Wind Tunnel Testing Techniques for a Tandem Tilt-Wing, Distributed Electric Propulsion VTOL Aircraft

Wind Tunnel Testing Techniques for a

Tandem Tilt - Wing, Distributed Electric

Propulsion VTOL Aircraft

1 2 3

Ronald C. Busan , Patrick C. Murphy , David B. Hatke , and

Benjamin M. Simmons

NASA Langley Research Center, Hampton, VA, 23681 Urban Air Mobility (UAM) vertical takeoff and landing (VTOL) aircraft designs frequently include multiple distributed propulsors, complex wing - propulsor aerodynamics, significant airframe configuration changes during normal flight operations, and no historical database regarding t he best ways to transition between vertical and horizontal flight. This paper describes the methodology used for wind tunnel testing of the Langley Aerodrome No.

8 (LA - 8 ) in the NASA Langley 12 - Foot Low - Speed Tunnel during multiple test entries in 2019 and 2020. The LA - 8 is a tandem tilt - wing aircraft with 4 motor - propeller units and 4 control surfaces distributed across each wing, plus an inverted V - tail with 2 ruddervators on the fuselage. An initial tunnel entry used one - factor - at - a - time (OFAT) testing t o (1) define candidate trimmed transition corridors between vertical and horizontal flight, (2) assess whether there was adequate control authority, and (3) define appropriate test factor ranges for subsequent design of experiment (DOE) wind tunnel testing . The total number of independent variables for these wind tunnel tests (23) made DOE testing an efficient option for assessing the large number of potential interactions associated with the LA - 8. The general advantages and disadvantages of OFAT and DOE wi nd tunnel testing techniques are also discussed – along with the benefits of a combined approach.

I. Nomenclature dPT1 = distance from wing 1 propeller thrust to CG C = lift coefficient L dPT2 = distance from wing 2 propeller thrust to CG C = drag coefficient D PIA1 = wing 1 propeller induced aerod ynamic force C = pitching moment coefficient m PIA2 = wing 2 propeller induced aerodynamic force  = angle of attack, deg dPIA1 = distance from wing 1 propeller induced aero to CG  = sideslip angle, deg dPIA2 = distance from wing 2 propeller induced aero to CG PT1 = wing 1 propeller thrust AIA1 = wing 1 airstream induced aerodynamic force PT2 = wing 2 propeller thrust AIA2 = wing 2 airstream induced aerodynamic force dAIA1 = distance from wing 1 airstream induced aero to CG dAIA2 = distance from wing 2 airstream induced aero to CG II. Introduction Technology developments in distributed electric propulsion, higher capacity lightweight batteries, and im proved stability and con trol augmentation has spawned development of a wide range of new vertical takeoff and landing ( VTOL ) configurations. Wind tunnel testing has historically played an important role in any aircraft development program, and the variety and complexity of many o f the new VTOL configurations pose new challenges for how wind Senior Research Engineer, Flight Dynamics Branch, MS 308, AIAA Member.

Senior Research Engineer, Dynamic Systems and Control Branch, MS 308, AIAA Associate Fellow.

Research En gineer, Flight Dynamics Branch, MS 308, AIAA Member.

Research Engineer, Flight Dynamics Branch, MS 308, AIAA Member.

tunnel testing should be conducted . In 2013 , testing of the GL - 10 tilt - wing tilt - tail 10 - motor airframe with 29 independent variables in the Langley 12 - Foot Low - Speed Wind Tunnel highlighted th e need to incorporate design of experiment (DOE) and response surface methods (RSM) into the testing process. In this paper references to DOE are intended in the general sense to include both DOE and RSM. Lessons learned from the GL - 10 wind tunnel tests were incorporated into the 2019 and 2020 testing of t he LA - 8 airframe shown in Fig. 1. Figure 2 shows a side - by - side comparison of t he LA - 8 and GL - 10 airframes as tested in the 12 - Foot Tunnel. Reference [ 1] documents the 2013 GL - 10 wind tunnel test, describes the test faculty capabilities, gives a history of the DOE testing conducted in the 12 - Foot Tunnel since 2002, provides example s of the GL - 10 data obtained, and discusses model require ments and the facility hardware/ software modifi cations needed to implement the DOE testing approach . In many ways the current paper can be considered a direct follow - on to that prior document. Refere nces [ 2 ] and [ 3 ] provide additional details about the GL - 10 20 13 DOE data collection and subsequent modeling and simulation work done with that data.

(a ) View from front l eft (b) View f rom t op a ft , showing all control e ffecto rs Fig. 1 LA - 8 m o unted in the NASA Langley 12 - Foot Low - Speed Tunnel .

A number of conference papers and NASA technical memorandums provid e extensive documentation about the LA - 8 research testbed and the recent testi ng and modeling work associated with the aircraft . Design , fabrication, load testing, and mass property assessments of the LA - 8 airframe are described in Refs. [ 4 ] and [ 5 ]. The full one - factor - at - a - t ime (OFAT) test results are presented in R ef. [ 6 ] and isolated propeller te sting is described in R ef. [ 7 ]. Use of the LA - 8 p owered airframe DOE test data and development of the LA - 8 aerodyn amic mo dels are described in Refs. [ 8 ] , [ 9 ], and [ 10 ].

The primary focus of the current paper is to provid e a better understanding of h ow the LA - 8 was tested, how the testing built upon the lessons learned from the GL - 10 tests , and describe the practical benefits and potential issues . A secondary goal is to provide some physics - based insight into LA - 8’s unusual pitching moment characteristics.

Fig. 2 Side - by - side c omparison of t he LA - 8 and GL - 10 a irframes in the 12 - Foot Tunnel .

Thi s paper is organized as follows: Section III looks at the advantages and disadvantages of several wind tunnel testing techniques. Section IV lists the major lessons learned from wind tunnel testing of GL - 10 in 2013. Section V describes the initial OFAT testing of LA - 8 done in the spring of 2019. Section VI includes discussion of LA - 8’s pitching moment characteristics and control options, with an emphasis on establishing a better understanding of the ph ysics involved. Section VII describes the LA - 8 configuration changes made immediately after the spring of 2019 OFAT testing. Section VIII includes some clarifying comments about th e DOE testing done in the summer of 2019.

Section IX briefly discusses compo nent testing and the buildup approach as it applies to LA - 8 and other VTOL configurations. Section X describes potential issues with the LA - 8 initial hover testing results. Section XI includes an initial comparison of LA - 8 OFAT and DOE test results plus a description of planned additional wind tunnel testing, and Sec. XII has concluding summary remarks.

III. Wind Tunnel Testing Approach Tradeoffs OFAT and DOE wind tunnel testing approaches each have their own advantages (pros) and disadvantages (cons).

The fol lowing sections highlight some of the most important pros and cons for each method to consider when developing a wind tunnel test plan, with the best option often being a combination of both approaches.

A. OFAT Testing Pros: OFAT testing is a traditional exp eriment design approach where all known factors, except the one currently under test, are held constant. OFAT wind tunnel testing has the following advantages: 1. OFAT testing is typically easy to set up and implement for systems with limited complexity. Swee ping through the range of a single factor while holding all other factors constant is a simple process to implement at most wind tunnel facilities. Exploratory work to capture a broad understanding of the test article response features, and detailed invest igations relating to individual factors are therefore conveniently handled with the OFAT approach.

2. OFAT data collection typically allows for immediate viewing and interpretation of the results. Most wind tunnels currently include computer software that can produce data plots of the results either in real time (i.e.

point by point) or immediately upon completion of an individual OFAT sweep through a range of test points.

Depending on the operational procedures for the particular wind tunnel, this may allow t est engineers the ability to quickly alter subsequent test runs based upon prior run results.

