performance analysis of a shrouded rotor for a wind powered vehicle

8/6/2019 Performance Analysis of a Shrouded Rotor for a Wind Powered Vehicle

1/12

Performance Analysis of a Shrouded Rotor for a

Wind Powered VehicleKoen Boorsma

Energy research Center ofthe Netherlands, ECN

P.O. Box 1, 1755 ZG [email protected]

Leo MachielseEnergy research Center of

the Netherlands, ECNP.O. Box 1, 1755 ZG Petten

[email protected]

Herman SnelEnergy research Center of

the Netherlands, ECNP.O. Box 1, 1755 ZG Petten

[email protected]

Abstract

This paper presents some design aspectsof a wind powered vehicle. Specialattention is paid to the specification of thediffuser-rotor combination. Theperformance of this combination is verifiedby means of wind tunnel testing. Theoptimal pitch angle is determined and theinfluence of roughness and rotormisalignment is investigated. In addition tothat, several additions to enhance rotorperformance have been researched. Thesuccessful test campaign has resulted in avaluable database for use in wind poweredvehicle design and beyond.

Keywords: Wind powered vehicles, rotoraerodynamics, diffuser, wind tunnel

1 Introduction

In August 2008, ECN participated in a racefor wind propelled cars in Den Helder, theNetherlands. Cars of various designscompeted for the highest speed in a raceagainst the wind.The ECN car - called Impulse waspropelled by an engine consisting of apurposefully designed 3-bladed rotor

within a diffuser.The Impulse has been very successful inthe race. In head wind with an averagevelocity of 9.3 m/s a vehicle speed of 53%of the wind velocity was reached, goodenough for the second prize in the speedcompetition.Later on, the same configuration ofdiffuser and rotor has been tested in theOpen Jet Facility (OJF) of Delft Universityof Technology.

Firstly some design requirements are

addressed. Special attention is paid to thedesign of the diffuser-rotor combination,including predicted performance. Then the

wind tunnel measurements are describedand the results are compared with thedesign predictions. Finally the measureddata is used to predict the car

performance, which is compared to theactual performance during the race. Onthis basis an estimate is given of theperformance potential using the currentrotor diffuser combination.

Figure 1: The Impulse driving on the racetrack, a sea dike in Den Helder, the

Netherlands.

2 Wind powered vehicle

The rotor of a wind powered vehicleproduces power but also adds to the total

drag. On a horizontal track and at constantcar velocity U in head wind V the usefulpower Pu is at equilibrium with the powerlosses due to the trust T, the aerodynamicdrag of the car body D and the rollingresistance R in the wheels and the frictionwith the ground.

Pu= U (T + D + R) (1)

where:

Pu = CP 0.5 (V + U)3

A (2)

T = CT 0.5 (V + U)2

A (3)

TORQUE 2010: The Science of Making Torque from Wind, June 28-30, Crete, Greece 325


2/12

D = 0.5 (V + U)

2(AD + Ao) (4)

R = f M g (5)

with:

A rotor areaAD drag area of the car bodyAo drag area of other components

(mesh, mast, ..)CP power coefficientCT thrust coefficientf friction coefficient (wheels and

ground contact)g gravitational constantM mass of car and driverU car speedV wind speed

drive train efficiency air density

Assume that the aerodynamic drag of thecar body and the rolling resistance arenegligible compared to the rotor thrust. Inthat case the equilibrium condition resultsin:

U/V = ( CP/CT ) / ( 1/ - CP/CT ) (6)

With this assumption high vehicle speedsare reached at high drive train efficiency

and high Cp/CT values. So an importantgoal in the design is to achieve highvalues for these parameters within therestrictions that were imposed by theorganizer of the race like restrictions forthe width and height of the car to 2 m and3.5 m respectively and the rotor area to 4m

2. Further the safety requirements

demanded containment of the rotor in asafety cage or net with a maximum meshof 0.1 m.

A more elaborate investigation into the

performance of wind driven vehiclesincluding a simple one point optimizationmethod, based on the Blade ElementMomentum theory, can be found in [1].

