DISTRIBUTION STATEMENT A: Distribution approved for public release; distribution is unlimited.

Deconstructing Hub Drag: Final Report on Experiments

Alexander H. Forbes, Vrishank S. Raghav, Michael Mayo, Narayanan M. Komerath,Daniel Guggenheim School of Aerospace Engineering

Georgia Institute of TechnologyAtlanta, Georgia 30332-0150

phone: (404) 894-3017 fax: (404) 894-2760 email: komerath@gatech.edu

Award Number: N00014-091-1019http://adl.gatech.edu/research/hubdrag/

1

SUMMARY

A progression of experiments deconstructed the hub drag of a helicopter. Starting with a cylindricalshaft, drag contributions from different elements were built up on a generic model of the hub of a4-bladed rotor. The experiments were conducted in the 2.13m x 2.74m (7 ft x 9 ft) wind tunnel atGeorgia Tech. The model included a generic hub, pitch links, blade shanks and scissor links. Load cellsseparating the 6 components of airloads, laser Particle Image Velocimetry (PIV), and hot film constanttemperature anemometry were used to capture the values and physics of drag contributions at the levelneeded to refine predictions. The generic model had dimensions approximating a 25-percent scalemodel of a modern rotorcraft hub. Identified issues including the role of hub rotation in airloads andflow deviation were examined in greater detail. Mounting the motor driving the rotating hub directly toa 6-DOF load cell enabled azimuth-resolved measurement of all 6 airload components. This in turnenabled identification of interaction effects. Various hypotheses were explored to investigate reasons forcited discrepancies in the published literature. The side force loosely ascribed to the Magnus effect wasstudied for its contribution to the vehicle power demand, indistinguishable to the engine from powerdemand due to drag. The side force was found to be highly unsteady depending on flow conditions, andits azimuth-averaged value could change sign with advance ratio. The torque required to drive the hubincreased with rotation speed and freestream speed, but was found to be of the same order as thatexpected from the drag difference between shanks on the two sides. The azimuth-resolved force andtorque data suggest that transient torque values peak at certain portions of the azimuth that slightly leadthose of highest drag, for the configurations studied. The addition of a scissors linkage to the generichub resulted in a 10 percent increase in drag above the prior generic configuration, but reduced theamplitude of periodic drag variation. Wake surveys using a hot-wire anemometer probe, and usingParticle Image Velocimetry, captured the flow deviation, wake deficits and frequency content of thewake. The large-amplitude fluctuations posed challenges to hot-film probe anemometry in the sharpestgradients of the wake shear regions. The frequency content was surveyed using both Fast FourierTransform and Wavelet Transform techniques. The total power demand attributed to the hub is stillmostly due to drag. Scaling to vehicle flight speed and size should improve this approximation, unlessinteractions with pylons, nacelles or other components cause asymmetries sufficient to change theresult. In general, the results showed that the drag of individual components behaves as expected, and isexplained to good accuracy by relatively simple estimation techniques. Interaction effects betweenflowfields of different components pose substantial issues to accurate prediction. An example was thefinding that addition of the shanks did not increase drag, because the wake of the shanks reduced thedynamic pressure experienced by downstream components, and the flow deviation reduced the drag dueto base separation. Diagnostic advances included high-accuracy, reliable 6-DOF load measurements inwind tunnels using load cells, and the ability to capture finely-resolved azimuthal variations of all 6 loadcomponents with a rotating setup. These set the stage for deeper studies of interaction effects. Acomputational effort was conducted under the same project using Navier-Stokes solvers and a PhDthesis is in progress from that effort at this writing. Those results are expected to be reported separately.

2

CONTENTS

NOMENCLATURE 4

INTRODUCTION 5

LONG-TERM GOALS AND OBJECTIVES 6

PRIOR WORK 7

MODEL CONFIGURATIONS 10

AERODYNAMIC LOADS 16

VELOCITY FIELD MEASUREMENTS 33

WAKE VELOCITY FLUCTUATION MEASUREMENTS 39

SUMMARY OF DATA AND RESULTS 49

CONCLUSIONS 56

REFERENCES 58

WORK COMPLETED 62

IMPACT/APPLICATIONS 62

TRANSITIONS 62

RELATED PROJECTS 63

PUBLICATIONS 63

HONORS/AWARDS/PRIZES 63

DEGREES EARNED 64

ACKNOWLEDGMENTS 64

3

NOMENCLATURE

b Cylinder radius, m

CDC Cylinder drag coefficient

D Drag, N. Also used to denote hub diameters as in 1D, 2D etc.

Dh) Hub diameter Y Side Force, N

τz Torque around axis of rotation, N-m

Da Drag of advancing side, N

Dr Drag of retreating side, N

FM Side force due to Magnus effect, N

Fs Data Sampling Rate, Hz

f f Fourier Frequency, Hz

R Hub radius, m

Re Reynolds number

s Wavelet Scale

U∞ Free stream velocity, m/s

α Spin Ratio

µ Advance Ratio: ratio of freestream speed to shank tip speed, (1/α)

Ω Hub rotation, RPM

ω Hub rotation, RPS

ρ Free stream density, kg/m3

Ψ Azimuthal angle, degrees

P Power, W

4

INTRODUCTION

Parasite drag is a limiter of helicopter forward flight speed and efficiency. The rotor hub is a challenge

to designers and aerodynamicists alike: it contributes up to 50 percent of the parasite drag of the

vehicle Charles N. Keys (1975). It poses a large and complex obstacle in achieving, predicting and

reducing drag. The drag contribution due to individual elements can be calculated to a reasonable

approximation for conceptual design from the methods given by Hoerner (1965). However, experience

in the literature appeared at the outset to suggest that the effect of a rotating hub assembly on the

parasite power of a helicopter in flight, was substantially different from those obtained using the above

approximations. This project sought to investigate these discrepancies using generic configurations and

basic research.

The experimental work was conducted in 3 stages. Initial experiments measured drag of a non-rotating

hub assembly as it was built up piece by piece starting with a cylindrical shaft. A drag deconstruction

process was used to gauge the difference between the drag on the assembly versus a linear superposition

of the drag of individual components. Next, flowfield measurements were conducted. A Pitot probe

rake was initially used to survey the dynamic pressure deficits in the wake. These results are not

presented in this report, since they were superseded by Particle Image velocimetry maps. Interaction

effects appeared for instance where the wakes of blade shanks reduced the onset flow dynamic pressure

for separation at the aft portion of the hub, thereby reducing the drag below that expected from

superposition. The flow features in the wake of a rotating hub were captured using Particle Image

Velocimetry. As reported in Raghav et al. (2012), once the loads agreed with computations, the detailed

wake structure behind a rotating hub agreed with surprising fidelity with computations, showing at least

that any unmodeled support interference was negligible.

The next area of major uncertainty was the effect of rotation, especially as relevant to high-speed flight

of compound helicopters. Wind tunnel and flight test lore suggested much stronger effects of rotation

on the overall power increment ascribed to drag, than cursory analysis would explain. Our experiments

examined the question of whether the side force, or the power required to generate the torque to

overcome asymmetric loads, may be substantial contributors to the vehicle power requirement

(indistinguishable from power required due to drag). The magnitude and relevance vary widely with

vehicle size and advance ratio, which translate to large changes in relative velocities and Reynolds

numbers. Regimes with steady wakes as well as those with periodic shedding were encountered. Prior

work had also shown the importance of the horseshoe corner vortices generated where the hub mast

meets the vehicle hub fairing. These vortices can interact with the flowfields of the blade shanks,

5

generating periodic effects and pulsing effects on the drag and side force. Such vortices are not

expected in the present experiments as there is no hub fairing, nacelle or fuselage present.

Major challenges in measuring and predicting hub drag include the following:

1. Complex Geometry

(a) The complex geometry leading to flow interactions over the hub create interference drag

which is hard to interpret.

(b) Interactions augmented by presence of fine structures such as tubes, wires, linkages and

fasteners.

2. Effect of Rotation

(a) Consideration of all hub structures in a typical assembly requires the analysis of a large

range of Reynolds number effects.

(b) Measurement and prediction of drag of rotating components is a challenge in itself.

(c) For high speed applications, the hub may experience compressibility effects, adding a new

dimension of complexity.

LONG-TERM GOALS AND OBJECTIVES

The long-term goal was to develop methods to understand the airloads on complex-shaped objects and

vehicles.

The objectives of this project were:

1. Isolate and quantify the different sources of hub drag.

2. Tighten tolerances of empirical upper-bound predictions suitable for conceptual design, by

reference to the basic experiments.

3. Enable first-principles prediction of drag through computational aerodynamics, suitable for the

preliminary design stage and beyond.

4. Advance the state of knowledge on flows around complex configurations involving flow

separation and a wide range of Reynolds numbers.