3. OFAT testing is better understood and the results are often more readily accepted by researchers used to working with wind tunnel data. The OFAT approach is often used in wind tunnels at NASA Langley and many other academic and commercial facilities for decades – and is therefore the default preference for many researchers.

B. OFAT Testing Cons: 1. The presence of any unknown systematic errors present during the test will be incorporated into the data and cannot be corrected by post - processing after the test measurements are taken. Uncharacterized drifting of sensors or test apparatus settings would be examples of this type of error. Considerable time is spent at some wind tunnel facilities attempting to identify and characterize sources of systematic errors – after which they become “known” sources of bias or variance.

2. Known sources of bias and variance (nuisance errors) such as variability among operators, testing on different days or times of day with large temperature differences, etc., can add to measurement errors. Test engineers typically attempt to minimize these contributions through various testing strategies and procedures.

Always approaching an angle of atta ck setting from a particular direction to minimize certain components of angle position error and aerodynamic hysteresis would be an example of this.

3. Repeat runs in an OFAT test are not statistically guaranteed to improve the modeling process. Repeat runs during OFAT testing are usually made to detect hysteresis or to check that the mechanical machinery and sensors associated with the wind tunnel and test article are working “close” to what they did during an earlier run – which may have been an hour earli er, the prior day, or as part of a prior wind tunnel test. Since the errors described in the previous two paragraphs are included in any repeated measurements (particularly if the repeats are taken sequentially or in close proximity to the original points) , the repeated measurements may lead to incorrect conclusions about the resulting models and/or the fidelity of the test machinery.

4. Test times increase geometrically for cases where many factors are involved, making OFAT testing impractical for complex v ehicles with significant factor interactions. Designing OFAT experiments to adequately capture the dependencies on a large number of factors (independent variables) as well as their interaction effects can be extremely difficult. An OFAT test with only 3 t est points (high, low, and mid) to k cover the range of each factor, would require 3 test points to cover all the possible interactions. (“k” is the number of factors or independ ent variables.) As shown in Fig. 1b the LA - 8 airframe had a total of 23 indepen dent variables for the wind tunnel testing done to date ( 8 motor RPM s, 2 wing angles, 4 elevon deflections, 2 ruddervator deflections, 4 flap deflections, angle of attack, angle of sideslip, and dynamic pressure) . At 5 seconds per point, 3 test points wo uld take 14,926 years (running 24 hours per day, 365 days per year). Making good engineering judgements based on symmetry, proximity of particular control effectors, etc. can significantly reduce the number of OFAT test points – but this may still result i n unreas onable test times for a VTOL vehicle like the LA - 8 . Using assumptions about left/right symmetry, and ganging motors to gether one could potentially reduce the number of i ndependent variables to about 12 for LA - 8 – although you would have to double the result to test symmetric and antisymmetric control effectors. The resulting 2*3 test points would still take over 200 days of OFAT testing (assuming 5000 points per day) – while only sweeping through 3 values to cover the range for each individual fa ctor.

C. DOE Testing Pros: 1. The DOE approach provides run - efficient designs and adds statistical rigor to the initial wind tunnel test design process, to the acquisition of data, and in the final data analysis. During the wind tunnel test design process, test points are selected to (1) orthogonalize regressors that improve modeling, (2) reflect the desired levels of model fidelity required, (3) minimize certain criteria such as prediction errors, and (4) provide adequate statistical power to t he test. During e xecution of a wind tunnel test, various modeling statistics, including fit errors and prediction errors, can be monitored.

2. Unknown systematic errors are addressed and mitigated in the DOE approach by randomizing the test matrix. Randomization prevents tim e - varying systematic errors from directly adding into the measurements and instead “transfers” the errors into a component of the measurement variance. In other words, estimated model parameters will not be corrupted, but the estimated parameter uncertaint ies will increase. For OFAT testing, estimated model parameters have a significant risk of being biased due to non - stationary systematic error – whereas those same errors will show up as higher parameter uncertainties in DOE testing.

3. Replication is used in a DOE design to allow assessment of pure error. Pure error and lack of fit are two sources of error that together define the total measurement error. A lack - of - fit test is commonly performed in a regression analysis of variance (ANOVA) table to determine if the estimated model presents a statistically significant lack of fit.

4. The DOE body of theory provides a broad range of approaches/options for many different experimental situations. “Screening” has not been used in the LA - 8 study, but it is a DOE appro ach used to quickly determine important factors when those factors are not obvious to subject matter experts. This is likely to become increa singly important for complex VTOL vehicles like the LA - 8 for which there is not a historical database of knowledge about what control effector interactions are important – or what the transitional concept of operations (CONOPS) should be.

5. The DOE approach’s testing efficiency allows for characterization of responses and interaction effects for test articles with large numbers of independent variables that would be impossible to obtain using the OFAT approach. Simple calculations show that with even a modest number of factors with interactions present can result in OFAT test times quickly approaching years to accomplish. The LA - 8 test had 23 factors.

The GL - 10 wind tunnel test had 29 factors – and there are other distributed electric prop ulsion configurations that ha ve even more ( e.g. some designs include collective and cyclic blade pitch for each propeller blade).

As not ed earlier, engineering judgement can be used to reduce the number of independent variables enough to provide some results from OFAT testing, but the danger is that you may miss some critical interaction – whereas DOE methods can capture all the interactio ns and identify those that are statistically significant.

D. DOE Testing Cons: 1. The wind tunnel facility hardware and control software must be capable of setting the successive random values of tunnel and airframe independent variables required by DOE testi n g – including things like angle of attack, angle of sideslip, control surface deflections , and motor RPM s. For both the GL - 10 and LA - 8 wind tunnel tests in the 12 - Foot Tunnel, the dynamic pressure was held constant for each DOE run – but between 22 and 28 other motor settings and control surface positions were changed between each static test data point. Since OFAT testing has generally been the default methodology used in wind tunnels, many facilities are not currently set up to easily implement DOE testin g.

2. DOE wind tunnel testing generally puts more short term “wear and tear” on the facility and airframe hardware on a point - by - point basis. DOE makes re latively larger changes in angle of attack, angle of sideslip , cont rol surface deflections , and motor RPM values between successive test points – subjecting the related hardware to higher transient stresses compared to the smaller incremental changes between test points typically seen in OFAT testing. There are certain cases, however, where DOE testing can pr oduce less stress on certain components. This could occur if an OFAT test run holds motor RPM s or control surface servos at higher values for longer periods (multiple successive test points) than a corresponding DOE run. If the motors or servos are operati ng near or above their continuous rated capacity, then the OFAT runs may produce more component failures than the DOE runs , where higher values are typically held for shorter periods (less successive test points). Facility model support hardware (stings, t urntables, etc.) making larger and more rapid changes in angle of attack and sideslip angle will almost certainly see more wear and tear from DOE testing – and this should be assessed for particular facilities considering the use of DOE testing.

3. DOE and fl ight dynamics subject matter experts are required to set up and generate DOE wind tunnel test schedules - and to interpret DOE wind tunnel test results after test data has been gathered. The availability of such subject matter experts may impact whether pe rforming a DOE test is even possible.

4. DOE wind tunnel testing at many facilities does not currently allow for physical insight into (and interpretation of) the data as it is collected. Software at many wind tunnel facilities does not allow for plotting of DOE run results (forces, moments, coefficients, etc.) in a form comparable to what can be done for OFAT testing. DOE testing ultimately provides the same information as OFAT testing but requires an extra modeling step to extract equivalent information. If DOE testing becomes used more frequently, then the ability to look at the results after each DOE run in a more understandable way will become more available.

There is a research and development effort called Rapid Aero Modeling (RAM) focused on closing the loop around wind tunnel testing and computational studies so that the results of individual DOE runs are used to guide and set parameters for subsequent runs to achieve desired levels of modeling fidelity – and would also provide the test engineers more i mmediate feedback about the test results. References [ 10 ] and [ 11 ] cover recent work in this area.