3 Rotor and diffuser

3.1Design

Despite the advantages of a VAWT(absence of a yaw mechanism and thepossibility to realize the largest single rotorarea with straight blades), the HAWT

concept was chosen. The main reasonswere the superior performance coefficient

that should compensate for the differencein rotor area and the possibility to use adiffuser for performance enhancement andas supporting ring for the required safetycage as well. Furthermore it was decidedto use a 3-bladed rotor for smooth running,

less vibrations and low rotational speed.

As mentioned above, safety requirementsfor the race led tot the use of a shroudaround the rotor. This shroud was giventhe shape of a diffuser, to augment themass flow through the rotor. The diffusergeometry was chosen based on thefollowing considerations. A long diffuser would lead to a small

rotor, due to the maximum diameter ofthe combination;

A NACA 44 series aerfoil was chosen

because of the decent behavior at lowReynolds numbers;

A small opening angle of the diffuserwas chosen, because of the lowReynolds number and the danger ofseparation;

The final choice was made for a diffuserchord length of 25 cm and an openingangle of 5 degrees. The chord basedReynolds number at 10 m/s speed is then1.7 x 10

5.

The flow augmentation due to the diffuserwas calculated with a lifting surface

program for the circular geometry and theNACA 44 series camber line, using 10vortex elements on the chord. Figure 2

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0 0.2 0.4 0.6 0.8 1 1.2

relativevelocityaugmentationfactor[-]

radial position [m]

Figure 2: The calculated flow field at 10cm behind the opening of the diffuser.

shows the result of the calculations. Thearea reduction due to the diffuser isapproximately 4%, but the calculatedaugmented volume flow equals 22%, sothat the net effect should be positive. Forthe adapted design of the rotor, theaugmented velocities were multiplied by a

conservative, but rather arbitrary factor of0.7, to account for decambering effectsresulting from the low Reynolds number.



3/12

Generally, such arbitrary approximationwas driven by the need of results in theminimum time possible.

Preliminary analysis revealed an optimalinduction coefficient of 0.15 for the

assumed drive train efficiency and otherlosses. The rotor geometry was designedfor this coefficient at a tip speed ratio of 5and including the flow augmentationfactor, in order to have a large CP/CT ratio,needed for the specific purpose. As adesign effective incoming speed, a valueof 11 m/s was used, composed of freewind speed, the speed of the car and theshroud augmentation. This would give arotor blade chord Reynolds number ofalmost 3 x 10

5. A NACA 4415 profile was

chosen for the entire rotor, and the

aerodynamic data for the Reynoldsnumber were obtained from RFOIL [5].The resulting geometry was adapted forthe manufacturing process into the resultshown in Figure 3. The blades wereshortened to 950 mm, to ensure sufficientdistance from the shroud surface.

radius chord twist

[mm] [mm] []

975 45 4.5

878 52.5 5.7

780 57 7.1

683 62 8.8

585 69 10.8

488 78.9 13.7

390 95 17.7

293 112 23.4

195 132 32.5

98 140 46.5 Table 1: Chord and twist distribution of therotor blades. The blade tips have been cut

off at a length of 950 mm.

3.2Manufacture

The blades were cut from a solid 6061 T6 Aluminum cylinder in a CNC millingmachine according to the calculatedgeometry. There was no time foroptimization of the milling process which isdifficult for such a slender product. As aresult the thickness and chord distributionof the blades deviated from each other.The resulting chord and thicknessdistributions are shown in Figure 3 and 4.The rotor mass distribution has beenbalanced but the blade geometry has not

been corrected. For simplicity the bladeswere fixed during the race though theblade pitch angle could be adjusted in

advance depending on the expected windconditions.The diffuser is CNC milled out of a ring oflaminated plates of AxonProlab 65 mouldmaterial.

chord distribution

- design vs. measured -

40

60

80

100

120

140

160

0 200 400 600 800 1000

radius [mm]

localchord[mm]

design

blade 1

blade 2

blade 3

Figure 3: Measured chord distribution of

the blades compared to the design values

thickness distribution

- design vs. measured -

0

5

10

15

20

25

0 200 400 600 800 1000

radius [mm]

localthickness[mm]

design

blade 1

blade 2

blade 3

Figure 4: Measured thickness distribution

of the blades compared to the designvalues

4 Experimentalarrangement

4.1Model

For the wind tunnel test the assemblyconsisting of rotor-diffuser, upper gear boxand yaw tube, have been mounted on acylindrical mast of equal diameter (130mm) to the Impulse.

