Wind Tunnel and Hover Performance Test Results for Multicopter UAS Vehicles

Wind Tunnel and Hover Performance Test Results for

Multicopter UAS Vehicles

Carl Russell Jaewoo Jung Gina Willink Brett Glasner Aeromechanics Office Aviation Systems Division Aero mechanics Office Universities Space Research NASA Ames Research NASA Ames Research NASA Ames Research Association Center Center Center Moffett Field, CA Moffett Field, CA Moffett Field, CA Moffett Field, CA ABSTRACT There is currently a lack of published data for the performance of multicopter unmanned aircraft system (UAS) ve hicles, such as quadcopters and octocopters, often referred to collectively as drones. With the rapidly increasing popularity of multicopter UAS, there is interest in better characterizing the performance of this type of aircraft. By studying the performan ce of currently available vehicles, it will be possible to develop models for vehicles at this scale that can accurately predict performance and model trajectories. This paper describes a wind tunnel test that was recently performed in the U . S . Army’s 7 - b y 10 - ft Wind T unnel at NASA Ames Research Center. During this wind tunnel entry , five multicopter UAS vehicles were tested to determine forces and moments as well as electrical power as a function of wind speed, rotor speed, and vehicle attitude. The test is described here in detail, and a selection of the key results from the test is presented.

and Revolutionary Vertical Lift Technology (RVLT) — have NOMENCLATURE an interest in determining the aerodynamic performance BDAS Basic Data Acquisition System characteristics of multicopters.

DELIVER Design Environment for Novel Vertical Lift Vehicles DELIVER and RVLT are focused on advancing the state of ESC Electronic Speed Controller the art in design and analysis tools for vertical lift vehicles FFT Fast Fourier Tran sform [Ref. 1 ]. DELIVER is particularly focused on small (less NDARC NASA Design and Analysis of Rotorcraft than 2 - passenger) alternative propulsion concepts. Both RVLT Revolutionary Vertical Lift Technology DELIVER and RVLT use a rotorcraft design software tool UAS Unmanned Aircraft System called NDARC (NASA Design and A nalysis of Rotorcraft), UTM UAS Traffic Management which has been extensively validated for large (above 2 - passenger) rotorcraft [Ref . 2 ] . Recently completed A Rotor disk area, ft 2 modifications to NDARC extended the propulsion models to C Thrust coefficient, T / ρ A ( Ω R ) T include electric propulsion systems [Ref. 3 ]. The NDARC F , F , F x - , y - , and z - force, lb x y z models s hould, in theory, be extensible to small M , M , M x - , y - , and z - moment, ft - lb x y z multicopters, but have not yet been calibrated to existing q Dynamic pressure, lb/ft aircraft. A research effort is currently underway to obtain R Rotor radius, ft data on small (up to approximately 10 lb) multicopters to aid T Thrust, lb in this calibration. The necessary data include rotor α Pitch angle, deg performance, airframe aerodynamics, battery and motor ρ Air density, slugs/ft performance, and detailed component weights. The data Ω Rotor rotation speed, rad /s gathered as part of this test will provide calibration data for rotor performance and airframe aerodynamics.

INTRODUCTION The UTM Project i s developing a prototype system Multi copter unmanned aircraft have seen a rapid rise in consisting of technologies that will enable safe, efficient production and utilization i n the past few years. As the use operations in low - altitude, where a significant increase in of these vehicles continues to increase , research questions unmanned aircraft flights is anticipated [Ref. 4 ] . Curren tly, have emerged concerning their design, handling qualities, the system shown in Fig. 1 is being consid ered as a and control. In particular, three NASA projects /sub - prototype . One of the key re quirements in implementing the projects — UAS Traffic Management (UTM), Design system in Fig. 1 is an understanding of the capabilities of Environ ment for Novel Vertical Lift Vehicles (DELIVER) , unmanned aircraft under a rang e of environmental conditions. In particular, there is a need to determine nd whether a given aircra ft will be able to maintain control Presented at the AHS 72 Annual Forum, West Palm Beach, and/or station - keeping within a pre - defined area of airspace FL, May 16 - 19, 2016. This is a work of the U.S. Government under forecasted weather condition s . Information that is and is not subject to copyright protection.

critical to this type of understanding is the interaction wind tunnel at NASA Ames Research Center . Measurements between r oto r RPM, vehicle air speed, and a ttitude on the included forces and moments on the multicopter rotors and forces and moments on the vehicle. The data collected from airframes as well as electrical power as a function of speed, the test described here will be used to help populate the attitude, an d rotor RPM. Complementary testing was vehicle performance database shown in the top left of Fig. 1. performed to measure hover performance of multicopter rotors and complete vehicles. This paper summarize s the wind tunnel and hover performance tests and presents a subset of the data collected .

TEST OVERVIEW Five ai rcraft were tested as part of this wind tunnel entry: 3D Robot ics (3DR) SOLO, DJI Phantom 3 Advanced, 3D Robotics Iris+, Drone America x8, and Straight Up Imaging (SUI) Endurance [Refs . 9 - 13 ] . All five aircraft are commercial l y available multicopters whose primary mission is photographic surveillance. Table 1 summarizes the pertinent details of the five aircraft, and Figs. 2 - 6 show pictures of the vehicles installed in the wind tunnel. In addition to the full vehicles, three o f the isolated rotors were tested in the wind tunnel, and this configuration is shown in Figure 1. UTM system schematic.

Fig. 7 for the SUI Endurance. 10 , 11 , 12 , 13 , The vast majority of publi shed research on multicopters All of the vehicles were modified from their typical focuses on dynamics and control of the vehicles, with little consumer configuration to facilitate testing in the wind on experimental performance data. In recent years, t here tunnel, with inte rnal electronic components being the most ha ve been some tests on the small propellers used for notable change. The brushless motors employed by all of the multicopters . In particular, Brandt and Selig [Ref s . 5 and 6 ] multicopters tested here require a three - phase switched DC published a database of propeller performance data including power signal provided by an electronic speed controller thrust and shaft power in axial flow in a wind tunnel as well (ESC). On production versions of the ve hicles, the ESCs are as at static conditions . There have also been limited results typically embedded in custom circuit boards. Directly published for small quadcopter propellers in edgewise flight, controlling the ESCs would therefore require access to the s uch as Ref. 7 . Flight test data has also been used to estimate programming interface for the custom circuitry for each quadcopter performance and enhance simulation models vehicle, which was an unnecessary complication for the test .

[Ref. 8 ].

Thus , the custom circuitry was removed and was replaced by off - the - shelf ESCs — one for each motor — for the wind There is significant overlap betwee n the data needs of the tunnel test .

UTM, DELIVER , and RVLT research efforts , so a jointly planned wind tun nel test was executed to determine the The other primary modification was the removal of the aerodynamic performance characteristics of five different camera gimbals in order to use the gimbal mounting holes to multicopter aircraft. The primary goal of the test was to install the vehicles i n the wind tunnel test section. Both the generate a high - quality set of data for multicopter interface with the ESCs and the physical interface with the performance that can be used for model developmen t and test vehicles are discussed further in the following section.

validation. The test took place in the U.S. Army 7 - by 10 - ft Table 1. Multicopter Test Vehicles .

Rotor - to - Rotor Rot or - to - Rotor Rotor Nominal Flight Make/Model Configuration ★ Length, in Width, in Diameter, in Weight, lb 3DR SOLO Quadcopter 11.5 11.5 10.0 3.3 DJI Phantom 3 Quadcopter 9.8 9.9 9.4 2.8 3DR Iris+ Quadcopter 10.4 16.1 9.6 2.8 Drone America x8 Octocopter 31.4 31.4 11.0 12.8 Straight Up Imaging Quadcopter 20.1 20.1 15.0 6.0 Endurance ★ Without additional p ayload Figure 2 . 3D Robotics SOLO . Figure 5 . Drone America DAx8 .