E. Combining OFAT and DOE Wind Tunnel Testing Approaches A well thought out wind tunnel test can combine the best features of both the traditional OFAT and DOE testing approaches. The OFAT approach can initially be used to: 1. Get a broad understanding of the vehicle’s response features. For a VTOL vehicle this might include doing enough OFAT testing to answer the following questions: Does the vehicle have adequat e thrust and control authori ty in the different axes? Can the vehicle be trimmed in pitch going from a dynamic pressure of 0 (hover) up to dynamic pressures (speeds) consistent with forward flight? What are the vehicle’s stability and control characteristi cs?

2. Define the ranges for individual factors (independent variables) appropriate for each portion of the flight envelope to be covered by particular DOE runs. For example, vehicle angle of attack, wing angle, and motor RPM ranges can be limited for subseq uent DOE runs at a particular dynamic pressure to a range where the vehicle could maintain trimmed forward flight at a constant airspeed and altitude (quasi steady state transition points) – or for the vehicle accelerating in forward speed and altitude whi le remaining trimmed in pitch (representing a climbing out and accelerating transition between hover and forward flight). How this was done for the LA - 8 testing is described in Sec. V .

IV. Lessons Learned f rom Static GL - 10 DOE Wind Tunnel Test ing in 2013 A. Rec ordin g of individual motor actual RPM s at each test point is necessary . Commercially available RC hobbyist hardware has become sophisticated enough in recent years to allow for frequent use in subscale airframes for flight or wind tunnel testing. These sys tems typically use pulse - width - modulated (PWM) control signals. The actual RPM resulting from a particular PWM command will change based on the loading on the prop/motor at particular test points – and also because the performance characteristics (as deter mined by a motor calibration run) can change over the course of just a few hours of continuous running in the tunnel. (This is particularly true if the motors are running at or near their maximum continuous rating.)

B. Testing at multiple dynamic pressure set tings is necessary to realistically assess transition performance between hover and forward flight . The incr eased importance of propulsion effects and propulsion - airframe interactions for VTOL configurations like the GL - 10 and LA - 8 requires testing the veh icle at multiple dynamic pressures due to the different scaling parameters required for different vehicle components. For example, airframe components scale with freestream dynamic pressure, whereas propeller aerodynamics scale proportional to the local dy namic pressures experienced by individual propeller blades.

C. Trying to understand and model a complicated blown - tilt - wing configuration based only on data for the full configuration is very difficult. Having individual prop eller /motor performance data and b are airframe data can greatly facilitate developing a usable model – al though implementing a traditional “ buildup ” approach to aerodynamic modeling may not be well - suited for certain airframe configurations like the GL - 10 and LA - 8 . This is di scussed more f ully in Sec. IV .

D. Averaging 10 seconds of static data (the historical default for the 12 - Foot Tunnel) is not necessary . Results were just as good whe n data was taken/averaged over 2 seconds for each test point ( Ref. [ 12 ] ) . To be conservative data was taken for 3 seconds per point after a 2 - second settling time.

V. Initial OFAT LA - 8 Wind Tunnel Test ing in the Spring of 2019 A significant number of OFAT test runs previously defined by the LA - 8 team were made for particular parts of the flight regime from hover to forward flight mode . Early test results raised some questions about the longitudinal stability of the airframe and where the airframe center of gravity ( CG ) ideally s hould be located. T he airframe was also significantly heavier than originally designed, which prompted the test engineer to change the overall test plan and pursue a procedure that would answer the following questions:  Can the LA - 8 airframe produce adequate lift and thrust to fly at all dynamic pressures (forward velocities) between hover an d wings - level forward flight?

 Can the LA - 8 airframe be trimmed in the longitudinal axis at all the dynamic pressures between hover and wings - level forward flight?

 Does the LA - 8 airframe have sufficient control authority throughout the h over - to - forward - flig ht transition?

 What wing angles, thrust settings, control surface settings, and concept - of - operations will allow the previous questions to be answered with a “yes”?

Initial hover testing with the tunnel dynamic pressure at zero indicated that a stationary hover could be achieved with both wings set to approximately 82°. Wing angles higher than 82° produced a net force in aft direction due to the blown - wing aerodynamics. Wing angles less than 82° produced a net force in the forward direction due to propelle r thrust in the forward direction exceeding the aft component of the blown wing aerodynamic force .

For a tunnel airstream dynamic pressure of 0.50 pounds per square foot (psf) both wings were swept from 85° down to 0° and the propeller thrust on both wings was adjusted until the normal (vertical) force seen by the strain gage balance matched the predicted vehicle flight weight of about 60 pounds, and the axial force was approximately zero.

These condition were satisfied with both wings at an angle of 56°. T hen the motor commands to the front and aft wing propellers were adjusted incrementally in opposite directions until the pitching moment coefficient (C ) was near zero m for a given CG location . (This front wing/aft wing differential thrust proc edure is high lighted in Fig. 4 later in this section.)

Using the process just described, “t rimmed ” pitch points were determined at dynamic pressures of 0, 0.50, 0.75, 1.00, 1.25, 1.50, 1.75, 2.00, 2.25, 2.50, 2.75, 3.00, 3.25, 3.5, 3.75, and 4.00 psf. At each trim poin t the wing angles (the same for both wings) and motor thrust settings were manually adjusted to determine the values requ ired to have lift = weight, drag = forward thrust, and pitc hing moment = 0 with all other control surfaces set to their nom inal zero po sitions. Table 1 lists the vehicle settings associated with each of these points. Motor RPM or thrust settings were not included because there were changes to the LA - 8 motor and propeller configuration made after the initial OFAT testing. This is discussed in Sec. VII of this paper.

Table 1 Dynamic pressures, wing angles, and control surface settings for “trim” points through transition where normal force = flight weight, axial force = 0, and C = 0.

m Wing 1 Wing 2 All Four All Fo ur Both Q Alpha Angle Angle Flaps Elevons Ruddervators (psf) (D eg) (Deg) (Deg) (Deg) (Deg) (Deg) 0 0 82 82 0 0 0 0.5 0 56 56 0 0 0 0.75 0 51 51 0 0 0 1 0 47 47 0 0 0 1.25 0 43.5 43.5 0 0 0 1.5 0 38 38 0 0 0 1.75 0 35 35 0 0 0 2 0 32 32 0 0 0 2.25 0 30 30 0 0 0 2.5 0 27 2 7 0 0 0 2.75 0 23 23 0 0 0 3 0 21 21 0 0 0 3.25 0 18.5 18.5 0 0 0 3.5 0 16 16 0 0 0 3.75 0 16* 16* 0 0 0 4 0 14 14 0 0 0 *Although wing angles at Q=3.75 psf are the same as at 3.5 psf, the motor thrust settings were different.

The settings summariz ed in Table 1 (plus the motor settings) were combined into a single semi - automatic test run where the dynamic pressure for the tunnel and all other independent parameter settings were changed bet ween successive points. Figure 3 includes plots of normal for ce, axial force, and C as a function of the tunnel dynamic m pressure (Q) . Although the plot shows all negative C values, it was very easy to flip the C positive with very small m m di fferential motor thrust setting changes between the front and aft wings. Th is sensitivity was particularly evident at hover and the low dynamic pressures settings .

The vehicle angle of attack and sideslip were kept at zero for all the point s. Note that the wings were still tilted at 14° for Q = 4 psf in Table 1 , indicating that transition to the forward flight configuration is not quite complete. ( Testing with wing angles at 0° is described later in this section.) Theoretically each point in Table 1 represents a steady - state non - accelerating condition tha t the LA - 8 airframe could fly , showing that longitudinally trimmed flight should be possible from hover up to slow forward flight.