The tower is supported by four legs on aground plate and provided with a yawingmechanism for directing the rotor atdifferent angles with the tunnel flow.

Within the rotor area the yaw tube isprovided with a tail featuring a 36 cm longtriangular cross-section extension forstreamlining purposes. The diffuser issupported by 2 streamlined Aluminumstruts of 31.75 mm thickness and 79.5 mmchord at 60 degrees with the vertical and avertical steel pole of 10 mm diameter. At

the front and the back the diffuser isprovided with of a mesh of 2 mm diameter



4/12

steel wire. The spacing between the wiresis about 0.1 m.The length of the wire at one side of thediffuser is 29 m. In front of the rotor cableterminals are used for fastening the wire tothe nose of the diffuser. The connections

between wire and diffuser are flush at theback side.The straight axis of the upper angular geardirectly drives a permanent magnet DCgenerator. The generator is mounted on asupport. The electric load for the DCgenerator was realized in an array ofresistors. By using parts of the resistorsand applying series and parallelconnections different values for theresistance could be obtained in order tovary the rotational speed of the rotor.

Figure 5: Measurement set-up with test

model positioned in front of the windtunnel nozzle.

4.2Test set-up

The model was positioned in the testsection of the newly built Open Jet Facility(OJF) of Delft University of Technology.The OJF is a closed loop tunnel with a testsection (h x w x l = 6.0 m x 6.5 m x 13.5m) featuring an open jet configuration. Anoctagonal nozzle with an equivalentdiameter of 3 m produces a jet that iscollected at the porous rear test sectionwall. The setup is illustrated in Figure 5.

The vertical position of the rotor centerwas aligned with the nozzle center (3 mabove the test section floor) and the axialdistance between rotor plane and nozzleexit amounted to 3 m. This ensuredminimum nozzle blockage whilst

preventing interference of the jet shearlayer with the model. At the time of writingthe flow quality of the tunnel (i.e. uniformityand turbulence intensity) had not beenmeasured yet, although the application ofanti-turbulence screens makes for anexpectation of intensities below 0.5%.

4.3Apparatus

4.3.1 Tunnel variablesThe tunnel speed has been determined byusing the pressure difference over twopressure sensors located in the nozzlecontraction. A calibration has beenperformed with a pitot tube at the modelcenter location (for an empty tunnel) todetermine the relationship between tunnelspeed and pressure difference.To quantify the dynamic head of the flow,the air density is determined using thebarometric pressure and temperature ofthe tunnel.

4.3.2 RPM and torqueBetween gear box and generator a torquetransducer and a pulse generator wereinstalled for measuring rotor torque androtational speed. Because of this set-upthe net rotor torque can be obtained bycorrecting for the bearing losses in thegear box.

4.3.3 Axial forceThe tower of the test model is providedwith a strain gauge bridge for measuringthe bending moment. Only the moment inflow direction was measured. The bridge

has not been calibrated but the bendingmoment is calculated directly from thetower dimensions (diameter 130 mm, wallthickness 5 mm) and the bridge and straingauges properties.

4.3.4 Tufts (Tell tales)Small dark blue colored woolen threads ofapproximately 6 cm length were mountedaround the circumference of the innersurface of the shroud, the wired mesh andthe tower extension. This allowed formonitoring separated flow features,

especially in the case of yawed flow.



5/12

Figure 6: Breakdown of axial force contribution

The signals for torque, rotor speed and

bending moment were sampled withDANTE data acquisition system of ECN,converted into physical units, recordedand displayed.