Figure 6 . SUI Endurance .

Figure 3 . DJI Phantom 3 Advanced .

Figure 7 . Isolated rotor – SUI Endurance.

Figure 4 . 3D Robotics Iris+ .

each ESC. These signals were passed through a Pacific TEST DETAILS Instruments signal conditioning system to the BDAS system , The objective of the test was t o measure forces, moments, which collected data at a rate of 1,024 samples per second and motor power as a function of wind speed, attitude, and and then computed th e average over a 30 second data record.

rotor RPM for each test vehicle . Each of the five multicopter aircraft was run through a range of test conditions , including The second ary data system contained both the software for the full airframe, bare airframe, and propell er only.

commanding the ESC and a second LabVIEW program for recording rotor RPM. This system recorded the rotor RPM The following sections detail the data acquisition and model and also provided a real - time display of RPM for each control systems that were used to meet the test objectives.

motor. R otor speed control inputs were made by a human in Next, the model mounting hardware is described, followed the loop (usually the test engineer), to match the target RPM.

by an overview of the testing procedures. Finally, the steps Radio control of the test vehicles was undesirable due to taken to reduce and correct the data are described.

concerns about radio interference. Control was therefore accomplished v ia a servo controller with a wired connection Test Control and Measurement Systems to the test article. The Pololu Mini Maestro servo controller has outputs for up to 12 servos, which can be controlled A n overview of the test systems used for data acquisition, independently. The pulse - width modulated (PWM) signal model control, and power delivery is shown in the schematic used by RC aircraft servos is the sam e as that used by the in Fig. 8 , and is described below . These aircraft nor mally fly ESCs, so this servo controller provides a convenient means either autonomously or under radio control with batteries of sending commands to the ESC.

providing power. To eliminate battery safety concerns and eliminate testing down time that would be required for Rotor RPM measurements were made with Eagle Tree battery charging , power was provided by a 3,000 - watt brushless motor RPM sensor s. These sensor s tap into th e Sorensen DC power suppl y. The power supply is capable of motor power leads and provide a sinu soidal output voltage supplyin g up to 20 V DC power at up to 150 A, which is signal with a frequency equal to the commutation frequency sufficient to power each of the test vehicles up to their of the motor. All of the motors used for this test are 7 - pole respective maximum power.

motors, with a commutation frequency of 7 times the rotation speed. The LabVIEW program mentioned above Two different data acquisition systems were used to collect provides filt ering on the signal and uses a Fast Fourier data for this test. Th e primary data system was the Basic T ransform (FFT) to isolate the commutation frequency and Data Acquisition System (BDAS), which is a LabVIEW - record rotation speed for each motor. Based on the based system that can be configured to record tunnel resolution of the FFT as well as comparison with a handheld conditions in addition to user data signals [Ref. 14 ] . For this optical tachometer during test preparati ons, the accuracy of test, the user data included the model angle of attac k, the model forces and moments , and the voltage and current to Control'Room' Test'Sec/on' Tunnel' Primary'Data'System' Conds .'

Test'Ar/cle' Forces'&' Moments' Amp/' Voltage'&' Signal' Current' Cond.'

ESC' Trigger'Signal' Control' DC'Power' Rotor'RPM' Secondary'Data'System' Power'Supply' Below'Test'Sec/on' Figure 8 . Schematic of data acquisition, control, and power systems .

the RPM measurements was estimated to be ±30 rpm. A is shown in gray, and the model interface plates were bolted typical operating rotor speed for the vehic les tested here is on top of the load cell as shown. Each model has its own around 5,000 rpm, so the relative accuracy of th e RPM mounting plate with a 1 .25 - inch - diameter boss that inserted measurements is better than ±1%. Acquisition of the motor into the interface plate and was secured in place with four RPM data was triggered by a signal from the primary data bolts. Fig ure 11 shows an example model mounting plate.

system so that the data collected by both systems was Because each model has the same 1 .25 - inch boss, model properly synchronized. changes could be made very quickly during the test .

The forces and moments were measured using six - axis load Yaw changes were made by rotating the yaw stub (shown in cell s manufactured by JR3, Inc. The majority of testing was dark blu e) below the load cell inside of the yaw tube, shown carried out using a JR3 50 - lb six - axis load cell, which has in orange. Two bolts could then be inserted into a series of maximum loads of 50 lb in the x - and y - directions, 100 lb in threaded holes, allowing model yaw angles of 0, 5, 10, 30, the z - direction, and maximum moments of 150 in - lb in all 45, 60, and 90 deg.

three directions. Manufacturer - stated accuracy of the load cells is 0.25% of full - scale . Sting balances typically have accuracies no worse than 0.1% of full - scale, so some Flow%Direc+on% accuracy was sacrificed in favor of lower cost and hardware risk. Measurement uncertainty is addressed later in the Results section. The test plan originally called for a smaller 10 - lb load cell for the lighter quadcopters, but early testing Yaw%Stub% revealed significant vibrations that exceeded the capacity of S+ng% Yaw% Load%Cell% the 10 - lb load cell. The 50 - lb model was therefore use d for A4achment% Tube% all five vehicles. These vibrations will be discussed later in the results section.

Mounting and i nterface hardware Mx % Each test vehicle was installed using the “sting stand” as shown in Figs. 2 - 7 . In order to expedite pitch movement Fx % using existing win d tunnel hardware and control systems , the Interface% My% vehicles were installed at a 90 - degree right roll angle , Plate% Fz % Mz % allowing t he tunnel turntable to be used to vary the vehicle angle of attack . This allowed parameter sweeps with Fy % arbitrary values of angle of attack without requiring access to the model . Changing the model yaw angle required Figure 9 . Model mounting hardware .

manual adjustment, but only a handful of yaw angles were tested per vehicle , so making yaw angle changes did not have a significant impact on testing time .

The hardware used to mount t he vehicles on the sting stand Fx % allowed quick model changes by using an interface that was Fz % common to all five multicopters. A drawing of the mountin g hardware is shown in Fig . 9 and a close - up photograph of the mounting hardware is shown in Fig. 10 . Note th at the viewing angle is fli pped 180 degrees between Figs. 9 and 10 to show the details of the hardware. Also shown in both Fy % images are the directions of the forces measured by the load cell and the direction of airflow relative to the model . The forces were measured using typical aircraft body axes, with Flow%Direc+on% F positive downstream, F positive up, and F positive to the x z y right. Moments were defined as M positive roll right, M x y positive nose up, and M positive nose right. Again, the z model is mounted at a 90 deg roll angle, so the y - and z - directions of the load cell do not align with a typical wind tunnel coordinate system.

The yellow sting attachment bar parallel to the x - axis in Figure 10 . Close - up of DJI Phantom 3 mounted on sting Fig. 9 was inserted into the sting tube on the sting stand and stand .

was bolted in place for the duration of the test. The load cell Test matrix and t esting procedures For each model, the test plan called for three different configurations to be tested in the wind tunnel: full vehicle, bare airframe (rotors removed), and isolated rotor. By testing the vehicles in these three configurations, it will be possible to determine the performance of the rotors in isolation as well as the full vehicle performance, where rotor - to - rotor an d rotor - to - airframe interactions have an effect.

