Exploring the Flow around a Savonius Wind Turbineltces.dem.ist.utl.pt/lxlaser/lxlaser2012/upload/281_paper_ywrcws.pdfIn order to confirm this supposition, experimental investigation

16th Int Symp on Applications of Laser Techniques to Fluid Mechanics Lisbon, Portugal, 09-12 July, 2012

- 1 -

Exploring the Flow around a Savonius Wind Turbine

Ivan Dobrev1,*

, Fawaz Massouh1

1: Fluid Mechanics Laboratory, Arts et Métiers ParisTech, 151 bd de l’Hopital, 75013 Paris, France

* correspondent author: [email protected]

Abstract This paper presents the study of flow through a vertical axis wind turbine of Savonius type. For this kind of wind turbines the axis of rotation is perpendicular to wind and the flow is always unsteady. Due to operation as a drag type wind turbine and the continuous variation of flow angle with respect to blades, strong unsteady effects including separation and vortex shedding are observed. In these conditions, it is supposed that CFD analysis with turbulence modeling by means of detached eddy simulation can give good results. In order to confirm this supposition, experimental investigation of flow around a Savonius wind turbine is carried out in the wind tunnel of Arts et Metiers ParisTech.

The studied rotor has two blades and a height which is approximately equal to the rotor diameter. Initially, the aerodynamic performance of the rotor is measured. Then, for optimal point of operation, the flow field through the rotor is investigated using particle image velocimetry. By means of conditional sampling technique, the velocity fields around the blades and in the wake are obtained for several angular positions of the rotor. The investigation permits to determine the structure of the real flow and to create a database which is helpful for CFD validation. The numerical simulation is carried out by means of three-dimensional unsteady solver which gives comparable results with experiments. These results confirm the capability of detached eddy simulations modelling to represent well the turbulent detached flow. Thus this kind of turbulence modelling can be applied to analyse and to optimize the Savonius wind turbine as well as another drag type wind turbine.

1. Introduction

Wind turbines are generally classified as two families: horizontal axis and vertical axis machines.

This classification refers to the position of rotor axis relatively to wind. The Savonius wind turbine

is thus classified as vertical axis wind turbine like the Darrieus, Gyromill or H-rotor etc. The basic

version of this rotor has S-shaped cross-section formed by two semi-circular blades with a small

overlap between them. The rotor has good starting characteristics and operates at relatively low

wind speeds. It does not need an orientation device and can work for all wind directions. This wind

turbine operates at low tip speed ratio and presents real interest for small or miniaturized machines.

Akwa et al (2012) publish recently an interesting review which presents the Savonius wind turbine

as and the influence of different parameters on its aerodynamic performance.

There are several experimental studies concerning a flow structure around Savonius rotor. One of

the first attempts to study the torque mechanism and the flow field is carried out by Fujisawa (1992)

and Fujisawa & Gotoh (1992). The authors study the influence of the blade overlapping on

aerodynamic performance of two-blade rotor. The flows visualizations allow to authors make

conclusion that a Coanda like flow on a convex side of advancing blade permits to improve the

torque. Lately Fujisawa (1996) performs the similar study carried out by means of particle image

velocimetry (PIV). The author compares the phase-averaged velocity distributions around Savonius

rotors for different overlap ratio with numerical calculations by a discrete vortex method. The

author found a significant difference between the velocity distribution on a rotating rotor and that


- 2 -

for a stationary rotor. However, the results of flow simulation are found to be quite satisfactory in

the case of rotation.

Interesting study is presented by Murai et al (2007). Here the authors study velocity field obtained

by means of particle tracking velocimetry (PTV) in order to calculate the pressure field around the

turbine and on the blade, and also to evaluate the rotor torque. The authors confirm a supplementary

effect of lift force on torque and the importance of the overlap effect of the blades to increase

torque. Nakajima et al (2008) investigate the aerodynamic performance of different rotor types

using a pigment streak-line method. The flow visualization and the torque measurement permit to

obtain information about single and two-stage Savonius rotor.

McWilliams & Johnson (2008) present a study of flow around different model of Savonius wind

turbine by means of PIV. The flow fields were obtained at six constant phase angles of the rotor.

Examination of the flow shows for all rotor positions vortices that shed from blade surfaces. The

authors show a considerable interaction between flow over the forward blade and the wake of the

trailing blade and note the importance of lift forces on the torque.

Another PIV exploration in the case of one or two Savonius rotors is presented by Shigetomi et al

(2011). Here the authors present the study of flow field around two Savonius turbines in close

configurations, which permits to understand quantitatively the interaction mechanism and to

revealed two types of power-improvement. The presented visualization and measurement of flow

through the Savonius rotor permit to understand the torque mechanism and the influence of gap

between blades. However, the obtained flow field results did not permit comparison between

experiments and simulation, which is needed to choose an appropriate turbulence model.