4.4Data reduction

4.4.1 Breakdown of axial forceIn order to determine the axial force on therotor, the contributions of shroud, towerand the several struts have to be removed.Figure 6 gives an estimate of the dragbreakdown of these components, using

the book of Hoerner [2].In addition to the estimate, axial forcemeasurements have been performed atrotor standstill. The blades were turned tovane position and the rotor azimuth anglewas fixed at 60 degrees. Hence the bladeswere positioned in front of the tower andthe struts holding the diffuser respectively.The measured drag is considerably higherthan the estimated contribution. Mostprobably this is mainly due to interferenceeffects. The velocity dependency visible inFigure 7 can be attributed to the cylinderthat crosses the critical Reynolds numberregion. Since the measured data isbelieved to represent the contribution ofshroud, struts and cylinder better than theestimation, this data is used for correctingthe axial force.

4.4.2 Wind tunnel boundary effectsSince the rotor area is significantcompared to the jet area, a correction forwind tunnel boundary effects is applied.The correction can be divided into nozzleblockage and solid blockage effects asdescribed in [3].

0.10

0.12

0.14

0.16

0.18

0.20

0.22

4.00 6.00 8.00 10.00 12.00 14. 00 16.00 18.00

U [m/s]

Ct[-]

with wire mesh

without front wire mesh

Figure 7: Axial force at rotor standstill

The proximity of the model to the nozzleresults in nozzle blockage and changesthe dynamic-pressure measuring systemfrom the empty-tunnel value. Byapproximating the model as a bluff bodywith frontal area equal to the rotor area, anestimate of approximately 1% tunnelspeed increase is obtained.

For the solid blockage, the method ofGlauert for a propeller in a closed wallwind tunnel as described in [4] is adopted. According to the authors of [4], the solidblockage of a bluff body in an open jetroughly amounts to a quarter of the solidblockage in a closed tunnel. However thesign of the blockage should be inverted,since the model induced jet expansionlowers the tunnel speed at the modellocation. The axial force coefficient CT thatwas used in this correction does notinclude the contribution of the shroud,

tower and struts. The solid blockage dueto these components is neglected.

n component Cd, Cdo Refe-rence

length Dimen-sion

Area Dragarea

[-] [m] [m] [m2] [m

2]

1 diffuser 0.015 XFOIL 6.180 0.250 1.545 0.023

2 cylinder

with tail

0.890 [2] 0.960 0.130 0.125 0.111

3 cylinder 1.200 [2] 0.450 0.130 0.059 0.070

4 streamlinedstrut

0.500 [2] 0.890 0.032 0.057 0.028

5 cylinder 1.200 [2] 0.880 0.010 0.009 0.011

6 nacelle 0.820 [2] 0.210 0.035 0.028

7 wire mesh 1.140 [2] 2.835 0.140

Total drag coefficient(ref rotor area)

0.145



6/12

Since the collector inlet approximately hasthe dimension of the rear test section wall,collector blockage effects are not expectedto be of importance here. In addition tothat, empty tunnel pressure gradients arenot considered.

Combining the above describedcorrections yields a change of tunnelspeed between -1.5% and 1.3%,dependent on the axial force.

4.4.3 Processing of corrected quantitiesThe corrected torque and axial force arereduced to non dimensional CP, CT andCP/CT values using the correctedfreestream velocity U, rotor area and airdensity. The rotor area used for thisreduction does not include the shroud andhence is based on a rotor diameter of 1.9

m instead of 2.0 m. The ideal gas law isused to process the measured pressureand temperature values to air density.The values are presented as a function oftip speed ratio lambda, which is the ratio ofrotational speed at the tip and thefreestream velocity.

4.5Test matrix

A variety of configurations was tested. Themain aim was to determine the rotorcharacteristics. An important subject of

parameter variation is the pitch angle ofthe blades. In addition to that the effect ofroughness and yaw on rotor performanceis assessed. The effect of the wire meshand the tower streamline has also beensubject of investigation. Unfortunately thetime schedule did not allow for removal ofthe shroud in order to determine its effecton rotor performance. All of the tests havebeen performed for a variety of tunnelspeeds ranging from 5 to 15 m/s. Using anaverage driving velocity of half the windspeed for the Impulse, this corresponds towinds between 3 and 5 Beaufort. TheRPM of the rotor for these wind speedsvaried between 250 and 850 to obtain tipspeed ratios between 3 and 8.To determine the rotor characteristics, therotational speed was varied using theresistors described in section 4.1, for aconstant wind tunnel speed and pitchangle. Prior to the measurement of eachdata set, the rotor was kept spinning atconstant wind speed until the measuredtorque stayed constant. This ensured thebearing losses not to vary due to heatingup throughout the measurements.