Additionally, the bare airframe runs will provide measurements of the forces and moments on the vehicle structure with the airframe drag being the primary measurement of interest.

In hover, both the full ve hicles and isolated rotors were tested . Again, these measurements will make it possible to quantify the isolated rotor performance as well as the effects of aerodynamic interactions between the rotors and airframe.

Rotor RPM was varied in two ways for the full vehicle configurations. First, the rotor speed was changed uniformly for all of the rotors on the vehicle in order to quantify the effects of RPM on the vehicle forces and moments, particularly lift and drag. Second, non - uniform rotor speeds Fi gure 11 . Model mounting plate for DJI Phantom 3 .

were tes ted, where there was a difference in RPM either from front to back or from side to side. These differential For the DJI and 3DR quadcopters , shown in Figs. 2 - 4 , the RPM runs will allow quantification of the pitch and roll camera gimbal mounting holes were used to support t he moments that can be induced by the rotors as well as provide vehicles and serve as attachment points to the interface estimates of the different rot or speeds required to trim hardware . The DA x8 octocopter (Fig. 5 ) and SUI Endurance pitching moment in forward flight.

quadcopter (Fig. 6 ) are heavier vehicles, so the camera mounting holes did not provide sufficient strength to support The primary variables for each test run were model angle of the vehicle weight and aerodynamic loads. For the latter two attack and rotor RPM. Each run started with a set of static vehicles , the mounting hardware was attached directly to the and housekeeping points with wind off and wind on that vehicle chassis. For the Endurance, it was possible to attach were repeated at t he end of the run to check for repeatability the DJI Phantom mounting plate to the bottom chassis plate of measurements . T o facilitate the application of of the vehicle. For the DAx8, a set of metal struts had to be aerodynamic tares, speed was set based on dynamic fabricated to attach t o the vehicle inner structure as sh own in pressure, q . With speed and vehicle pitch fixed, RPM was Fig. 12 .

swept from approximately 80 to 120 percent of the base line RPM. All target rotor speeds were rounded to the nearest 100 rpm to make RPM adjustments easier. Baseline RPM was selected based on a thrust approximately equal to the manufacturers’ specified nominal flight weight. Table 2 gives the baseline thrust a nd rotor speed for each of the five test vehicles .

Table 2 . Baseline vehicle thrust and rotor RPM .

Baseline Baseline Vehicle Thrust (lb) Rotor RPM 3DR SOLO 3.3 5,700 3DR Iris+ 2.8 5,400 DJI Phantom 3 Advanced 2.8 5,300 Drone America DAx8 12.8 7,000 SUI Endurance 6.0 3,500 Figure 12 . Drone America x8 internal mounting hardware.

RPM for each rotor was adjusted manually on the secondary a nd is discussed further in the Measurement Uncertainty data acquisition system as described earlier . After testing the section .

uniform RPM data points , differential rotor speed was tested to measure moments on the vehicles. The proced ure was The linear fit based on tunnel temperature was applied to the then repeated for a series of pitch angles. The test m atrix is force measurements using the recorded temperature for each summarized in Table 3 and presented in full in Table s A1 - data point. T he results of these corrections for an example A3 in the Appendix. The bulk of testing concentrated on the run are s hown in Fig. 14 . The data shown are for the 3DR full vehicle configurations at nominal speeds of 20 and Iris+ at q = 1.9 lb/ft , with pitch ranging from - 5 to 0 deg and 40 ft/s, wit h a limi ted number of runs at 60 and 80 ft/s. Due RPM from 4,300 to 6,500, plus differential RPM. In this to time constraints, yaw sweeps as well as isolated rotor particular case, the temperature increased from 54.6° F at the testing were limited to the 3DR SOLO, DAx8, and the SUI beginning of the run to 62.9° F at the end of the run. This Endurance. temperature change resulted in the uncorrected z - force measurement for the final static point being 0.2 lb below the Table 3 . Test matrix summary . initial static point reading, or about 4% of the maximum force measured during this run. After applicat ion of the Pitch Yaw Airspeed RPM temperature drift correction, the final static point only differs Configuration (deg) (d eg) (ft/ s) (% baseline ) from the initial static point by 0.003 lb.

Full Airframe - 40 – 40 0 – 90 20 – 80 80 – 120 Bare Airframe - 40 – 40 0 – 90 20 – 80 N/A Load cell a erodynamic tares for the full vehicles and bare Rotor Only - 40 – 40 N/A 20 – 80 80 – 120 airframes were measured with just the model interface plate Hover N/A N/A 0 70 – 130 installed and ta ped over as shown in Fig. 15 (note that this photo was not taken in the wind tunnel, but does show the For the DAx8 and the Endurance, h over testing was done configuration as tested for the aerodynamic tares) . Not with the model mount ed in the wind tunnel due to limited availability of those two vehicles. The remaining three vehicles were hover tested in a lab . In the wind tunnel, the 0.8 rotors are approximately six feet from the walls, so there is 0.6 potential for recirculating air to affec t the measurements. In 0.4 the lab, the models were mounted on the sting stand , with 0.2 the nearest wall approximately 30 feet away. Recirculation should therefore not be a problem for the data collected in the lab. The effects of recirculation for the hover test s in the Change, lb -0.2 z tunnel have yet to be quantified. F -0.4 Data reduction and corrections -0.6 -0.8 Most of the data processing was performed in real time by the BDAS system. The two post - processing tasks required -1 -5 0 5 10 15 were temperature drift corrections and application of Temperature Change, ° F aerodynamic tares. During the test, it was observed that the Fi gure 13 . Variation of z - force static measurements with final static points were not matching well with the initial temperature .

static points. In particular, the z - force for the final static point was often approximately 0.1 – 0.3 lb less than that for 5 the initial static poi nt on a given run. This drift appeared to be well correlated with temperature changes in the test section. After the test was completed, a linear fit was generated for each of the six force and moment measurements relating temperature change to drift in th e static point force measurement s . The z - force measurement , lb z F drift with temperature is plotted in Fig. 13 along with the linear fit to the data.

Corrected Data Uncorrected Data The temperature typically increased from th e beginning of a run to the end because testing was carried out in t he morning and early afternoon when exterior temperatures were rising.

-1 As the plot shows, there is additional scatter in the data not 0 10 20 30 40 50 Point related to temperature drift. This scatter can provide a first approximation of the uncertainty in the test measurements Figure 14. z - force fo r Run 63 – 3DR Iris+ with and without temperature drift corrections.

accounted for in the tares were the effect of the individual RESULTS model mounting plates and the cables that supplied power A selection of the test results is presented in this section. A and RPM commands to the vehicles. The mounting plates follow - on data report will be published containing additional were not includ ed, because putting them in the airstream in data . Forward flight results for the full airframes will be isolation would not be representative of their effects on the presented first in this section including trim analysis, full model. The cables were not included, because with the followed by bare airframe and isolated rotor results. Nex t, a vehicle removed, there was nothing to attach them to. A subset of the hover test results will be presented. The separate set of aerodynamic tares was measured for the vibrations of the models will then be discussed, followed by isolated rotors, and this configuration is shown in Fig. 16.

a discussion of measurement uncertainty.

For the isolated rotor tares, al l of the hardware shown in Fig. 7 was installed except for the rotor blade.

