictive hypotheses in
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
many numerical studies although it was identified as a key component in draft tube simulations [1, 5]. It is
also part of the circumferential non-uniformities and contributes to dynamic phenomena taking place
downstream the runner. These issues motivated to push forward with the acquisition of this component in
the third measurement campaign.
Fig.1. Measurement sections and main turbine components
Considering the available measurement area of the system that will be presented afterwards, the available
data of the first campaign (see Tab.1 – right) and the interference of the system with the draft tube, the
measures where made on four radiuses with 41 points each, on one cross section of the conical diffuser
bellow the runner. The angles of the radiuses with the positive x axis are 33.5°, 213.5°, 273.5° and 337.5°.
Fig.2 illustrates the position of the radiuses with respect to the reference coordinate system. This
configuration of the conical diffuser and the access window was used for the radiuses 2 and 3. In order to
acquire the radiuses 1 and 4, the conical diffuser was rotated of 120° counter-clockwise.
Fig.2. Rotating access measurements radius position and LDV configuration
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
4. METHODOLOGY
As the LDV system measures two components perpendicular to the laser propagation direction, getting
directly the radial velocity component on different radial locations means moving the LDV probe parallel
to the measured radius. With the parallelism constraint on the optical access, this method would imply an
unacceptable modification of the hydraulic profile in the conical section concerned. The method proposed
in this paper is overcoming this problem by performing the measure through a fixed location rotating
optical access. For each spatial position measured, the optical access window angle varies and so the angle
of the absolute velocity components perceived. The probe is oriented so the axial component is measured
as validation with the first campaign. The other component acquired (Cr*) corresponds to a projection of
the r-θ velocity vector on planes proper to each position of the optical access. Successive measurement
campaigns then provide two dependent partial velocity vectors that can be combined through planar
rotation to compute Cr on each spatial position. The equations system includes Cr, Cθ in the reference
coordinate system, Cr* in the optical access coordinate system and the angle β between the two systems.
Fig.3 illustrates the coordinate system rotation in one point. One can observe that for the third campaign to
be sufficient to get Cr from this coordinate rotation, Cr* has to be in the r-θ plane. Otherwise, a third partial
velocity vector would be required to solve the equations system and no common velocity component (i.e.
Cz) would be available to validate the results.
Fig.3. Measured velocity components in the local reference frame of the translational (left) and rotational (right)
configuration
5. MECHANICAL DESIGN
One of the challenges of this method is to position the probe with respect to the optical access, and the
optical access with respect to the reference coordinate system. For all spatial positions, the probe has to be
kept perpendicular to the window surfaces at a distance fixed by the optical path of the LDV beams to the
measurement point. The realisation of such an experimental setup is constraining and sensitive to many
parameters. The design developed in this study will be presented briefly. In order to allow the rotation of
the window on the measurement plane while preventing leaks, a spherical joint similar to a ball valve was
conceived to be mounted on the acrylic conical diffuser of the model. A fixed frame insures sealing with
the conical diffuser and squeezes the joint in place by means of two teflon rings. The frame is made of two
threaded parts that allow adjusting the squeeze on the joint, which has to be sufficiently tight for sealing
and loose enough to allow rotation. A shouldered cylindrical glass window is inserted in the spherical joint
and pressed on an o-ring by a threaded tubular part. Fig.4 illustrates the rotating optical access assembly.
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
Dimensioning of this mechanism is subject to a compromise between the resulting measurement area and
the modification of the hydraulic profile. As first constraint, the diameter of the window is given by the
maximum laser beams separation. Also given a minimum distance between the teflon rings (related to the
diameter of the joint) to insure adequate joint sealing, the angular stroke of the joint is primarily function
of the sphere diameter. On the other hand, covering a wider angle also requires a greater modification of
the hydraulic profile. In the present context, a sphere diameter of 0.21·RRef procured a stroke of 40°.