The research presented in this paper aims to create database for flow around a Savonius wind

turbine. This database contains the flow fields for different phase angles, which are needed for the

comparison with flow simulation. Initially, the PIV exploration is carried out, and then numerical

simulations are performed for the experimentally studied cases. Finally, the analysis and

comparison of experimental and numerical results permit to validate the applied numerical

modeling.

2. Test bench

The aim of this experimental study is to obtain the aerodynamic performance of studied wind

turbine and also the detailed information about velocity field around the rotor. The experimental

aerodynamic performance permits to qualify the validity of numerical tests. The PIV measuring is

needed to obtain detailed data concerning the mechanism of torque creation and the vortex structure

downstream of the rotor.

The experiments are conducted in the wind tunnel of Arts et Métiers - ParisTech, fig.1. The wind

tunnel is a closed circuit type and has a three-blade axial fan with a rotor diameter of 3m. The fan is

driven by frequency controlled asynchronous motor with a power of 120 kW. Behind the fan, flow

is decelerated in a settling chamber, which is equipped with honeycomb straighteners and wire

mesh in order to smooth the flow. The tunnel nozzle accelerates the wind, from settling chamber to

test section, up to 40 m/s. The nozzle has contraction ratio of 12.5, which ensures a uniform

velocity profile with a turbulence ratio less than 0.25%. The semi-guided test section has a cross-

section of 1.35 m x1.65 m and a length of 2 m. The static pressure in the test section is equal to

atmospheric pressure. Hence, the upstream velocity depends on the stagnation pressure in the


- 3 -

settling chamber which is measured by pressure transducer Furness Control FC20.

The tested wind turbine has a two-blade rotor with a diameter D=219.5 mm and the height

H=200 mm. As result, the rotor aspect ratio AR = H/D is low, in order of 0.91. The blade has

circular arc form of 180° and a thickness of 1mm. The blade radius r is 57.5 mm and the gap width

s is 11.5 mm. Hence, the gap width ratio s/2r is 0.1. The endplates with diameter of 300 mm reduce

the flow leakage from pressure to suction blade sides. Also, the endplates permit to strengthen the

rotor structure. The blades and endplates are fabricated from thin polymethylmethacrylat sheet in

order to ensure the transparency, which is needed for PIV investigation, fig 2.

Fig. 1 Test bench Fig. 2 Flow field around the rotor

The rotor is mounted on a shaft coupled with a DC generator. The rotor load is controlled with a

rheostat connected to the generator. The coupling between the rotor shaft and the generator is made

via a contactless torque transducer HBM T20WN. This transducer also emits 360 pulses per

revolution. Because the measured power is quite low, the seals of the rotor ball bearings are

removed, and the grease is replaced by thin silicon oil. A fiber optic sensor Keyence FS20V, which

detects a reflective target on the shaft, permits to locate the passage of the blade considered to be

the reference. Thus, by counting the number of square signals delivered by the torque-meter, after

the passage of the reference signal, it is possible to obtain the rotor's angular position with an

accuracy of 1°. The acquisition of data from the sensors is carried out by data acquisition card,

which emits a TTL signal for triggering the PIV measurements at a desired rotor angular position.

During the test, the wind turbine rotational speed varies from 800 rpm to 1000 rpm and the

upstream flow velocity V∞ varies from 9 m/s to 15 m/s. Thus, Reynolds number calculated with

rotor diameter and rotor peripheral velocities during the tests is greater than 140,000. Data

acquisition is carried out by computer equipped with acquisition card. At each operating point, all

the torque T, rotational velocity n and upstream wind velocity V∞ are measured continually with a

rate of 10 samples per second. To minimize the torque fluctuations, before to calculate the rotor

power, the results are averaged for a period of 1 minute. Then power coefficient Cp, which

represented the ratio of obtained and available power is calculated:

AV

PCp

2

3

∞

=ρ

(1)

The test permits to obtain the variation of the power coefficient Cp when the tip speed ratio TSR


- 4 -

varies. The TSR is calculated as the ratio between peripheral velocity and upstream wind velocity:

∞

=V

UTSR (2)

The variation of power coefficient with respect to tip speed ratio is presented on fig. 3. The curve

presents typical characteristics of Savonius wind turbine, for similar Reynolds number, and for

similar aspect ratio. The curve is quite flat and the power coefficient has a maximum of 0.18 for a

TSR equal approximately of 0.8, which is chosen to perform PIV exploration.