Full Airframe Lift and drag results for all five vehicles are presented in Figs. 17 - 26 . For the three smaller quadcopters (Figs. 17 - 22 ), the scales on the lift and drag plots are the same in order to show any differences between these similar vehicles. T est condition s shown include variations in RPM and model pitch at zero deg rees yaw and q = 0.4 8 lb/ft ( approximately 20 ft/s ) . F or the 3DR SOLO and SUI Endurance , pitch angles were tested from - 40° to +40° (positive pitch defined as nose - up) at this airspeed. DAx8 pitch ranged from - 40° to +20°. Due to time constraints, positive angles of attack w ere not tested for the DJI Phantom or 3DR Iris+, with pitch ranging from - 40° to 0°.

In general for the quadcopters, lift increases from α = - 40° to Figure 15 . Full vehicle and bare airframe aerodynamic 0° and then levels off. For the DAx8 octocopter, lift continues to increase above zero pitch, most likely because tare con figuration .

the DAx8 airframe generates a significant amount of lift at positive angles of attack. F or all five vehicles, the drag increases monotonically with angle of attack. This is not surprising, since at negative angles of attack, the rotors are thrusti ng forward, and at positive angles of attack, they are thrusting aft.

The variation in drag with rotor RPM at zero angle of attack is not consistent across the five different vehicles. The DAx8 shows the largest dependence of drag on RPM, with the drag al most doubling as RPM is increased from the lowest to the highest rotor speed. For the other vehicles, this variation is much less pronounced at zero pitch . The cause of this discrepancy is yet to be determined, but may be aerodynamic interactions or differ ences in rotor geometries.

Figure 16 . Isolated rotor aerodynamic tare Trim configuration.

By searching for the regions where the lift values equal the weight of the vehicle and where the net drag force and Typically in this type of wind tunnel test, weight tares would pitching moment are zero, it is possible to estimate at what be required to account for changes in the model center of RPM and pitch value the tr im condition is rea ched.

gravity relative to the load cell during model pitch changes.

Figure 18 shows that for all rotor speeds, the drag values for For this test, however, weight tares were not required, the SOLO pass through zero at approximately α = - 6 ° . At because th e model was mounted on its side and the load cell α = - 6 ° , the RPM for a lift value of 3.3 lb (the nominal flight moved with the model. In this configuration, the gravity weight of the SOLO including camera and battery), is vector of the model does not move relative to the load cell approximately 5,400 rpm.

during pitch changes. The gravity vector does move relative to the load cell for model yaw changes, but a new data run was started any time the model yaw was changed, at which To trim moments, a higher rotor speed would be required on the aft rotors. Pitching moment for the differential RPM runs point the load cell measurements were re - zeroed.

with the SOLO at q = 0.48 lb/ft is shown in Fig. 27 . The 6 5 5 4600 RPM 5100 RPM 3 5700 RPM 6300 RPM 4 6800 RPM 3 1 Lift, lb Drag, lb 4600 RPM -1 5100 RPM 5700 RPM 6300 RPM -2 6800 RPM 0 -3 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 17 . Lift for 3DR SOLO, q = 0.48 lb/ft , Figure 18 . Drag for 3DR SOLO, q = 0.48 lb/ft , yaw = 0° .

yaw = 0° .

4200 RPM 4200 RPM 4800 RPM 3 4800 RPM 5300 RPM 4 5300 RPM 5800 RPM 5800 RPM 6400 RPM 2 6400 RPM Lift, lb Drag, lb -1 -2 -50 -40 -30 -20 -10 0 10 20 30 40 50 -3 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 19 . Lift for DJI Phantom 3, q = 0.48 lb/ft , yaw = Figure 20 . Drag for DJI Phantom, q = 0.48 lb/ft , 0° .

yaw = 0° .

4300 RPM 4900 RPM 5400 RPM 4300 RPM 5900 RPM 4900 RPM 6500 RPM 5400 RPM 5900 RPM Lift, lb 6500 RPM Drag, lb -1 -2 -3 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 21 . Lift for 3DR Iris+, q = 0.48 lb/ft , yaw = 0° .

Figure 22 . Drag for 3DR Iris+, q = 0.48 lb/ft , yaw = 0° .

18 10 5000 RPM 5600 RPM 6200 RPM 6800 RPM 7400 RPM 8 Lift, lb Drag, lb 5000 RPM 5600 RPM 6200 RPM -2 6800 RPM 7400 RPM -4 -6 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 23 . Lift for DAx8, q = 0.48 lb/ft , yaw = 0° .

Figure 2 4 . Drag for DAx8, q = 0.48 lb/ft , yaw = 0° .

10 8 Lift, lb 0 Drag, lb 2800 RPM 2800 RPM 3200 RPM -2 3200 RPM 3500 RPM 3500 RPM 3800 RPM 3800 RPM 4200 RPM 4200 RPM -4 -6 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 26 . Drag for SUI Endu rance, q = 0.48 lb/ft , yaw = Figure 25 . Lift for SUI Endurance, q = 0.48 lb/ft , yaw = 0° .

0° .

moments are expressed about a point at the center of the four rotors at the height of the rotor plane. Based on Fig. 27, at α = - 6 ° , a pitch moment of zero would be reached for a rotor speed of approximately 6,200 rpm on the aft rotors and 5,200 on the forward rotors. Assuming changes in pitching moment are only dependent on the difference in rotor speeds (and not on the averag e rotor speed, which is not a bad approximation), the trim condition for the SOLO at -5 q = 0.48 lb/ft would be a pitch of - 6°, forward rotor RPM of Fwd 6800, Aft 4600 Pitch Moment, in-lb 4,900 and aft rotor RPM of 5,900.

Fwd 6300, Aft 5100 -10 Fwd 5700, Aft 5700 Fwd 5100, Aft 6300 Fwd 4600, Aft 6800 Bare Airframe -15 Lift and drag results for the bare airframe runs are shown i n -20 -15 -10 -5 0 5 10 15 Figs. 28 and 29. The solid lines show results for q = 1.9 lb/ft Pitch, deg (40 ft/s), and the dotted lines show results for q = 0.48 lb/ft Figure 27. Pitching moment for the 3DR SOLO, (20 ft/s). The results are shown normalized by dynamic q = 0.48 lb/ft , yaw = 0°, differential RPM.

pressure to expose any Reynolds number effects. The quadcopter airfra mes all generate very little lift. The DAx8 octocopter airframe, however, is a large flat disk with a curved top, so it actually does generate a significant amou nt of lift. The curves for the two airspeeds are very similar, indicating very little effect of Reynolds number on lift.

Because the DAx8 is a much larger vehicle than the quadcopters, its drag is much higher and makes it difficult to 2 interpret the result s for the smaller vehicles in Fig. 29. The SOLO q = 0.48 -1 same drag results are shown in Fig. 30 but with the DAx8 Lift/q, ft SOLO q = 1.9 Phantom q = 1.9 results removed to increase the scale and better show the Iris q = 0.48 Iris q = 1.9 -2 quadcopter results. The Rey nolds number effects on drag of DAx8 q = 0.48 DAx8 q = 1.9 t he DAx8 and SUI Endurance airframes ar e minimal. For the Endurance q = 0.48 -3 Endurance q = 1.9 smaller quadcopters, lowering the speed results in a higher q - normalized drag.

-4 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Isolated Rotor Figure 28 . Lift/ q for bare airframe runs .

In addition to the full airframe runs, the isolated rotors of the 3DR SOLO, DAx8, and SUI Endurance were tested in forward flight. This c onfig uration was previously shown in Fig. 7 for the SUI Endurance rotor. A small metal adapter SOLO q = 0.48 was fabricated to attach a motor to the model interface plate SOLO q = 1.9 4.5 with a standoff distance of 2.1 in from the interface plate.

