Effect of Turbulence on Savonius Rotor

Efficiency
Final Repoit
ME241FluidsLaboratory
Group7

Jeff Whalley
Matt Johnson
Brian MacMillin

4/29/2009

We tested a Savonius rotor, a common vertical axis wind turbine, in order to determine turbulent
performance properties. Previous research has show the Savonius rotor is less efficient than the
more common horizontal axis wind turbines in laminar flow, but there is little research in
turbulent flow. Several orientations of rotors were used to simulate the turbulence that would be
created from a field of generators. By constructing a 1/6 scale model and attaching it to a DC
motor, we were able to obtain voltage outputs along with RPM data. Using these outputs we
calculated a 20% to 55% decrease in power generation based on rotor orientation.

TableofContents
1.Introduction
..3
2.ModelConstraints
..4
3.Construction
..5
4.HysteresisIssues
..7
5.Results
..8
6.ExperimentalRotorFailure
..10
7.Conclusions
..10
8.References
..12
AppendixA:Figures
..13
AppendixB:OrientationofRotorsintheWindTunnel
..18

Introduction:
Previous research has concluded that the Savonius rotor and other vertical axis generators are not
as efficient as horizontal axis generators in laminar flow (Appx A, Fig 2) [1, 2]. Our goal is not
to prove this research incorrect, but rather to accept this fact, and look at Savonius rotors in
turbulent flow. Our theory is that Savonius rotors are actually more effective in turbulent flow
than horizontal axis rotors. Because turbulent flow is much more common than laminar flow in
many real world situations, we hope that by showing the effectiveness of the Savonius rotor in
turbulent flow, we can shed light on a possible real world implementation.

In a realistic setting, multiple rotors would be combined to create a field of rotors to maximize
power output. This field would create additional turbulence from the rotors themselves. In order
to simulate turbulence in the wind tunnel, we experimented with several different orientations of
rotors to explore how their own turbulence affected power outputs. By finding the most efficient
orientation, unnecessary losses in power generation can be avoided.

To help better understand how turbulence affects these rotors we measured voltage outputs from
a DC motor attached to the bottom of a designated test rotor. In addition revolution per minute
(RPM) data was collected using a digital bike speedometer mounted to the bottom of the test
rotor axis. This speedometer gave us a resolution of about 10 RPM. These two values helped us
understand how the turbulence from the system of rotors was affecting the test rotor and gave us
a better understand how Savonius rotors preformed under turbulence compared to vertical axis
turbines.

All experimentation took place in the Wind Tunnel Laboratory and made use of scaled
experiments using Reynolds similarity. Throughout this report, wind speeds are reported as a

percentage of the maximum tunnel speed of 147 miles/hr (66 m/s). Using a one sixth scale
model, we simulated wind speeds of up to 24.5 miles/hr (11.0 m/s) being applied to our rotors.
Our baseline data consisted of one test rotor located in the back of the test panel, while each of
the following test results were measured off the base line data.

Model Constraints:
The wind tunnel assembly has several testing limitations. Most important is the size restriction of
the wind tunnel, consisting of a box of 45.7 cm (18 inches) square and 40.6 cm (16 inches) in
height, which comprises the total useable volume inside the testing area. This means that we
must scale down our rotor design to work within these bounds. Given the need to experiment
with the effects of turbulent flow flowing from several rotors in series, the 45.7 cm (18 inch)
depth is the more critical dimension for our testing purposes. We were able to slightly extend this
depth by allowing the dummy rotors to overhang outside of the observable area into the wind
tunnel. For our experiment, we have received a wooden board (henceforth referred to as the test
board) with which to mount both our testing apparatus on top in the airstream as well as our
voltage generator underneath (Fig 3).

We decided to make a rotor five and a half inches diameter, to still have greater than three
diameters worth of space behind the rotor model to work with (Appendix A) Our initial
hypothesis was that this would provide enough of a buffer for turbulent effects to propagate and
spread downstream, impacting other rotors in the system. The various test positions were chosen
based on the limitations of the testing space we chose the most obvious ones based on
orientation in the wind tunnel.