Fig. 3 Power performance Fig. 4 Coordinate system and azimuth angles

3. PIV investigation

The PIV system is managed by the Dantec DynamicsStudio 2.30. The taking of images is done by

implementing a Litron Nano-L 200-15 laser, with an impulse power of 2x200 mJ, a two cameras of

2048x2048px Dantec FlowSense 4M, equipped with the lens Micro-Nikkor AF 60 mm f/2.8D, a

frame grabber cards and a synchronization system. The last synchronizes images and laser flashes

with the blade angular position. The seeding of flow is made with micro droplets of olive oil created

by a mist generator 10F03 Dantec. The droplets diameter is supposed to be 2-5µm.

The measuring of flow field inside and downstream of the rotor is carried out in the case of low

rotation velocity n=500 rev/min. The upstream velocity is 7.2 m/s, which corresponds to a

TSR=0.8.

The study is carried out for 6 different azimuth positions of the rotor θ = 0°, 30°, 60°, 90°, 120° and

150°, and for each point a sequence of 200 double images is taken. The image capture is

synchronized with the rotation of the rotor and the PIV system is triggered when the blade is

positioned at the desired angular position.


- 5 -

Fig. 5 Raw PIV image

Fig. 6 Raw image and instantaneous flow field

The flipped vertically and horizontally image, taken by first camera is presented in fig. 5. The

results show a good cross-correlation, except in the region near the blade surface. Despite image

saturation near the blades due to reflections and geometric perceptive, the velocity field is

exploitable and can be used for comparison with CFD simulation, fig. 6. Before the tests, the

calibration images are taken in order to calculate the scale and to synchronize spatially the two

cameras. Here the first camera takes the images around the rotor and the second with slight

overlapping takes in the downstream. Finally, the treatment of all PIV images has permitted to

establish a database containing the instantaneous velocity fields, and the average velocity field for

each of the 6 positions of the rotor, fig. 7.

PIV 0°/180° CFD 0°/180°

PIV 30°/210° CFD 30°/210°


- 6 -

PIV 60°/240° CFD 60°/240°

PIV 90°/270° CFD 90°/270°

PIV 120°/300° CFD 120°/300°

PIV 150°/330° CFD 150°/330°

Fig. 7 Averaged velocities fields around the rotor

4. Numerical modeling

Because the aspect ratio AR=0.91 of the rotor is low, there exists a flow leakage near the blade tip,

between the suction and pressure surfaces. The blade circulation variation during rotation creates a

periodical vortex structure downstream of the rotor. This vortex structure is similar to vortex


- 7 -

structure of an oscillating wing with low aspect ratio. As results a three dimensional modeling is

carried out Dobrev&Massouh (2011).

The grid is created by means of ANSYS Gambit 2.4.6. Because the rotor changes its position with

respect to upstream wind direction, the concept of sliding mesh is applied. The grid has two distinct

parts: an external stationary, which represents the flow around the turbine and an internal, which

rotates in order to represent the rotor blades.

The external part of the grid is composed by multiply structured blocks with 5.7 million cells. The

internal part is composed by 2 unstructured blocks with more than 2.3 million parallelepiped cells.

In order to have the parameter y+ <10, 10 layers of brick cells are created around the blades. All the

computations are performed using the Navier-Stokes solver ANSYS Fluent 12.1. Detached eddy

simulation (DES) technique is used in order to improve the flow simulation. In this study the DES

treats near-wall region by means of k-ω turbulence model. Because the flow field does not have

such periodicity like those obtained by other kinds of turbulence modeling, the averaged flow fields

and rotor torque are obtained from 100 revolutions. The numerical results presented on Tab. 1 are

closer to the experimental data for all TSR and confirm the capability of DES to represent well the

detached flow.

Table 1. Experimental and numerical results for power coefficient, Dobrev&Massouh (2011)

Power coefficient Cp TSR=0.6 TSR=0.8 TSR =1.05

Cp experimental data 0.176 0.180 0.167

Cp simulation, unsteady 3D, DES/k-ω SST 0.172 0.176 0.149

Fig.8 Blade and rotor torque

It is useful to show how the torque varies with respect to azimuth position of the rotor. In the

coordinate system shown on fig. 4, the rotor turns in clockwise direction, the flow is from left to

right and by convention the useful torque has negative sign. The torque of the blade is created by

the pressure applied on the convex and concave side. To obtain dimensionless values, the blade

torque is divided by the rotor torque averaged over one revolution. The results show that maximum

useful torque is produced when the blade angular position becomes 27° and the minimum useful

torque at 270°, fig. 8.