The reproducibility of each obtained dataset was checked by finishing with arepetition of the first rotational speed.

5 Results

5.1Reynolds number effects

Comparing the non-dimensional curves forvarious freestream velocities yields thetrends given in Figures 9 to 11.The effect of Reynolds number on theprofile aerodynamics is investigated usingRFOIL [5] as illustrated in Figure 8for therange under consideration.

NACA4415

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

0 2 4 6 8

alpha [deg]

Cl[-],Cdx45[-]

10

Cl, Re=1x10^5

Cl, Re=3x10^5

Cd, Re=1x10^5

Cd, Re=3x10^5

Figure 8: Influence of Reynolds number on

NACA4415 profile characteristics

The increase in frictional drag by forwardcreeping transition is overshadowed by thepressure drag decrease due to thethinning boundary layer. However the lift ishardly affected by the decreasingboundary layer displacement. Hence theCT-lambda curves coincide for differentfreestream velocities, whilst the CP-lambdacurves are shifted upward for increasingtunnel speed.

0.30

0.40

0.50

0.60

0.70

0.80

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Ct[-

Pitch=5deg, U=10m/s Pitch=5deg, U=12m/sPitch=3deg, U=8m/s Pitch=3deg, U=10m/s

Pitch=3deg, U=12m/s Pitch=0deg, U=8m/sPitch=0deg, U=10m/s Pitch=0deg, U=12m/sPitch=4deg, U=10m/s Pitch=4deg, U=12m/sPitch=4deg, U=15m/s

Figure 9: Measured CT values



7/12

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp

[-

Pitch=0deg, U=8m/s

Pitch=0deg, U=10m/s

Pitch=0deg, U=12m/s

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp[-

Pitch=3deg, U=8m/s

Pitch=3deg, U=10m/s

Pitch=3deg, U=12m/s

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp[-

Pitch=4deg, U=10m/s

Pitch=4deg, U=12m/s

Pitch=4deg, U=15m/s

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp[-

Pitch=5deg, U=10m/s

Pitch=5deg, U=12m/s

Figure 10: Measured CP values

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/C

t[

Pitch=0deg, U=8m/s

Pitch=0deg, U=10m/s

Pitch=0deg, U=12m/s

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[

Pitch=3deg, U=8m/s

Pitch=3deg, U=10m/s

Pitch=3deg, U=12m/s

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[

Pitch=4deg, U=10m/s

Pitch=4deg, U=12m/s

Pitch=4deg, U=15m/s

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[

Pitch=5deg, U=10m/s

Pitch=5deg, U=12m/s

Figure 11: Measured CP/CT values



8/12

5.2Pitch angle

The influence of pitch angle is depicted inFigures 12 and 13. The trend for U=10 m/sis similar to the trend in the remainder ofthe tunnel speeds. The Figures clearlyillustrate that although the power issignificantly higher for a pitch angle of 0degrees, the accompanied increase inrotor thrust diminishes the CP/CT ratio. Apitch angle of 5 degrees seems to yieldthe best ratio, although the difference issmall with 3 and 4 degrees. The maximumratio is achieved at a relatively low tipspeed ratio (~4) compared to the lambdaat maximum CP (~5.5).