Fig.4. Rotating optical access assembly
To position the optical access and the probe, a specific mechanism was also designed, based on the
traverse system used in the other experimental campaigns. Aiming to be precise and easy to align, the
system uses a linear actuator to guide another one in rotation around the access window rotation axis. Both
actuators are characterised by a precision of 0.02 mm, traduced in a 0.0014° angular precision for the
mobile motor position. The second linear actuator is supported by two circular motion guides and is fixed
to the threaded tubular part showed on the left of Fig.4. This mechanism allows the probe to be moved
perpendicularly to the window and both the window and the probe to be rotated for each measurement
point. Fig.5 shows a picture of this system used in the third experimental campaign.
Fig.5. LDV experimental setup for measurement of the radial velocity component at the runner outlet
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
6. LIMITATIONS AND UNCERTAINTY
The method presented so far is subject to some specific issues. First, the alignment process of the LDV
system results in an uncertainty on the spatial position of each measurement point. In the case of the
rotating optical access calibration, the position of the system has to be precisely adjusted regarding more
degrees of freedom than usual. Misaligned laser beams can result in phase shift, spatial shift and crosscontamination of velocity components [3]. The fact that two measurement campaigns were made, and
therefore two alignment processes, implies that this uncertainty is increased. Simultaneous acquisition of
the two vectors by means of a three components LDV system would allow verifying the coincidence of the
laser beams on the same spatial point thus reducing the number of necessary alignments to one, but would
also augment the relative uncertainty because of the low amplitude of the third component [4]. On the
other hand, common 3D-LDV systems, unless using two 2D probes and four laser wavelengths, allow
measurement of three components without the possibility to validate each of them directly. Performing two
2D campaigns allowed to validate the axial velocity and to insure good orientation between the measured
components. Furthermore, the principal uncertainty issue in computing Cr is related to the angle β used in
the coordinate system rotation from the optical access reference to the global reference. One can note that
for a specific measurement point where β is 0°, the radial velocity Cr is obtained directly from Cr*. At the
opposite, a value of β close to 90° leads to a value of Cr* similar to Cθ without much information on Cr.
Increasing the angle β results in progressively amplifying the uncertainty on the radial velocity computed;
Tab.2 gives the resulting radial velocity error for an initial error of 1% on measured velocities (Cr* and Cθ).
The impact of these issues tends to increase in low velocity or recirculation regions, for which the relative
error inherent to the LDV system increases significantly. Also, operating conditions such as flow rate, head
and temperature are subject to fluctuations. The actual conditions from one campaign to the other are then
affected by variations of less than 1% on the model efficiency and the unit speed (n11) value.
β (°)
Relative error (%)
0
1.00
20
1.42
40
2.14
60
3.73
80
11.43
Tab.2. Effect of angle β on the radial velocity accuracy
7. RESULTS
In the scope of this study, nine operating regimes are covered by each experimental campaign, combining
three different net heads and seven wicket gate openings. In the context of this paper, the operating
conditions presented are listed in Tab.3. Time averaged data will be presented for all spatial points of the
inlet and outlet sections. Lengths and velocities are normalized with respect to the runner reference radius
RRef, and the mean velocity in the reference section CRef, respectively.
#
8
3*
4
5
6
Q11
(m3/s)
0.7905
1.2945
1.3862
1.2762
1.3155
n11
(RPM)
124
124
124
120
130
Guide vanes
opening
17°
33°
38°
33°
33°
Relative
efficiency
62.8 %
100.0 %
99.1 %
100.1 %
97.6 %
Fig. Id.
17°
33°Hnom
38°
33°Hmax
33°Hmin
Tab.3. Operating conditions analysed (* nominal point)
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
The effect of varying the wicket gates opening at constant head on the radial and axial velocity profiles is
shown in Fig.6. The 33° opening at nominal head, best efficiency point, leads to relatively uniform profiles
with a narrow stalled region bellow the hub, as one can observe on left side of Fig.6. A minor velocity
deficit is notable over radius 2 (θ = 213.5°) and attests of a slight flow unbalance towards positive x axis. A
similar tendency can be observed in radial velocity profiles, as radius 2 presents a wider negative velocity
region in the center of the section. One could interpret this tendency as the combined effect of the upstream
velocity non-uniformities and the draft tube elbow. Fig.6 (right side) illustrates radial velocity profiles for
radiuses 2 and 4 in nominal (33°Hnom) and overload (38°) operating conditions. The radial asymmetry is
found for both conditions, with greater velocity in the overload case.
