You are on page 1of 15

F LOW F IELD OF A ROTATING -W ING M ICRO A IR V EHICLE

Manikandan Ramasamy∗ J. Gordon Leishman† Timothy E. Lee‡

Alfred Gessow Rotorcraft Center


Department of Aerospace Engineering
Glenn L. Martin Institute of Technology
University of Maryland
College Park, Maryland 20742

Abstract CT Rotor thrust coefficient, = T /ρAΩ2 R2


DL Disk loading, = T /A or T /Ae
An experiment was conducted to measure the hovering FM Figure of merit
performance of a rotor typical of that used on a rotating- PL Power loading, = T /P or W /P
wing micro air vehicle. The rotor was shown to have rela- r Radial distance
tively low hovering efficiency that can be traced, at least in rc Core radius of the tip vortex
part, to its significant viscous wake and the relatively large
r Non-dimensional radial distance, = r/rc
aerodynamic losses that are associated with the wake.
R Radius of the blade
High-resolution flow visualization images have divulged
several interesting flow features that appear unique to Rev Vortex Reynolds number, = Γv /ν
rotors operating at low Reynolds numbers. The vortex t Time
sheets trailing the rotor blades were found to be much T Rotor thrust
thicker and also more turbulent than their higher chord Te Equivalent time
Reynolds number counterparts. Similarly, the viscous vh Hover induced velocity
core sizes of the tip vortices were relatively large as a vi Induced velocity
fraction of blade chord compared to those measured at Vr Radial velocity
higher vortex Reynolds numbers. However, the tip vor- Vθ Swirl velocity
tices themselves were found to be laminar near their core W Vehicle weight
axis with an outer turbulent region. Particle image ve- α Lamb’s constant, = 1.25643
locimetry measurements have been made at various wake Γ Circulation, = 2πrVθ
ages that have quantified the structure and strength of the Γb Bound circulation
wake flow, as well as the tip vortices. An analysis of
Γc Circulation at the core radius
the vortex aging process has also been conducted, includ-
Γv Circulation at large distances
ing the development of a new non-dimensional equiva-
lent time scaling parameter to normalize the core growth δ Ratio of apparent to actual viscosity
of tip vortices generated at substantially different vortex ζ Wake age
Reynolds numbers. ν Kinematic viscosity
νt Eddy or tubulent viscosity
νT Total kinematic viscosity, = ν + νt
Nomenclature ρ Air density
σ Rotor solidity
A Rotor disk area ψ Azimuthal position of blade
Ae Effective disk area Ω Rotational speed of the rotor
c Blade chord
Cd Drag coefficient
Cl Lift coefficient Introduction
∗ Assistant
Research Scientist. mani@glue.umd.edu Aerodynamic research on hover capable micro-air vehi-
† Minta Martin Professor. leishman@eng.umd.edu cles (MAVs) that have good flight endurance and can also
‡ Minta Martin Undergraduate Intern. telee@umd.edu perform desirable maneuvers has been gaining significant
Presented at the 62nd Annual Forum and Technology Display interest from the research community over the past few
of the American Helicopter Society International, Phoenix, AZ, years. This is because MAVs have potential advantages
May 9–11, 2006. 2006
c by M. Ramasamy, J. G. Leishman, & for performing critical military operations at low risk,
T. Lee, except where noted. Published by the AHS International such as surveillance over enemy territories, various covert
with permission. operations, or remote sensing in hazardous environments