6

PRIOR WORK

Early work on hub drag is described in Churchill and Harrington (1959); Linville (1972); Sheehy

(1977). Churchill and Harrington (1959) did full scale tests on a direct tilting hub with no blade shanks

at 0, 100, 150, and 200RPM at a velocity of 50 m/s. Rotation was found to have no effect on parasite

drag. Tests were also done on a fully articulated hub with a swash plate and control system but no

shanks, at 50RPM and at velocities of 24, 29, 35 and 42 m/s. A very slight increase in drag was found

with increase in advance ratio. Linville (1972) tested a Sikorsky S-65-200 compound helicopter hub

with no fairing, a floating fairing that incorporated a pylon and a boundary layer control system, and a

rigid fairing with a sealed cover for the rotor head. The gross model drag increased slightly as RPM

increased from 0-100%, for some configurations, attributed possibly to the Magnus effect. The data did

not reveal any consistent significant effect of RPM on the model or rotor head drag. Sheehy and Clark

(1976) and Sheehy (1977) reviewed available hub drag data, and pursued reduction of the interference

drag between the hub and the pylon. Charles N. Keys (1975) laid out guidelines for reducing the

parasite drag due to rotor hubs. Kerr (1975) studied the relationship between helicopter drag, loads and

component life. Hoffman (1975) studied the relationship between helicopter drag, stability and control.

Williams and Montana (1975) laid out a comprehensive NASA plan to reduce helicopter drag.

The bluff body wakes due to rotor hub components have a large detrimental impact on the performance

of many types of rotor powered vehicles. Work on the topic has used analogies with simple model

geometries such as cylinders Roshko (1993); Chng and Tsai (2006). More complex hub geometries

were studied recently by Bridgeman and Lancaster (2010a,b) using an unstructured Navier-Stokes flow

solver for the bluff body problem, and sought to identify a physics based analysis methodology capable

of accurately predicting drag on realistic geometries. Wind tunnel data showed that the typical

computational resolution used at Bell Helicopter for fuselages was able to predict drag to within 5% of

measured values. Much finer meshes were needed to achieve the same results for hub configurations.

The addition of hub fairings is meant to reduce separation and interference drag from the rotor hub and

fuselage, see Young et al. (1987). Drag reduction on a 1/5 scale Bell Textron Model 222 Helicopter of

20% was shown by Martin et al. (1993) by using a small hub fairing combined with a non-tapered

pylon fairing. Without the small hub fairing the model’s drag was reduced by 14.5%. Martin et al.

(1993) additionally showed that for an unfaired hub there was negligible influence on the model forces

due to the hub’s rotation rate. This agrees with the Linville (1972) result that Ω had no noticeable or

consistent effect on total model drag.

Felker (1985) tested an XH-59A advancing blade concept helicopter rotor hub at four different

7

configurations. The hub had very few protuberances compared with conventional, articulated helicopter

rotor hubs, with no flap hinges, lead-lag hinges, or lead-lag dampers. Pitch horns and pitch links for the

upper rotor were installed in the rotor hub. The faired hub drag was not significantly affected by

rotation. The unfaired hub drag was significantly influenced by rotation. The aerodynamic cleanliness

of the rotor hub may have been a factor in how the rotation rate reduced the drag. The drag of a circular

cylinder is initially reduced by rotation about its axis, and a similar effect could occur on the inter-rotor

shaft of the XH-59A. Felker found that for the unfaired hub’s of the XH-59A the drag area was actually

reduced by the rotation of the hub, possibly as a result of the cleanliness of the XH-59A relative to other

hub designs.

Increasing the distance between the rotor hub and pylon has shown no significant reduction in drag due

to an offsetting increase in rotor shaft and control drag, per Charles N. Keys (1975). This offsetting

increase in rotor shaft and control drag could be countered by placing the pitch links in the shaft and

adding a fairing to the shaft. Keys and Rosenstein showed that at low advance ratio, an increase in

average shank dynamic pressure explains the drag increase. However, wind tunnel tests with and

without rotation do not conform to expectations. Wind tunnel tests performed on compound rotorcraft

show large uncertainty in source and magnitude of what is called interference drag, which must be

well-understood if reductions in hub drag are to be consistently applied.

While gaps in the understanding of the nonlinear physics remain, advances have been made in hub drag

design. Fairing designs have been explored by industry, see Wake et al. (2009a), to reduce flow

separation and interference drag between the hub and fuselage. To date, frontal swept area of the hub

design has been the leading parameter tied to hub drag per Sheehy and Clark (1976), therefore the

fairing of an existing hub design does not address the issue directly. This is especially true for

articulated hubs, where empty space is required for the control hinges. While empirically corrected

analytic estimates have been developed to predict hub drag based on frontal area, there is no consistent

trend when accounting for interference effect and frontal swept area, per Sheehy (1977). NASA and

industry researchers showed that a cambered, flat-bottomed hub fairing was effective due to elimination

of separated flow between the hub fairing and pylon, and elimination of shedding from upper corners of

the pylon. Gaps between the pylon and cambered fairing negate benefits from having a cambered flat

lower surface. A survey by Carter Copter shows complex flow interactions. Hub/pylon gap interference

can cause a superflow region where Mach numbers can reach high, even supersonic, levels in the case

of compound rotorcraft. Components encountering this flow experience increased drag. More recently,

in-flight data on the RSRA vehicle were compared to wind tunnel test data with some success. Bluff

body drag has been studied on automobiles, however the complexities due to relative motion of

8

components and the rotor-airframe-wake interaction of hub drag are unique to rotorcraft.

Configuration-based factors affecting hub drag include the center section, the shanks, hoses and cables,

the pylon/ hub gap, the fuselage attitude, the drive shaft, swash plate, pitch link rods, the pylon opening

beneath the hub, the number of blades, and fairing leakage. Several of these components are in the

transitional regime of Reynolds number under typical full-scale rotorcraft operating conditions, and

have large drag coefficients. The confusion regarding the effect of rotation becomes evident when

dealing with compound rotorcraft that operate at high advance ratios. Under these conditions, the

predicted contribution from rotation should be smaller than the data apparently indicated.

Flow separation due to interactions between the hub, blade shanks and pylon is a large contributing

factor to rotor hub drag, per De Gregorio (2012b). Recently, the FLUENT code was applied to dual

rotor hubs for validation with wind tunnel data, per Wake et al. (2009b). Acceptable correlation was

achieved for bluff-body geometries, but a hexahedral mesh was needed. Methods for drag reduction

were presented including fairing reshaping to reduce flow separation and interference drag, splitters and

vanes to reduce flow separation, flow control using steady and unsteady blowing concepts in the

non-rotating frame of reference, streamwise vorticity generation to help the flow remain attached on the

fairings, and reshaping of the pylon/fuselage to redirect flow away from the bottom hub fairing. The

vane and splitter methods showed marginal benefits, but the fairing reshaping resulted in more than

70% relative drag reduction. Larger shaft fairings reduced drag on the top and bottom hub fairings.

Reshaping of the shaft fairing airfoil shape reduced drag on all components of the hub, an hour glass

shape of the shaft fairing in the vertical direction reduced top and bottom hub drag, and improved

shaping to reduced interference drag.

Several experimental efforts on hub drag have been directed towards improving drag characteristics of

current hub designs by the addition of fairings. Considerations for hub displacement from the fuselage

have been made, weighing the effects of increased frontal area to decreased interference Sheehy and

Clark (1976).

A typical helicopter hub is comprised of a myriad of bluff body components, mainly due to

compromises in manufacturing procedures. The distinction between streamlined and bluff bodies is

associated with the poor performance characteristics (such as lift to drag ratio) of the latter

configurations. The primary drag component in these bluff bodies is due to flow separation rather than

viscous effects. The pressure drag of bluff bodies results from vortices in the wake that are shed from

the structure.

Bluff body wakes associated with rotor hub components affect the performance of both commercial and

military air vehicles Gregory et al. (2008). While many studies of two-dimensional (2D) bluff bodies

9

have been performed in the past, three-dimensional (3D) studies have been typically restricted to

spheres and cylinders of varying aspect ratios and/or Reynolds numbers. Reynolds number sweeps

provide a sequence of distinct bifurcations which produce significant and measurable differences in the

flow field per Roshko (1993). Seidel et al. (2006) experimentally found two counter rotating helical

modes in the wake of bullet shaped bodies.

The current objective is the quantification of different sources of hub drag aimed at tightening the

tolerances of empirical upper-bound predictions suitable for conceptual design by reference to basic

experiments. Recent experimental investigation on a complete helicopter by De Gregorio (2012a) and

coupled experimental and CFD investigations by Antoniadis et al. (2012) suggest that the main hub

wake formation as a reason for the unsteady forces observed on the horizontal stabilizers. Hence, the

flow field details could provide useful insights into the well known problem of helicopter “tail

shake” Ishak et al. (2008); de Waard and Trouve (1999).