Fig. 3 LA - 8 normal f o rce, axial force, and C for “ trim” points through transition.

m Starting with the previously defined transition longitudinal t rim points , additional sweeps were made to assess the adequacy of pitch, roll, and yaw control authority resulting from elevon, flap, ruddervator, and differenti a l motor thrust settings. These sweeps included angle of attack ranging from - 6° to +6° in 2° i ncrements (t rim points were taken at 0°) . All four elevons were deflected antisymmetrically to roll right and roll left . To test pitch control authority, either t he elevons on the front and aft wing s were actuated in opposite directions or the ruddervators were deflected together up and down . Yaw control authority was tested by moving both ruddervators to the left and then to the right . Flap deflections included a ll flaps on both wings at 10° and 20°, front wing flaps only at 10° and 20°, back wing flaps on ly at 10° and 20°, right hand (RH) flaps (both wings) at 10° and 20°, and left hand ( LH ) flaps (both wings) at 10° and 20°. Combinations of elevon, flap, and ruddervator control surfaces were not checked. Motor differential thrust variations included sweep ing the front wing motors below and then above the pitch trim point thrust values – while doing the opposite on the back wing. All control surface and motor thrust variations were done at sideslip angles of 0° and 5°. Because of the time required to make t hese OFAT control authority assessment sweeps for LA - 8, they were done at a smaller number of dynamic pressure settings than was used to establish the trimmed points defining a transition flight corridor.

Differential wing tilt angles between the forward a nd aft wings was not part of thi s control authority assessment because it had been shown to be relatively ineffective for pitch control in earlier runs. Some of t he reasons for this are discussed in Sec. VI of this paper.

Figure 4 is an example of an OFAT LA - 8 test run where all 8 motors (4 per wing) were initially set at identical thrust settings , and both wings were held at 0°. This produces a nose - down pitching moment. Decreasing the thrust on the front wing and increasing thrust on the aft wing makes th e C even more negative, but increasing the thrust on m the front wing (W1) and decreasing thrust on the aft wing (W2) allows the C to cross zero for all the variations of m alpha and wing flaps that were also included in this run. The dynamic pressure for th e data in Fig. 4 was 6 psf, which is the highest we were able to test the LA - 8 airframe to in the 12 - Foot Tunnel.

Fig. 4 OFAT run showing e ffectiveness of W1/W2 differential thrust for pitch c ontrol .

Near the end of the OFAT testing in the spring of 2 019 another series of runs was made to r oughly define the LA - 8 major control effector settings for an accelerating climb - out transition corridor from hover to forward flight. The difference between this assessment and that described previously is that lift was set to exceed weight and thrust was set to exceed drag to match a condition where the vehicle in flight would be accelerating. (Pitching moment was still set to zero.) This was an initial assessment for a transition that might be more typical for UAM operations – wh ereas the prior steady - state “trim” settings would likely be closer to what is used during initial flight testing and flight envelope expansion. Table 2 includes all th e vehicle settings obtained. Unlike the prior assessment, the settings f or an accelerating climbout include the wings rotating all the way down to 0°.

Table 2 Dynamic pressures, alphas, wing angles and control surface settings for the LA - 8 test run simulating an accelerating climbout.

Wing 1 Wing 2 All Four Both Q Alpha Angle Angle Flap 1 Flap 2 Flap 3 Flap 4 Elevons Ruddervators (psf) (D eg) (Deg) (Deg) (Deg) (Deg) (Deg) (Deg) (Deg) (Deg) 0 0 84.5 84.5 0 0 0 0 0 0 0 0 83.5 83.5 0 0 0 0 0 0 0.5 0 56 56 0 0 0 0 0 0 0.75 0 49.5 49.5 0 0 0 0 0 0 1 0 44 44 0 0 0 0 0 0 1.25 0 38.5 38.5 0 0 0 0 0 0 1.5 0 33.5 33.5 0 0 0 0 0 0 1.75 0 29 29 0 0 0 0 0 0 2 0 25 25 0 0 0 0 0 0 2.25 0 21.5 21.5 0 0 0 0 0 0 2.5 0 15 15 0 0 5 5 0 0 2.75 1 12 12 5 5 10 10 0 0 3 2 9 9 10 10 15 15 0 0 3.25 3 6 6 15 15 20 20 0 0 3.5 4 4 4 20 20 25 25 0 0 3.75 5 2 2 15 15 20 20 0 0 4 6 0 0 20 20 25 25 0 0 For lift to exceed or at least equal weight at the higher dynamic pressure settings , positive angle of attack and partial flap deflection were required. The normal force plot in Fig. 5 shows the lift exceeding weight for a good portion of transition. But in order to get the wing angles down to zero while maint ain ing enough lift eventually requires positive angle of attack and partial flap deployment . This reflects the fact that at its current weight, the nominal LA - 8 cruising forward flight speed is higher than could be tested in the 12 - Foot Tunnel .

The axial force plot shows a gradual buildup in force to about 25 pounds in the forward direction. An initial calculated estimate of the time required to execute this accelerating climbout yie lded 18 seconds to go from hover to 4 psf ( ~40 kts ), which seemed reasonable. Since the motor thrust settings were limited in the wind tunnel by the need to ensure the wind tunnel strain - gage balance did no t get o verloaded, this probably underestimates the performance LA - 8 could achieve in actual flight testing.

The C plot shows small negative values near zero . As was noted in the discussion about Fig. 3, at hover and low m dynamic pressures it only takes a s mall change in the relative motor thrust settings between the front and aft wings to make the C switch signs . There was sufficient pitch control power available to trim C to zero or drive it significantly m m positive or negative at a ny dynamic pressure settings tested . For all wind tunnel testing of the LA - 8 the individual motor and control surf ace values for each test point were initially read from a text file table by the tunnel system software, which also generate d the open loop commands to move the surfaces and drive the motors . (No closed loop controllers were used.)

Fig. 5 Normal force, axial f orce, and C m for LA - 8 run s imulating accelerating c limbout .

A primary reason for attempting to determine indepen dent parameter settings definin g potential transition corridors was to establish ranges to be used in the planned subsequent DOE testing. Unlike the GL - 10 DOE wind tunnel test where the full range for each control effector was tested at a dynamic pressure of 3.2 psf, the LA - 8 control effector ranges for each test dynamic pressure were limited to values that are practical for that portion of flight envelope . For example, the wing angle range for DOE hover testing was 75 to 95 degrees (non - translating hover r equired the wings to be set between 82 and 85 degrees) – and for forward flight the wings were both locked at zero degrees. The ranges for the wing angles and control surface position were set primarily by the level flight pitch trim assessment – with some margin to provide overlap. For the motors, a somewhat higher range was generally used to allow assessment of the impact of losing particular motors. (For the hover DOE runs the range wa s 1900 to 6500 RPM .)

Detailed descriptions and extensive data plots fo r the spring OFAT testing are included in Ref. [4 ] .

VI. Understanding LA - 8’s Pitching Moment Characteristics, Stability, and Control Options Previous wind tunnel and flight testing of the GL - 10 configuration identified some unusual pitching moment characteri stics, with the vehicle undergoing at least one noticeable reversal in the pitching moment direction as it transitioned from hover to forward flight. GL - 10 had a tilt - wing with eight electrically driven propellers and a tilt - tail with two additional propel lers, which gives it a number of similar ities to the LA - 8 configuration (r eference Fig.

2 ) . Although the OFAT testing summarized in Sec. V established that the LA - 8 airframe could be trimmed in pitch throughout a transition from hover to forward flight, so me of the pitching moment characteristics were “unusual” in comparison to conventional airplane s . The following figures and accompanying discussion are an attempt get an initial understanding of the physics and fund amental aerodynamics that might account f or those characteristics.