1
when airborne chemicals or biological agents exist. While the ratio of vehicle weight to power required to hover, i.e.,
various definitions of MAVs exist, generally speaking they W /P = T /P = PL. The induced (ideal) power required to
are defined as flight vehicles that have a maximum size di- hover is given by P = T vh , where vh is the minimum av-
mension that does not exceed 6 inches (15 cm). erage induced velocity through the plane of the rotor disk
Biomimetic (insect based) flapping as well as rotat- (or normal to the effective stroke plane of wing flapping)
ing wings are two different concepts being considered for to produce the thrust T . This means that the ideal power
hovering types of MAVs. Both mechanisms have their loading will be inversely proportional to the induced ve-
own relative advantages and disadvantages. Flapping- locity; this is a fundamental result that comes about based
wing mechanisms are known to have relatively poor me- on the solution to the momentum theory, which invokes
chanical efficiency, but there have been several hypothe- the principles of conservation of mass, momentum, and
sis forwarded that claim that flapping wing concepts offer energy in the flow.
better aerodynamic efficiency compared to rotating wings Using the momentum theory, the average ideal (mini-
when operated at extremely low chord Reynolds numbers, mum) induced velocity can be written in terms of the ef-
say below 50,000. Numerous attempts have been made fective disk loading as
through computational fluid dynamic simulations, as well s s
as experiments, to verify these hypothesis and to under- T DL P
stand the basic physics of flapping wing flight (Refs. 1, vh ≡ vi = = = = (PL)−1 (1)
2ρAe 2ρ T
2). This work, however, has resulted in only limited suc-
cess. Computationally, one difficulty is tracking vortic- where DL is the effective disk loading, T /Ae . For a rotor
ity to long times. Experimentally, there are difficulties Ae = A and in the case of a flapping wing the effective disk
in measuring accurately the low thrust generated and the area Ae is based on the net swept area in the stroke plane
corresponding small power required, not to mention the over one complete wing stroke. The power loading can
difficulties of both measuring and predicting the com- also be written in terms of the figure of merit FM (i.e., the
plex three-dimensional, highly unsteady, viscous domi- aerodynamic efficiency) of the system as
nated vortical flows that are present in this type of flow
field. √
T 2ρ FM
As a result, a comprehensive knowledge of flapping PL = = √ (2)
P DL
wing flight has not yet been established to fully explain
the aerodynamic performance of flapping-wing insects or where the FM accounts for all sources of non-ideal losses.
birds, or to prove the postulated gains in relative aero- This means that the best hovering efficiency (i.e, the max-
dynamic efficiency of flapping-wing based MAVs over imum power loading) is obtained when the effective disk
rotating-wing MAV concepts. Therefore, the present au- loading is a minimum and also when the FM is a maxi-
thors have been conducting a research program, funded mum.
by the U.S. Army Research Office, to better understand According to the results in Fig. 1, the power loading for
the fluid dynamics and relative performance merits of hovering flight increases quickly with decreasing effective
rotating-wings versus flapping-wing MAVs. Initial work disk loading (note the logarithmic scales). The best theo-
on the flow diagnostics of a flapping-wing concept is re- retical hovering performance under the stated assumptions
ported in Ref. 3, and the focus of the present paper is on a is given by the FM = 1 line. It will be apparent that hov-
rotating-wing concept. ering concepts that have low effective disk loadings will
The primary objective of the present work was to ex- always require relatively low power per unit of thrust pro-
amine the wake from a MAV-size, micro-rotor, and to un- duced (i.e., they will have high ideal power loading) and
dertake a baseline experiment that helps further the un- will require less power (and consume less fuel or energy)
derstanding rotor behavior at low blade Reynolds num- to generate any given amount of thrust.
bers. This experiment principally sets the groundwork for Therefore, the key to hovering efficiency for any type of
further research. The experiment was performed in two MAV concept (rotating-wing or flapping-wing) is always
stages. First, the performance of the micro-rotor was mea- to have a low effective disk loading, although it must also
sured at various rotational speeds and blade pitch angles. have good aerodynamic efficiency (i.e., a high FM). No-
This was followed by flow visualization and PIV measure- tice that insects and hummingbirds generally have very
ments. The work reported in this paper describes the var- low effective disk loadings and so have good hovering
ious aerodynamic structures that are present in the flow efficiency, even although their effective FM values may
field and which contribute to the net aerodynamic perfor- not always be that high. Measurements made on rotating-
mance of a micro-rotor. wings at similar disk loadings show similar hovering per-
formance efficiency to that achieved by insects.
Because of the relatively low mechanical efficiency of
Background existing flapping-wing MAV concepts, a proven concept
such as rotating-wing may be the best short-term solu-
The efficiency of any hovering vehicle can be quantified tion towards successfully developing a hovering-efficient
in terms of effective power loading, which is defined as MAV, but only if the rotor can be designed to have low

2
Experiment
Rotor Performance Measurements
The rotor blades for the micro-rotor were made of com-
posite carbon fiber, and had circular arc, cambered airfoil
sections. The radius of the blade was 86 mm, with a uni-
form chord of 19 mm, giving a blade aspect ratio of 3.7.
The blades had no twist or taper. The rotor had a solidity,
σ, of 0.14 with two blades attached.
The rotor was mounted in a test fixture with one load
cell to measure thrust, and another load cell was used to
measure torque. The performance of the rotor was mea-
sured at different combinations of rotor rpm and blade
Figure 1: Power loading versus effective disk loading pitch angles. Tares were measured at different rpms with
the blades detached from the hub, and the tares were
for biological and mechanical systems. Low effective
subtracted from the measured thrust and torque with the
disk loading always leads to high hovering efficiency blades attached.
(high power loading). The measurements were then converted to standard
thrust and power coefficients, which are shown in Fig. 2.
3/2 √
disk loading and also be given good aerodynamic effi- The ideal power is given by CPideal = CT / 2, and with a
ciency. While good aerodynamic efficiency requires the figure of merit FM the predicted power is
design of blade airfoil sections with low drag and high   3/2
lift-to-drag ratios, a major source of performance loss for 1 CT
a rotor is contained within the structure of its blade wake, CP = √ (3)
FM 2
i.e., the induced losses.
Comprehensive rotor wake measurements have been Notice that the measurements suggest a relatively low
carried out to help understand the source of these losses aerodynamic efficiency, with an average figure of merit
using various scales of rotors, from model-scale to full- of about 0.5 describing the measurements.
scale (Refs. 7–14). However, no detailed wake mea- A figure of merit plot versus blade loading coefficient
surements have yet been performed on MAV-scale rotors, is shown in Fig. 3. For reference purposes, FM curves
where the blade tip chord Reynolds numbers lie in the predicted on the basis of the equation
range of 10,000 to 50,000. This lack of data is not only 3/2
because of the experimental complexities associated with CT

