Preliminary performance assessment from towing tank testing of a horizontal-axis turbine

Abstract

Following the Norwegian government’s goal to be a ‘low emission society’ by 2050, there is a significant push to develop offshore renewable energy; principally offshore wind, but also wave and tidal stream. The MarinLab towing tank at the Western Norway University of Applied Sciences, Norway is 50 m long with a 3.0×2.2 m section and was inaugurated in 2016 as an educational and research facility, supporting local industry to achieve a low-carbon transition. To advance knowledge of aero- or hydrodynamic interaction of future turbines for both wind and tidal stream energy, a model scale test-bed horizontal axis turbine has been developed. This turbine will enable testing of rotor diameters typically in the range of 500-600 mm. The turbine is instrumented with a torque-thrust sensor of 5 Nm/100 N capacity, custom-manufactured by Marin in the Netherlands. The turbine is speed regulated via a 4096 counts angular encoder connected to a 200 W Maxon EC-i brushless motor. Before testing new rotor geometries, a benchmarking study is being undertaken with an existing known geometry. The benchmark rotor has diameter, D=700 mm and employs the same NACA 63418 airfoil as Mycek et al. (2014), allowing comparison. Unlike Mycek et al. (2014), the nacelle has a nominal diameter of 90 mm and length 760 mm. The size of the rotor risks exceeding the capacity of the torque sensor, such that maximum tow-speed is limited to 0.8 m/s. The benchmark blades are machined from solid aluminium in a four-axis CNC milling machine, following a method similar to Payne et al. (2017). Due to machining limitations, the minimum trailing edge thickness is specified as 0.2 mm. The final surface finish is achieved by manual polishing, and the final dimensions, quantified using a Hexagon ROMER Absolute laser scanner, are found to be within ±0.2 mm accuracy. Future blade sets will typically have a smaller diameter, allowing for 3D printing and consistency of surface finish. A blade-element momentum model has been run with Prandtl tip and hub losses and 0° pitch (Fig. I). Lift and drag coefficients were calculated for a range of chord Reynolds numbers using XFOIL, for angles of attack between -5 to 16° and extrapolated to ±180° with Viterna’s method. Due to difficulties in estimating laminar-turbulent transition around critical Reynolds numbers, there is uncertainty in the XFOIL coefficients. At peak TSR, the chord-based Reynolds number is approximately 280k at 3/4 span. Curves for both Rec =250k (solid) and 500k (dashed) with Ncrit = 9 are presented, showing a significant drop in performance for lower Reynolds numbers, due to a collapse in Cl/Cd. Initial tests were conducted in near- zero ambient turbulence, though passive turbulence grids are available for future testing. Reasonable agreement on CT is observed when corrected for blockage according to Bahaj (2007), and further assessment will provide results on CP, tower loads and wake recovery.

Full text

PROCEEDINGS OF THE 15TH EUROPEAN WAVE AND TIDAL ENERGY CONFERENCE, 3–7 SEPTEMBER 2023, BILBAO 545–1
Preliminary performance assessment from
towing tank tests of a horizontal-axis turbine
David Lande-Sudall, Sondre Tolleifsen, Kjetil Gravelsæter, Harald Moen and Jan Bartl
Abstract—Accurate and reliable experimental test data
remains a vital part of prototype development for new tidal
stream and offshore wind turbines, whereby the numerical
models built for verifying device performance and loading
need to be validated thoroughly. A test-bed turbine has
been developed for this purpose, for use at the MarinLab
towing tank in Bergen, Norway. In this paper, a standard
rotor geometry with diameter 0.7 m has been used for initial
benchmark testing against existing experimental data from
similar hydrodynamic test facilities. The rotor blades are
3D printed to enable rapid prototyping of future optimised
blade geometries. Inflow speeds at the rotor plane were
varied between U∞=0.4−0.8m/s, in both zero and 5.5%
ambient turbulence intensity. Good agreement on power
and thrust coefficients were found across a tip-speed ratio
(TSR) range between 0 to 5. Above TSR =5, the turbine
gave higher output than existing test data, however, both
the blockage ratio and turbulence intensity was higher in
this study. Large torque oscillations were observed between
the stall and design TSR regions, though in future this
behaviour can possibly be dampened by improving the
regulation tuning of the dynamometer controller.
Index Terms—Experimental tidal turbine, Performance
testing, Towing tank, Hydrodynamic test facility.
I. INTRODUCTION
DESPITE numerous limitations in testing at re-
duced scale, laboratory testing remains an es-
sential step in device design; informing numerical
modelling methods and their associated design coeffi-
cients which can be tuned and validated against mea-
surements from a controlled laboratory environment.
Whether for wind turbine design [1] or tidal stream
turbine design [2]–[5], water tank testing has provided
valuable data on rotor loading [6] and [7], device wake
recovery [8] and [9], and farm/array interactions [10]
and [11]. New experimental apparatus nevertheless
requires characterising, to assess the facility’s accuracy
and determine the levels of uncertainty associated with
results. No two facilities are identical, with testing
procedures differing between them, thus there has been
a push to standardise the methodologies used [12], [13].
In order to quantify the effect of hydrodynamic test-
ing facility on model-scale tidal turbine performance,
the authors in [14] conducted identical measurements
© 2023 European Wave and Tidal Energy Conference. This paper
has been subjected to single-blind peer review.