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

0.38

0.40

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp

[-

Pitch=0deg, U=10m/s

Pitch=3deg, U=10m/s

Pitch=4deg, U=10m/s

Pitch=5deg, U=10m/s

Figure 12: Influence of pitch angle on CP

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[

Pitch=0deg, U=10m/s

Pitch=3deg, U=10m/s

Pitch=4deg, U=10m/s

Pitch=5deg, U=10m/s

Figure 13: Influence of pitch angle on

CP/CT

5.3Roughness

The influence of roughness on the bladeswas assessed by application of twodifferent types of roughness strips. Thefirst type with a width of 16 mm and athickness of 0.8 mm was cut out of P60sandpaper. The second type was cut outof 0.29 mm thick smooth surfaced tape

featuring 45 degree angle serrations onthe side facing the flow. The width of theserrations was 5 mm. The strips wereapplied along the span of the blades at achord wise distance of 15% chord on bothpressure and suction side. The results are

shown in Figure 14 to 16for a pitch angleof 4 degrees.

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

3.00 3.50 4.00 4.50 5.00 5.50 6.00 6.50

lambda [-]

Cp

[-

Pitch=4deg, U=12m/s

Pitch=4deg, U=12m/s rough1


Figure 14:Influenceofroughnesson

CP

0.20

0.25

0.30

0.35

0.40

0.45

0.50

3.00 3.50 4.00 4.50 5.00 5.50 6.00 6.50

lambda [-]

Ct[-

Pitch=4deg, U=12m/s



Figure 15: Influence of roughness on CT

0.30

0.40

0.50

0.60

0.70

0.80

0.90

3.00 3.50 4.00 4.50 5.00 5.50 6.00 6.50

lambda [-]

Cp/Ct[-

Pitch=4deg, U=12m/s



Figure 16:Influence of roughness on CP/CT



9/12

The effect of the first strip is rather large,owing to its large thickness. The effectiveairfoil shape and thus profilecharacteristics are altered significantly.Increase in profile drag and decrease in liftis causing a decrease in CP and CT

respectively. The optimum Cp locationshifts from a tip speed ratio of 5.5 to 4.The maximum value of CP/CT is decreasedby 20%.The effect of the second strip is lessstrong but the observed trends are inqualitative agreement with the first strip.The strip is comparable to moreconventional tripping devices, merelycausing the boundary layer to be turbulentwithout inducing an excessive boundarylayer momentum loss.

5.4Rotor misalignment

The effect of rotor misalignment wasassessed by turning the rotor stepwise intothe wind for a constant resistance at

lambda4. The results are illustrated inFigure 17. A misalignment of 5 degreesdoes not significantly affect rotorperformance. Although the powercoefficient stays approximately constantup to a yaw angle of 12 degrees, the axialforce increases above 5 degrees and

thereby reduces the CP/CT ratio.Increasing the misalignment above 20degrees kills the rotor speed and thus thepower, while only a standstill axial forcecontribution remains comparable to Figure7.

Pitch=4deg, lambda=4, U=12m/s

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

0 5 10 15 20 25

Yaw angle [degree]

Cp,

Ctor

Cp/Ct

Ct, tower extensionCtCp/Ct, tower extensionCp/CtCp, tower extensionCp

Figure 17: Influence of rotor misalignment

Streamlining the tower by means of the taildescribed in section 4.1 negatively affectsthe power in yaw. Since the tail is notaligned with the undisturbed flow direction

in yaw, the power decreases although thedifference in axial force (in rotor axialdirection) is hardly affected. Tufts on thetowertail suction surface indicated flowseparation from 5 degree yaw angleonwards.

5.5Performance enhancement

5.5.1 Shroud As outlined in section 3.1, the openingangle of the shroud was chosen ratherconservatively at 5 degrees because ofthe danger of separation. Flow separationfor various yaw angles was monitored bytufts on the inside surface of the upwindside of the shroud. These tests have beenperformed at U=10 m/s and a pitch angleof 5 degrees, for both rotor stand still aswell as a spinning rotor (around 400 rpm).The tufts at rotor center height areobviously most representative formimicking an increase of the shroudopening angle.It appeared that flow separation based onthese tufts exists only from 19 degreesand 23 degrees yaw onwards for rotorstandstill and the spinning rotorrespectively. However, the tufts locatedabove or below rotor center height featureflow separation from 7 and 16 degreesyaw onwards for rotor standstill and thespinning rotor respectively. Most probablythe spinning rotor energizes the boundarylayer, resulting in higher separationangles.Although extrapolation of these test resultsto a new opening angle is notstraightforward, they indicate thatincreasing this angle certainly is an option. Although the shroud is expected toincrease the mass flow through the rotorand hence augment the power, the sameholds for the axial force. According to

Jamieson [6], the effect of the shroudwould cancel out for the CP/CT ratio, whichis the dominant performance indicator forthis application.