The locations of aerodynamic centers, wing rotation points, the vehicle overall CG, a nd all distances shown are approximate , but they are clo se enough to help illustrate why certain changes occur through transition flight . This se ction will describe and discuss: 1) how the different forces produce pitching moment increments, 2) how those increments change as the vehicle transitions from hover to forward flight, 3 ) why differential wing tilt is less effective and predictable for con trolling pitch than diffe rential thrust between the front and aft wing propellers, and 4 ) how a nose - down fuselage attitude from hover throug h most of the transition may provide better pitch control and less wing - to - wing interaction than a nose - up fuselage attitude.

Figure 6 shows a simplified version of the aerodynamic forces acting on the LA - 8 airframe in hover with no forward velocity. Propeller thrust includes a forward - facing component that just offsets the aft - facing component of the blown - wing aerod ynamic forces. For the front wing (Wing 1) the propeller thrust acting about the CG produces a nose - up (positive) pitching moment, whereas the blown - wing aerodynamic force acting about the CG for the same wing produces a nose - down (negative) pitching momen t. For the aft wing, (Wing 2) the directions of the pitching moments produced by the propeller thrust and the blown - wing aerodynamic force are again in the opposite directions – with each also being opposite the correspond component on the front wing.

Fi g. 6 LA - 8 f orces relative to CG while in h o ver mode (no forward velocity, w ings at 82 °) .

Tilting an individual wing will therefore produce two incremental changes in pitching moment that will be opposite in direction. Determining the net effect of tilt ing a wing on the pitching moment then becomes a function of which combination of force magnitude and offset distance produces the larger incremental change.

Before any testing was done on LA - 8 there was an assumption that differential wing tilt might be a n effective method for controlling pitch attitude. Test sweeps to assess this showed far less control authority than expected. Figure 6 gives some indication wh y this may be true in hover, and additional figures will show how things get even more complicat ed as LA - 8 transitions toward forward flight.

Figure 7 shows the forces relative to the vehicle CG for LA - 8 roughly midway through transition. For discussion purposes the wing aerodynamic force induced directly by the propeller is shown separately from th e wing aerodynamic force induced by the airstream. Both wings are at approximately 40° and the fuselage is nominally level (vehicle angle of attack = 0°) . For this case the pitching moment increments due to the direct propeller thrust and the propeller in duced aerodynamic forces are now in the same direction – not opposing each other like they were for the hover case. Somewhere between a wing angle of 82° and a wing angle of 40° the line of action for the propeller - induced aerodynamic forces for each wing passes through the CG, reversing the direction of the resulting pitching moment increment.

Fig. 7 . LA - 8 f orces relative to CG while in transition mode (moderate forward velocity, w ings at ~40°) .

Although wing aerodynamic force components due to the a irstream have relatively small moment arms as shown in Fig. 7 , it is clear that these components will also produce a reversing increment of pitching moment as the wings rotate. Since the reversal of pitching moment increments would be expected occur at dif ferent wing angles for the front and aft wings, this will create a complicated changing character to the net pitching moment as the vehicle goes through transition between hover and forward flight.

Figure 8 shows LA - 8 in the forward flight position with wi ngs nominally zero and the fuselage at an angle of attack of 0°.

Fig. 8 LA - 8 f orces relative to CG while in forward f lig ht mode (max forward velocity, w ings at 0°) .

In Fig. 8 the propeller thrust is aligned with the airstream so that the wing aerodyn amic force components will be aligned. Front wing forces all act to produce a nose - up pitching moment, and aft wing forces all act to produce a nose - down pitching. If all 8 motors were run at the same thrust setting, then the expectation would be that the net pitching moment would be nose - down, since the aft wing has a larger area , the distance dPT 2 is slightly larger than dPT 1, and there is little difference between the incidence angles. This expectation was confirmed by the OFAT run previously shown in Fi g. 4. Figure 4 data is from an OFAT run made after the vehicle configuration change described in the following section, so that all motors and propellers were identical and set to the same initial thrust settings. With all motors at the same thrust setting C was negative. To trim the aircraft, the front wing motors thrust m setting had to be increased relative to the aft wing motors.

Figure 9 a shows the LA - 8 in a mid - transition position where both wings are tilted at approximately 40°. The mom ent arms for th e aft wing are dPT2 for the propellers, dPIA 2 for the propeller - induced component of wing aerodynamic forces, and dAIA 2 for the airstream - induced aerodynamic forces. For t he configuration in Fig. 9 a PT2*dPT2, PIA2*dPIA 2, an d AIA2*dAIA 2 all produce a nose - d own pitching moment, although the last two component are small.

(a) Both w ings at ~40° (b) Aft wing r otated CW to ~55 ° Fig. 9 Impact of wing tilt change on LA - 8 forces r elative to CG .

In Fig. 9 b the af t wing has been rotated appr oximately 15° leading edge up r elative to Fig. 9a. T he changes in the aft wing (W2) moment arm lengths will slightly increase the nose down pitching moment due to W2 propell er thrust, reduce the W2 pro peller - induced wing aerodyn amic force contribution to nose pitching moment , and cause the airstream - induced wing aerodynamic forces to produce a nose - up pitching moment.

Changing the wing tilt angle of either wing can produce incremental changes to the pitching moment about the CG i n different directions – depending on which forces are being considered and what the particular values are for forward velocity, propeller thrust, wing tilt angle, and the vehicle angle of attack.

Differential wing tilt was the method originally proposed f or controlling pitching moment for LA - 8. OFAT testing in the spring of 2019 indicated that using differential wing tilt as a method for controlling pitch ing moment was not as effective or as predictable as originally expected. This was identified early on in the test, and alternate approaches for controlling pitch attitude were investigated. Keeping both wings at the same tilt angle and using differential propeller thrust from the f ront and aft wings produced far better control authority and was much more p redictable – since the direction of the pitching moment change remains constant throughout the transition between hover and forward flight.

VTOL configurations that undergo geometric changes like LA - 8 can present additional options for how transitioning be tween hover and forward flight is accomplished. Figure 1 0 a shows LA - 8 with both wings rotated fully down to their forward flight positions, but the fuselage angle of attack is relatively high. This could represent slow forward flight where the higher alpha is required to get adequate overall lift.

(a ) Wings at 0°, v ehicle AO A at +15° (b) Wings at 30 °, v ehicle AOA at - 15° Fig. 1 0 LA - 8 forces relative to CG for different fuselage a ngles .

Figure 1 0 b shows LA - 8 with the wings at the same angle relative to the airstream as the left side, but the fuselage has been angled nose down. If the thrust settings are the same for both configurations, then we would expect both configurations to produce about the same wing lift. The follow ing possibilities may warrant further investigation:  The configuration on the right has longer moment arms from the propeller thrust to the CG, which may provide for better pitch control.

 The configuration on the right will probably have less adverse inter actions between the front and aft wings.

(The configuration on the left looks like shed vortices and unsteady wake effects from the front wing may directly impact the aft wing.

 The previous potential benefits would have to be weighed against the reduce fus elage lift and higher fuselage drag the configuration on the right would produce.

In this section the figures and discussion imply that the LA - 8 overall vehicle CG location is constant. In reality the overall vehicle CG location sweeps through an arc as th e wings rotate from vertical to horizontal. For wings vertical the overall CG is approximately 1.1 inches aft and 1.1 inches higher than when the wings are horizontal in the forward flight configuration. (See Ref. [ 4 ] for a more detailed discussion of the LA - 8 mass property changes associated with the tilting wings.) The overall vehicle CG location change is small relative to the aerodynamic force re orientation and distance changes , so it was not deemed important enough to include in the previous discussion .

However , it should be noted that the amount of CG shift associated with wing tilting could be significantly higher for other tilt - wing vehicles , depending of the wing rotation points in relation to the wings’ ma ss distributions.