Prior aerodynamic experiments in wind tunnels on rotating models (see Horanoff (1969); Mehta (1985);

Smits and Smith (1994); Sayers and Hill (1999); Asai et al. (2007)) were unable to resolve the loads

with high azimuthal precision. We were able to acquired azimuthally resolved, 6-DOF airloads on

rotating hubs.

MODEL CONFIGURATIONS

Experiments were conducted in the 2.14m x 2.54m (7’ x 9’) test section of the John J. Harper low speed

wind tunnel located at the Daniel Guggenheim School of Aerospace Engineering of Georgia Institute of

Technology. A generic four bladed hub model was assembled to approximately one-quarter scale to that

of a 10-ton helicopter (Figure 1). The model consists of the hub plates, shanks for the rotor blades, a

swashplate with pitch links, drive shaft, and supports. The effect of capping the shanks was studied

initially, and then blade shanks were capped for the rest of the experiments performed. In the final stage

of experiments, two models of the scissors linkages used on modern hubs were also attached to the

swash plate, shown in Fig. 5. A complete description of the model can be found in Raghav et al. (2012).

The model includes structures to represent hub plates, blade shanks, a swashplate, pitch links, drive

shaft and the required hardware for assembly. Though greatly simplified from the complexities of

full-scale hubs with hydraulic lines and intricate mechanisms, this model provides the interference and

flow separation attributed to the full-scale system while maintaining the principal geometric

characteristics. The model was adapted into three configurations, as shown in Figure 2.

Configuration 2(a) has unplugged blade shanks, configuration 2(b) has plugged shanks, while

configuration 2(c) has plugged shanks and a capped region between the two hub plates

10

Figure 1: Hub model setup inside the John J. Harper wind tunnel test section

(a) Unplugged (b) Plugged (c) Hubcapped

Figure 2: Three model configurations for static hub experiments and computational simulations

11

Full Hub No Shanks No Hub No Pitchlinks No Swashplate

Full hub + scissors

Figure 3: Flowchart depicting progression of deconstruction.

Figure 3 shows the progression of deconstruction tests, followed by final experiments with the full hub

with and without a pair of scissors components added.

In the final stage of experiments for rotating hub airload measurements and hot-film anemometer

surveys, the hub was mounted on a stepper motor as seen in Fig. 4. The motor was mounted on a load

cell for force and torque measurements. The load cell was supported by a tripod attached to the wind

tunnel support structure. Unless otherwise stated on any of the following figures, a circular fairing was

placed around the motor.The hub was accelerated to the desired Ω using the stepper motor. After the

hub reached the desired Ω, the wind tunnel was turned on and taken to a specified speed. When finding

azimuth resolved loads the hub was first set to 0 yaw and the tunnel was set to the desired speed. Once

at the test speed the hub was then accelerated to Ω. Knowing the time taken to reach Ω permitted the

instantaneous counter reading to be related to the instantaneous azimuth of the hub based on the yaw

reference.

Magnus force, the force perpendicular to U∞ around the rotating hub is also of interest. The Magnus

force on a cylinder can be negative over a small range of parameters per Fletcher (1972). The

possibility of this occurring on the rotor hub was investigated. In addition to the Magnus force, the

effect of varying Ω and tunnel speeds on drag and torque were investigated. The effects of the advance

ratio, µ described in Eqn. 1 on drag, side force, and torque around the axis of rotation were also

investigated. Some test conditions were selected to coincide with the work of Fletcher (1972), who

described a critical Reynolds number regime in which negative Magnus effect was shown to occur over

rotating cylinders. His results specified a Reynolds number range from 1 x105 to 5 x105. The critical

Reynolds number regime corresponds to wind tunnel speeds ranging from 6 to 13m/s. In addition to

focusing on the critical Reynolds number regime, critical spin ratio (α) values were also investigated.

The critical α value is approximately 0.2, corresponding to an advance ratio of 5. The side force in

these two sets of critical regions was specifically investigated.

12

Test Configuration Azimuth/RPM Increments VelocityStatic 2(a),2(b),2(c) 0−90 15 0-70 mphRotating 2(b) 4, 80, 240 RPM - 0-70 mph

Table 1: Summary of Experimental Test Conditions

The progression of data acquisition was initiated with static hub tests performed on all the

configurations to understand the difference in drag behavior. The model was swept through azimuthal

orientations throughout a series of runs performed at zero angle of attack (pitch angle). Six-axis force

transducer data were obtained for a 90 azimuthal sweep in 15 increments at a range of tunnel speeds.

Subsequently, tests were performed on a rotating hub with the configuration 2(b). A range of rotation

rates from 4 to 240 rpm was explored. The test conditions are summarized in Table 1.

In addition, the plugged shank configuration was broken down in stages to measure the individual

contribution of each structural component. Since most theoretical drag prediction methods have been

developed for the interference effects between two streamlined bodies, or one streamlined body and one

less so, for example a wing joint with a fuselage, measuring the interference effects of the hub

components was left to experimental trials. Therefore, the hub drag was measured in progression as

depicted in the flowchart (Figure 3). The effect of rotation on the deconstructed model was also

measured and is presented in this report.

Intermediate results were presented in several publications. Ortega et al. (2011) presented

measurements and computations on the deconstruction of the drag of a generic hub in a wind tunnel

freestream. The scaling of loads between model scale and full scale helicopter hubs was

investigated Shenoy et al. (2011, 2012). Raghav et al Raghav et al. (2012) reported loads, wake velocity

fields, turbulence spectra and rotation effects on the same generic hub model.

µ =60U∞

2πΩR(1)

bf Support and probe interference

Support interference is always present in wind tunnel experiments. In the experiments conducted here,

standard precautions were taken, such as ensuring that maximum local blockage of test section frontal

area was well below the accepted 5 percent. Fairings were used around the stepper motor to minimize

azimuthal variations in interference. Tare readings without the model mounted, and with the model

mounted and no flow on, were used for linear subtractions, which are of course not perfect ways if there

is substantial interference present. Comparisons with computations presented in the papers published

13

Figure 4: Rotor hub model and motor

Figure 5: Swash plate and scissors link assembled with drive shaft

14

Figure 6: Sketch of the support structure for wake PIV measurements

from this effort, for instance Shenoy et al. (2011, 2012) and Raghav et al. (2012) indicate that the

support interference effects were indeed minimal, or were adequately accounted with the given support

geometry.

In the case of the wake measurements, downstream and in-plane interference are both serious concerns.

Two different setups were used, one for the Particle Image Velocimetry data acquisition, and the other

for the hot-film anemometer surveys. The PIV set up is considered first. It is sketched in Figure 6. The

items placed inside the test section included a traverse system to move the laser sheet generating optics

vertically and horizontally, and the PIV camera mounted on its standard camera rails and attachments,

attached to the tunnel floor directly below the horizontal measurement plane. Concern about

interference would be justified in this case. However, as shown in Raghav et al. (2012), near-perfect

agreement was shown by the computational group with the shear-layer profiles provided to them,

without considering the cameras and light sheet generating setup in the computational grid.

In the case of the hot-film anemometer probe surveys, probe interference is again a concern, and

unavoidable. This was minimized by pointing the probe upstream so that the flow encountered the

sensor with minimal disturbance. However, the support attachment for the probe, and the traverse

system, were in this case far less intrusive than the setup for the PIV wake surveys, as shown in Figure

7. Based on the above reasoning, we submit that there is no need to model any support interference in

correlating with any of the data in this report.

15

Figure 7: Sketch of the support structure for wake hot-film probe anemometry

AERODYNAMIC LOADS

Force measurements used an ATI Gamma load cell, whose specifications are shown in Table 2. The hub

setup was mounted on the load cell. Force and torque data from the load cell were recorded at the

sampling frequency described in Equation 2, which is far above the expected 100Hz frequency response

range of the load cell. The goal of the force measurements was to find the side force, drag, and torque

around the axis of rotation for several and tunnel speeds varying from 0 to 23m/ s.

Fs = × 3600Hz (2)

Measurement Range Sensitivity LimitX and Y Force 130 N 0.025 N

Z Force 400 N 0.05 NX and Y Torque 10 N-m 0.00125 N-m

Z Torque 10 N-m 0.00125 N-m

Table 2: LOAD CELL SPECIFICATIONS

For rotating hub experiments, load cell data acquired at a high digitization rate were phase-averaged

with 1 degree azimuthal resolution, indexed to a selected position of the hub. The formula for number

of data points in each set is shown in Equation 3. The FS for each was chosen in Equation 2, such that

there would be 10 data points per rotation per degree for phase averaging. The technique was verified

16

through a static test case comparison to the quasi-steady rotation of an aspect ratio = 1 circular cylinder.