This work was supported in part by the Research Council of
Norway under grant No. 324388
All authors are with the Western Norway University of Applied
Sciences, Inndalsveien 28, Bergen 5063, Norway (e-mail: dla@hvl.no).
S. Tolleifsen is also at the University of Bergen, Department of
Physics and Technology, All´
egaten 55, 5020 Bergen .
Digital Object Identifier:
https://doi.org/10.36688/ewtec-2023-545
using a standard protocol at three facilities; CNR-
INSEAN, IFREMER and the Kelvin Hydrodynamics
Laboratory (KHL). Here, differences in time-varying
thrust and power were believed to arise from dif-
ferences in turbulence inflow between recirculating
flumes and carriage vibrations in towing tanks. Sim-
ilarly, [15] and [16] conducted ‘Blind test’ comparisons
for wind turbine aerodynamic performance and wakes.
These datasets have been invaluable for informing in-
dustry comparisons of numerical modelling [17], [18].
The MarinLab towing tank at the Western Norway
University of Applied Sciences was inaugurated in
2016, and has since been used for a range of model
testing of floating wind turbines [19], installation ves-
sels [20], and multi-rotor wind turbine systems [21].
Recently, a horizontal-axis turbine test-bed has been
designed for use in the facility. In order to assess the
suitability of the device and facility for conducting
performance tests of future novel rotor geometries, the
test-bed platform has been tested with the same rotor
geometry as [11], [14], [22], with and without ambient
turbulence inflow. In this paper, the results of these
initial tests are presented, firstly with a summary of the
device design and instrumentation, before introducing
the experimental methodology, and subsequently re-
sults on tower drag, rotor performance, torque and
structural response.
II. DESIGN AND MANUFACTURE
A test-bed hydrokinetic turbine has been designed
and manufactured in-house for use in the MarinLab
towing-tank at the Western Norway University of Ap-
plied Sciences, Bergen. This facility is a 50 m long tank,
with 3.0 ×2.2 m section and is fitted with two towing
carriages (master and slave), capable of towing with
a maximum velocity of 5.0 m/s and acceleration at
1.2 m/s2, with 0.001 m accuracy in position. In order
to keep blockage effects to a minimum (<5%), the
test-bed dynamometer is designed for use with rotor
diameters, D= 2R, typically in the range 0.5-0.6 m,
for a maximum tip-speed ratio, T SR (Eq.(1)), around
10 at U∞= 1 m/s inflow velocity. These parameters
define the design envelope for the turbine drive-train,
with maximum foreseeable thrust, Tand torque, Q, in
the region of 120 N and 4 Nm, respectively.
T SR =ωR
U∞
(1)
In Equation (1), ωis the rotational speed of the rotor
in rad/s.
The test-bed has been heavily inspired from the
design and personal discussions with the authors in

545–2 PROCEEDINGS OF THE 15TH EUROPEAN WAVE AND TIDAL ENERGY CONFERENCE, 3–7 SEPTEMBER 2023, BILBAO
[23]. In their study, the turbine was 1.2 m diameter
and bed-supported, with transducers measuring thrust
and torque on the rotor and streamwise root bending
moment on each blade. However, due to the smaller
rotor diameter and hence smaller torque requirements
in the MarinLab, a narrower nacelle and corresponding
motor/gearbox diameter has been incorporated. The
dynamometer consists of a 200 W Maxon EC-i brush-
less motor (No. 634043) with 12:1 planetary gearbox
(No. 223083) providing 15 Nm maximum continuous
torque and an EPOS4 50/15 positioning controller (No.
504383) with 4096 step digital encoder (No. 575827).
A torque-thrust (QT) transducer of 5 Nm and 100 N
capacities, with similar design to [24] has been custom-
made by Marin, Netherlands. A 480 W Mean Well 48V
DC power supply (HEP-480-48A) is used for powering
the dynamometer, with any excess power generated
being absorbed by a 50 Ωresistor with 200 W rating
(RS200). After two design iterations of the aluminium
nacelle housing, the final outer diameter around the
central tower-nacelle connection boss is 96 mm, with
an overall length of 760 mm. Prior to assembly with
the dynamometer electronics, a water-tightness test
conducted under a compressed air supply, found a
leak at the top of the tower-nacelle boss and required
a new O-ring groove to be machined. A humidity
sensor (DHT11) was subsequently added at the tower-
nacelle interface, connected to LabView via an Ardunio
microprocessor to provide a warning alarm should the
humidity level exceed 60%, indicating possible water
ingress.
A. Rotor design
In order to benchmark the test-bed, the well-
documented rotor geometry from [11], [14], [22] with
diameter, D= 0.7m and NACA 633-418 airfoil section
is chosen, despite being larger than the diameter range
of the design envelope. In order to not exceed the
torque capacity from the QT-sensor, the inflow speed is
limited to max{U∞}= 0.8m/s. The foil co-ordinates
from [25] are shown in Figure 1 and the chord and
twist distribution is identical to those given in [11]. The
overall intended design parameters are summarised
against those implemented in this benchmark study in
Table I.
Fig. 1: Airfoil co-ordinates normalised by chord-length, cfor the
NACA 633-418 profile from [25].
To enable rapid prototyping of future model-scale
blades, such as those optimised for particular perfor-
mance requirements e.g. [26], a FlashForge Adventurer
4 [27], 3D printer has been used with ABS plastic at
100% infill to minimise water ingress. The available
TABLE I: Intended design parameters for the test-bed turbine, versus
those tested in this benchmark study.