Construction:
In order to understand how turbulence affects the Savonius rotor we constructed a system of
three rotors consisting of one test rotor that was kept in a fixed position and two dummy rotors
which we moved into different positions along the test board. All the rotors were made out of
three inch diameter PVC pipe attached to a six inch diameter acrylic disks using PVC cement
along with a 14 inch carriage bolt that ran down the center of the rotor. Our three inch diameter
PVC came in a four foot pipe and required some cutting to produce the desired size. Using the
band saw we cut the PVC into the three eight-inch sections. Again using the band saw, we cut
these sections vertically in half to create two symmetric semicircle sections. The top and bottom
surfaces of these pieces were then milled down in order to create a perfectly level surface that
could be bonded to the acrylic disks. This level surface proved to be crucial in creating a smooth
rotation. Rotors that were not milled down had significantly more wobble at high speeds then
those that were not.

Our original design called for spacing the blades a distance of 20% of the diameter apart. This
did not leave enough room for the carriage bolt and air circulation to occur though the rotor. As a
result we moved the blades to 25% of the diameter apart which allowed for circulation and the
carriage bolt to be placed in between the rotors. This distance still falls within the range for an
optimized rotor as defined by previous research [1].

The two blades were attached to the acrylic disk using the PVC cement. After ample time for
drying, the second acrylic disk was attached. The carriage bolt was then placed through holes
that had been milled out of the acrylic disks. The bolts were then screwed on to help keep the
rotor in compression while the cement dried over night.

To allow the rotors to spin freely while in the wind tunnel, each rotor was placed on top of a
system of bearings and bushings that were attached to the test board. The bearings that we
initially purchased were intended for a high torque application and had a substantial amount of
resistance to spinning. This led to the rotor requiring higher wind speeds to self start and caused
a significant loss of power. In addition, these bearings had slack in them of about two or three
degrees from the vertical axis. This led to excessive wobble induced vibrations that reduced
potential power output, as well as sending visible vibrations throughout the wind tunnel structure
and viewing panel.

In order to correct this issue we switched to a bushing and bearing combination - the bushing to
alleviate the wobble and instability, and a bearing that has less play in it as well as a lower
resistance to spinning. We were able to salvage a mounted bearing that had much less resistance
and provided ideal support for the turbine axle. Along with this bearing, metal bushings were
placed lower on the axis between the test board and carriage bolt to help remove any remaining
play.

The majority of our attention was given to the fixed test rotor. It was connected to the test board
using the less resistive mounted bearing. Because rotational speed and thus bearing friction in the
dummy rotors were inconsequential to the power output from the test rotor, the dummy rotors
were assembled using the remaining high torque bearings and metal bushings.

Our main source of data for this experiment was the voltage output from a DC motor attached to
the bottom of the rotors axle. We constructed a belt driven system consisting of a DC motor
connected in a triangular pattern between the motor, the rotor axle, and a third gear acting to
tension the belt. In order to collect voltages we wired a DAQ in parallel with a large resistor to

the two leads on the DC motor. The DAQ input was collected by a VI through labVIEW and
exported the results to an excel file (Appx A, Fig 4).

Hysteresis Issues:
At wind tunnel speeds approximately 50% and 80% for the dummy rotors and test rotor,
respectively, the rotors exhibited a rapid ramping up of RPMs. For the dummy rotors, this RPM
rate was estimated to be approximately 2500 RPM (estimated due to a lack of rotation measuring
equipment on the dummy rotors), while the test rotor was measured to increase from a stable 960
RPM to 1400 RPM going from 75% to 80% wind speeds instead of a fairly linear ramping up as
had followed in previous wind speed increases (Appx A, Fig 5). This increase was attributed to
the centering of each rotor based on the inertial effects of sufficient RPMs to overcome the play
and imbalances in each rotor system. While these hysteresis effects were not expected, the
inability to recreate these effects without first ramping the wind speed very high and then
backing it down led to issues with data collections of our test rotor. Hysteresis effects of the
dummy rotors did not have any measurable effects on the test rotor, indicating that the rotating
speed of the dummy rotors played little effect on downstream turbulence and thus power
generation of the system.