5.5.2 Tower extension and wire meshFrom Figure 17 it can be deduced that foraxial flow conditions, the application of thetowertail yields an approximate CTreduction of 0.02. It was hypothesized thatthe reduced blockage in front of the towerwould also result in an increased massflow through the rotor and thus a higher CP

value. However the Figure clearlyillustrates by the constant CP value thatthis is not the case. Most likely the



10/12

increase in flow velocity at the cylindersides compensates for the blockage andhence the total mass flow through the rotoris not affected.Figure 7showsthat a reduction of 0.04 inCT can be achieved by removal of the wire

mesh in front of the rotor. Thecorresponding effect on CP has not beenmeasured although this is expected to besignificant. The positioning of the mesh infront of the rotor will redirect extra massflow outside of the rotor and hence affectthe CP value as well. Using 1D momentumtheory a rough estimate can be made ofthe increase in mass flow through the rotorand hence in CP. For a CT value (withmesh) of 0.7, this would indicate anincrease in mass flow rate of 2.3%. Thenet increase of the CP/CT ratio would then

be 8.5%.The figures above indicate that asignificant performance increase can beachieved if both front and rear meshes canbe omitted.

5.6Comparison with predictions

Performance curves have been estimatedfor various pitch angles using BEM theory.It was found that application of Jamiesonstheory [6] on the calculations would overpredict the experimental results (C

Pand

CT) consistently by more than 50%.Therefore the effect of the shroud is nottaken into account. Since it can be arguedthat the effect of the shroud on CP/CT ratiowould cancel out (see also section 5.4.1),the comparison is made using thisvariable.

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[

Pitch=0deg, U=8m/s

Pitch=0deg, U=10m/s

Pitch=0deg, U=12m/s

Prediction, pitch=0deg, Re=1x10 5

Prediction, pitch=0deg, Re=3x10 5

Figure 18: Comparison for pitch=0

degrees

The results for two pitch angles are shownin Figure 18 and 19. The Reynolds effect

is simulated by inputting two different setsof profile coefficients. They yield an effectof the same order as measured in thetunnel.

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

3.00 4.00 5.00 6.00 7.00 8.00

lambda [-]

Cp/Ct[-

Pitch=5deg, U=10m/s

Pitch=5deg, U=12m/s

Prediction, pi tch=5, Re=1x10^5

Prediction, pi tch=5, Re=3x10^5

Figure 19: Comparison for pitch=5

degrees

For low tip speed ratios the agreement issatisfactory, but the shape of the curveand hence the values deviate significantlyfor increasing lambdas. It can bequestioned whether the effect of theshroud on the displayed CP/CT ratiocancels out. As noted in section 3.1 thealteration of the velocity field by the shroudis not uniform over the rotor plane andlocal effects might cause different trends inpower and thrust. A more detailedinvestigation is recommended (e.g. anintegral simulation of the rotor-diffusercombination using vortex line theory) tocome to a more final conclusion on thismatter.

5.7Model parameters of the car

The model parameters for the Impulsehave been estimated with the performancemodel of Chapter 2 from measurements

during the race, the rotor characteristics(CP, CT) for 5 degrees pitch angle at 12m/s and the measured drag area of theother components (Ao). Ao is determinedfrom the axial force coefficient at rotorstandstill (Figure 7) and corrected for thecontribution of part 3 from Figure 6 whichis shielded by the car body. The mass ofthe car with driver was 430 kg. The resultsare presented in Table 2. The calculatedcar velocities are compared with themeasured velocities during the race inTable 3.



11/12

A rotor area 2.835

AD drag area of the car body 0.244

Ao drag area other components 0.440

drive train efficiency 90%

M mass of car and driver 430

f friction coefficient 0.017

Table 2: Model parameters for the Impulse

wind speed measured calculated

8.5 3.8 3.8

9.3 4.85 4.7

Table 3: Car velocities measured duringthe race and calculated with the

performance model.