Data points per degree =FS

ω(3)

Selected results are presented in a progression from integrated loads to flow field details. The results are

broken down into three main categories:

1. Static and dynamic force measurements

2. Wake velocity characterization.

3. Wake frequency spectra

First, the measurement of integrated drag and side force is presented from experimental results. In

addition, the effects of rotation are inferred. Next, the wake velocity field is examined. Finally, spectra

of speed fluctuations in the wake are presented. Figure 8 illustrates the static, rotating and deconstructed

hub drag for a range of wind tunnel speeds.

Several configurations are shown together to highlight the many qualities of the drag contributions. The

unplugged configuration had the shank tubes open for through-flow: this was rejected as

non-representative of helicopter applications. The capped configuration closes the remaining

through-flow gaps in the hub. This is seen to have only a minor effect on drag. The capped hub

configuration creates complete flow blockage through the center of the hub plates and results in the least

drag at the 45 static orientation. Therefore, it appears that the drag due to flow separation on the

capped configuration is less than the interference drag caused by the channel-like flow through the

center of the hub on the non capped configuration. The maximum drag is obtained for the static hub

when it is oriented at 0 azimuth (blade shanks normal and perpendicular to the free stream). This

orientation corresponds to the maximum frontal area of the hub. The 45 static orientation and the hub

in rotation at 240 rpm result in nearly the same drag. This result is not unexpected, as the 45

orientation is equivalent to the average frontal area of the hub in rotation. The result is also in

agreement with previous findings by Sheehy (1977) for unfaired hubs.

The drag forces for each test condition are shown in Figure 11. For the values of Ω tested, rotation did

not have any noticeable effect on the drag of the rotor hub. Figure 12 displays the roughly 10 percent

increase in drag caused by the scissors at each U∞.

Drag normalized by the tunnel dynamic pressure, as shown in Figure 9, shows little variation due to

change in Reynolds number in the speed range covered. The complete model at 0 orientation shows

17

(a)

(b)

Figure 8: Variation of the hub drag with azimuth and free stream speed.

18

Figure 9: Drag scaled by tunnel dynamic pressure.

slight Reynolds number effects, however the scaling of the vertical axis may be understating the

variation.

Collecting load data for a rotating model presents the challenge of decoupling the drag component from

the anti-torque of the rotating shaft and motor system. The force transducer is fixed just below a

compact high torque stepper motor which drives the hub shaft. Data indicate negligible coupling of the

measured torque about the shaft axis and the tunnel axis. This was verified by detailed load cell

calibration with the mounted hub model and by comparison of torque measurements with static wind

tunnel tests. In order to clarify trends, torque measurements are shown as positive values in

Figure 10(a). However, due to the orientation of the force transducer with respect to the model,

measured torques along the drive shaft axis were aligned along the negative z-axis of the load cell.

The torque measured along the drive shaft axis is observed to increase in magnitude with increased

19

(a) Measured torque along drive shaft axis (b) Side force variation with hub rotation

Figure 10: Variation in forces with changing rotational and free stream speeds.

rotation rate as delivered by the motor (Figure 10(a). The static model also measures an increasing

torque for greater tunnel speeds. The model is nearly axially symmetric, however the connection joints

of the pitch links to the blade shanks provide some asymmetry, thus generating the measured torque.

Further investigation with an axially symmetric model is planned to confirm this assumption.

Initial measurements showed little variation in side force due to rotation, and even a reduction of

measured side force at greater tunnel speeds (Figure 10(b).

The side forces observed for each test condition are shown in Fig. 13. The side force on the hub appears

to have some dependence on Ω, however there is no discernible trend due to Ω. Raghav et al. (2012)

found there was a non-zero side force for the static rotor hub, which was attributed to model asymmetry

(the pitch links are offset from the diameter). The side force followed a seemingly linear trend, with a

shift in side force at 13 m/s followed by a return to the linear slope. This shift shows a sign reversal in

side force for some range of Reynolds numbers between 2.61×105 and 3.66×105. While the side

force changes sign for certain test conditions there is no evidence that this is due to Magnus effect.

The torque was demonstrated to increase with not only Ω but also with U∞. This effect is seen in

Fig. 14. Both Ω and U∞ have a large effect on the torque required of the motor to maintain rotation. The

reason for the torque dependance on Ω was needed to move forward.

To simplify the problem, the blade shanks were considered independently to see if a reason for the

20

Figure 11: Drag Force Vs Freestream Speed for Various Ω

rising torque could be found. The drag on the blade shanks were assumed to be major torque

contributors due to their distance from the hub axis. Figure 15 shows the direction of drag at the shank

tip for advance ratios less than 1. For µ <1 the drag on the shanks on the retreating side is aiding the

rotation of the hub, however, after µ becomes >1 the drag on the retreating side opposes the direction

of rotation. Torque due to the shanks is described by Equation 4

τzshanks = (Da−Dr)R (4)

Drag on the cylinder shanks was calculated using the average velocity that would be experienced on

each shank. Since both shanks are cylinders, CDC needs to be adjusted depending on the Reynolds

number, which is near the transition between laminar and turbulent flow for many of the test cases.

Velocity used in the shank drag calculations is dependent on µ . The difference in the average velocity

between the advancing and retreating shanks is fixed based on the hub Ω. However, since drag on each

shank scales with U2, the difference in drag between the two shanks, and by extension the torque,

increases with increasing U∞ or Ω. The drag difference experienced by the advancing and retreating

sides was seen in previous experiments on rotating hubs by Raghav et al. (2012).

21

Figure 12: 120 RPM Drag Comparison with/without Scissors with no Motor Fairing and no TareRemoved

22

Figure 13: Side Force Vs Freestream Speed for Various Ω

23

Figure 14: Average Torque Around Axis of Rotation Vs Freestream Speed for Varying

Figure 15: Direction of Drag Forces of Shanks ( µ < 1)

24

Figure 16: Aerodynamic Loads Through A Hub Cycle (80 RPM 8.94 M/S)

25

Figure 17: Drag During Two Hub Cycles With No Motor Fairing Phase Averaged Through 10Seconds(120 RPM 11.2 M/S )

26

The removal of the blade shanks to examine the deconstructed model indicates that the blade shank

contribution to the total drag is negligible. Furthermore, removal of the hub plates reduced the total drag

by approximately one-third. Consequently, two-thirds of the hub drag is due to the contributions of the

drive shaft, swashplate and pitch link drag. Rotating each of these deconstructed models confirms that

there is little variation in drag from the static configuration. The pitch link drag contribution may be

computed from the shift in measurements from the prior deconstructed model.

Experimental measurements indicate that the time averaged force transducer data do not vary with the

onset of hub rotation, as noted previously. However, azimuthally varied static hub measurements do

capture a variation in the drag with orientation of the model. Note that the measurements shown in

Figure 8(a) do not include the drag due to the motor used in the drive shaft.

Using the work of Raghav et al. (2012) on the same generic hub model the drag breakdown of each

component was found. The addition of scissors was found to add 10 percent, leading to the drag

breakdown shown in Fig. 18. Interference drag could impact the amount of drag from each component.

To find if there was any impact of interference on the drag breakdown a simple estimate of drag of each

component was made. The hub shaft, pitch links, and shanks were easily simplified as plain cylinders.

The swash plate was considered to be a bluff body with a small thickness. Using these simplifications

the drag could be estimated to a reasonable degree. The comparison between calculated and actual drag

is shown in Fig. 19. The swash plate + shaft is largely under predicted, the reason for which is

unknown. The differences in drag for the shanks may be due to a reduction in drag of the shanks due to

interference with the rotor hub.

Figure 16 shows the aerodynamic loads through 1 rotation of the hub. The resolution of the drag and

side force is 1 degree, where as the resolution of the yaw moment is 5 degrees. The lower resolution is

due to poor signal to noise ratio of the data resulting from the high stepping frequency of the motor.

From the figure there are 4 orientations per rotation where the drag, side force and torque exhibit a

maximum value. These 4 peaks correspond to the azimuthal orientations of the hub where a set of

shanks is perpendicular to the flow. Additionally, at the 0 orientation the hub scissors are parallel to the

flow. The drag variation with time through 2 revolutions of the hub, with and without scissors is shown

in Fig. 17. The presence of the scissors linkages appears to cause higher drag during the azimuth

intervals where drag was low in the absence of the scissors. The drag was previously shown to increase

by 10 percent with the addition of the scissors. This is demonstrated in Fig. 17 by comparing the

average drag across the 2 cycles.

The average torque is plotted against advance ratio in Fig. 20. This plot shows a very interesting

27

Figure 18: Drag Contributions of Components

Figure 19: Component Drag Expected Vs Measured

28

Figure 20: Average Torque Vs µ for varying Ω

29

behavior. At the low rotation speed of 80 rpm, the torque variation with advance ratio has a slope

exponent less than one, i.e., the slope decreases and the curve flattens out as advance ratio increases. At

120 rpm, the slope is constant, so that the variation is linear with advance ratio. Above 120 rpm the

torque rises with increasing slope as advance ratio increases. At the highest rpm tested, 240 rpm, the

slope is very steep.