Symbol Parameter Design range Benchmark
U∞Inflow speed 0-1 m/s 0.4-0.8 m/s
dTower diameter 50 mm
DRotor diameter 500-600 mm 700 mm
Dnac Nacelle diameter 96 mm
lnac Nacelle length 760 mm
QTorque rating (sensitivity) 5 Nm (0.006 Nm/V)
TThrust rating (sensitivity) 100 N (0.23 N/V)
print volume is limited to 220 ×200 ×250 mm whilst
the total blade length is 296 mm, so the inner third
of the blade (root) is printed separately to the outer
two-thirds of the blade (tip). The two sections are
joined with universal water-resistant glue and two
M5 threaded rods, to prevent excessive torsion and
bending of the blade tip. The minimum trailing edge
thickness was defined to 0.2 mm. The three blades took
approximately 36 hours to print and were lightly hand-
sanded with 600 g/m2paper to remove minor surface
defects at the leading edge (see inset to Figure 2).
The complete assembled turbine is shown in the main
image to Figure 2.
Fig. 2: Complete turbine assembly, mounted with machined alu-
minium blades and showing load cell attachment to the top of
the tower. The image inset shows the 3D printed ABS blades with
NACA633-418 foil section used in this study.

LANDE-SUDALL et al.: PRELIMINARY PERFORMANCE ASSESSMENT FROM TOWING TANK TESTS OF A HORIZONTAL-AXIS TURBINE 545–3
2.20 m
0.70 m
x=0 m~42 m50 m
Beach
Towing carriage
Turbine attachment
Passive turbulence
grid
Load cell
QT sensor
2.00 m
Fig. 3: Schematic diagram (not to scale) of the 50 m long and 3 m wide MarinLab towing tank, showing the turbine attached to the master
towing carriage, placement of the tower load cell, torque-thrust transducer and passive turbulence grid.
B. Dynamometer control
Torque and thrust signals from the QT-sensor are
passed via a 56 mm outer-diameter, 12-wire Senring
H1256-12S slip ring, to a National Instruments (NI)
9237 AI Bridge module housed on a NI cDAQ-9184
data acquisition system and recorded into NI LabView
at 2000 Hz. The encoder output signal is connected via
RS232 communication, which has a minimum refresh
rate of approximately 10 ms. Due to further processing
time of the encoder signal inside LabView, the encoder
is sampled four times with 80 ms intervals, inside the
first half of a 0.5 s timed-loop, in order to obtain the
time-averaged rotational speed over each 0.5 s interval.
Whilst this limits the time-resolution of the rotational
speed, this is only a limitation on the collected signal
as opposed to the motor controller feedback rate itself,
which receives all 4096 pulses per revolution and is
controlled with a constant velocity controller. Future
improvements will include communication to the mo-
tor via a CAN bus interface to allow up to a 1 ms
refresh rate.
III. EXPERIMENTAL METHODOLOGY
The MarinLab towing tank at the Western Norway
University of Applied Sciences has a total length of
50.0 m, width of 3.0 m and depth of 2.2 m, with qui-
escent fluid of nominally 0% turbulence intensity, I.
The test-bed turbine is mounted to the towing car-
riage, as shown in Figure 3, with a HBM 10 kgf U9C
miniature tension-compression load cell measuring the
combined tower drag and rotor thrust. Unfortunately,
due to a wiring fault from the QT-sensor, only the
torque could be measured and the rotor-only thrust
was not available for this study. Thus all thrust mea-
surements referred to herein are from the load cell at
the top of the turbine tower.
A. Grid-generated turbulence
The turbine’s performance was tested with both zero
ambient turbulence and I= 5.5% turbulence intensity
generated by a passive grid. The grid was mounted at
the front of the towing carriage (see Figure 3). The grid
has a total submerged cross section of 1.5 m ×1.4 m,
covering about 32% of the towing tank’s cross sectional
area.
The main dimensions of the grid are sketched in
Figure 4(a). It consists of a 50 ×100 mm rectangular
stainless steel frame and an aluminium grid of 14 ver-
tical and 14 horizontal cylindrical tubes. The tubes have
a diameter of 20 mm and a centre-to-centre distance of
M= 0.1m, resulting in a total grid solidity of σ= 0.36.
The ambient velocity behind the grid is reduced with
respect to the towing velocity, while ambient turbu-
lence is created. The mean and turbulent flow be-
hind the grid was measured by traversing a Nortek
Vectrino+ Acoustic Doppler Velocimeter (ADV) [28] in
streamwise, transverse and vertical directions on an
automated traverse. The ADV measures the Doppler
frequency shift scattered back from seeding parti-
cles distributed in the tank and calculates the three-
dimensional velocity vector, (U, V , W ), at 200 Hz sam-
pling frequency from a small cylindrical measurement
volume of length 15 mm. More details about ADV flow
measurements in the MarinLab towing tank can be
found in [21].
The flow characteristics behind the grid were mea-
sured at a towing speed of U∞= 0.4m/s, with tur-
bulence intensity calculated according to Equation 2.