measuring rotor flows at any scale, but also from the spe- 2
cific measurement challenges that are unique at the MAV- FM = 3/2
(4)
C σCd0
scale level. This includes, but is not limited to, the phys- κ √T +
ical size of the flow structures that are present, which are 2 8
often too small to be sufficiently resolved with most types are shown, which are marked simply to represent bounds
of flow diagnostic methods. rather than the actual predicted performance. An average
The substantial difference in the operating chord
Reynolds numbers between a MAV-size rotor and even
moderately larger scale rotors (say, 1/4 to 1/6 of full-
scale) raises immediately several scaling issues that need
to be addressed. Clearly, viscous forces are more impor-
tant for determining the characteristics of the flow field
at these low operating Reynolds numbers. Also, existing
experimental evidence suggests that the hover efficiency
of rotating-wing MAVs are much lower, with FM val-
ues of no more than 0.5 when compared with their high
Reynolds number counterparts (Refs. 4–6). This clearly
suggests that the recovery of aerodynamic efficiency to
levels comparable to full-size rotors stems, in part, from
an understanding and minimization of the various sources
of losses in the rotor wake. This includes the vortex sheets
trailed from each blade, as well as the blade tip vortices Figure 2: Power polar for the micro-rotor showing its
and their evolution at low vortex Reynolds numbers. relatively low aerodynamic efficiency.

3
Flow Field Measurements
This experiment included flow visualization and two-
component PIV measurements in the rotor wake. The
two-bladed rotor was placed on specially made rotor
stand, as shown in Fig. 5, and was tested in a flow con-
ditioned test cell. For these sets of experiments the rotor
was operated at a rotational frequency of 50 Hz with a tip
speed of 27.02 m/s. The operating tip Mach number and
Reynolds number based on chord were 0.082 and 34,200,
respectively. The measured CT /σ for the test conditions
was 0.0867.
For both the flow visualization and PIV measurements,
the flow at the rotor was seeded with a thermally pro-
duced mineral oil fog. The average size of the seed
Figure 3: Figure of merit curve for the micro- particles were between 0.2 to 0.22 microns in diameter,
rotor versus blade loading coefficient with theoretical which was small enough to minimize the particle track-
bounds shown. ing errors for the vortex strengths found in these experi-
ments (Ref. 15). For the PIV experiments, the entire test
area was uniformly seeded before each sequence of mea-
surements. For the flow visulization, judicious adjustment
of the seeder was required to introduce concentrations of
fog at the locations needed to clearly identify specific flow
structures.
A laser light sheet from Nd:YAG pulsed laser source
capable of frequencies up to 15 Hz was used in synchro-
nization with the rotor frequency to illuminate planes in
the flow field. A fully articulated optical arm was used
to locate the light sheet in the required region of focus.
A 6.1 mega-pixel digital still camera was used to acquire
all of the images. Digitizing the images relative to a cali-
bration grid provided the required spatial locations of the
various observed flow structures, such as the wake sheets
and the tip vortices. A schematic of the experimental set
Figure 4: Rotor figure of merit in terms of wake in- up is shown in Fig. 6.
duced loss and blade section efficiency. The PIV system included dual Nd:YAG lasers that were
operated in phase synchronization with the rotor, the opti-
cal arm to transmit the laser light into the region of inter-
sectional drag coefficient Cd0 = 0.03 was used (typical of rogation, a digital CCD camera with 1 mega-pixel reso-
airfoil section drag coefficients at blade chord Reynolds lution placed orthogonally to the laser light sheet, a high-
numbers of 50,000), with induced power factors of 1.5 and speed digital frame grabber, and PIV analysis software.
2.0. The predicted results show reasonable bounds on the The laser could be fired at any blade phase angle, enabling
measurements, and besides significant blade profile losses PIV measurements to be made at any required wake age.
also suggest that relatively high induced losses are present
at the rotor compared to those obtained at higher Reynolds
number.
The problems associated with the design of an efficient
rotating-wing MAV now becomes immediately apparent.
By plotting the rotor FM as a function of induced power
factor (which is a measure of induced losses) and blade
3/2
section Cl /Cd (which is a measure of rotor efficiency),
as shown in Fig. 4, it will be apparent that reductions in
both induced losses and blade section losses are required
to improve FM. Clearly with high induced losses (in-
duced power factors > 1.5) then no amount of improve-
3/2
ment in airfoil section Cl /Cd can lead to increased val- Figure 5: Photograph of the upper part of the rotor
ues of FM. test fixture.