Power required from the motor was investigated in this study. The power due to each aerodynamic

load (drag (D), side force (Y) and torque (τz)) at each condition was computed using the following

expressions:

PD = D×U∞ (5)

PY = S×U∞ (6)

Pτz = τz×2πω (7)

The comparison between power usage due to each aerodynamic load is shown in Fig. 21. The Power

requirement to overcome drag on the rotor hub is significantly larger than either of the other 2

components, due to drag power scaling with U∞3. The drag power requirement is not changed by the

hub’s µ and side force power, as with side force seen in Fig. 13, does not change consistently with µ .

Torque, however, does have consistent changes in power requirement with µ , shown in Fig. 22.

30

Figure 21: Power Comparison at 240 RPM

31

Figure 22: Torque Power Comparison at Varying RPM

32

VELOCITY FIELD MEASUREMENTS

A detailed velocity map of the hub wake was obtained via Particle Image Velocimetry. The light sheet

illumination was provided with a Litron double-pulsed Nd:YAG laser(532 nm, 200mj/pulse) and the

scattering from seed particles entrained in the flow was captured by a LaVision Imager Intense CCD

camera (Imager Pro X 2M-1600×1200 pixels, 14bit). The flow was seeded with 2µm-3µm droplets

produced with a aerosol seeder. The recording rate yielded 14 velocity fields per second. The time

separation between the two light flashes used to capture particle displacements was set based on the

velocity range expected and adjusted as needed based on the velocity range measured. One hundred

PIV image pairs were acquired for each field captured. The data presented were obtained by averaging

the 100 velocity vector fields. In cases where the hub rotated, the averaging was done with phase

locking to obtain azimuth-resolved velocity fields. Single realizations of the velocity field are usually

inadequate, because the seeding may not be present adequately in all parts of the image in every

instance, resulting in several areas where velocity vectors could not be obtained. This would hinder

construction of derivative fields such as vorticity.

Velocity deficits in the wake at a location one-half of the hub diameter (Figure 23) downstream of the

hub axis were profiled for several model orientations. Figure 24 illustrates the typical variation

observed in the wake profile for the hub in two static configurations and one case in rotation. The data

show a contraction of the momentum deficit in the hub wake when the model is oriented at 45 azimuth,

corresponding to a reduction of hub drag based on frontal area (Figure 24). The largest deficit at the

static hub centerline is clearly defined by the strong velocity deficit just behind the hub main shaft. The

two secondary wake deficits appear behind the pitch links to the left and right of the hub shaft. When the

hub rotates in a counter clockwise direction, the primary velocity deficit is translated upward and to the

right, appearing behind the right (aft looking forward) blade shank. The velocity deficit due to the main

drive shaft appears to have coalesced with that of the right pitch link, leveling only a secondary velocity

deficit from the left pitch link, which has also translated in the positive y-direction (to the right) of the

flow induced by the counter clockwise rotation. The rotated wake is seen to shift with the hub rotation

direction. Also, the free stream region for the rotating case shows an increase in velocity with respect to

the tunnel speed setting of 20mph. The static cases of 0 and 45 show substantial asymmetry, which is

interesting as the only asymmetry in the model arises from the pitch link joint with the blade shanks.

The wake shows interesting features. One is the apparent convection of the low-speed zone (presumably

due to separation and recirculation) that occurs downstream of the shank: this convects in a periodic

manner and appears in phase-resolved, ensemble averaged contours of the flow speed in the horizontal

33

(a) Top view

(b) Isometric View

Figure 23: Map of PIV data collection plane comprised of overlapping stitches

34

Figure 24: Wake tunnel axis velocity profiles captured by PIV and scaled by free stream velocity of20mph

35

Figure 25: PIV result ihowing a color map of the speed, and velocity vectors, in the horizontal (X-Y)plane downstream of the hub, at a hub phase of 30 degrees

plane. This is illustrated in Figures 25, 26 and 27, taken at 30, 45 and 60 degrees phase with respect to

the zero reference of azimuth. The position coordinates are referred to the PIV measurement locations

shown in Figure 23. The primary axis of vorticity in the flowfield downstream of a shank is expected to

be parallel to the shank axis. The plane being viewed here is probably parallel to the primary axis of the

vorticity vector field, and hence vorticity contours in this plane do not produce useful visualization.

However, it is interesting to note the large discrete region of low speed. As this convects, it may be

expected to cause severe fluctuations in the speed being sensed by hot-film anemometer probes,

addressed in the next section.

The vorticity in the wake flowfield was studied using the velocity field shown in Figure 28. The

Normalized Angular Momentum (NAM) algorithm was used to capture centers of vortical structures. A

test case of the NAM is shown in Figure 29. A numerical test case velocity field was generated, with a

few discrete vortices. This is shown in the left side of Figure 29. The vorticity field computed by the

NAM is shown on the right of Figure 29. The technique was applied to the portion of the hub wake

shown in Figure 23. The result is shown in Figure 30. The plot on the left side shows the vorticity field

computed directly, while the right side shows the vorticity field computed using the NAM.

36

Figure 26: PIV result ihowing a color map of the speed, and velocity vectors, in the horizontal (X-Y)plane downstream of the hub, at a hub phase of 45 degrees

Figure 27: PIV result showing a color map of the speed, and velocity vectors, in the horizontal (X-Y)plane downstream of the hub, at a hub phase of 60 degrees

37

Figure 28: PIV field showing

Figure 29: Validation of the Normalized Angular Momentum algorithm to capture vorticalstructures in a velocity field.

38

Figure 30: Vortical structures in the indicated portion of the hub wake, computed directly as vorticity,and using the Normalized Angular Momentum algorithm.

WAKE VELOCITY FLUCTUATION MEASUREMENTS

A hot-film constant temperature anemometer probe was used to measure flow fluctuations in the wake

at locations listed in Table 3. A hot-film sensor uses the principle of a hot-wire anemometer sensor.

Feedback-controlled electrical resistance heating is used to set the sensor external temperature at a

constant level in the range of 520K, well above the ambient flow temperature. The voltage needed to

maintain this setting through a Wheatstone Bridge circuit, is empirically related to the flow speed over

the sensor. In the case of the hot film, the active resistance is that of a thin film of metal coated over a

strong non-conducting substrate rod, instead of using an extremely thin metal wire. Thus the sensor has

greater mechanical ruggedness, and is appropriate where the micron-sized sensor wires are not needed.

The anemometer voltages were digitized and sampled at 5000Hz, for 30 seconds at each station. A

filtered signal was simultaneously recorded, with a high pass filter set at 3Hz and a low pass filter set at

2000Hz. An amplifier gain, typically 20dB, was used to optimize the amplitude of the fluctuating signal

for digitization. Using these voltages the instantaneous flow speed could be calculated using the

unfiltered voltage, while the small velocity fluctuations could be amplified for good resolution, and

calculated from the filtered voltages. The full-signal speed calculation uses the nonlinear voltage-speed

39

relationship obtained from calibration of the anemometer probe against a Pitot-static probe. The

conversion of the filtered signal uses the instantaneous slope of the voltage-speed calibration. This

method still poses substantial issues when the fluctuation amplitude approaches the mean value of the

speed. In regions where the flow may encounter transient stagnation or reversal, the use of such probes

is extremely difficult, and prone to large error.

Axis Locations ResolutionX 1Dh to 3Dh 1DhY -0.4955 Dh to 0.502 Dh 0.0208 DhZ +0.0675 Dh, 0, -0.0675 Dh and -

0.204 Dh

–

Table 3: Summary of wake velocity measurement locations

The full database of hot-film anemometer measurements is given on the Internet at the project website

given on the front cover of this report. NOTE: The gain applied to the signal has been

accounted (removed) for in the data presented. The data are presented in two columns, the first column

is the mean velocity and the second column is the fluctuation velocity. Each column consists of 5000

samples/s × 30s = 150,000 samples. The data were first collected as voltages and then converted to

velocity using the calibration coefficients, and the data set is presented in m/s. Please refer to Table 3

for the specific file names corresponding to each measurement location. For example: X = 1D, Y = 1, Z

= 1 corresponds to X = 1 Dh, Y = -.4955 Dh, Z = +0.0675 Dh.

Time-averaged wake profiles obtained using the hot-film probe are given below. These are not expected

to be the same as the velocity profiles obtained using PIV because the hot-film anemometer is sensitive

to all components of velocity, although with dominant sensitivity to the components that are

perpendicular to the axis of the cylindrical sensor element, whether wire or film substrate. Figures 31

and 32 compare the effect of having scissors present, on the time-averaged velocity profile 1 diameter

downstream, at 30 mph freestream speed. Figure 33 presents the wake profile at the same location (1

diameter downstream) at 20 mph, and Figures 34 and 35 show what happens at 2 and 3 diameters

downstream, respectively. The expected spreading of the wake is seen. Curiously, the case with no

scissors has more features in the wake profile. This agrees with the azimuth-resolved load

measurements, which show that the presence of the scissors causes the dips in the drag variation to be

be filled in, reducing the frequency content of the azimuth-resolved loads.