Although small, t his CG s hifting is another factor that further complicates the physics of the pitching moment changes as LA - 8 transitions between hover and forward flight.

VII. LA - 8 Configuration Changes Made After the Spring 2019 OFAT Testing Figure 1 1 a is a front view of the LA - 8 configuration that shows the propeller configuration originally installed and tested during the OFAT wind tunnel entry in the spring of 2019. All eight tractor pro pellers wer e located ahead of leading ed ge of the wings to which they are mounted . The six smaller units were 3 - bladed folding propeller s , and t he aft wing (W2) tip propellers we re larger n on - folding propellers mounted on bigger motors than the other six. The original intent was that the aft wing tip motors and propellers would be used for cruise flight, and all the other propellers would be s hut down and folded back. All propellers on the left - hand side of the airframe rotated clockwise ( CW ) as seen from the front, and all propellers on the right - hand side rotated counter - clo ckwise (CCW).

The smaller - diameter CCW propellers were custom - made. All the other propellers were c ommercial off - the - shelf (COTS) items.

(a) S pring of 2019 OFAT t est configuration (b) S ummer of 2019 DOE test to p resent configuration Fig. 1 1 Front view s howing LA - 8 motor and propeller c o nfigurations .

Early testing revealed the following issues with the 11a configuration : 1) the custom - made folding CCW propellers produced noticeably less thrust than the COTS CW folding propellers at identical RPM s, 2) the blade pitch on the larger non - folding aft wing tip propellers was too shallow to produce thrust a t the higher vehicle velocities requi red for cruising forward flig ht , and 3) a wings level forward flig ht configuration with equal thrust on all eight propellers produces a vehicle net nose - down pitching moment, indicating that front wing motors might be better for cruise f light. (Using two front wing motors for cruise will tend to counteract the vehicle ’ s nose - down pitching moment , whereas using two aft wing motors would increase the nose - down moment that would have to be coun tered to maintain longitudinal trim. R eference Sec. VI.)

After considering the issues with the 11a configuration, a decision was made to: 1) replace the aft wing tip motors and propellers with ones identical to those at the other six locations, and 2) split the CW and CCW propellers evenly between the left - hand and right - hand sides of the airframe. This is the motor and propeller configuration shown in Fig. 11b and used for the summer 2019 powered DOE testing , The 11b configuration simplifies the assessment of the subsequent wind tunnel test results, and allow s for any subset of the propeller s to be used for cruise forward flight.

VIII. Powere d DOE Wind Tunnel Tes t ing of LA - 8 in the Summer of 2019 During August and September of 2019 a DOE test of the LA - 8 airframe was completed in the 12 - Foot Low - Speed Wind Tunnel. All eight motors were the same type, and folding propellers of the same diamete r were used at all eight locations. Sets of DOE run blocks were completed at dynamic pressures of 0, 0.5, 1.0, 1.5, 2.5, 3.5, 4.5, and 5.0 psf.

All planned DOE testing was successfully completed with the exception of several runs which had to be conducted with elevons locke d out due to burned out servos. Details of how the data from this test was assessed and used in development of aerodyn amic models is described in Ref. [6] Comments highlighting particular aspects of the powered LA - 8 DOE testing: 1. At each dynamic pressure tested, the ranges for wing angles and all the control surfaces were limited to values appropriate for the portions of the flight envelope corresponding to that dynamic pressure. This was already discussed in Se c. V in relation to the OFAT testing, but it bears repeating to state that for the LA - 8 DOE testing the data was taken only for component settings that are insi de the vehicle's nominal flight envelope. This was not true for DOE testing of GL - 10 in 2013 , wh ich did include taking data for the full range of all variables at all dynamic pressures tested. Limiting the LA - 8 independent variable ranges based on the projected transition envelope obtained from the OFAT “trim” point studies was a major improvement in the wind tunnel testing efficiency in comparison to the GL - 10 testing. Table 3 lists the independent variable setting ranges that were applied t o each of the dynamic pressures included in the LA - 8 DOE testing. Although flaps, elevons, ruddervato rs, and motors are listed together in the table, each individual control effector was treated as an independent test factor. ( Although outside the range of normal flight operations, allowing all individual motor RPMs to drop to idle allowed for the resulting aerodynamic models to include an initial estimate of the impact of losing any motor or combination of motors. ) Table 3 Independent variable ranges for each dynamic pressure for the LA - 8 powered DOE test.

Wing 1 Wing 2 All Four All Four Both Q Alpha Bet a An gle Angle Flaps Elevons Ruddervators All Eight Motors (PSF ) (D eg) (D eg) (Deg) (Deg) (Deg) (Deg) (Deg) (RPM ) 0 - 9.9 to +15 - 5 to +5 75 to 95 75 to 95 0 to 40 - 25 to +25 - 30 to +30 1900 to 6500 * 0 .5 - 9.9 to +15 - 5 to +5 45 to 85 45 to 85 0 to 40 - 25 to +25 - 30 to +30 1900 to 6500 * 1.0 - 9.9 to +15 - 5 to +5 35 to 55 35 to 55 0 to 40 - 25 to +25 - 30 to +30 1900 to 6500 * 1.5 - 8 to +8 - 5 to +5 25 to 45 25 to 45 0 to 20 - 25 to +25 - 30 to +30 1900 to 6500 * 2.5 - 6 to +6 - 5 to +5 0 to 35 0 to 35 0 to 20 - 25 to +25 - 30 to +30 1900 to 6500 * 3. 5 - 6 to +6 - 5 to +5 0 to 25 0 to 25 0 to 20 - 25 to +25 - 30 to +30 1900 to 6500 * - 25 to 4 .5 - 6 to +6 - 5 to +5 0 to 15 0 to 15 0 to 20 +25 ** - 30 to +30 1900 to 6500 * - 25 to 5.0 - 6 t o +6 - 5 to +5 0 0 0 to 20 +25 ** - 30 to +30 1900 to 6500 * *Motor lower limit set to slow idle. Motor upper limit determined by strain gage balance loading limits.

** Burnt - out elevon servos required the elevon deflection angles to be set to zero a t Q = 4.5 and 5.0 psf Figur e 12 shows in gray the LA - 8 Table 3 DOE wing angle range as a function of dynamic pressure. The Table 1 steady - state trim points and the Table 2 accelerating climbout trim points are also plotted. This figure demonstrates how th e OFAT testing results were used to define DOE test ranges for critical independent variables.

Fig. 12 LA - 8 OFAT trim point wing angles and DOE wing angle range limits.

2. The test point settings for independent variables generally cover the entire ran ges for that DOE run block – and not just the endpoints and middle values. Based on prior discussions about DOE with other rese archers , it is evident that there is a misconception of only testing at the end points and middle of the range for each factor. This is not the case. There are many options for experiment design. For the LA - 8 DOE tests , designs were based on concepts being considered for the R apid A ero M odeling ( RAM ) approach described in [10]. The first block is a face - centered des ign (FCD), which explicitly ensures th at endp oints are well - captured. T he second block is a “nested FCD” block that combined with the initial FCD guarantee s at least 5 points for each factor [1 3 ] . Th e third and fourth DOE blocks provide a significant number of additional optimized test points that minimize prediction error . For prior DOE wind tunnel tests at NASA Langley , the entire range between the endpoints have typically been covered well when looking at all the data run blocks associated with a particular dynamic pressure. For example, during the GL - 10 DOE wind tunnel test in 2013 the wing angles actually set/tested included points at about 2 - degree increments for the entire range between 0 and 90 degrees. F igure 1 3 includes 2D plots showing the distribution of test points with a series of five DOE run block s specified from RAM for a 22 - factor experiment [7,8] .

Fig. 1 3 2D plot s howing the d istribution of test p oints w ithin a s eries of five RAM run b lock s .