4
locity vectors across the entire region of interest. The
processing was performed in such a way that the maxi-
mum number of interpolated vectors allowed in each im-
age (which has 50-by-50 nodes) was less than 10. The
interrogation window was chosen such that the maximum
displacement of the seed particles within the interrogation
window was less that one-quarter of the window size. It
should be noted that the vortex has nearly zero rotational
(swirl) velocity at its center (axis of rotation), and has
maximum swirl velocity at its core boundary. The core
center, however, has a significant convection velocity. Be-
cause the interrogation window size used was uniform (no
adaptive grid has been developed yet) and is optimized
for peak velocity measurements, the particle displacement
near the vortex center will be a very small fraction of the
window size. This may not yield accurate results near the
vortex core axis.
Figure 6: Schematic showing the experimental PIV set Furthermore, the combination of centrifugal and Cori-
up. ollis forces affect the trajectories of seed particles away
near the vortex core center (Ref. 15), which gives a clear
The two lasers were fired with a pulse separation time seed “void” with a low concentration of seed particles, as
of 15µs; this corresponds to less than 0.1◦ of blade mo- can be seen in Fig. 7. This further complicates the prob-
tion. The interrogation region was focused to a particular lem of making measurements, because without enough
region of interest within the image using 50-by-50 nodes seed particles, the PIV analysis may not yield many ve-
on either side. This corresponded to an interrogation size locity vectors in those regions. As a result, some of the
of 38-by-44 mm with 0.75-by-0.88 mm between adjacent interpolated vectors obtained during data reduction come
nodes. from this seed void region.
Determining the center of the vortex and estimating the
core radius at which the swirl velocity is a maximum, are
PIV Image Processing two fundamental requirements for understanding the evo-
Processing the acquired PIV images to obtain reliable ve- lutional properties of tip vortices at any scale. Because
locity vectors at this scale was found to be relatively chal- the flow velocity vectors are measured through a spatially
lenging. The biggest advantage of PIV is its ability to digitized grid over a given plane in the flow field, the prob-
measure the velocity field (over a plane) at a given instant ability is relatively high that the grid lines will not pass
of time rather than the point-by-point measurements typi- through either the center of the vortex or its core boundary.
cal of laser Doppler velocimetry. This advantage is gained Identifying the center of the vortex by estimating its cen-
for many flow conditions where the flow is not highly troid of vorticity is a procedure followed by many compu-
three-dimensional and/or unsteady, such as an unidirec- tational fluid dynamic analysts (e.g., Ref. 16) as well as by
tional flow, or a steady or periodic flow. However, for a some experimentalists (e.g., Ref. 9). However, for the PIV
flow that has substantial three-dimensional velocity gra- measurements made in the current study it was found that
dients and/or is combined with large strain or rotational the grid resolution at this small scale was not completely
effects (e.g., for vortex flows, as in the present case), the sufficient to accurately determine the centroid of vorticity
processing of raw PIV images is more complicated. by this approach. Therefore, the center of the vortex was
This is further complicated by the aperiodic movement defined as the mid-point of the peak swirl velocity that
of the tip vortices; this requires individual PIV images to exists on either side of the vortex flow.
be spatially orientated so that the center of any one vortex This process brings in yet another complication of re-
(i.e., the centroid of vorticity) in all of the images coin- moving the local convection velocities of the tip vortices
cide with each other before phase-averaging to obtain the through the wake flow. In the current experiment, this was
mean velocity field. This correction for aperiodicity, how- done by choosing equal number of points on either side of
ever, is based on the assumption that the inner region of the vortex core, and averaging the tangential component
the vortex flow rotates like a solid body, and so that all of velocity measured at these points. The velocity compo-
the measurement points inside the vortex are displaced by nents associated with the rotation of the tip vortex is then
equal amount and at the same velocity. cancelled, leaving only the convection velocity.
The following describes the procedure used in the cur- The next step was to increase the resolution of the grid
rent experiment for determining the velocities in the flow to more accurately determine the tip vortex core bound-
field from the images acquired from the PIV system. The aries. This was done using Kriging, which is a standard
raw images were first processed using a commercially technique followed to accurately interpolate the available
available PIV analysis software, which yielded the ve- data with less uncertainty. Kriging interpolates a value for