Once the rotor had centered itself, the rapid increase in RPMs led to a spike in the power
generation, of which would lead to uneven results throughout the experiment even after the wind
tunnel was backed below the 80% threshold of the initial occurrence. Without braking the test
rotor, which would have skewed our other data collection, we decided that collecting useful data
up to the hysteresis speeds would be adequate. Any hysteretic data was disregarded during our
data analysis of power generation.

For purposes of real world application however, we feel that the ramping up of the rotors would
have been a more accurate measurement of power generation, given the increase in overall
performance of the test rotor after the hysteresis point has been passed.

Results:
We found a correlation between turbulence and power generation of a Savonius rotor. The
location of the rotors led to distinct decreases in power on our test rotor, with several strong
trends emerging.

Through our results, several predictable patterns were observed. In relation to the baseline power
output, the least loss in power generation occurred while in the Far Center configuration (Appx
A, Fig 6). Contrary to our predictions of Far Split being the best configuration, the Center rotor
was the superior choice.

The strange behavior of several rotor configurations in the initial two data points worth of
testing is the result of some configurations starting to rotate and others not at these 30-40% wind
tunnel settings, essentially creating infinitely greater/lesser percentages (Appx A, Fig 7). At 30%,
several of the configurations produced sporadic power generation without constant revolution.
The exception to this was the Far Center configuration, which actually started the test rotor at
30%. We believe this was due to a funneling of wind speed and pressure to the test rotor, causing
it to start.

In general, having the rotors further away was better than closer, which is expected based on the
flow reorganizing into less turbulent flow over time and distance. The eddies and vortices being
shed by the rotor decreased in strength with distance, negating the effects of the rotors further

away. The Split configurations had a 24% decrease over the baseline power generation, which
was a full 10% better than the next best configuration, Center. The configurations of Center Left
and Center Right were far worse on a performance basis, on average 52% less than the Baseline
power generations, due to the sheer volume of turbulence causing material in the way of the path.
(Appx A, Fig 8).

As seen in the graphs of RPM and voltage versus wind tunnel power percent (Appx A Fig 5 and
Fig 6), at percentages that exceeded 75% on two of the tests (Appx A, Fig 5), the centering of the
rotor impact caused a strong but non-linear increase in power generation and RPM's. While this
occurred within a relatively narrow frame of wind tunnel percentages, we have not included the
data collected from these points in our power calculations.

We also calculated the effective power output of our rotor as a percentage of the theoretical
efficiency of our rotor system for our baseline data.
Powcr
tubnc
=
1
2
p A :
tunncI
3
C
p
(1)
Equation (1) was used to calculate the maximum power and comparing it to our measured
voltage; we found that our system exhibited rational properties based on its wind speed
correlation. At intermediate wind tunnel speeds (50-74%), our system had an efficiency of
approximately 12% of the theoretical maximum for Savonius rotors of our design in laminar
flow. At low speeds (30-45%), we achieved only five to eight percent efficiency, due to the high
initial friction in the test rotor. At 80% and 85%, after the rotor had centered itself, the efficiency
climbed to around 14%, a direct result of less friction and a higher coefficient of performance
ratio in relation to the tip speed velocity (Appx A, Fig 2 and 9) [1]. The overall performances
ratios, ranging from 5% to 14%, are based on the many inefficiencies in our test system,

10

including bearing friction, losses in the belt/gear assembly, resistance from the DC motor, and
turbulent air flow. A better balanced system with proper gearing and with an inline power
generator could have increased these numbers significantly.

Experimental Rotor Failure:
During a test with the dummy rotors in the far row, after hysteresis of the back row had been
achieved, one of the dummy rotors came apart during a test. With the wind tunnel speed at 85%,
we estimated that it was spinning at approximately 2500-2800 RPM. After inspection, the test
rotor first failed at the keyed portion of the top acrylic disk with a crack originating at one of the
corners (Appx A, Fig 10) A stress concentration factor at the squared end led to the failure of the
disk, leading to a separation of the PVC portions and the bottom acrylic disk from the carriage
bolt axle due to the high rotational speeds in combination with the 120 miles/hr (53 m/s) wind
speeds in the tunnel.