Phantom q = 1.9 Iris q = 0.48 The adapter only allowed for attaching a mo tor with a 4 Iris q = 1.9 DAx8 q = 0.48 16 mm x 19 mm bolt pattern (the standard bolt pattern used DAx8 q = 1.9 3.5 Endurance q = 0.48 by the motors on the Iris+, SOLO, and Phantom 3).

Endurance q = 1.9 Therefore, an Iris+ motor was used for all of the isolated 2.5 rotor tests. Due to time constraints on testing, only a limited number of con ditions wer e tested for the isolated rotor cases .

Drag/q, ft These conditions are listed in Table A3 in the Appendix.

1.5 Figures 31 and 32 show the lift and drag for the 3DR SOLO rotor at q = 0.48 lb/ft .

0.5 By comparing the isolated rotor results with the full airframe -50 -40 -30 -20 -10 0 10 20 30 40 50 results, it is possible to estimate the effect of rotor - airframe Pitch, deg and rotor - rotor interactions. For example, at 0 degrees pitch and 5700 RPM, the isolated rotor generates 1.0 lb of lift. For Figure 29 . Drag/ q for bare airframe runs .

the same test condition with the full airframe, the lift shown in Fig. 17 is 3.8 lb, and the bare airframe lift (calculated by multiplying the value given in Fig. 28 by 0.48 lb/ft ) is 0.8 0.01 l b. With no aerodynamic interactions, the total expected lift would be the number of rotors (4) times the isolated rotor SOLO q = 0.48 0.7 SOLO q = 1.9 lift, plus the isolated airframe lift, for a total of 4.01 lb. This Phantom q = 1.9 2 Iris q = 0.48 indicates that at q = 0.48 lb/ft and zero pitch , approximately 0.6 Iris q = 1.9 Endurance q = 0.48 5 percent of the SOLO rotor lift is lost to rotor - airframe and Endurance q = 1.9 0.5 rotor - rotor interactions. Similar calculations can be made for the other vehicles and different test conditions.

0.4 Drag/q, ft Because there are no flap hinges on the rotors used for these 0.3 small vehicles, it can be expected that the rotors will carry 0.2 hub moments, e specially at higher airspeeds. As an example, t he roll moment at the SOLO ro tor hub at q = 0.48 lb/ft is 0.1 given in Figs. 33 and 34 , first as the measured roll moment, and second as the lift offset. The lift offset is given by: -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg !

!

!""#$% = Figure 30 . Drag/ q for bare airframe runs ( DAx8 results ! !

!

removed ) .

Roll moment varies much more with shaft pitch than with Finally, isolated rotor tests enable measurement of rotor RPM. The lift offset peaks at a pitch angle of - 2° for all individual rotor shaft torque, allowing calculation of the rotor speeds and varies from 5 to 8 percent of the rotor mechanical power produced by t he rotor at each test radius. As expected, the lift offset generally decreases with condition. Results for both mechanical and electrical power increased rotor speed (and decreased advance ratio). are given in Fig. 35 for the 3DR SOLO rotor at Advance ratios range fro m 0.07 at 6,800 RPM to 0.10 at q = 0.48 lb/ft . The motors and ESCs have not yet been 4,600 RPM. It is unclear why the lift offset is higher for bench tested for their individual efficiencies, so dividing the 5,100 RPM than it is for 4,600 RPM; however, it could be value fo r the mechanical power (solid line) by the electrical due to measurement error, since the measured r oll moments power (dotted line) gives the electrical efficiency of the are very small. The discrepancy could be due to Rey nolds combined motor and ESC. In this case the efficiency is number effects. between 72 and 78 percent, depending on rotor speed. Note that a motor and ESC from the 3DR Iris+ were used for all For q = 1.9 lb/ft (not shown), the measured lift offset was as three isolated rotors that were tested. As the results in high as 13% for an RPM of 5,700; however, a resonance Fig. 35 show, the power is largely independent of shaft pitch issue encountered with the test stand prevented testing the angle, and is much more dependent on rotor RPM. The two lower rotor speeds at this higher airspeed. results for the other two isolated rotors tested (not pr esented here) showed similar trends.

4600 RPM 5100 RPM 0.8 5700 RPM 2.5 6300 RPM 0.6 6800 RPM 0.4 0.2 1.5 Lift, lb Drag, lb 4600 RPM -0.2 5100 RPM 5700 RPM -0.4 6300 RPM 6800 RPM -0.6 0.5 -0.8 -1 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Fi gure 31 . Lift for 3DR SOLO rotor, q = 0.48 lb/ft .

Fig ure 32 . Drag for 3DR SOLO rotor, q = 0.48 lb/ft .

0.5 0.15 4600 RPM 5100 RPM 5700 RPM 0.1 6300 RPM 6800 RPM 0.05 Lift Offset -0.5 Roll Moment, in-lb 4600 RPM 5100 RPM 5700 RPM 6300 RPM 6800 RPM -0.05 -1 -50 -40 -30 -20 -10 0 10 20 30 40 50 -50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Pitch, deg Figure 34 . Lift offset for 3DR SOLO rotor, q = 0.48 lb/ft .

Figure 33 . Roll moment for 3DR SOLO rotor, q = 0.48 lb/ft .

4600 RPM Mech.

Power, W 4600 RPM Elec.

5100 RPM Mech.

5100 RPM Elec.

5700 RPM Mech.

20 5700 RPM Elec.

6300 RPM Mech.

6300 RPM Elec.

6800 RPM Mech.

6800 RPM Elec.

-50 -40 -30 -20 -10 0 10 20 30 40 50 Pitch, deg Figure 35 . Power for the 3DR SOLO rotor, q = 0.48 lb/ft .

Hover Test Results Figure 36 . 3DR Iris+ hover test configuration .

The hover performanc e was measured for all five complete vehicles and all five isolated rotors. As previously mentioned in the Test Description section, the DAx8 and SUI Endurance were hover tested in the test section of the 7 - by 10 - ft wind tunnel mounted at zero degrees in the wind - on test configuration . The remaining three vehicles and all five Phantom SOLO isolated rotors were tested on the sting stand in a lab , as Iris shown in Fig. 36 . A clone of the BDAS system used in the DAx8 Endurance wind tunnel was used for data acquisition in the lab. The range o f hover test conditions is included with the full Thrust, lb airframe and isolated rotor test matrices in Tables A1 and A3.

Results for the full vehicle hover tests are given in Figs. 37 and 38. The first figure shows the thrust as a function of rotor RPM and the se cond shows the thrust coefficient, 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 defined as T / ρ A ( Ω R ) . As expected, the thrust follows a RPM quadratic trend with rotor speed. Thrust coefficient for all Figure 37 . Full vehicle hover thrust .

five vehicles shows a slight increase with RPM, and the Phantom operates at the highest thrust coefficient of the five 0.02 vehicles. The variation with roto r RPM indicates that non - Phantom 0.018 SOLO dimensional rotor lift increases slightly with increasing Iris 2 DAx8 Reynolds number over the range of rotor speeds tested. For 0.016 Endurance R) reference, the Reynolds number at the tip of the DJI + 0.014 A( Phantom rotor blade at 5,000 RPM is approximately 50,000.

; 0.012 T hrust coefficient results for the isolated rotors are given in 0.01 Fig. 39. The results are similar to the full vehicle results, but 0.008 there is more noise in the thrust coefficient curves. This is likely due to higher relative measurement uncertainty for the 0.006 iso lated rotors. The thrust being measured is only a quarter 0.004 Thrust Coefficient, T/ of that of the full vehicles for the quadcopters (and an eighth 0.002 of that of the octocopter), so the relative error in the measurement is four times higher, which carries into the 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 thrust coefficient c alculation. Regardless, the same trend is RPM observed as for the full vehicle, with thrust coefficient increasing slightly with RPM.