At the location of the turbine rotor, 2.0 m downstream
of the grid, the velocity was reduced by about 14%
from the towing speed, and found to vary by not more
than 2% over the cross sectional area covered by the
grid. As depicted in Figure 4(b), the velocity reduction

545–4 PROCEEDINGS OF THE 15TH EUROPEAN WAVE AND TIDAL ENERGY CONFERENCE, 3–7 SEPTEMBER 2023, BILBAO
Fig. 4: (a) Front view of the passive turbulence grid, and (b) Variation in turbulence intensity, I[%] (blue) and velocity reduction U∞/Utow
[-] (red) with downstream locations measured along the centreline of the turbulence grid.
tends to an asymptote from approximately 2 m to
10 m downstream. Along the downstream centreline
of the grid, the measured turbulence intensity decays
from over I0.4m = 12% at 0.4 m downstream, to about
I10m= 4% by 10 m. At the location of the turbine
rotor an ambient turbulence of I2m = 5.5% was created.
Additional measurements in crosswise and vertical
directions confirmed homogeneous turbulence levels
within ±0.5%, with the exception of a local increase
of +2% behind the frame holding the grid.
I=s1
3(u2+v2+w2)
U2+V2+W2(2)
In Equation 2, bar-notation represents the mean part,
and lower-case characters the fluctuating part.
B. Test-cases
A total of three inlet velocities at the rotor plane,
U∞=0.4, 0.6 and 0.8 m/s are tested, with and without
passive turbulence generation. Additionally, the tower-
only drag was measured for each towing speed in zero
ambient turbulence. All test-cases are summarised in
Table II.
TABLE II: Test-cases used in this study. Utow is the carriage towing
speed and U∞is the velocity observed at the rotor-plane without
the turbine present.
Utow, m/s U∞, m/s I, % Tower-only Rotor & tower
0.4 0.4 0 ✓ ✓
0.6 0.6 0 ✓ ✓
0.8 0.8 0 ✓ ✓
0.45 0.4 5.5 - ✓
0.68 0.6 5.5 - ✓
0.90 0.8 5.5 - ✓
Due to a finite tank length, all tests were specified
with a maximum run distance of 37 m, with accelera-
tion/deceleration of 0.3 m/s2, giving a total acquisition
duration varying between approximately 45-90 s. After
cropping the inertial ends from each run, the typical
sample duration was between 25-60 s. Due to time
limitations, repeat measurements were only conducted
in the regions of TSR 2.5-3.5 for all cases and for cases
where anomalous time-averages were first obtained.
Two repeat measurements were taken for every TSR
point in the U∞= 0.6m/s case with turbulence. All
time-averaged values presented here are an ensemble
average of all available repetitions.
IV. RESULTS
Results for the tower-drag are first presented, before
discussing performance curves, time-varying torque
and structural response.
A. Tower-drag
Since the thrust measurements from the QT-sensor
were not available, thrust could only be measured from
the tower load cell. To establish the contribution of
the tower drag towards the net thrust, the tower-only
drag (i.e. without nacelle) and tower-with-nacelle drag
were measured for the case of zero ambient turbulence
and no rotor, with towing speeds ranging from 0.2
to 1.0 m/s, as presented in Figure 5. Quadratic least-
squares fit lines to each data set show good agreement.
Compared to the cases with nacelle, the larger standard
deviations found at low speeds for the tower-only cases
are expected to be due to the overall lighter mass of the
tower-only being more sensitive to carriage vibrations.
For U∞= 0.8and 1.0 m/s, the standard deviation of
the tower-with-nacelle case grows, which was visibly
observed as the onset of some tower vibrations. These

LANDE-SUDALL et al.: PRELIMINARY PERFORMANCE ASSESSMENT FROM TOWING TANK TESTS OF A HORIZONTAL-AXIS TURBINE 545–5
vibrations were possibly vortex-induced, presumably
due to the extra mass of the nacelle providing a lower
natural frequency which then may approach the vortex
shedding frequency for the tower with nacelle. More
frequency analysis is considered for the cases with the
rotor in Section IV-D.
In dimensionless form, the mean drag coefficient for
the tower-only is CD,twr = 1.20 and tower with nacelle
is CD,twr+nacelle = 1.11, where the referenced projected
areas are that of the tower-only and tower-with-nacelle,
respectively. Since results in [14] are presented with
tower drag included, and due to the uncertainty in
methods of trying to isolate the rotor thrust, all further
results on thrust in this paper are presented with
tower-drag included.
Fig. 5: Time-averaged drag force for the tower-only (black) and
tower with nacelle (red) with zero ambient turbulence. Shaded areas
represent ±σstandard deviation, and solid lines are quadratic curves
least-squares fit to the data and through the (0,0) origin.
B. Rotor power coefficient, CPand thrust coefficient, CT
In this section, results of the rotor performance and
thrust coefficients, CPand CTare presented, where
we follow the standard definitions in Equation (3) and
Equation (4), respectively:
CP=8ωQ
ρπD2U3
∞
(3)
CT=8T
ρπD2U2
∞
(4)
where ρ=1000 kg/m3is taken as the density of fresh-
water.
Data from KHL and CNR-INSEAN towing tanks [14]
are used for comparison. The global blockage ratio of
rotor swept-area to tank cross-section is B= 5.8% in
MarinLab. This is larger than the 3.3% and 1.2% block-
age ratios of KHL and CNR-INSEAN, respectively.
Whilst various methods of blockage correction exist
[29], [30], we present the results without any correction
applied to allow direct comparison to non-corrected
results from [14] and to avoid any ambiguity of the
post-processing..