5.8Performance improvement

ECN was the first subscriber to the racebut concrete design activities started muchlater. Therefore design choices in theproject were dominated by rough insteadof sophisticated analysis and priority forshort realization paths for production andassembly. As a result many model

parameters have not been optimized.So it is interesting to estimate the vehiclespeed of an improved car with the samerotor-diffuser combination.Assume that the weight of the car could bereduced to 130 kg - just like the Inventusof the University of Stuttgart, one of thecompetitors in the race and that the totaldrag area could be reduced by about 30%through reduction of the frontal area andinterference, mainly of the terminalconnections between mesh and diffuser.Then the total mass reduces to about 200

kg and the drag area to about 0.48 m2

. As mentioned in section 5.5 theperformance of the rotor-diffusercombination will be increased by omittingthe front and rear meshes. As a result thecar velocity will increase to 8.2 m/s at 9.3m/s average wind speed.Further increase of the car velocity has tobe sought by optimizing the design of therotor-diffuser combination.

6. Conclusions

A successful wind tunnel test campaign ofa shrouded rotor has resulted in a valuable

design and beyond. The results indicatethat for this specific rotor the optimumperformance is achieved for a pitch angleof 5 degrees at relatively low tip speedratios of around 4. Reynolds numbereffects are significant for the present rotor,

showing an increase of up to 10% in CP/CTratio in the range of freestream velocitiesbetween 8 and 12 m/s. The removal of thefront wire mesh has the potential ofincreasing the CP/CT ratio by 8.5%. Theperformance at small yaw misalignment (5degrees) is hardly affected, but at 20degrees misalignment the rotor speed iskilled. Furthermore flow visualizationimplies that a larger opening angle of theshroud can be applied resulting in asignificant increase in CP but also CT. Acomparison with BEM predictions shows

good agreement of the CP/CT ratio for lowtip speed ratios. However largediscrepancies are present betweencalculations and measurements for highlambdas. The absence of modeling theeffects of the shroud is presented aspossible cause for the discrepancy.However, a more detailed investigation isrecommended to come to a more finalconclusion on this matter.

database for use in wind powered vehicle

nalysis with the model for the car

press their

and ye, S. and

[2] S ., Fluid-dynamic drag,Published by the author, 1965

Aperformance shows that 88% of the wind

speed can be achieved with the samerotor-diffuser combination at realisticreductions of the total weight of the carand its frontal area and by improving theCP/CT ratio only by omitting the wire mesh.

Acknowledgements

The authors would like to exgratitude to Delft University of Technologyand more specifically to Nando Timmer foropening up the possibility to use the OJF.Furthermore the contribution to the

realization of the test model by RudolfHoep, Ton Ruiter, Erwin Werkhoven andGeert Hanraads of ECN was greatlyappreciated.

References

[1] Gaunaa, M.Mikkelsen, R. Theory and Design ofFlow Driven Vehicles Using Rotors forEnergy Conversion, ProceedingsEWEC 2009, March 16-19, Marseille,France.

Hoerner, .F



12/12

[3] Ewald, B.F.R., AGARDograph 336,Wind tunnel wall correction, Canada

[4] A., Low-speed wind tunnel

[5]d Piers, W.J. and Van

[6]nstrained flow

communication group Inc., October1998

Barlow, J.B. and Rae, W.H. jr. andPope,testing, third edition, John Wiley &Sons Inc., 1999

Snel, H. and Houwink, R. andBosschers, J. anBussel, G.J.W., Sectional predictionsof 3-d effects for stalled flow onrotating blades and comparison withmeasurements, Conferenceproceedings European Wind EnergyConference, Lbeck-Travemnde,Germany, March 1993

Jamieson, P.M., Beating Betz: energyextraction limits in a cofield, Journal of Solar EnergyEngineering, Vol 131, Issue 3, 2009


performance analysis of a shrouded rotor for a wind powered vehicle

Documents