The frequency content at the above locations is next examined using spectra obtained from the hot-film

data. The specific wake location examined below is X=1Dh and Z=0 at Ω=120 and U∞=8.9 m/s from

40

Figure 31: Time -averaged speed profile with no scissors, at 1 diameter downstream.

Figure 32: Time -averaged speed profile with scissors, at 1 diameter downstream.

41

Figure 33: Time -averaged speed profile with scissors, at 1 diameter downstream.

Figure 34: Time -averaged speed profile with scissors, at 2 diameters downstream.

42

Figure 35: Time -averaged speed profile with scissors, at 2 diameters downstream.

Y=−0.5Dh to 0.5Dh. Figure 42 illustrates the FFT power spectral density of the velocity fluctuations

across the span of the hub wake. The power spectral density data presented are for the energy content in

the wake at 8Hz, 4/rev. The 4/rev content shows a clear asymmetry between the advancing and

retreating sides of the rotor hub. This was also observed in the investigation by Roesch and Dequin

(1985).

The effect of the scissors components added to the hub for the final round of experiments, can be seen

from the following figures. Figure 36 shows the spectrum at X=1D (1 diameter downstream of hub

centerline, Z=2 denoting the measurement location where Z is -0.204 hub diameters as shown in Table

3, with 240 rpm rotation speed and freestream speed of 30 mph (13.41m/s). The frequency axis is

expressed in units of per revolution. Thus the major spectral content is at 8 per rev, with large peaks

downstream of the advancing side. When the scissors are added, the spectra change as seen in Figure

37. The fluctuation intensity at 8 per rev rises dramatically further inboard on the advancing side, but

drops at outboard locations.

The next set of sample figures trace what happens at a given Z location (Z0 as given in Table 3 as we

move downstream, with the scissors present. The first, Figure 38, shows the spectra at Z0 and X=1D,

which is directly above the location shown in the prior two figures, but the data were acquired at a lower

43

Figure 36: Spectral features in the hot-film anemometer data obtained across the wake behind thehub with no scissors, at 1 diameter downstream.

Figure 37: Spectral features in the hot-film anemometer data obtained across the wake behind thehub with scissors added, at 1 diameter downstream.

44

Figure 38: Spectral features in the hot-film anemometer data obtained across the wake behind thehub with scissors added, at 1 diameter downstream, and Z=0, at 20mph.

speed of 20mph (8.94m/s). At this point the 8 per rev peak occurs inboard, on the retreating blade side.

Going downstream to 2 diameters (X=2D), we see from Figure 39 that the 8 per rev fluctuation intensity

is spread across the entire retreating side. Figure 40 shows that a X=3D there is much more broadband

content as the wake gets more chaotic.

In an attempt to better comprehend the temporal nature of the hub wake, the wavelet transform was

applied to the same wake turbulence data set. Table 4 shows some of the differences between the

Wavelet and FFT approaches. The purpose of the wavelet transform was to find discrete fluctuations

occurring in the wake Farge (1992). The wavelet used in the transformation was the complex Morlet,

visualized in Fig. 41, which has harmonic parts, useful when comparing results to FFT, and is often

used to analyze turbulent flow using constant temperature anemometry Hoa (2009). The wavelet

transform was carried out using the wavelet toolbox built into Matlab. The signal is easily loaded in the

interface into the 1-dimensional continuous wavelet feature. From there the wavelet type, complex

Morlet in this instance, and the number of scales, 600 were used for the analysis here, and then the

wavelet transformation was performed and generated the wavelet coefficients at each integer scale.

The Wavelet transform coefficients are compared to the FFT of the signal, at the Y=-0.3915Dh, in

Figure 43 for 2 seconds of a 30 second signal. The color legend indicates the correlation of the wavelet

45

Figure 39: Spectral features in the hot-film anemometer data obtained across the wake behind thehub with scissors added, at 2 diameters downstream, and Z=0, at 20mph.

Figure 40: Spectral features in the hot-film anemometer data obtained across the wake behind thehub with scissors added, at 3 diameters downstream, and Z=0, at 20mph.

46

Figure 41: Real and Imaginary Components of Complex Morlet Wavelet

Feature FFT Wavelet TransformHarmonic Signals High Quality Low QualityTime Resolved Frequency content Ensemble Averaged Frequency Time Resolved Content

Table 4: Fourier and Wavelet Transform Comparison

coefficients to the signal. The wavelet transformation shows the highest correlation with the complex

Morlet at, or near, the low frequency peaks in the FFT. As the frequency increases the power spectral

density computed from the Fourier Transforms decreases, at the same time the correlation with the

complex Morlet also decreases significantly. From the wavelet transform it can be observed that the

high correlation frequencies line up well with the peak frequencies seen in the FFT. The largest peaks in

frequencies occurred at 4/rev and 8/rev, corresponding to the passage of the 4 shanks of the hub. There

are also peaks offset from these, at 6/rev and 12/rev, due to parts of the hub passing by twice per

rotation. It is conjectured that the two bolts connecting the hub plates could cause the 6/rev and 12/rev

energy content in the hub wake.

Once the wavelet transform is performed the scales of the wavelet must be related to a domain that is

easier to understand. Because the comparison to the Fourier transform is presented here the scale is

converted to the Fourier frequency( f f ). Equation 8 relates the scale of the wavelet transformation to f f .

f f ∝1s

(8)

47

UNCERTAINTY ESTIMATES

The main errors in constant temperature anemometry measurements arise due to the following

reasons (adapted from Yavuzkurt (1984)):

1. Calibration error, εC - caused due to inaccuracies during calibration measurements, curve fitting

and calibration drift error.

2. Approximation error, εA - Error caused by approximation techniques, mainly caused by standard

deviation deviation.

3. Temperature variation, εT - Error due to temperature variations in the flow. It can change air

density slightly and affect the heat transfer between the flow and the wire.

4. Pressure uncertainty, εP - Uncertainty due to not knowing ambient pressure exactly. Pressure

uncertainty causes an uncertainty on assumed air density.

The total error (ε) is then calculated by ε = k√

εC2 + εA

2 + εT 2 + εP2, where k ≈ 2 for a confidence

interval of 95%. Error estimates for wake velocity measurement are summarized in Table 5, along with

uncertainties in all the parameters.

Parameter Worst case estimateDynamic Pressure 0.02%Dynamic Viscosity 0.01%Density 0.03%Velocity 0.03%Load Cell Calibration 1.37%Calibration 0.77%Temperature variation 0.45%Approximation 0.43%Ambient pressure 0.03%Total ±1.98%

Table 5: Summary of experimental uncertainties

48

Figure 42: 4/Rev Energy content in the hub wake at X = 1D and Z = 0

SUMMARY OF DATA AND RESULTS

Air loads data are given on the Internet at the project url given on the front cover of this report. The data

organization is summarized in Table 6.

The airloads data are published with a README text file. All force measuremements are reported in

lbs, and all moment measurements are reported in lb-in. The loads data are in 4 categories of test cases:

1. General Tests. These give loads at specified rotation rates (RPM) and freestream speed (Meters

per second).

2. No Scissor Tests. These are tests done with no scisssors and no motor fairing, taken at a

digitization rate of 10,000 Hz.

3. Azimuth-resolved.

4. Fairing + Tare. Load measurements on the motor with a fairing on the motor.

The report ADLR2013070101.pdf summarizes the hot-wire anemometer data. Wake velocity

measurements were conducted at locations summarized in Table 7 and the axis are illustrated in Fig. 45.

The locations are normalized with hub diameter Dh. The measurements were made across the span of

the hub wake (Y axis) at a resolution of 10.16 mm (0.0208 Dh). Specific Z planes were chosen in

49

Figure 43: Wavelet transform coefficients compared to FFT at Y=-0.3915Dh

50

Figure 44: Downstream hot-wire measurement locations

order to understand the overall nature of the unsteady wake and in specific to investigate the effect of

shanks and the scissors on the unsteady rotor hub wake. The measurement planes where

0. 0675 Dh Z 0. 0675 Dh were chosen in order to investigate the flow over the main hub assembly,

whereas the Z =0. 204Dh measurement plane is to investigate the effect of scissors and pitch links.

The X locations (1Dh X 3Dh) were chosen to investigate the evolution of the wake on moving

downstream. The downstream locations are also referred to as near wake (1Dh), mid wake (2Dh) and far

wake (3Dh).

Some significant results are reiterated in the figures below. Figure 46 shows azimuth-resolved airloads

data, with drag, side force and torque shown. Drag is the dominant airload, but there are measurable

azimuthal variations in side force and torque. This result is not expected to change in scaling to

full-scale helicopter cases; in fact drag is expected to dominate in a much stronger way. This aspect is

studied in Figure 47.