5
each output cell by calculating a weighted average of the that trail from blades operating at higher vortex Reynolds
values at nearby points. Closer points are weighted more numbers (Ref. 8). In this case, a tip vortex generally shows
heavily than more distant ones. The Kriging method ana- three distinct regions: 1. An inner laminar region where
lyzes the statistical variation in values over different dis- there is no mixing interactions between adjacent layers of
tances and in different directions to determine the shape fluid, 2. A transitional flow region where there are eddies
and size of the point selection area as well as the set of of several scales and, 3. An outer turbulent region where
weighting factors that will produce the minimum error in the flow is mostly turbulent but relatively free of any larger
the interpolated estimate. The grid resolution was then eddies. This multi-region structure arises mainly because
increased by an order that resulted in the maximum dis- of a Richardson number effect (Ref. 17), which is a ro-
tance between adjacent nodes to be 0.05 mm, leaving the tational stratification effect, and so the relative size of
maximum possible uncertainty as less than 0.0125 mm. the three regions in the vortex flow depend on the vortex
Once the vortex center was estimated, the procedure was Reynolds number Γv /ν.
repeated for multiple number of images that were obtained
at the same wake age. Upon identifying the center of
vorticity for all the images, the images were appropri- PIV Measurements
ately collocated so that the center of the tip vortices co-
incided with each other. This allowed a determination of Representative PIV results in the rotor wake that were ob-
the phase-averaged flow properties. tained at 33◦ wake age are shown in Fig. 12. The resul-
tant velocity vectors ~Vθ + ~Vr shows the slipstream bound-
ary from the rotor. The flow is well-organized and has a
Results higher velocity inside the wake slipstream boundary with
essentially a quiescent flow outside the boundary. The
mean velocity field shown in this figure was obtained by
Flow Visualization phase-averaging 80 individual PIV vector fields without
Flow visualization images were acquired at different wake correcting for aperiodicity. As previously discussed, it is
ages by slipping the phase of the blade passage relative to difficult to apply aperiodicity corrections to larger regions
the plane occupied by the laser sheet. Figure 7 shows an of the flow because the individual parts of the vortices (at
example of the concentrated vortices trailing from the tips different wake ages) exhibit aperiodic displacements with
of blades. Because this was a two-bladed rotor system, slightly different magnitudes and frequencies. As a result,
every second tip vortex that appears in sequence on same correcting the flow field based on the aperiodic properties
side of the image are 180◦ of wake age apart. at one point in the flow simply introduces a biased and
A closer view of the trailed vortex sheet that corre- incorrect correction at other locations
sponds to the “boxed” region in Fig. 7 is shown in Fig. 8. The vorticity contours, which are shown in the back-
This image clearly reveals the presence of more organized ground of these images identifies the presence of three
trailed eddies. These eddies identify regions of local con- strong tip vortices. The tip vortex that is present closer
centrated vorticity that result from the merging of bound- to the rotor blade (33◦ wake age) has not yet rolled up
ary layers from the upper and lower surfaces of the blade. completely, resulting in relatively lower levels of vortic-
In this case, the boundary layers are relatively thick be- ity. However, by a wake age of 213◦ , the vortex clearly
cause of the low chord Reynolds numbers (< 50, 000 at shows strong vorticity at its core, which then diffuses ra-
the blade tip). dially away from the core as wake age further increases.
The trailed vortex sheet rolls up into a concentrated vor- It is apparent from this image that the tip vortices move
tex near the blade tip, as shown in Fig. 9. Notice that radially inward and axially downward as their wake age
the inboard parts of the helical vortex sheets convect ax- increases. Also, from the axial location of the vortices
ially below the rotor at a much faster rate compared to below the rotor plane, it can be observed that the axial
the tip vortices. This is because they are well inside the convection velocity of the tip vortex increases after the
slipstream boundary of the rotor wake, and so the vortex first blade passage. Of course, these are fundamental fea-
sheets take on increasing inclinations to their initial (al- tures common to the wake development on rotors at larger
most parallel) orientation to the rotor plane, as seen in scale.
Fig. 10. It should be noted that the downstream vortex Figures 13(a) and (b) show the average velocity field ac-
sheets are clearly much thicker and are also more turbulent quired before and after correcting for the effects of aperi-
than would be obtained with a rotor operating at higher odicity, respectively. On comparing the vorticity contours
blade Reynolds numbers. The net effect is that the helical on the background of both figures, it is apparent that the
vortex sheets are folded down on top of each other, ulti- ensemble phased-averaged vector plot identifies the max-
mately occupying a substantial part of the vena contracta. imum vorticity at the vortex axis. The results of simple
Even though the parts of the vortex sheets that roll up averaged vector plot shows the presence of a tip vortex,
into the tip vortices appear more turbulent, the flow near however, the vorticity is not a maximum at its center.
the core axis of the tip vortex is clearly laminar – see The images in the left column of Figs. 14 and 15 show
Fig. 11. This is similar to that found with tip vortices the ensemble phase-averaged velocity vector field for six

6
Tip vortex
ζ = 0 deg
Rotor blade

Tip vortex
ζ = 180 deg
Turbulent Tip
vortex vortices
sheet

Rotor shaft

Figure 7: Flow visualization image acquired at 0◦ wake age showing the trailed blade tip vortices and the inner
vortex sheet.

Rotor blade

Tip
vortices
Merging
boundary
layers
(vortex sheet)

Figure 8: Closer view of the vortex sheets trailing from the rotor blade at 0◦ wake age.

wake ages. The swirl velocity and vorticity distributions pected based on known tip vortex behavior at higher vor-
were determined by making a horizontal cut across the tex Reynolds numbers.
center of the vortex, which are shown in the middle and
right columns of these figures, respectively. The tan- The classical swirl velocity signature of a tip vortex can
gential velocity distribution was normalized using the tip be seen in all of the measured velocity profiles. Compar-
speed of the rotor. It can be seen from the vorticity con- ing the swirl velocity profiles measured at 26◦ and 63◦
tours that the maximum value of vorticity is near the vor- of wake ages reveals that the peak swirl velocity initially
tex axis at all wake ages, as would be expected. This is increases with increasing time. Such an observation has
confirmed through the vorticity distribution plotted across been reported earlier using tip vortex measurements that
the vortex. Also, it is apparent that the peak value of were obtained from a sub-scale rotor (Ref. 18). However,
vorticity reduces (hence diffusing vorticity radially out- the initially lower peak swirl velocity in the present case is
ward) with increasing wake age, again as would be ex- likely attributed to the increased boundary layer thickness
on the blade at this low chord Reynolds number, which

7
Tip vortex
ζ = 10 deg

Tip vortex
ζ = 190 deg
Tip
Turbulent vortices
vortex sheet

Rotor shaft

Figure 9: Flow visualization image obtained at 10◦ of wake age showing the roll up of the vortices at the tip of the
blade.

Rotor blade Tip vortex


ζ = 73 deg

Turbulent Tip
vortex sheet vortices

Rotor shaft

Figure 10: Flow visualization image obtained at 73◦ wake age showing the higher axial convection velocities of the
vortex sheets when compared to the convection velocity of the tip vortices.

could more significantly affect the initial development of ages, which thereafter exhibit a continuous reduction of
the tip vortex. This can also be seen from Fig. 12, where peak swirl velocity with increasing wake age. This is ac-
the tip vortex at a wake age of 213◦ exhibits higher values companied by an increase in the vortex core size (defined
of vorticity than at 33◦ . The effects of viscous diffusion as the distance between two swirl velocity peaks), which
can be seen from the swirl velocity profiles at older wake is consistent with conserved core circulation.