Figure 6 (a) shows the CPand CT(T SR)curves
without turbulence, compared to curves from KHL and
CNR-INSEAN towing tanks [14] with U∞= 0.8m/s.
For the MarinLab turbine, at U∞= 0.4m/s, there is a
significant reduction in both CPand CT. As reported in
[14], low-Reynolds number effects are likely generated,
with unstable laminar separation bubbles forming for
such low towing speeds [31], where the chord-based
Reynolds number at 0.7Ris only Rec∼50k, for
TSR=3.5. Such effects act to destroy the lift generated
by the foil and thus the local torque and power. For
U∞= 0.6m/s, this drop in CPis not observed and
very close agreement to [14] is found for T SR values
from 0 to 5. For T SR > 5, the MarinLab test-bed
generates higher values of CP. A similar result, but to a
greater extent, is also visible for the U∞= 0.8m/s case.
This could be due to possible local twisting of the 3D
printed blades, where even small angles (<−5◦) would
benefit CP, although this was not visually confirmed.
Alternatively, since the measurement points have not
been corrected for blockage, it could be a result of this,
where blockage would tend to have a greater effect at
higher T SRs. Lastly, the carriage speed was not mea-
sured directly, and whilst previous ADV measurements
in [21] have shown good agreement with the target
carriage speed, power is particularly sensitive to small
discrepancies.
In general, mean CTis slightly lower than for the
KHL and CNR-INSEAN results, however, the submer-
sion depth used in MarinLab was only 0.7 m compared
to 1.0 m in [14], which would account for much of this
discrepancy. 0.7 m depth was chosen here due to the
need to centralise the turbine behind the available tur-
bulence grid (Fig. 4). The standard deviation in thrust
coefficient is generally of a similar, low magnitude for
all towing speeds. However, between 2< T S R < 3the
standard deviation increases significantly. In Section
IV-D it is discussed if this oscillation is related to the
turbine’s structural response. Alternatively, this region
is in the transition from where much of the blade is
predominantly in stall to it then generating maximum
lift at slightly higher T SR. It may also be related to the
possibility of a critical Reynolds number being reached
for the local airfoil section, where the Reynolds number
is affected by both towing and rotational speeds. This
T SR region is thus very unstable and hence leads to a
large temporal oscillations in thrust. Further discussion
with regards to torque is found in Section IV-C.
In Figure 6(b) with turbulence inflow, the CP(T S R)
curves from U∞= 0.4to 0.8m/s collapse on top of
each other. Ambient turbulence will hinder the forma-
tion of laminar separation bubbles, hence improving
the performance for U∞= 0.4m/s compared to no
turbulence. This contrasts with the results in [11] for
U∞= 0.4m/s and I= 3%, where there was still
a significant drop in CP. However, in MarinLab the
turbulence intensity was greater, at I= 5.5% and [11]
also reported for I= 15% that the CPcurves increased
compared to I= 3% for U∞= 0.4m/s. In terms of
thrust, CTfor U∞= 0.8m/s is consistently lower for
T SR > 2.5, which may be an indicator of some coning
of the 3D printed rotor for high towing speeds.

545–6 PROCEEDINGS OF THE 15TH EUROPEAN WAVE AND TIDAL ENERGY CONFERENCE, 3–7 SEPTEMBER 2023, BILBAO
(a) (b)
Fig. 6: CPand CTversus tip-speed ratio, TSR, for (a) zero ambient turbulence and (b) 5.5% ambient turbulence intensity. Mean (◦) and
standard deviation ±σ(shaded areas) for U∞= 0.4m/s (blue), U∞= 0.6m/s (green) and U∞= 0.8m/s (red), along with reported values
from KHL ( ) and CNR-INSEAN ( ) at U∞= 0.8m/s from [14].
C. Rotor torque, Q
With the benchmark turbine diameter being 0.1 m
larger than the maximum 0.6 m intended for the QT-
sensor rating, it was necessary to pay close attention to
the torque magnitudes, especially for the highest tow-
speed with turbulence generation.
Figure 7 shows a sample time-series of Q(t)for
U∞= 0.6m/s and T SR = 2.5and 3.5, with the latter
representing the design T SR of the rotor. Evident, is
a large difference in mean rotor torque for T S R = 2.5
compared to 3.5, which is expected given the transition
from the stall-region of the T SR curve to optimal
performance. However, of more significance are the
large oscillations in torque for T SR = 2.5; cycling
between approximately −1Nm to +3.5 Nm. From vi-
sual inspection, this appeared as the rotor turning
before abruptly braking to nearly zero rotational speed,
before continuing to rotate again. The cause of this is
not immediately apparent. As mentioned, T SR = 2.5
is at the transition between the rotor operating pre-
dominantly in stall, to generating maximum power
output at T SR = 3.5. From a mechanical perspec-
tive, at U∞≥0.6m/s, T SR = 3.5is where the
turbine transitions from requiring a net power input,
to generating a net power output, which is passively
diverted into the electrical resistor. However, torque
oscillations were still noted for U∞= 0.4m/s, when
there was no net power output. Given that the torque
oscillations are well within the maximum continuous
torque rating of the motor gearbox, and within the
QT-sensor capacity, it seems to be a product of the
motor controller not being able to respond adequately
to the large fluctuations in torque which arise in this
transition region. Nevertheless, the mean CPand CT
values still match those of [14] well. Thus, as long as
the turbine is studied outside of this transition region,
it still provides the anticipated load characteristics.