Figure 48 shows the azimuth-resolved drag, as well as the variation of drag with freestream speed.

There is about a 10 percent drag increase due to these two elements. The azimuth-resolved drag shows

that the effect of the scissors elements (which may be unique to how we placed them on the hub) is to

fill in azimuthal intervals where the drag was otherwise lower. Note that the widely differing

magnitudes of the loads is accommodated by using different vertical scales. Power demand due to side

force and torque are referred to the right side axis.

51

Folder Name RPM Velocities(mph)Azimuth Resolved 80 20

120 30240 20, 30, 50

Fairing + Tare 0 10, 15, 20, 25, 30, 35, 40, 45, 50General Tests 80 10, 15, 20, 25, 30, 35, 40, 45, 50

120 10, 15, 20, 25, 30, 35, 40, 45, 50160 10, 15, 20, 25, 30, 35, 40, 45, 50240 10, 15, 20, 25, 30, 35, 40, 45, 50

No Scissors Tests 120 20, 25, 30, 35, 40, 45, 50, 55With Scissors Comparison 120 25, 30, 35, 40

Table 6: Rotor Hub Aerodynamic Load Test Cases

Figure 45: Illustration of axis referred to in the measurement locations

Figure 49 shows an attempt to gauge the significance of interaction effects. The theoretical calculations

are in fact empirical, using methods such as those given in Hoerner (1965). The actual ones are

obtained from the drag deconstruction tests. The shaft in this case includes the swashplate. The reduced

drag on the shanks when installed, is real: it is attributed to the shank wakes reducing the onset dynamic

pressure and perhaps redirecting the flow that would otherwise cause large flow separation at the base.

The next result is that there is substantial 4 per rev fluctuation content in the wake on the retreating

blade side of the hub. This is indicated in Figure 50 where the power spectral density at 4 per rev (units

are arbitrary) is shown, as the probe location was moved across the flow from the retreating blade side

52

Figure 46: Drag is the dominant airload on the hub

Figure 47: Implications for the total power demand on a helicopter

53

Figure 48: Effect of the scissors elements on azimuth-resolved drag

Figure 49: Significance of interactions between component flowfields, as gauged from simplepredictions of component drag

54

Axis Locations Resolution File name in data folderX 1 Dh to 3 Dh 1 Dh X=1D–3DY -0.4955 Dh to 0.502 Dh 0.0208 Dh 1.txt–49.txtZ +0.0675 Dh, 0 Dh, -0.0675 Dh and -0.204 Dh Discrete Z=1,0,-1,-2

Table 7: Summary of wake velocity measurement locations

Figure 50: Variation of the power spectral density at the 4 per rev frequency, at points across thewake.

to the advancing blade side. As discussed before, this is probably due to each shank leaving behind a

region of separated flow, which then convects downstream. Such a feature would appear as discrete

pulses of large velocity fluctuation.

55

CONCLUSIONS

A generic hub model of a 4-bladed rotor was used in a low-speed wind tunnel to deconstruct the sources

of hub drag, and explore the effects of rotation on the hub aerodynamic loads, as well as the effect on the

wake. The use of a 6-DOF load cell as a primary aerodynamic load sensor in wind tunnel experiments

is no longer unique, but the level of accuracy and reliability that are now achieved was not possible even

a few years ago, and involves detailed calibration and nonlinear (actually piecewise linear) correction

for cross-couplings. In this effort all components of airloads could be captured, and specifically, it could

be ensured that the drag reading was not affected by side load or torque, even under rotating conditions.

The geometry is completely defined, allowing for analysis and computational modeling at any level of

precision ranging from single component aerodynamics to high-fidelity computation. Detailed CAD

files are also available for the experimental setup. The findings are summarized below:

1. The use of a 6-DOF load cell permitted unambiguous separation and capture of all 6 components

of the airloads on the hub.

2. 6DOF load cell resolution and accuracy now good enough to be a primary airloads sensor for

wind tunnel aerodynamic measurements.

3. Hub/shank airloads behave generally as expected in both the static and rotating cases.

4. The drag measurements showed no significant increase with Ω in the test cases studied, agreeing

with prior literature on rotor hub drag.

5. Torque rises with freestream speed with weaker than linear slope.

6. Side force and torque total ¡ 10% of total power demand. Extrapolation to full-scale contributions

should result in an even greater dominance of the drag.

7. The azimuthal variation of drag generally corresponds to projected frontal area.

8. Addition of two scissors linkages increased the drag by roughly 10 percent.

9. In the case studied, the drag due to the scissors has the effect of reducing the amplitude of

azimuthal drag variation by filling in the azimuthal regions of lower drag.

10. Wake spectra show peaks at even harmonics of the rotation speed.

11. 4 per rev fluctuations correspond to the wakes of the shanks; 6 per rev components appear to be

due to bolts connecting the hub plates

56

12. Interactions between components is significant. The best example is that the shank wakes appear

to reduce base drag on the hub.

13. Side force seen on the hub is primarily due to asymmetry of the pitch links and other components.

14. Hypotheses of Magnus effect causing side force were not supported by experiments.

15. Particle image velocimetry detects sideward wake shifts due to rotation; however the side forces

due to this are not large. The flow distortion is more a drag-induced effect than a case of smooth

attached flow.

16. There is substantial unsteadiness in the side force, and its azimuth-resolved value changes sign

during the course of rotation.

17. The azimuths where torque is highest correspond roughly to, but slightly lead, the azimuths

where drag is highest.

18. Substantial 8 per rev content seen in the wake spectra on the retreating blade side appear to be due

to the regions of low speed flow left behind shanks.

57

REFERENCES

A. Antoniadis, D. Drikakis, B. Zhong, G. Barakos, R. Steijl, M. Biava, L. Vigevano, A. Brocklehurst,

O. Boelens, and M. Dietz. Assessment of CFD methods against experimental flow measurements for

helicopter flows. Aerospace Science and Technology, 19(1):86–100, 2012.

T. Asai, K. Seo, O. Kobayashi, and R. Sakashita. Fundamental aerodynamics of the soccer ball. Sports

Engineering, 10(2):101–109, 2007.

J. Bridgeman and G. Lancaster. Predicting hub drag on realistic geometries. In Proceedings of the

American Helicopter Society Specialists’ Conference, San Francisco, California, 2010a.

J. O. Bridgeman and G. T. Lancaster. Physics-based analysis methodology for hub drag prediction. In

Proceedings of 66th Annual Forum of the American Helicopter Society, 2010b.

R. W. Charles N. Keys. Guidelines for reducing helicopter parasite drag. Journal of the American

Helicopter Society, 20:31–40, 1975.

T. Chng and H. Tsai. Effects of spanwise geometric disturbances on a bluff body wake. In 3 rd AIAA

Flow Control Conference; San Francisco, CA. American Institute of Aeronautics and Astronautics,

1801 Alexander Bell Drive, Suite 500, Reston, VA, 20191-4344, USA,, 2006.

G. B. Churchill and R. D. Harrington. Parasite-drag measurements of five helicopter rotor hubs. Report

19980228291, NASA, 1959.

F. De Gregorio. Flow field characterization and interactional aerodynamics analysis of a complete

helicopter. Aerospace Science and Technology, 19(1):19–36, June 2012a.

F. De Gregorio. Flow field characterization and interactional aerodynamics analysis of a complete

helicopter. Aerospace Science and Technology, 19(1):19–36, June 2012b.

P. de Waard and M. Trouve. Tail shake vibration in flight- objective comparison of aerodynamic

configurations in a subjective environment. In Proceedings of the 55th Annual Forum of the American

Helicopter Society ,, pages 2306–2316, Montreal, Canada, 1999.

M. Farge. Wavelet transforms and their applications to turbulence. Annu. Rev. Fluid Mech., 24:

395–457, 1992.

F. F. Felker. An experimental investigation of hub drag on the xh-59a. In AIAA 3rd Applied

Aero-dynamics Conference, AIAA Paper, pages 85–4065, 1985.

58

C. A. J. Fletcher. Negative magnus forces in the critical reynolds number regime. Journal of Aircraft, 9,

1972.

J. W. Gregory, C. O. Porter, and T. E. McLaughlin. Circular cylinder wake control using spatially

distributed plasma forcing. In 4th AIAA Flow Control Conference, Seattle, WA, 2008.

L. T. Hoa. Use of continuous wavelet transform-based tools for studying turbulence, aerodynamic

forces and coherence. VNU Journal of Science, Mathematics - Physics, 25:74–88, 2009.

S. F. Hoerner. Fluid Dynamic Drag. Sighar Hoerner, 1965.

J. Hoffman. The relationship between rotorcraft drag and stability and control. In Proceedings of the

31st Annual Forum of the American Helicopter Society, Washington, DC, May 1975.