For future experiments, it is our recommendation that the dummy rotors be constructed out of a
tougher, more durable material such as aluminum. Another option is to prevent the dummy rotors
from ramping up their RPMs through hysteresis effects by applying a brake or other load to the
axle of each rotor. This would not have any impact on the power generation of the system
downwind.

Conclusions:
Through data analysis of our measured outputs we were able to draw several conclusions on how
turbulence affects Savonius rotors in order to determine which orientation would be the most
practical in a real world application. According to our data the far center orientation had the least
power losses and would be the idea setup for a field system. The far center and the other

11

orientations were all based on the wind tunnel limitations and trying to test the widest range of
rotor combinations. There remains room for further research in testing more precise changes in
the distance between rotors, especially for orientations that have already proven to be more
efficient.

12

References:
[1] Menet J.-L., A double-step Savonius rotor for local production of electricity: a design
study Renewable Energy 29 (2004)1843-1862
[2] Altan, B.D. , Atilgan M. An experimental and numerical study on the improvement of the
performance of Savonius wind rotor Energy Conversion and Management 49 (2008) 3425-3432

13

Appendix A: Figures

Figure 1 Constructed rotor 8 inches tall by 5.5 inches in diameter

Figure 2 Coefficients of performance in relation to tip velocity/wind speed

14

Figure 3 Schematic representation of test system

Figure4VirtualInstrument(VI)PanelinlabVIEW

15

Figure5RPMforgivenrotorconfigurations

Figure6Voltagegeneratedfromdifferentrotorconfigurations

0
200
400
600
800
1000
1200
1400
1600
30 40 50 60 70 80
R
e
v
o
l
u
t
i
o
n
s

p
e
r

M
i
n
u
t
e
WindtunnelPower(%)
RPMvsWindTunnelPower
baseline
closecenter
closesplit
close2left
close2right
farcenter
farsplit
far2left
0
0.5
1
1.5
2
2.5
3
3.5
30 40 50 60 70 80
V
o
l
t
a
g
e

(
v
o
l
t
s
)
WindTunnelPower(%)
VoltageVsWindTunnelPower
baseline
closecenter
closesplit
close2left
close2right
farcenter
farsplit
far2left

16

Figure7Powerlostforeachrotorconfigurationintermsofbaselinepower

Figure8Overlalimpactofeachrotorconfiguration,inpercentofbaselinepowergenerationthrough
alldatapointscollected
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
140.00%
30 40 50 60 70 80
%

o
f

B
a
s
e
l
i
n
e

P
o
w
e
r

P
r
o
d
u
c
t
i
o
n

(
%
)
WindTunnelPower(%)
%PowerLostVsWindTunnelPower
baseline
closecenter
closesplit
close2left
close2right
farcenter
farsplit
far2left

17

Figure9Modelperformanceinrespecttotheoreticalpowerthatwasexpectedtobegenerated

Figure10Crackeddiskafterfailurewithpointofstressconcentrationinupperrightcorner

0
5
10
15
20
25
30
30% 40% 50% 60% 70% 80% 90% 100%
A
c
t
u
a
l

p
e
r
f
o
r
m
a
n
c
e

o
f

m
o
d
e
l

(
W
a
t
t
s
)
Windspeed
Modelperformancevs.WindSpeed

18

AppendixB:
WindFlowsfromRighttoLeft,LefthandCircleistheTestRotor

Figure1 Figure2
Close2Right Close2Left
Rotorcenterdisplacementsare9horizontal Rotorcenterdisplacementsare9horizontal
0and7vertical 0and7vertical

Figure3 Figure4
CloseCenter CloseSplit
Rotorcenterdisplacementis9horizontal Rotorcenterdisplacementsare9horizontal
0vertical 7and7vertical

19

Figure5 Figure6
Far2Left FarCenter
Rotorcenterdisplacementsare18horizontal Rotorcenterdisplacementsare18horizontal
0and7vertical 0vertical

Figure7
FarSplit
Rotorcenterdisplacementsare18horizontal
7and7vertical