Fig. 7: Sample time-series measurements of torque, Q(t)for TSR=2.5
(blue) and TSR=3.5 (red).
D. Structural response
Although the turbine attachment (Fig. 3) features
a large support bracket to increase bending stiffness,
clear vibrations of the tower were still observed for
some of the test-cases. To investigate this further we
follow Welch’s method [32], to obtain the Power Spec-
tral Density (PSD) for U∞= 0.6m/s at T SR = 2.5and
3.5 in Figure 8 using four windows, with 50% overlap
and a Hann window to remove any spectral artefacts
arising from having an otherwise discontinuous win-
dowed signal. The first blade passing frequency, 1P, is
identified for each T SR case. For T SR = 2.5, there is
a clear peak in the signal at 1.09 Hz, which is greater
than the expected 1P frequency (0.68 Hz). However,

LANDE-SUDALL et al.: PRELIMINARY PERFORMANCE ASSESSMENT FROM TOWING TANK TESTS OF A HORIZONTAL-AXIS TURBINE 545–7
Fig. 8: Power-spectral density of the tower thrust response, for
T SR = 2.5(blue) and T S R = 3.5(red). Showing the tower
structural eigenfrequency, f0and range of 1P (green shaded) and cor-
responding 3P (orange shaded) blade-passing frequencies between
1< T SR < 7.5.
as discussed for Figure 7 for this T SR case, the rotor
was oscillating between rotation and braked, resulting
in significant oscillations in torque which necessarily
impacts thrust. Given that during each oscillation the
rotor is momentarily braked, then when the turbine ro-
tates again, the motor controller requires the rotational
speed to be greater than the target 0.68 Hz in order to
maintain a steady average. Unfortunately due to time
limitations and a concern of overloading the QT-sensor
from the large inertia of the blades whilst mounted to
the hub, the motor’s PID controller was not tuned to
provide slower feedback, which could potentially help
dampen these severe oscillations.
In contrast, T SR = 3.5has a less significant peak
near the 1P frequency, which aligns with the visual
observations of minimal tower vibrations occurring for
this run, and indeed is indicative of all cases outside
of 2≤T SR ≤3. Figure 8 also highlights the ranges
expected for 1P and 3P frequencies for the tested T SR
range. In both T SR = 2.5and 3.5, the 3P frequency
is evident as the largest peak in the shaded (orange)
region. The natural frequency of the tower, f0is iden-
tified with a common peak in both signals at 7.81 Hz.
Given that this frequency is outside of the 1P and 3P
ranges, then the tower is a stiff-stiff construction and
will not be excited to resonance. However, should for
example heavier aluminium blades be used in future
with this rotor diameter, there is a risk that the tower
natural frequency will reduce, possibly to inside the 3P
operating range.
V. CONCLUSION AND FUTURE WORK
A test-bed horizontal-axis turbine has been designed
and tested at the Western Norway University of Ap-
plied Sciences, MarinLab towing tank in Bergen, Nor-
way. The turbine is instrumented with thrust and
torque sensors and an encoder to resolve turbine
power. A NACA 633-418 foil was used for the rotor
geometry to be directly comparable to tests in [11], [14],
[22] for benchmarking purposes. For tip-speed ratios,
T SR < 5.5, the turbine gives comparable mean perfor-
mance (CPand CT) to [14] with U∞= 0.4−0.8m/s.
However, large oscillations in torque around the stall
transition at T SR = 2.5were observed. Future tests
will investigate tuning the motor controller to provide
slower feedback and maintain better speed regulation
in this region. Also, low Reynolds number effects were
apparent for the case of U∞= 0.4m/s without
turbulence and so the minimum tow-speed for this
rotor should not be less than U∞= 0.6m/s in order
to avoid this. Use of a passive turbulence grid allowed
testing with turbulence intensity of 5.5%. Here it was
found that low-Reynolds number effects vanished at
U∞= 0.4m/s, and there was closer agreement on CP
for all towing speeds, whilst CTwas marginally lower
for U∞= 0.8m/s. A second turbulence grid available
in the lab, with 50 mm cylindrical tubes and centre-
centre spacing of M= 0.25 m can enable testing with
turbulence intensities of >6% in future.
A method of rapid prototyping the turbine blades us-
ing a 3D printer has proven effective, especially around
the design T SR rating, and future tests with Kriging-
optimised rotors [26], representing both model scale
tidal stream turbines and offshore wind turbines, are
planned. A LaVision Shake-the-Box [33] tomographic
particle tracking velocimetry system is also available in
MarinLab and future testing will make use of this for
providing time-resolved wake measurements down-
stream of the turbine, potentially helping to isolate
some sources of performance discrepancies observed
here, e.g. at T SR > 5, where there was uncertainty
on whether the blades were coning or if this was due
to blockage. Finally, the QT-sensor is equipped with
additional wiring through to the hub, to allow blade-
root bending to be measured in future, along with re-
establishing the hub thrust signal.
VI. ACKNOWLEDGEMENTS
The authors are thankful for constructive discussions
with Gr´
egory Payne and Robert Larsson on design
considerations for the turbine, Nafez Ardestani and
Gloria Stenfelt for lab assistance, and to BSc students
Bendik Weltzien, Hans Joakim Jakobsen and Nikolai
Arntzen for the CAD drawings of the final design iter-
ation. Thank you also to Carola Buness for performing
flow measurements behind the passive grid. This work
has been partly supported by the Research Council of
Norway, grant No. 324388).