E. V. Horanoff. Wind tunnel balance, June 3 1969. US Patent 3,447,369.

I. S. Ishak, S. Mansor, and T. Mat Lazim. Experimental research on helicopter tail shake phenomenon.

In Proceedings of the 2nd Regional Conference on Vehicle Engineering & Technology, Hotel Istana,

Kuala Lumpur, July 2008.

A. Kerr. Effect of helicopter drag reduction on rotor dynamic loads and blade life. In Proceedings of the

American Helicopter Society Symposium on Helicopter Aerodynamic Efficiency, Washington, DC,

March 1975.

J. C. Linville. An experimental investigation of high-speed rotorcraft drag. Technical report, DTIC

Document, 1972.

D. M. Martin, R. W. Mort, L. A. Young, and P. K. Squires. Experimental investigation of advanced hub

and pylon fairing configurations to reduce helicopter drag. NASA Technical Memorandum 4540, 1993.

R. D. Mehta. Aerodynamics of sports balls. Annual Review of Fluid Mechanics, 17(1):151–189, 1985.

F. Ortega, R. Shenoy, V. Raghav, M. Smith, and N. Komerath. Deconstructing hub drag. In Proceedings

of the 29th AIAA Applied Aerodynamics Conference, number 2011-3821 in AIAA Paper, Honolulu, HI,

June 2011. AIAA.

V. Raghav, R. Shenoy, M. Smith, and N. Komerath. Investigation of drag and wake turbulence of a rotor

hub. Aerospace Science and Technology, 2012.

D. Reisch, B. Elbing, C. Berezin, and S. Schmitz. Water tunnel flow diagnostics of wake structures

downstream of a model helicopter rotor hub. In Proceedings of the AHS Annual Forum, Washington

DC, May 2013. American Helicopter Society.

59

P. Roesch and A.-M. Dequin. Experimental research on helicopter fuselage and rotor hub wake

turbulence. Journal of the American Helicopter Society, 30(1):43–51, 1985.

A. Roshko. Perspectives on bluff body aerodynamics. Journal of Wind Engineering and Industrial

Aerodynamics, 49(1):79–100, 1993.

A. Sayers and A. Hill. Aerodynamics of a cricket ball. Journal of Wind Engineering and Industrial

Aerodynamics, 79(1):169–182, 1999.

J. Seidel, S. Siegel, K. Cohen, and T. McLaughlin. Simulations of flow control of the wake behind an

axisymmetric bluff body. In 3rd AIAA Flow Control Conference, San Francisco, CA, 2006.

T. W. Sheehy. A general review of helicopter rotor hub drag data. Journal of the American Helicopter

Society, 22(2):2–10, 1977.

T. W. Sheehy and D. R. Clark. A method for predicting helicopter hub drag. Technical report, United

Technologies Corporation, Stratford, CT - Sikorsky Aircraft Division, 1976.

R. Shenoy, M. Smith, and N. Komerath. Computational investigation of hub drag deconstruction from

model to full scale. In Proceedings of the 37th European Rotorcraft Forum, Milan Italy, September

2011.

R. Shenoy, M. Holmes, M. J. Smith, and N. Komerath. Scaling evaluations on the drag of hub system.

Journal of the American Helicopter Society - Accepted with revisions, 2012.

A. Smits and D. Smith. A new aerodynamic model of a golf ball in flight. In Science and Golf II:

Proceedings of the World Scientific Congress of Golf, pages 340–347, 1994.

B. Wake, E. Hagen, S. Ochs, and C. Matalanis. Assessment of helicopter hub drag prediction with an

unstructured flow solver. In Proceedings of 65th Annual Forum of the American Helicopter Society,

Fort Worth, TX, May 2009a.

B. Wake, E. Hagen, S. Ochs, and C. Matalanis. Assessment of helicopter hub drag prediction with an

unstructured flow solver. In Proceedings of 65th Annual Forum of the American Helicopter Society,

Fort Worth, TX, May 2009b.

R. M. Williams and P. S. Montana. A comprehensive plan for helicopter drag reduction. Technical

Report 19750054284, NASA Report, 1975.

S. Yavuzkurt. A guide to uncertainty analysis of hot-wire data. ASME Transactions Journal of Fluids

Engineering, 106:181–186, 1984.

60

L. Young, D. Graham, and R. Stroub. Experimental investigation of rotorcraft hub and shaft fairing drag

reduction. Journal of Aircraft, 24:861, 1987.

61

WORK COMPLETED

This report describes all the work done to-date on the experimental tasks. The proposed sub-task where

a pylon assembly was to be used in the experiments, was deferred, and substituted with more extensive

hot-wire measurements, and further focus on the effects of rotation on the torque and side force

contributions to power. The primary reason for this change was that the rotation effects on drag, side

force and torque had to be established unambiguously, following initial experimental results and

hypotheses based on those and prior work, without the aerodynamic interference from the rest of the

rotorcraft. This had to be done before an effective experiment could be designed to focus on the

interaction effect. One more tunnel entry is planned in late 2013/early 2014 with a pylon or other

interaction geometry below the hub to resolve the issue of whether there is a large effect on vehicle drag

induced by a rotating hub, due to transient flow separation and attachment on other surfaces such as the

pylon. Recently, true a priori computational predictions followed by single-component drag load cell

and PIV data in a water tunnel have been published by Reisch et al. (2013). Preliminary indications

from there are that no such effect has been seen. The design of the experiment to capture or negate any

such effects must include these results as well, to focus on what has not yet been investigated.

The computational tasks are to be reported in a separate volume as decided by Dr. M. Smith, the co-PI

responsible for those tasks.

IMPACT/APPLICATIONS

This study addressed several unresolved questions that arose from prior work on the drag of complex

configurations. The results succeed in allowing the future designer to rule out some of the fears

generated by these questions. Examples are the so-called Magnus Effect of hub rotation, and any large

rise in power required to overcome torque increases due to interaction of the hub with the freestream. It

shows that simple models can be developed for parts of the interaction drag, for instance the reduction

in hub base drag when shanks are present.

The detailed data sets permit future developers of computational methods to exercise their codes in an a

posteriori refinement mode against the geometries and multifaceted database placed on the Internet.

TRANSITIONS

None.

62

RELATED PROJECTS

If none, so state. Task 10 of the Vertical Lift Rotorcraft Center: The load-cell-mounted testing

technique developed here was used to break through the problem of measuring airloads on objects of

arbitary shape, encountered in the bluff body aeromechanics problem. A continuous-rotation method

was developed, to obtain all 6 components of quasi-steady airloads very efficiently, compared to the

prior technique of aligning objects for each attitude.

PUBLICATIONS

1. Forbes, A., Raghav, V.S., Mayo, M.G., Komerath, N.M., Rotation Effects on Hub Drag.

IMECE2013-63547. ASME Congress & Exposition, San Diego, Nov. 2013

2. Shenoy, R., Holmes, M., Smith, M. J., and Komerath, N., Scaling Evaluations on the Drag of a

Hub System. Journal of the American Helicopter Society, to appear.

3. Raghav, V., Shenoy, R., Smith, M. J., and Komerath, N. M., Investigation of Drag and Wake

Turbulence of a Rotor Hub, Aerospace Science and Technology, Vol. 28, Issue 1, July 2013, pp

164-175.

4. Ortega, F., Shenoy, R., Raghav, V., Smith, M. J., and Komerath, N., Exploration of the Physics of

Hub Drag. AIAA-2012 -1070, Nashville, TN, Jan.9-12, 2012

5. Shenoy, R., Smith, M. J., and Komerath, N., Computational Investigation of Hub Drag

Deconstruction from Model to Full Scale. 37th European Rotorcraft Forum, Italy, Sept. 12-15,

2011

6. Shenoy, R., Raghav, V., Ortega, F., Smith, M. J., and Komerath, N., Deconstructing Hub Drag,

AIAA-2011-3821, June 2011

HONORS/AWARDS/PRIZES

Omit if none.

1. Vrishank Raghav, Vertical Flight Foundation 2012 Scholarship

2. Vrishank Raghav, United Technologies Research Center Fellowship

3. Michael Mayo, 2013 Vertical Flight Foundation 2013 Scholarship

63

4. Michael Mayo, 2013 Roberto Goizeuta Scholarship

DEGREES EARNED

1. Felipe Ortega, MSAE,December 2012

2. Ryan McGowan, BSAE May 2013

3. Rafael Lozano, MSAE May 2012

4. Michael Mayo, MSAE, expected December 2013

5. Alex Forbes, MSAE, expected December 2013

ACKNOWLEDGMENTS

The authors gratefully acknowledge the support of the Office of Naval Research and valuable

discussions with the technical monitors Judah Milgram and John Kinzer. Assistance from Felipe Ortega,

Ryan McGowan and Rafael Lozano for the experimental data acquisition is gratefully acknowledged.

64