3. The resolution or “fine detail” of response characteristics and factor interactions is a function of data density – with DOE testing being able to achieve the same resolution as OFAT testing for a given data density. O ne concern or misconception is that DOE tests may not capture fine details at a particular test condition as well as an OFAT test. To make it ea sier to see why this is not the case, let’s first consider a wind tunnel test run that is a single long OFAT sweep through angle of attack in 1° increments. This run could also be done with DOE techniques ( such as randomizing the test point order, but taki ng the same number of test points) - and the DOE results would include information about e rrors that the OFAT run did not. T he mathematical interpolation options available to estimate responses between data points would be equivalent for both methods, and the DOE data points in this case would be just as fine as the OFAT results with regard to resolution. However, in many cases DOE /RSM data is collected to ultimately fit models specifically because it is not practical to collect the data at as high a densit y as typical OFAT testing of all the independent variables would provide . Models generated from a lower density of data points will miss some of the fine detail, regardless of whether the collection te chniques involved OFAT or DOE testing. F or any subset o f factors and a given number of test points the DOE approach should always be able to match or provide better resolution than can be achieved for the same number of test points taken using the OFAT approach.

DOE /RSM simply provides a sequential process of addi ng data points, only as needed, to match the observed model complexity.

IX. Component Testing and Buildup Approach for Aerodynamic Modeling for LA - 8 Past experience has shown that conducting bare airframe test s and combining the results with is olated propeller test results can produce reasonable aerodynamic models for conventional airplanes . Recent work testing quadcopter s at NASA Langley indicates that this approach is also applicable to quadcopters [1 4 ] .

Configurations like LA - 8 present cha llenges for using the component testing and buildup approach. For example, Fig. 1 4 depicts the bare airframe aerodynamic forces acting on LA - 8 in both the forward flight configuration and midway through transition. In the forward flight configuration the f low remains attached, resulting in typical airplane wing aerodynamic forces. As the wings are rotated up to transition from forward flight to hover, at some point each wing will st all, potentially not at the same angle. Figure 1 4 b shows fully developed sta ll on both wings, and it is obvious that vortices being shed by the front wing will impact the flow field for the aft wing.

( a) Wings in forward flight configuration (b) W ings at ~40° ( transition flight configuration ) Fig. 1 4 LA - 8 aerodynamic forces for all motors off and props folded Obtaining physics - based models for VTOL configurations that combine component testing results together is a worthwhile goal, particularly if it allows for configuration changes that produce accurat e predictions even if that particular configuration has not been tested. However, t he discussion in Sec. VI would suggest that using a buildup approach for initially developing an accurate aerodynamic model for LA - 8 may be very difficult.

Isolated propell er testing of the LA - 8 propellers was conducted in the 12 - Foot Tunnel in January of 2020 [ 8 ], and bare airframe DOE testing was completed in February of 2020. Attempts were made to combine these test results with the powered LA - 8 DOE test r esults fro m the summer of 2019 . However, the best approaches were found to be either (1) to use only the powered airframe DOE data for model development, or (2) combine the isolated propeller models [ 9 ] with the powered airframe DOE data for model developm ent , using the methods described in Ref. [ 8] . Bare airframe data was found to be challenging to superimpose with the isolated propeller and powered - airframe data , particularly in the transition flight regime.

Several of this paper’s coauthors are involved with computational flow analysis of the Lift + Cruise VTOL configuration shown in Fig. 1 5 . (See Ref. [ 11 ] for more details.) Figure 1 5 b highlights the rotor wake interactions with the wing and tail during transition to horizontal flight. These interactio ns were found to be airspeed dependent and highly unpredictable, making the buildup approach very difficult to implement for this configuration.

(a) (b) Fig. 1 5 Recent computational flow analysis image for the Lift + Cruise c onfiguration .

S ome of the new VTOL configurations currently being investigated by the aerospace community may allow the traditional component testing and buildup approach for aerodynamic modeling to be used successfully , but based on recent experience the authors of this paper would caution against assuming that the approach will work well for all VTOL configurations.

IX. Potential Issues With LA - 8 Initial Hover Testing Results The LA - 8 2019 OFAT and DOE hover testing results need to be assessed carefully - keeping some of the limitations of the test facility in mind. Figure 1 6 is a cutaway of the NASA Langley 12 - Foot Low - Speed Tunnel showing some of the major featur es of this facility. Reference [1 5 ] provides overview of dynamic test techniques used at NASA Langley Research Center on scale models to obtain a comprehensive flight dynamics characterization of aerospace vehicles – including testing in the 12 - Foot Tunnel.

Fig. 1 6 Facility c utaway for the NASA Langley 12 - Foot Low Speed Tunnel .

T he LA - 8 was mounted horizontal ly in the test section for all testing done to date. Therefore when the hover testing had the wings tilted between 80 ° and 90° the propeller flow was directed down towa rd the floor . Figure 1 7 a shows a properly scaled LA - 8 aft wing in the octa gonal 12 - Foot Tunnel test section as it was for the hover testi ng. Figure 1 7 b shows a depiction of the flow field for a 2 - rotor tilt wing vehicle operating in ground effect.

(a) (b) Fig. 1 7 LA - 8 potential hover t est ing i ssues .

Near the ground , t ilt - wing V TOL configurations with multiple rotors or propellers can exp erience an increment of lift that is positive (i.e. if deflected flow from the ground pushes up on the fuselage), ne gative (i.e. if propeller flow creates a lower pressure area under the airfra me), or neutral. The 2 - rotor configuratio n in Fig. 1 7b is shown experiencing a positive lift increment. Assessing the impact of ground effect on a VTOL vehicle can be very important – particularly for configuration s where wind gusts or control oscillations can cause sudden changes to the effect by shifting how the flow gets deflected. In general , a larger number of lift rotors or propellers results in a reduction in the unsteady nature of the ground effect, which should be favorable for the LA - 8 configurati on.

However, LA - 8 is relatively large compared to the 12 - Foot T unnel test section dimensions ( Fig. 1a) – and the thrust produced by all the motors running at once will change the flow field in the test section. (In the forward flight configuration wit h wings level, running all the motors produces enough thrust to turn the tunnel fan blades when the tunnel drive motor is off.) As shown in Fig. 1 7a , the outboard propellers on LA - 8 produce flow that will be deflected off the angled sides of the lower half of the te st section. This will direct more of the flow back toward the center of the tunnel and limit the spreading out of the flow away from the vehicle. This may increase the positive ground effect and give a higher prediction for total lift than would be seen in a less confined test area.

Another potent ial consequence is that changes to the angle of attac k or sideslip angle may produce unexpected interactions between components widely separated on the vehicle . For example, adding sideslip to the LA - 8 configuration depicted in Fig. 17a will change the flow field asymmetrically . R elative to the vehicle cen terline, the angle d lower section s of the tunnel wall s will deflect the flow from the propellers on one side of the vehicle slightly toward the front, and fr om the other side of the vehicle slightly toward the back. Instead of meeting at the same point in the middle of the tunne l and getting deflected upward there , the flow s may pass by each other , hit the angle d lower wall section on the opposite side of the tunnel, and then deflec t up to inter act with components on the opposite side of the vehicle from the propeller producing that flow. This potential complicates assessments of whether particular regressor terms should be incl uded in the aerodynamic modeling. In still air the flow field under and around the hovering vehicle d epicted in Fig. 17b wi ll be the same regardless of the sideslip angle , whereas in the octagonal test section of the 12 - Foot Tun nel changes in sideslip angle can produce real measureable effects.

The significance of the wall, floor, and ceiling effects on VTOL configuration testing in the 12 - Foot Low - Speed Tunnel is an area for further investigation. As an initial step in this investigation, planned additional test ing of LA - 8 includes mounting the airframe both vertically and horizontally in the test section to compare the results obtained for hover testing. (If the LA - 8 airframe is mounted vertically, the flow from the propellers will be directed along the normal f low path for the tunnel , thereby eliminating the impact of flow being deflected of f the walls, floor or ceiling .