8
Figure 11: Flow visualization image obtained at 90◦ of wake age showing the three-region flow structure inside the
tip vortices.

Figure 12: PIV velocity vector plots obtained at 33◦ of wake age, also clearly showing the slipstream boundary.

9
(a) (b)

Figure 13: PIV velocity vector plots obtained at 26◦ of wake age before and after correcting for the effects of
aperiodicity.

The circulation distribution in the tip vortices can be de-


termined from the swirl velocity profiles given in Figs. 14
and 15. Based on the assumption that the flow inside the
vortex is closely axisymmetric, the circulation distribution
can be calculated using the expression
 
Γ Vθ  r 
= 2π (5)
ΩRc ΩR c

The tip vortex strength, Γv , is related to maximum bound


circulation on the blade, Γb , and can be estimated from
Ref. 6 using
 
Γv Γb CT
= =k (6)
ΩRc ΩRc σ Figure 16: Measured circulation distribution across
the tip vortices at various wake ages.
where k = 2 for a rotor blade with ideal twist and k = 3
for an untwisted blade. For the current operating condi- Lamb and Oseen (Refs. 19, 20) derived an exact solution
tion of this rotor, the estimated value of Γb /ΩRc is 0.256, to the Navier–Stokes equations to predict the core growth
which is consistent with the measurements of Γv . It can rate of laminar vortices, which is given by the simple ex-
be observed from Fig. 16 that the total vortex circulation pression √
approaches the value of bound circulation at large radial rc (t) = 4ανt (7)
distance away from the center of the vortex. where α is Lamb’s constant and ν is the kinematic viscos-
ity of the fluid. This result is based on the assumption that
Tip Vortex Core Growth the flow inside the vortex is completely laminar, although
this is really not a correct assumption based on the flow vi-
The relative viscosity of the fluid plays a substantial role sualization results shown previously in Fig. 11. The term t
in the evolution of all lift-generated tip vortices. This is es- is time, which represents in this case the real time elapsed
pecially true in the case of MAV-size rotors, which always since the tip vortex was trailed from the blade. The use
create tip vortices that have much lower vortex Reynolds of real time to compare the growth properties of laminar
numbers than achieved with even moderately larger rotors. tip vortices at any vortex Reynolds number is entirely ap-
The viscous spin-down and core growth behavior of these propriate. This is because the transfer of momentum be-
tip vortices are often explained using Lamb-like models. tween adjacent layers of fluid is only through molecular

10
Figure 14: Velocity and vorticity distributions across the tip vortex for wake ages of 26◦ , 63◦ , and 83◦ .

diffusion. This would mean that the time scale governing the vortex flow, on average. It is known that δ is a function
core growth is only dependent on the fluid properties (i.e, of vortex Reynolds number (Refs. 17, 21, 22); an increase
the effective kinematic viscosity), which is independent of in vortex Reynolds number results in an increase in turbu-
vortex Reynolds number if the flow is completely laminar. lence that, in turn, increases the average eddy viscosity in
For real vortices, the transfer momentum and vorticity the vortex flow and so increases the core growth rate.
between adjacent layers of fluid due to turbulence yields
the concept of eddy viscosity, which can be viewed as a Vortex models based on the Lamb–Oseen model can be
measure of turbulence in any given flow field. The corre- modified empirically to include the effects of turbulence,
sponding vortex core growth is then given by which then have different values of averge eddy viscosity
√ at different vortex Reynolds numbers. This means that the
rc (t, δ) = 4αδνt (8) time scale (which is dependent on the total viscosity) will
be different at different vortex Reynolds numbers. There-
where δ > 1 accounts for the increased eddy viscosity in fore, the comparison of vortex core growth measurements

11
Figure 15: Velocity and vorticity distribution across the tip vortex for wake ages of 123◦ , 163◦ , and 263◦ .

acquired at significantly different Reynolds numbers us- number of MAV-size rotor is about five times smaller than
ing the same time scale is probably not appropriate, and those even in the sub-scale measurements, when plotted
that the time scale must be normalized by an appropri- versus wake age the value of δ appears to be twice that
ate scaling parameter that must take into account vortex of the sub-scale measurements. This would imply that the
Reynolds number effects. tip vortices trailing the MAV-size rotor at the same age
contain on average twice the amount of turbulence com-
This concept can be better explained by plotting the pared to the vortices from the sub-scale rotor. This does
core size of the tip vortices measured in the current exper- not seem reasonable bearing in mind the much lower vor-
iment with the micro-rotor and comparing them to results tex Reynolds number at MAV-scale.
measured from a sub-scale rotor operated at much higher
vortex Reynolds numbers, which are shown in Fig. 17. It This problem can be further understood by plotting the
can be seen from this figure that the tip vortex core mea- tip vortex measurements from the micro-rotor in terms of
sured from the MAV-size rotor appears much higher at the equivalent peak swirl velocity and equivalent downstream
same wake age. Despite the fact that the vortex Reynolds distance using an Iversen-type correlation curve (Ref. 22),

12
Figure 17: Comparison of the measurements of tip Figure 19: Variation of the tip vortex core growth ver-
vortex core size at different Reynolds number with in- sus normalized time for the sub-scale rotor measure-
creasing wake age. ments.