Fig.6. At constant head: (Left) Axial velocity profiles, 33° opening at nominal head. (Right) Radial velocity profiles,
radii 2 and 4, at best efficiency point and overload, at nominal head
Fig.7 illustrates axial and circumferential velocities measured over radius 2, for the first three operating
conditions. Closer to the design hypothesis of pure axial velocity, the nominal condition profiles are
generally in between the two others. At partial load, a major recirculation region (up to R/RRef = 0.53) is
clearly visible in the center of the section, deflecting the flow towards the outer wall where high axial and
tangential velocities are found. Best efficiency point exhibits low tangential velocities and no recirculation.
Overload condition also shows a uniform axial profile but a major counter-rotating region in the center,
associated with a vortex downstream the runner. This structure is also visible on Fig.6 (right) with the
increased modulus of the radial velocity in this region. The opening of the overload conditions being closer
to the nominal 33° than the partial load 17° opening, the efficiency at overload is not greatly diminished.
The impact of these inlet flow behaviours can be investigated through the axial velocity contours at the
outlet section presented in Fig.8. The blurred contour lines identify zero velocity boundaries; the header of
each graph displays the operating condition and the integrated flow rate over the two measured areas
referenced to the operating condition. Similarly to the inlet section and although a flow rate unbalance is
measured, the best efficiency operating condition (center) leads to quite uniform velocity distributions on
both draft tube channels. Overload condition (top-left) exhibits a slightly more balanced flow repartition
between the channels but also a low velocity region in the center of the left channel.
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
Fig.7. Axial and tangential velocity profiles at the draft tube inlet for constant head, radius 2
Fig.8. Axial velocity contours at the draft tube outlet. Header: Integrated flow rate over each measured area
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
The presence of such structure implies higher velocity gradients, higher turbulence thus lower draft tube
energy recovery which can contribute to lower the efficiency of the whole machine (see Tab.3). Partial
load condition (bottom-left) results in even more irregular distributions with one major recirculation zone
near the pier in both channels. One can presume that this recirculation and the resulting flow unbalance
arise from the interaction between the central recirculation at the inlet section, the draft tube elbow
and the pier.
Now considering a fixed wicket gate opening of 33°, the effect of varying the unit speed (n11) can be
analysed through operating conditions 5, 3* and 6. As illustrated by Fig.9, similar profiles are obtained for
the three operating conditions, with variations in amplitude. Also implying a slight flow rate augmentation,
the increase of the runner rotating speed with respect to the net head is traduced at the draft tube inlet
section by an augmentation of the tangential velocity and the radial velocity (decreasing the flow tendency
to converge towards the center). This variation also affects the axial velocity, reducing in the central half of
the section and increasing near the outer wall.
As for operating conditions previously analyzed, asymmetry is found in radial velocity profiles of Fig.9
(right). A wider negative velocity region in the center of the section is still present over radius 2. Also, the
decrease in tangential velocity associated with lower unit speed (n11) leads to the apparition of a narrow
counter rotating region close to R/RRef = 0.23. Fig.8 (top-right, center and bottom-right) presents the axial
velocity contours at the outlet section for the constant wicket gate opening analysis. Witnessing the
similarity and uniformity of the nominal (33°Hnom) and maximum head (33°Hmax) velocity profiles, what
one can note is the flow rate unbalance of the minimum head operating condition (33°Hmin). The slightly
higher flow rate and the outward tendency of the inlet flow of this operating condition could trigger such
unbalance that develops in the draft tube elbow and at the pier. The resulting higher velocity region in the
bottom right corner of the right channel, the stalled region in the left channel and the high velocity
gradients associated with them contribute to lower the draft tube energy recovery which possibly lower the
model efficiency (see Tab.3).