Figure 38 . Full vehicle hover thru st coefficient .

0.02 150 1 0.018 0.9 Mech. Power Elec. Power 0.8 0.016 R) Efficiency + 0.7 0.014 A( 100 ; 0.6 0.012 0.5 0.01 Power, W 0.4 0.008 Phantom 0.3 Electrical Efficiency 0.006 SOLO Iris 0.2 DAx8 0.004 Thrust Coefficient, T/ Endurance 0.1 0.002 0 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 RPM RPM Figure 40 . Hover power fo r DJI Phantom rotor .

Figure 39. Isolated rotor thrust coefficient.

By comparing the full vehicle and isolated rotor results for 0.9 thrust coefficient, it is also possible to calculate the download, DL , in hover with the following: 0.8 0.7 ! − !

! , !" ! , !"

!" = 0.6 !

! , !"

Phantom 0.5 SOLO where the IR and FV subscripts represent isolated rotor and Iris 0.4 DAx8 full vehicle, respectively. The hover download across all five Endurance 0.3 vehicles for the baseline rotor speed ranges from a low of Electrical Efficiency approximately 5% for the SUI Endurance to about 15% for 0.2 the DAx8. The SUI Endurance has fairly thin arms 0.1 supporting the rotors, while the DAx8 rotor support arms are relatively thick, which is likely the driving factor behind the 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 difference in hover download between the different vehicles.

RPM As with the forward flight isolated rotor tests, the Figure 41 . Isolated rotor electrical efficiency .

mechanical and electrical power can be compared to compute the efficiency of the motor - ESC combination. The since the motor and rotor are normally paired to achieve same motor and ESC from the Iris+ that were used for maximum efficiency on the production vehicles. If the other forward - flight testing were used for the hover tests. As an rotors were hover tested with their appropriate motors, the e xample, Fig. 40 shows the mechanical and electrical power measu red electrical efficiencies would be expected to be of the DJI Phantom rotor on the primary y - axis. Electrical higher.

efficiency is plotted on the secondary axis. The combined motor - ESC efficiency is as low as 50% at the lowest RPM Finally, with the mechanical power and thrust tested and increases quic kly above 70% once rotor speed is measurements, it is possible to calculate the rotor figure of increased to 4,500 rpm. Maximum efficiency of 77% was merit, M , defined as the ratio of ideal induced power to total reached at the highest rotor speed tested. Efficiency results rotor mechanical power: were similar to the Phantom results for the other similarly sized quadcopter rotors (3DR Iris+ and 3DR SOL O).

! ! 2 !"

! = !

!"# !

For the DAx8 and SUI Endurance, which use larger rotors, the efficiency drops off significantly, as shown in Fig. 41.

In the above equation, T is the rotor thrust, and P is the mech Again, the Iris+ motor was used to test all five isolated mechanical power, equal to M Ω . The measured figure of z rotors, and it was very under - sized for the larger rotors. This merit for all five rotors is plotted in Fig. 42 . The highest is particul arly true for the 15 - inch SUI rotor, which is figure of merit measured is for the 3DR I ris + , which has a normally mated to a much larger motor on the vehicle. The maximum of 0.72 at 8,000 rpm. For each rotor , the figure of highest electrical efficiency for the isolated rotor hover tests merit is maximized at the upper end of the RPM ran ge. The was measured for the Iris+ rotor, which is not surprising, results show good agreement with results from Ref. 6 for the vibration magnitude varies significantly with time. If the plot APC Thin Electric 10x5 blade , which has a similar geometry showed a longer sample time, the beat frequency wou ld be to the rotors of the DJI Phantom 3. The figure of merit for much more apparent. An FFT of the same data is given in the APC blades ranges from 0.59 at 2,508 rpm to 0.66 at Fig. 44 .

6,708 rpm, while for the Phantom blade measured here, the figure of merit is 0.58 at 2,500 rpm and 0.67 at 6,5 00 rpm. At 5,700 rpm, the 1/rev frequency is 95 Hz. There is a low peak at 1/rev due to blade imbalance, but the primary The four smaller quadcopter rotors have very similar trends vibration frequencies are at 2/rev and 4/rev. The 2/rev is for figure of merit. The SUI Endurance rotor has a higher explained b y the fact that the rotors are 2 - bladed, with very figure of merit at the low RPM values than any of the other little flexibility in the lag direction. The alternating rotors. The Endurance rotor was limited to a maximum aerodynamic conditions seen by the blades between the speed of 4,500 r pm because of the power limit of the Iris+ advancing and retreating sides therefore lead to high 2/rev motor being used for isolated rotor testing. vibrations. Aerodynamic interactio ns between the forward and aft rotors are the likely cause of the higher harmonic content at 4/rev and 8/rev.

0.9 Vibrations were measured for the same test condition for the 0.8 isolated rotor, and the results are shown in Fig. 45. Because 0.7 there was only one rot or operating at a fixed speed, there is no beating as there is for the full vehicle results. Looking at 0.6 Phantom 0.5 SOLO Iris 0.4 DAx8 Figure of Merit Endurance 0.3 0.2 0.1 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Fx, lb RPM -5 Figure 42 . Isolated rotor figure of merit .

-10 Vibrations One of the initial ch allenges encountered during the wind -15 tunnel test was a high level of vibration . Vibration s in this 0 0.1 0.2 0.3 0.4 0.5 Time, s type of vehicle are not unexpected, and in fact, the designs for all of the vehicles show that the manufacturers actively Figure 43. F time history for 3DR SOLO, q = 1.9 lb/ft , x work ed to reduce vibrations by using rubber dampers on α = 0°, RPM = 5,700.

critical components, the camera gimbals in particular. The 3.5 level and character of vibration, however, was previously not well documented.

As discussed in the Test Setup section, the BDAS system 2.5 was configured to measure the mean values of the load cell measurements over a 30 - second data record. It was not set up to measur e dynamic data; however, the signal conditioning/amplifier system could be used at any time to |Fx|, lb 1.5 capture short data records at a rate of 100,000 samples per second. Vibration data were collected for several test points using this system. Sample results for t he 3DR SOLO are provided in this section. Figure 43 shows a 0.5 second time 0.5 history of F for the full vehicle for a rotor speed of x 5,700 rpm at a vehicle pitch of zero degrees and 0 200 400 600 800 1000 q = 1.9 lb/ft (approximately 40 ft/s).

Frequency, Hz At this condition, the mean drag i s 0.8 lb, but as the plot 2 Figure 44. FFT of F for 3DR SOLO, q = 1.9 lb/ft , x shows, the peak loads are as much as ±10 lb. Additionally, α = 0°, RPM = 5,700 .

due to the inconsistent phasing of the four rotors, the the FFT for the isolated rotor results, shown in Fig. 46 , the highest vibration levels are at the 2/rev frequency, but there is still a significant amount of 4/rev content. Since the data shown here were collected at a shaft pitch of zero degrees, blade vortex interaction is the likely cause of the higher harmonic content [Ref. 15 ].

Because vibration measurements were not in the original test plan, there were only limited measurements taken. Further Fx, lb testing that focuses on better characterizing the vibra tions of -5 these vehicles and exploring mitigation strategies would therefore be a good follow - on topic of research.