- 8 -

5 Analysis

The analysis of obtained numerical and experimental results shown on fig. 7 and fig. 9 permits to

understand the flow structure around the Savonius rotor. So called by Fujisawa (1992) “Coanda-like

flow”, which occurs on convex side of advancing blade for azimuth angles between 300° and 90°,

represents the flow acceleration caused by the curved blade surface, fig. 7. It must be noted that

most of useful torque is created by this blade surface, fig. 8. Similarly to flow on suction side of a

pitching airfoil, the flow on the convex side stays attached for angles of attack up to 60° degrees. It

must be noted that such high angles are attained because the blade surface moves in flow direction.

On the base of obtained results it is difficult to understand the role of flow leakage in blade gap on a

rotor torque. It seems that is not the flow leakage, but the blade enlarging due to high gaps which

produces a negative effect on torque. In this case the blade internal edge moves in opposite

direction to flow. This movement facilitates the flow detachment from convex side of advancing

blade. However, some additional studies are needed in order to elucidate the role of blade gap.

Another important flow structures are vortices created by advancing and retracting blades. The

advancing blade vortex appears on the blade tip for angle of 90°, stays close to the blade tip up to

120° and then moves downstream the rotor. The vortex emitted by retracting blade can be observed

downstream of the blade tip for an angle of 240°.

PIV 0°/180° PIV 0°/210°

PIV 60°/240° PIV 90°/270°

PIV 120°/300° PIV 150°/330°

Fig. 9 PIV average vector field around the rotor


- 9 -

Conclusion

The paper presents the experimental and numerical study of flow around a Savonius wind turbine.

The experiments were carried out in wind tunnel using a model turbine equipped with transparent

rotor. PIV was used to measure the instantaneous velocity field in the middle of the rotor normally

to the axis of rotation. The velocity measurements were synchronized with the rotor azimuthal

position. In order to enlarge the investigation plane, two parallel cameras were used simultaneously.

The PIV experiments were conducted at 6 rotor azimuth positions. Then the rotor performance was

also measured. The obtained experimental data are needed for comparison with the numerical study.

Simulations are carried out using 3D Navier-Stokes solver with DES/k-ω model. The analysis of

obtained results for the power coefficient shows that they are very close to experimental data. The

comparison of wake and shedding vorticity with experiments shows that the 3D/k-ω modeling gives

results quite similar to phase averaged velocity.

The results of simulation confirm the capability of DES model to represent well the turbulent

detached flow. This model can be used to study the flow around Savonius rotor in order to analyse

and to optimize the Savonius wind turbine as well as another type of wind turbines.

Acknowledgements

This research is undertaken in the framework of EURIPIDES program “Multi Low Power Energy

Source Packaging” ENERPACK EUR-09-704, with the financial support of the French Ministry of

Economy, Finance and Industry.

References

Akwa JV, Vielmo HA, Petry AP (2012) A review on the performance of Savonius wind turbines. Renew.

and Sust. Energy Rev, Vol. 16 (5), pp 3054-3064

D’Alessandro V, Montelpare S, Ricci R, Secchiaroli A (2010) Unsteady Aerodynamics of a Savonius wind

rotor: a new computational approach for the simulation of energy performance. Energy 35, pp. 3349-

3363

Dobrev I, Massouh F (2011) CFD and PIV investigation of unsteady flow through Savonius wind turbine.

Energy Procedia, Vol. 6, 2011, pp. 711-720

Fujisawa N (1992) On the torque mechanism of Savonius rotors. J. of Wind Eng. and Ind.

Aero. 40 (3), pp. 277-292.

Fujisawa N, Gotoh F (1992) Visualization study of the flow in and around a Savonius rotor. Exp. in

Fluids,vol 12(6), pp.407-412.

Fujisawa N (1996) Velocity measurements and numerical calculations of flow fields in and around Savonius

rotors. J. of Wind Eng. and Ind.l Aero., 59 (1), pp. 39-50

McWilliam M, Johnson, DA (2008) Velocity measurement of flow around model vertical axis wind turbines,

Int. J. of Green Energy, 5 (1-2), pp. 55-6

Murai Y, Nakada, T, Suzuki, T, Yamamoto, F (2007) Particle tracking velocimetry applied to estimate the

pressure field around a Savonius turbine. Meas. Sci. and Tech. 18(8), pp. 2491-2503.

Nakajima M; Iio S, Ikeda T (2008) Performance of Double-step Savonius Rotor for Environmentally

Friendly Hydraulic Turbine. J. of Fluid Sci. and Tech. 3(3):410-419

Shigetomi A, Murai Y, Tasaka Y, Takeda Y (2011) Interactive flow field around two Savonius

turbines, Renewable Energy, 36 (2), pp. 536-545

Documents

Exploring the Flow around a Savonius Wind Turbineltces.dem.ist.utl.pt/lxlaser/lxlaser2012/upload/281_paper_ywrcws.pdfIn order to confirm this supposition, experimental investigation