Fig.9. At constant opening: (Left) Axial and tangential velocity profiles, radius 2. (Right) Radial velocity profiles,
radii 2 and 4.
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
8. CONCLUSION
Current results of the study of the draft tube flow in a propeller turbine by Laser Doppler Velocimetry
were presented in this paper. A method to acquire three velocity components at draft tube inlet using a
rotating optical access has been presented, the radial velocity being obtained by the combination of two
partial velocity vectors. It is found that the method is subject to specific uncertainty issues regarding the
LDV system calibration and the radial velocity computation.
Time averaged results were presented at draft tube inlet for three velocity components and at draft tube
outlet for two components. Results at the outlet section show varied distributions, from clearly asymmetric
with high velocity gradients to best efficiency point uniform profiles. Inlet velocity profiles asymmetry and
diversity have been witnessed. Recirculation in the central region, secondary flows, counter rotating
regions and converging radial velocity in the central half of the section were identified. Future analysis of
the radial velocity component at draft tube inlet is expected to give more details on the impact of this
component and of asymmetry on numerical simulations of the draft tube flow.
9. ACKNOWLEDGEMENTS
The authors would like to thank the participants of the Consortium on Hydraulic Machines for their
support and contribution to this research project: Alstom Hydro Canada, Andritz Hydro LTD, Edelca,
Hydro-Quebec, Laval University, NRCan, Voith Hydro Inc. Our gratitude goes as well to the Canadian
Natural Sciences and Engineering Research Council who provided funding for this research.
10. REFERENCES
[1]
Payette, F.-A., De Henau, V., Dumas, G., Sabourin, M.: Sensitivity of draft tube flow predictions to
boundary conditions. IAHR - 24th Symposium on Hydraulic Machinery and Systems. Foz de Iguazu,
Brasil, 2008
[2]
Gagnon, J.-M., Deschênes, C., Ciocan, G.D., Iliescu, M.: Numerical simulation and experimental
investigation of the flow in an axial turbine. IAHR - 24th Symposium on Hydraulic Machinery and
Systems. Foz de Iguazu, Brasil, 2008
[3]
Tropea C., Yarin A., Foss J., (Eds.): Springer Handbook of Fluid Mechanics. 2007. pp. 296-309
[4]
Ciocan G.D., Avellan F., Kueny J.-L.: Optical Measurement Techniques for Experimental Analysis
of Hydraulic Turbines Rotor - Stator Interaction. Proceedings of the ASME Fluids Engineering
Division Summer Meeting, Boston, Massachusetts, USA, June 11-15, 2000
[5]
Page M., Giroux A.-M.: Turbulent Flow Computations in Turbine 99 Draft Tube with CFXTASCflow, FIDAP and FINE/Turbo. Turbine 99 – Workshop 2. The second ERCOFTAC Workshop
on Draft Tube Flow. Vattenfall Utveckling AB, Älvkarleby, Sweden, June 17-20, 2001
11. NOMENCLATURE
Cx, Cy, Cz (m/s)
Velocity components
LDV
Laser Doppler Velocimetry
3rd IAHR International Meeting of the Workgroup on Cavitation and Dynamic Problems in Hydraulic
Machinery and Systems, October 14-16, 2009, Brno, Czech Republic
in the global reference frame;
Cr, Cθ, Cz (m/s)
Radial, tangential and axial
velocity components;
Cr*
(m/s)
Radial velocity in the local
coordinate system of the
rotating optical access;
Cref
(m/s)
Mean velocity in the reference
section of the machine
Hmin, Hnom, Hmax(m) Minimum, nominal and maximum
head of the operating conditions
Radial position with respect to
the runner axis
Rref
(m)
Runner reference radius;
x, y, z (m) Cartesian position relatively to
the reference coordinate system;
β
(°)
Angle between rotating access
and reference coordinate systems
θ
(rad) Circumferential position with
respect to positive x axis
n11
(rpm) Unit speed
R
(m)