-10 Measurement Uncertainty -15 As previously mentioned, the manufacturer’s stated 0 0.1 0.2 0.3 0.4 0.5 Time, s uncertainty for the load cell is 0.25% of full scale. For the 50 - lb load cell that was used for the majority of testing, the Figure 45 . F time history for 3DR SOLO isolated rotor, x capacity in the x - and y - directions is 50 lb. In the z - direction, q = 1.9 lb/ft , α = 0°, RPM = 5,700 .

the capacity is 100 lb, and the moment capacity is 150 in - lb in all three directions. This translates to an uncertainty of 3.5 0.13 lb in the x - and y - force measurements, 0.25 lb in the z - force measurements, and 0.38 in - lb in all three moment directions. Based on the error analysis presented below, the uncertainty is believed to be less than the manufacturer’s 2.5 stated uncertainty for the load ranges measured during this test.

Check loads were performed both before and after the test in |Fx|, lb 1.5 order to quantify the accuracy of the load cell over the load range tested. The results for the post - test check loads in the x - direction are shown in Fig. 47. Measurements were taken with both increasing and decreasing loads to check for 0.5 hysteresis. If the load cell measurements were perfect, the measured loads would exactly follow the black line, with a 0 0 200 400 600 800 1000 slope of 1. The measured loads shown are slightly less than Frequency, Hz expected, with the absolute error increasing to a maximum of 0.19 lb as the load increases. The relative error for all of Figure 4 6 . FFT of F for 3DR SOLO isolated rotor, x q = 1.9 lb/ft , α = 0°, RPM = 5,700 .

the check loads is fairly constant for the non - zero measurements, indicating that the slope of the sensitivity curve over this l oad range is slightly off. Based on the check load data, the uncertainty in the x - direction due to calibration error is calculated to be ±0.020 F . As shown, the Expected Load x 10 Measured Load, Increasing hysteresis is very small, with a maximum value of 0.021 lb.

Measured Load, Decreasing The uncertainty in the x - direction due to hysteresis is therefore considered to be ±0.021 lb.

The third source of force and moment measurement uncertainty that can be applied across all of the data runs is Measured the repeatability of measurements. Ideally, the temperature - x F corrected forces and mom ents measured for the static point at the end of each run would be zero; however, there is scatter in these ending static points (shown for the z - direction in Fig. 13). The 95% confidence interval for the uncertainty due to non - repeatability of data points is two -2 times the standard deviation of the errors for the ending -2 0 2 4 6 8 10 12 F Applied static points. The contributions to uncertainty for all six x force and moment measurements are given in Table 4.

Figure 47. F check load results x The final source of uncertainty is the fact that the force and database for multicopter vehicles. This database can then be moment measurements are based on an average of unsteady used to accurately model trajectories of these types of loading measurements. For each measurement taken during aircraft. There is currently an effort underway to build a the test, the standard deviation was recorded, so for a 95% query - able database of the data collected in this test to make confidence interval, the uncertainty due to measurement it e asier to extract trends. In its current form, data can only be extracted by finding a desired test condition in the test log unsteadiness is 2 ! / ! , where σ is the standard deviation at and pulling the matching data from the dataset.

a given dat a point and N is the number of samples . For the 30 - second data records at 1,024 samples per second There are also follow - on experimental efforts planned to collected for this test, N equals 30,720.

further characterize sma ll multicopter vehicles. The motors and speed controllers used for this test will be bench tested The total measurement uncertainty is then calculated as the to better characterize their efficiency as a function of torque root - sum - squared of the four different uncertainty sou rces.

and motor RPM. These tests will follow the methodology Figure 48 shows an example of the z - force data with error described by Green and McDonald in Ref. 16 . Finally, in bars for Run 63, which is the same run shown in Fig. 14.

order to accurately characterize the performance of these The maximum calculated uncertainty in the z - force for this vehicles using a comprehensive analysis tool such as run is 0.163 lb.

CAMRAD II, it may be necessary to measure blade deflections in flight. An effort is underway to use a Table 4. Uncertainty contributions.

photogrammetry system to me asure the blade deflections in Source F F F M M M x y z x y z both hover and in forward flight.

Relative Cal.

0.020 0.012 0.002 0.004 0.003 0.008 Error CONCLUSIONS Hysteresis, 0.021 0.011 0.052 0.010 0.032 0.018 The goal of this test was to generate a high - quality set of lb or in - lb Repeatability, data for the performance of multicopter UAS vehicles. The 0.085 0.055 0.123 0.148 0.269 0.191 intent is to use this data to enhance software tools in use lb or in - lb both inside and outside of NASA to design and analyze multicopter vehicles. The applications targeted by this test are trajectory estimation and vehicle design and analysis, but there could certainly be others that will be able to make use of the data.

Th e test generated data for five different vehicles in both forward flight and in hover. The data include forces and moments and electrical power as a function of rotor RPM, 2 , lb airspeed, and vehicle attitude. This paper described the test z F setup and procedure as well as summarized the results of the test at a select number of operating conditions. A large amount of data from this test was not included in this paper and will be published in a technical report at a later date.

The data that were collected for the full vehicles as well as -1 for the bare ai rframes and isolated rotors will allow for 0 10 20 30 40 50 development and validation of analytical and numerical Point models at both the full vehicle and component l evels. This Figure 48. z - force for Run 63 – 3DR Iris+ including breadth of validation data was previously unavai lable in the uncertainty.

public domain. T he data produced by this research effort represent a significant step forward in advancing the FUTURE WORK understanding of multicopter UAS performance.

The next major research effort related to this test is to use ACKNOWLEDGMENTS the data to calibrate and validate the various models used to model multicopter vehicles. In particular, research under A large team was required to accomplish this test, and the DELIVER and RVLT will seek to validate CAMRAD II authors wo uld like to acknowledge the contributions of all models of sma ll rotors using the data collected here. These who were involved. In particular, the authors would like to models will then be used to calibrate rotor performance thank Nili Gold and the 7 - by 10 - ft Wind T unnel test crew models in NDARC to allow for design trade studies of for all the hard work they put in to ensure that the test ran multicopter vehicles. Research under UTM will use the data smoothly. The NASA machine shop teams, led by Robert collected here to help populate a vehicle performance Kornienko and Vincent Derilo, produced the vehicle mounting hardware, which was absolutely critical to the success of the test. Tom Norman provided invaluable 10 . “Phantom 3 Advanced Specs,” insights into wind tunnel testing and data reduction. Straight http://www.dji.com/product/phantom - 3 - adv/info#specs , Up Imagi ng provided the Endurance vehicle for the wind accessed March 28, 2016.

tunnel test. The assistance of all of those who acted as safety loads monitors during the te st is greatly appreciated. Three 11 . “3D R Iris+,” https://3dr.com/iris - plus/ , NASA projects supported this test: the UAS Traffic accessed March 28, 2016.

Management (UTM) Sub - p roject , und er the Safe Autonomous Systems Operations (SASO) Project; the 12 . “Drone America – DA x8,” Design Environment for Novel Vertical Lift Vehicles http://www.droneamerica.com/systems/dax8 , (DELIVER) Sub - p roject, under the Convergent Aeronautics accessed March 28, 20 16.

Solutions (CAS) Project; and the Revolutionary Vertical Lift Technology (RVLT) Project.

13 . “Products – Straight Up Imaging,” http://www.straightupimaging.com/products/ , REFERE NCES accessed March 28, 2016.

14 . Storms, B., Nishikawa, D., Mason, S, Hange, C., and 1 . Gorton, S., López, I., and Theodore, C ., “NASA Phillips, J., “BDAS 13.2 Use r Manual,” Rev. 2, Technology for Next Generation Vertical Lift February 2016.