Figure 20: Variation of the tip vortex core growth


with normalized time for the MAV-size rotor measure-
ments.
perienced by the filament as it convects in the flow. The
initial core size, r0 , takes into account the thickness of the
boundary layer on the blade (which also rolls up into the
Figure 18: Iversen-type of correlation function used to tip vortices) and the total viscosity addresses the effects
consolidate all tip vortex measurements. of scaling (vortex Reynolds number, Γv /ν). Using dimen-
sional analysis, an equivalent time parameter can be de-
as shown in Fig. 18. Tip vortex measurements acquired fined as
at different vortex Reynolds numbers (both rotating- and t ζ/Ω ζR Γv
fixed-wings) have been plotted along with Iversen’s solu- Te = = 2 = (9)
c2 /Γv c /Γv c ΩRc
tion. It is apparent from the figure that the measurements
from the micro-rotor correlates well with the Iversen with the assumption that any strain effects on the vor-
model. This suggests that the growth characteristics of the tex flow are negligible. This equivalent time parameter
tip vortices trailed from the micro-rotor should be substan- can also be derived from the non-dimensional similarity
tially similar to that generated by any type of tip vortex. A variable used in the Ramasamy–Leishman vortex model
substantially higher value of δ (growth rate) suggested by (Ref. 17), which is an exact solution to the Navier–Stokes
Fig. 17, therefore, cannot be physical. equations for a vortex flow.
The viscous core size and growth rate of the vortex at Using this equivalent time concept, the measured core
any given time depends at least on four parameters: 1. The growth of the tip vortex is compared in Figs. 19 and
initial core size as it leaves the blade, 2. The total viscos- 20 with measurements from the larger sub-scale rotor.
ity of the fluid (kinematic plus eddy, ν + νt ), 3. Time or Clearly, the initial core radius of the MAV-size rotor is
wake age and, 4. The strain (stretching or squeezing) ex- much higher. This is expected and, as mentioned earlier,

13
sions have been drawn from this study:

1. The hover performance of a MAV will directly de-


pend on its effective disk loading and its aerody-
namic efficiency (FM). Decreasing the effective disk
loading increases the power loading and attempting
to raise the FM by reducing both induced and profile
losses are the primary requirements for developing
any form of efficient hovering MAV.
2. The maximum FM values measured for this small
rotor were found to be substantially lower (no more
than 0.5) than for rotors operated higher chord
Reynolds number (which often approach 0.8). The
increased boundary layer thicknesses on the blades
Figure 21: Normalized circulation of the tip vortices and the more turbulent wake trailed from the blades
from the MAV-size rotor relative to Iversen’s solution. (both of which increase losses), seem to play an im-
portant role in reducing the FM of rotating-wings
is partly a result of the relatively thick boundary layers on that are operated at low chord Reynolds numbers.
the blade that determine the initial structure of the tip vor-
tices. Upon comparing the results in the two figures, it is 3. Viscous effects, which are relatively more important
apparent that when plotted in terms of equivalent time Te , at low Reynolds numbers, appears to affect the initial
the core growth of the MAV-size rotor is nearly identical formation and roll-up of the blade tip vortices, which
to that of the sub-scale rotor. Even though the growth rate do not reach their full circulation until some distance
of tip vortices is much smaller than that was predicted ear- distance downstream of the blade. This roll-up is fol-
lier based on Fig. 17, it is higher despite the lower vortex lowed by diffusion of vorticity, which results in an in-
Reynolds number. creased tip vortex core size and a reduced peak swirl
A further analysis of the micro-rotor tip vortex mea- velocity with increasing wake age.
surements is obtained by plotting the ratio of core circula-
tion to the total vortex circulation as a function of vortex 4. The properties of blade tip vortices that are measured
Reynolds number (as obtained from the Iversen’s exact at vastly different vortex Reynolds numbers must be
solution to the N–S equations (Ref. 22) – see the results properly analyzed on the same equivalent time scale.
in Fig. 21. A larger value for this ratio (0.707 for laminar It was shown than by using an equivalent time param-
flow) would mean that the vortex will diffuse its contained eter to compare the vortex core sizes the growth rate
vorticity more slowly. It is apparent that the value of this of the tip vortices is similar to that found at higher
ratio for the micro-rotor measurements is smaller when vortex Reynolds numbers.
compared with the Iversen’s solution at the measured vor-
tex Reynolds number (10,000). This may be because of
the observed increase in boundary layer thickness on the Acknowledgments
blade (because of the lower chord Reynolds number), al-
though further work must be done to examined this hy- This research was supported, in part, by the Multi-
pothesis. University Research Initiative under Grant ARMY
W911NF0410176. Gary Anderson is the technical moni-
tor. The authors wish to acknowledge the contributions of
Jayant Sirohi and Moble Benedict in helping to measure
Conclusions the performance of the micro-rotor.