REFERENCES
[1] V. L. Okulov, R. Mikkelsen, J. N. Sørensen, and I. V. Naumov,
“Power Properties of Two Interacting Wind Turbine Rotors,”
vol. 139, no. September, pp. 1–6, 2017.
[2] A. Bahaj, W. Batten, and G. McCann, “Experimental
verifications of numerical predictions for the hydrodynamic
performance of horizontal axis marine current turbines,”
Renewable Energy, vol. 32, no. 15, pp. 2479–2490, Dec. 2007.
[Online]. Available: https://linkinghub.elsevier.com/retrieve/
pii/S0960148107002996

545–8 PROCEEDINGS OF THE 15TH EUROPEAN WAVE AND TIDAL ENERGY CONFERENCE, 3–7 SEPTEMBER 2023, BILBAO
[3] L. Myers and A. Bahaj, “Near wake properties of horizontal axis
marine current turbines,” 8th European Wave and Tidal Energy
Conference, pp. 558–565, 2009.
[4] J. McNaughton, B. Cao, S. Ettema, F. Zilic De Arcos, C. Vogel,
and R. Willden, “Experimental testing of the performance
and interference effects of a cross-stream array of tidal
turbines,” in Developments in Renewable Energies Offshore,
1st ed., C. Guedes Soares, Ed. CRC Press, Oct. 2020, pp.
563–570. [Online]. Available: https://www.taylorfrancis.com/
books/9781000318739/chapters/10.1201/9781003134572-64
[5] G. S. Payne, T. Stallard, R. Martinez, and T. Bruce, “Variation
of loads on a three-bladed horizontal axis tidal turbine
with frequency and blade position,” Journal of Fluids
and Structures, vol. 83, pp. 156–170, Nov. 2018. [Online].
Available: https://www.sciencedirect.com/science/article/pii/
S0889974617306734
[6] T. Blackmore, L. E. Myers, and A. S. Bahaj, “Effects of turbulence
on tidal turbines: Implications to performance, blade loads, and
condition monitoring,” International Journal of Marine Energy,
vol. 14, pp. 1–26, 2016, publisher: Elsevier Ltd. [Online].
Available: http://dx.doi.org/10.1016/j.ijome.2016.04.017
[7] R. Martinez, B. Gaurier, S. Ordonez-Sanchez, J.-v. Facq,
G. Germain, C. Johnstone, I. Santic, F. Salvatore, T. Davey,
C. Old, and B. G. Sellar, “Tidal Energy Round Robin Tests:
A Comparison of Flow Measurements and Turbine Loading,”
Journal of Marine Science and Engineering, vol. 9, no. 4, p. 425,
2021. [Online]. Available: https://www.mdpi.com/2077-1312/
9/4/425
[8] T. Stallard, T. Feng, and P. Stansby, “Experimental study of
the mean wake of a tidal stream rotor in a shallow turbulent
flow,” Journal of Fluids and Structures, vol. 54, pp. 235–
246, Apr. 2014, publisher: Elsevier. [Online]. Available: http:
//linkinghub.elsevier.com/retrieve/pii/S0889974614002485
[9] D. Lande-Sudall, T. Stallard, and P. Stansby, “Experimental
Study of the Wakes due to Tidal Rotors and a Shared Cylindrical
Support,” in Proceedings of 12th European Wave and Tidal Energy
Conference, Cork, Ireland, 2017.
[10] T. Stallard, R. Collings, T. Feng, and J. Whelan, “Interactions
between tidal turbine wakes: experimental study of a group
of three-bladed rotors.” Philosophical transactions. Series A,
Mathematical, physical, and engineering sciences, vol. 371, no.
1985, p. 20120159, Feb. 2013. [Online]. Available: http:
//www.ncbi.nlm.nih.gov/pubmed/23319702
[11] P. Mycek, B. Gaurier, G. Germain, G. Pinon, and E. Rivoalen,
“Numerical and experimental study of the interaction
between two marine current turbines,” International Journal
of Marine Energy, vol. 1, pp. 70–83, Apr. 2013. [Online].
Available: https://www.sciencedirect.com/science/article/pii/
S2214166913000088
[12] D. Ingram, G. Smith, C. Bittencourt-Ferreira, and H. Smith,
“Protocols for the Equitable Assessment of Marine
Energy Converters,” University of Edinburgh, Tech. Rep.,
2011. [Online]. Available: https://www.pure.ed.ac.uk/ws/
portalfiles/portal/1726320/book1.pdf
[13] ITTC, “ITTC – Recommended Procedures and Guidelines:
Model Tests for Current Turbines 7.5-02 07-03.9,” ITTC, Tech.
Rep., 2021. [Online]. Available: https://www.ittc.info/media/
9749/75-02-07-039.pdf
[14] B. Gaurier, G. Germain, J. V. Facq, C. M. Johnstone, A. D.
Grant, and A. H. Day, “Tidal energy ” Round Robin ” tests
comparisons between towing tank and circulating tank results,”
International Journal of Marine Energy, vol. 12, no. December, pp.
87–109, 2015.