Vehicles,” AIAA SciTech, Kissimmee, FL, January 2015.

15 . Johnson, W., Rotorcraft Aeromechanics , Cambridge University Press, New York, NY, 2013, Chapter 18.3: 2 . Johnson, W., “NDARC: NASA Design and Analysis of Vibration, pp. 717 - 722.

Rotorcraft,” NASA/TP - 2015 - 218751, April 2015.

16 . Green, C, and McDonald, R. “Modeling and Test of the 3 . Johnson, W., “Propulsion System Models for Ro torcraft Efficiency of Electronic Speed Contr ollers for Brushless Conceptual Design,” AHS 5th Decennial DC Motors,” 15th AIAA Aviation Technology, Aeromechanics Specialists’ Conference, San Francisco, Integration, and Operations Conference, Dallas, TX, CA, January 2014.

June 2015.

4 . “UTM: Air Traffic Management for Low - Altitude Drones”, NASA facts, NF - 2015 - 10 - 596 - HQ .

5 . Brandt, J. and Selig, M., “Propeller Perfor mance Data at Low Reynolds Numbers,” 49th AIAA Aerospace Sciences Meeting, Orlando, FL, January 2011.

6 . Brandt, J., Deters, R., Ananda, G., and Selig, M., “UIUC Propeller Data Site,” http://m - selig.ae.illinois.edu/props/propDB.html, accessed March 28, 2016.

7 . Carroll, T., George, I. - R., Bramesfeld, G., and Raahemifar, K., “Design Optimization of Small Rotors in Quad - Rotor Configuration,” AIAA SciTech, San Diego, CA, January 2016.

8 . Huang, H., Hoffmann, G., Waslander, S., and Tomlin, C., “Aerodynamic s and Control of Autonomous Quadrotor Helicopters in Aggressive Maneuvering,” 2009 IEEE International Conference on Robotics and Automation, Kobe, Japan, May 2009.

9 . “SOLO Specs: Just the facts,” ht tps://3dr.com/solo - gopro - drone - specs/ , accessed March 28, 2016.

Table A1. Full airframe test matrix .

u Speed Nominal Yaw Pitch Angle v Model RPM ( q , lb/ft ) Speed, ft/s Angle, deg 3DR SOLO 0.48 20 0 - 40 – 40 4,600 – 6,800 + Δ 1.9 40 0 - 40 – 40 4,600 – 6,800 + Δ 7.7 80 0 - 40 – - 20 5,700 – 6,800 0.48 20 - 5 - 40 – 40 4,600 – 6,800 + Δ 1.9 40 - 5 - 40 – 40 4,600 – 6,800 + Δ 1.9 40 - 30 - 10 – 10 4,000 – 6,300 0.48 20 - 45 - 10 – 10 4,000 – 6,300 + Δ 1.9 40 - 45 - 10 – 10 4,000 – 6,300 1.9 40 - 60 - 10 – 10 4,000 – 6,300 0.48 20 - 90 - 10 – 10 4,000 – 6,300 + Δ 1.9 40 - 90 - 10 – 10 4,000 – 6,300 0 (hover) 0 0 0 3,500 – 8,000 + Δ DJI Phantom 3 0.48 20 0 - 40 – 0 4,200 – 6,400 + Δ 1.9 40 0 - 40 – 0 4,200 – 6,400 + Δ 0 (hover) 0 0 0 3,500 – 7,500 + Δ 3DR Iris+ 0.48 20 0 - 40 – 0 4,300 – 6,500 + Δ 1.9 40 0 - 40 – 0 4,300 – 6,500 + Δ 0 (hover) 0 0 0 2,500 – 8,000 + Δ Drone America x8 0.48 20 0 - 40 – 20 5,000 – 7,400 + Δ 1.9 40 0 - 40 – 0 5,000 – 7,400 + Δ ★ 0.48 20 - 5 - 40 – 20 5,600 – 8,400 + Δ ★ 1.9 40 - 5 - 40 – 0 5,600 – 8,400 + Δ ★ 0.48 20 - 30 - 10 – 10 5,600 – 8,400 ★ 0.48 20 - 45 - 10 – 10 5,600 – 8,400 + Δ ★ 0.48 20 - 60 - 10 – 10 5,600 – 8,400 ★ 0.48 20 - 90 - 10 – 10 5,600 – 8,400 + Δ 0 (hover) 0 0 0 4,500 – 9,000 + Δ SUI Endurance 0.48 20 0 - 40 – 40 2,800 – 4,200 + Δ 1.9 40 0 - 40 – 40 2,800 – 4,200 + Δ 4.3 60 0 - 40 – - 20 4,200 – 5,000 0.48 20 - 5 - 40 – 40 2,800 – 4,200 + Δ 0.48 20 - 30 - 10 – 10 2,800 – 4,200 0.48 20 - 45 - 10 – 10 2,800 – 4,200 + Δ 0.48 20 - 60 - 10 – 10 2,800 – 4,200 0.48 20 - 90 - 10 – 10 2,800 – 4,200 + Δ 0 (hover) 0 0 0 1,500 – 4,600 + Δ u Pitch angle is the angle of the tunnel turntable, so in terms of vehicle Euler angles, the pitch rotation would be executed first and the yaw angle second v + Δ next to the RPM entries indicates that differential RPM values were tested in addition to the u niform RPM sweeps ★ During testing of the DAx8, the baseline RPM was changed from 6,200 to 7,000 to better represent the baseline weight of the vehicle Table A2. Bare airframe test matrix .

Speed ( q , Nominal Yaw Angle, u Model Pitch Angle RPM lb/ft ) Speed, ft/s deg 3DR SOLO 0.48 20 0 - 40 – 40 N/A 1.9 40 0 - 40 – 40 N/A 7.7 80 0 - 40 – 40 N/A 0.48 20 - 90 - 10 – 10 N/A 1.9 40 - 90 - 10 – 10 N/A DJI Phantom 3 1.9 40 0 - 40 – 40 N/A 3DR Iris+ 0.48 20 0 - 40 – 40 N/A 1.9 40 0 - 40 – 40 N/A Drone America x8 0.48 20 0 - 40 – 40 N/A 0.48 20 0 - 40 – 40 N/A 0.48 20 - 45 - 10 – 10 N/A 1.9 40 - 90 - 10 – 10 N/A SUI Endurance 0.48 20 0 - 40 – 40 N/A 1.9 40 0 - 40 – 40 N/A 0.48 20 - 90 - 10 – 10 N/A u Pitch angle is the angle of the tunnel turntable, so in terms of vehicle Euler angles, the pitch rotation would be executed first and the yaw angle second Table A3 . Isolated rotor test matrix .

Speed ( q , Nominal Yaw Angle, Model P itch Angle RPM lb/ft ) Speed, ft/s deg 3DR SOLO 0.48 20 N/A - 40 – 40 4,600 – 6,800 1.9 40 N/A - 40 – 0 5,700 – 6,800 7.7 80 N/A - 40 – - 20 5,700 – 8,600 0 (hover) 0 N/A 0 2,500 – 8,000 DJI Phantom 3 0 (hover) 0 N/A 0 2,500 – 8,000 3DR Iris+ 0 (hover) 0 N/A 0 2,500 – 8,000 Drone America x8 0.48 20 N/A - 10 – 0 6,200 – 8,400 1.9 40 N/A - 20 – 0 6,200 – 8,400 0 (hover) 0 N/A 0 2,500 – 8,500 SUI Endurance 0.48 20 N/A - 20 – 0 2,800 – 4,200 0 (hover) 0 N/A 0 2,000 – 4,500