The performance of a small rotor typical of application to


a rotating-wing micro-air vehicle has been measured. This References
was accompanied by flow visualization and PIV measure-
ments in the wake of the rotor. It has been shown that 1 Ellington, C. P., and Usherwood, J. R., “Lift and
the wakes generated by the rotor blades are thicker and Drag Characteristics of Rotary and Flapping Wings,”
more more turbulent than compared to the wakes gener- Fixed and Flapping Wing Aerodynamics for Micro-Air
ated by rotors operating at higher chord Reynolds num- Vehicle Applications, Edited by Thomas A. Mueller,
ber. A closer examination of the vortex sheets revealed Vol. 195 of AIAA Progress in Aeronautics and Astronau-
a more organized series of discrete eddies along the blade tics, Chap. 12, AIAA, Reston, VA, 2001, pp. 231–248.
span. The tip vortex structure also showed some important
differences from that expected on rotors operated larger 2 Tarascio,M., Ramasamy, M., Chopra, I., and Leish-
chord Reynolds numbers. The following specific conclu- man, J. G., “Flow Visualization of MAV Scaled Insect

14
Based Flapping Wings in Hover,” Journal of Aircraft, 16 Duraisamy, K., and Baeder, J., “Studies on Tip Vortex

Vol. 42, (2), March 2005, pp. 355–360. Formation, Evolution and Control,” Ph.D. Dissertation,
3 Ramasamy, University of Maryland, April 2005.
M., Leishman, J. G., and Beerinder Singh.,
”Wake Structure Diagnostics of a Flapping Wing MAV,” 17 Ramasamy, M., and Leishman, J. G., “A Generalized
2005 SAE Transactions Journal of Aerospace, Vol. 114 Model For Transitional Blade Tip Vortices,” Journal of the
(1), February 2006, pp. 907–919. American Helicopter Society, Vol. 51, (1). January 2006,
4 Laitone, E. V., pp. 92–103.
“Aerodynamic Lift at Reynolds Number
Below 7×104 ,” AIAA Journal, Vol. 34, (9), September 18 Leishman, J. G., Baker, Andrew., and Coyne, Alan.,
1996, pp. 1941–1942. “Measurements of Rotor tip Vortices Using Three-
5 Laitone, Component Laser Doppler Velocimeter,” American He-
E. V., “Wind Tunnel Tests of Wings at
licopter Society International Aeromechanics Specialists
Reynolds Number Below 70,000,” Experiments in Flu-
Conference, Bridgeport, CT, October 11–13, 1995.
ids, Vol. 23, 1997, pp. 405–409.
19 Lamb, H., Hydrodynamics, 6th ed., Cambridge Uni-
6 Leishman, J. G., Principles of Helicopter Aerodynam-
ics, 2nd Edition, Cambridge University Press, New York, versity Press, Cambridge, 1932.
NY, 2006. 20 Oseen, C. W., “Uber Wirbelbewegung in einer reiben-
7 Martin,P. B., Pugliese, G., and Leishman, J. G., “High den Flüssigkeit,” Arkiv för Matematik, Astronomi och
Resolution Trailing Vortex Measurements in the Wake of Fysik., Vol. 7, (14), 1911, pp. 14–21.
a Hovering Rotor,” Journal of the American Helicopter 21 Squire,
H. B., “The Growth of a Vortex in Turbulent
Society, Vol. 49, (1), January 2003, pp. 39–52. Flow,” Aeronautical Quarterly, Vol. 16, August 1965,
8 Ramasamy, M., and Leishman, J. G., “Interdepen- pp. 302–306.
dence of Diffusion and Straining of Helicopter Blade Tip 22 Iversen,
J. D., “Correlation of Turbulent Trailing Vor-
Vortices,” Journal of Aircraft, Vol. 41, (5), September–
tex Decay Data,” Journal of Aircraft, Vol. 13, (5), May
October 2004, pp. 1014–1024.
1976, pp. 338–342.
9 McAlister,K., “Rotor Wake Development During the
First Rotor Revolution,” Journal of the American Heli-
copter Society, Vol. 49, (4), October 2004, pp. 371–390
.
10 Heineck,
J. T., Yamauchi, G. K., Wadcock, A. J., and
Lourenco, L., “Application of Three-Component PIV to
a Hovering Rotor Wake,” American Helicopter Society
56th Annual National Forum, Virginia Beach, VA, May
2–4, 2000.
11 Yu., Y. H., and Tung, C., “The HART-II Test: Ro-

tor Wakes and Aeroacoustics with Higher Harmonic Pitch


Control (HHC) Inputs,” American Helicopter Society 58th
Annual National Forum, Montréal, Canada, June 11–13,
2002.
12 Cook, C. V.,
“The Structure of the Rotor Blade Tip Vor-
tex,” Paper 3, Aerodynamics of Rotary Wings, AGARD
CP-111, September 13–15, 1972.
13 Kraft,C. C., “Flight Measurements of the Velocity
Distribution and Persistence of the Trailing Vortices of an
Airplane,” NACA TN 3377, 1955.
14 Boatwright, D. W., “Three-Dimensional Measure-
ments of the Velocity in the Near Flow Field of a Full-
Scale Hovering Rotor,” Mississippi State University Re-
port EIRS-ASE-74-4, 1974.
15 Leishman, J. G., “On Seed Particle Dynamics in
Tip Vortex Flows,” Journal of Aircraft, Vol. 33, (4),
July/August 1996, pp. 823–825.

15