[15] J. Bartl and L. Sætran, “Blind test comparison of the
performance and wake flow between two in-line wind turbines
exposed to different turbulent inflow conditions,” Wind Energy
Science, vol. 2, no. 1, pp. 55–76, Feb. 2017. [Online]. Available:
https://wes.copernicus.org/articles/2/55/2017/
[16] F. M ¨
uhle, J. Schottler, J. Bartl, R. Futrzynski, S. Evans,
L. Bernini, P. Schito, M. Draper, A. Guggeri, E. Kleusberg,
D. S. Henningson, M. H¨
olling, J. Peinke, M. S. Adaramola,
and L. Sætran, “Blind test comparison on the wake behind a
yawed wind turbine,” Wind Energy Science, vol. 3, no. 2, pp.
883–903, Nov. 2018, publisher: Copernicus GmbH. [Online].
Available: https://wes.copernicus.org/articles/3/883/2018/
[17] F. Pierella, P.- Krogstad, and L. Sætran, “Blind Test 2 calculations
for two in-line model wind turbines where the downstream
turbine operates at various rotational speeds,” Renewable Energy,
vol. 70, pp. 62–77, Oct. 2014. [Online]. Available: https://www.
sciencedirect.com/science/article/pii/S0960148114001815
[18] P.- Krogstad and L. Sætran, “Wind turbine wake interactions;
results from blind tests,” Journal of Physics: Conference Series,
vol. 625, p. 012043, Jun. 2015. [Online]. Available: https:
//iopscience.iop.org/article/10.1088/1742-6596/625/1/012043
[19] A. M. T. Ripe, “An Assessment of Extreme Mooring
Loads for Floating Offshore Wind Structures Using
Conditional Waves,” Ph.D. dissertation, University of Bergen,
Department of Physics and Technology, Aug. 2022. [Online].
Available: https://bora.uib.no/bora-xmlui/bitstream/handle/
11250/3014612/HTEK399 final.pdf?sequence=1&isAllowed=y
[20] D. R. Lande-Sudall, T. S. Høyven, K. Herfjord, and T. C. Thues-
tad, “Wave-induced collision loads and moments between a
spar-buoy floating wind turbine and an installation vessel,”
Journal of Physics: Conference Series, vol. 1669, no. 1, 2020.
[21] J. Bartl, C. H. Aasnæs, J. R. Bjørnsen, G. Stenfelt, and
D. Lande-Sudall, “Lab-scale measurements of wind farm
blockage effects,” Journal of Physics: Conference Series, vol. 2362,
no. 1, p. 012004, Nov. 2022. [Online]. Available: https://
iopscience.iop.org/article/10.1088/1742-6596/2362/1/012004
[22] P. Mycek, B. Gaurier, G. Germain, G. Pinon, and E. Rivoalen,
“Experimental study of the turbulence intensity effects
on marine current turbines behaviour. Part I: One single
turbine,” Renewable Energy, vol. 66, pp. 729–746, 2014,
publisher: Elsevier Ltd ISBN: 0960-1481. [Online]. Available:
http://dx.doi.org/10.1016/j.renene.2013.12.036
[23] G. S. Payne, T. Stallard, and R. Martinez, “Design and manufac-
ture of a bed supported tidal turbine model for blade and shaft
load measurement in turbulent flow and waves,” Renewable
Energy, vol. 107, pp. 312–326, 2017.
[24] J. Dang, J. Brouwer, R. Bosman, and C. Pouw, “Quasi-Steady
Two-Quadrant Open Water Tests for the Wageningen Propeller
C - and D -Series,” Gothenburg, Sweden, Aug. 2012, pp. 26–31.
[25] I. Abbott, H., Theory of Wind Sections - Including a Summary of
Airfoil Data. New York: Dover Publications Inc., 1959.
[26] T. H. Hansen and F. M ¨
uhle, “Winglet optimization
for a model-scale wind turbine,” Wind Energy,
vol. 21, no. 8, pp. 634–649, 2018, eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/we.2183.
[Online]. Available: https://onlinelibrary.wiley.com/doi/abs/
10.1002/we.2183
[27] FLASHFORGE, “FLASHFORGE 3D Printer Adventurer 4 Series
User Guide,” 2021.
[28] Nortek AS, “Comprehensive Manual,” NORTEK AS, Tech. Rep.,
Nov. 2015.
[29] A. S. Bahaj, A. F. Molland, J. R. Chaplin, and W. M. J. Batten,
“Power and thrust measurements of marine current turbines
under various hydrodynamic flow conditions in a cavitation
tunnel and a towing tank,” Renewable Energy, vol. 32, pp. 407–
426, 2007.
[30] G. T. Houlsby and S. Draper, “Application of Linear Momentum
Actuator Disc Theory to Open Channel Flow by,” University of
Oxford, Tech. Rep. OUEL 2296/08, 2008.
[31] M. Drela, Flight vehicle aerodynamics. Cambridge (Mass.): MIT
press, 2014.
[32] P. D. Welch, “The Use of Fast Fourier Transform for the Esti-
mation of Power Spectra: A Method Based on Time Averaging
Over Short, Modified Periodograms,” IEEE Trans. Audio and
Electroacoust., vol. AU-15, pp. 70–73, 1967.
[33] Y. J. Jeon, “4D Flow Field Reconstruction From Particle
Tracks By Vic+ With Additional Constraints and Multigrid
Approximation,” BRISK Binary Robust Invariant Scalable
Keypoints, pp. 12–19, 2018, iSBN: 8610828378018. [Online].
Available: https://doi.org/10.3929/ethz- a-010025751