Theoretically this should allow for hover testing that produces results more consistent with what would be se en at altitude (outside of ground effect).

X. Compari son of LA - 8 OFAT and DOE Test Results , Planned Additional Testing Although we cannot currently compare powered LA - 8 DOE and OFAT results because of the motor and propeller configuration changes between the initial two wind tunnel tests, we can make some compa ri sons for the unpowered data. Figure 1 8 shows m easured OFAT bare airframe data for 𝐶 and 𝐶 at Q = 4.5 psf with zero wing angle and zero 𝐿 𝐷 control surface deflections as a function of 𝛼 depi cted using blue circles. The dashed red lines represent an RSM aerodyn amic model developed from the February 2020 DOE bare airframe test data covering the same airframe configuration and tunnel test conditions.

Fig. 1 8 Comparison of OFAT data and the a ero model for the LA - 8 bare airframe DOE t est .

Even though the previous comparison is encouraging, the original intent was to include much more detailed and quantitative comparisons for powered OFAT and DOE tes ting of the LA - 8 configuration. To date the C OVID - 19 pandemic has delayed obtaining the additional test data required.

Planned additional near - term LA - 8 wind tunnel testing includes the following: 1) DOE h over testing with the LA - 8 airframe mounted both vertically and horizontally in the test section t o help resolve the questions regarding how the test section wall, floor, and ceiling flow interactions affect the results.

2) OFAT hover testing with the LA - 8 airframe mounted both vertically and horizo ntally in the test section.

P erforming these OFAT ru ns while the airframe is mounted for the DOE tests will minimize the time required to get this additional comparison data.

3) Repeat runs of the higher dynamic pressure forward flight mode DOE runs from the summer of 2019 test where some of th e elevons were inoperable. (Several of the original elevon servos burned out near the end of the DOE test , and the elevons had to be locked in their nominal neutral positions for some of the final DOE run blocks . All the elevon servos have since been replaced with h igher capacity units that should be able to handle the higher loading.)

4) As with the hover testing, a limited set of forward flight mode OFAT test runs at the same dynamic pressures will also be made to allow detailed comparisons with the DOE run results.

5) D epending on test times and priorities, consideration will be given to possibly including bare airframe test runs along with some of the previous cases. S witching to the “bare airframe” configuration would consist of folding back the propellers and leaving the motors unpowered. (No removal or replacement of hardware will be required.)

6) Depending on test times and priorities, consideration will be give n to possibly including some test runs exploring the potential benefits of keeping the fuselage angled nose do wn through most of the transition from hover to forward flight. (Referen ce the Sec. VI discussion about Fig. 1 0 . ) Longer - term wind tunnel testing plans for the LA - 8 airframe and other VTOL configurations include testing on a 3 - degree - of - freedom (3 - DOF) rig that will allow pitch, roll, and yaw motion. This method will provide a means to ob tain dynamic derivatives and to test control approaches by “flying” the vehicle in the wind tunnel.

XI. Conclud ing Summary Remarks New VTOL vehicle designs with complex aerodynamic characteristics and large numbers of independent variables present significant challenges for wind tunnel testing. Building from previous testing experience with the GL - 10 aircraft , several new test techniques were formulated and tested for the LA - 8 aircraft in multiple wind tunnel tests.

The findings and overall conclusions from this effort are summarized as follows: A. S tepping through dynamic pressure settings in small increments whil e adjusting propeller thrust and wing - tilt angles to achieve a “trimmed” condition where lift equals flight weight, net axial force is zer o, and pit c hing moment is zero was a simple wind tunnel test technique to : 1) establish that the LA - 8 airframe tested has sufficient thrust and control authority to maintain a constant trimmed attitude at any speed between hover (with wings near vertical) and forward flight mode (with wings nominally horizontal), and 2) define airframe configuration and control settings a ppropriate f or LA - 8’s transition corridor between hover and forward flight. This technique may prove useful for te sting other VTOL configurations in the future.

B. Sequentially sweeping each set of control effectors (control surfaces and motors) through a ran ge of values on either side of the “trim” setting values deter mined as described in Sec. V can provide an initial assessment of the vehicle’ s control author ity for all three axes.

C. The LA - 8 airframe tested also has sufficient thrust and control authority to perform a trimmed accelerating climbout when going from hover to forward flight. The same general procedure previously used for obtaining the steady - state longitudinal trim points was followed, except lift was required to exceed the predicted flight weigh t, and the axial force in the forward direction was required to exc eed drag. (Reference Sec. V .)

D. The LA - 8 OFAT testing results from the trim and control authority assessments were used to set the independent variable ranges for the subsequent DOE testing. This ensured that the DOE testing fully covered the likely flight envelope, without spending test time evaluating test conditions that were w e ll outside the practical flight envelope of the vehicle. (Reference Sec. VIII .)

E. Tilting the LA - 8 wings changes the moment arm distances and the resulting pitching moments generated by propeller thrust and wing aerodynamic forces. At some wing - tilt angle between the hover and forward flight configuration s the direction of the pitching moment increment about the CG from the wing aerodynamic forces will reverse. Pitching moment increments generated by propeller thrust acting about the CG will change magnitude, but will always be in the same direction as the wings tilt between the hover and forward flight configurations. ( Reference Sec. VI .)

F. Initial LA - 8 OFAT testing showed that using differential tilt between the two wings was not as effective for controlling pitch attitude as was originally expected. At hover and through a significant portion of the transition part of the flight envelope , changes in wing angle for either wing will tend to produce pitching moment increments in one direction for the prop eller thrust, and in the opposit e direction for the wing aerodynamic forces.

G. O FAT testing showed that keeping both the LA - 8 wings at the same tilt angle and using differ ential thrust between the front and aft wing propeller s is a much more effective way to control pitch attitude than by using differential wing tilt.

H. Isolat ed propeller testing done in January of 2020 1) confirm ed significant differences between the performance of the COTS and custom - made propellers, 2) provided d ata for a key component for possible future buildup aerodynamic modeling of LA - 8 , and 3) informed the overall aerodyn a mic model development work. (See R ef. [ 7 ] . ) I. Bare airframe unpo wered DOE testing of LA - 8 in February of 2020 provided data for possible buildup aerodynamic modeling, provided configuration - consistent comparison data with OFAT testing done in the spring of 2019, and informed the overall LA - 8 aerodyn amic model deve lopment work. However, bare airframe data is challenging to superimpose with isolated propeller and powered - airframe data for aerodynamic model development using a buildup approach , particularly in the transition flight regime.

J. The complicated and changing nature of the a erodynamics and stability characteristics of LA - 8 as the wings tilt through transition flight present significant challenges for using the component testing and buildup approach for aerodynami c modeling. Present ly the use of the powered - airframe DOE test data alone or in combination with isolated propeller data produces the best aerodynamic model [ 8 ] .

K. Because of 1) the large number of independent variable s, 2) multiple possible control allocation strategies for c ontrolling the vehicle or transitioning between hover and forward flight , and 3) the lack of a historical database for complicated VTOL designs like LA - 8, DOE testi ng is currently the only known practical way to experimentally obtain all the aerodynamic in teraction data without making major simplifying assumptions that could miss critical issues affecting safety, or conversely, opportunities to operate in ways not originally anticipated.

L. Although DOE theory and the LA - 8 bare airframe testing provide some ba sis for comparing results, the replacement of the aft wingtip motors and propellers with different performance characteristics, and the rearrangement of the other CW and CCW propellers means no major comparisons can be made between the spring 2019 OFAT tes ting results and the summer DOE testing results. (Additional OFAT/DOE comparison wind tunnel testing is planned for the next LA - 8 entry in the 12 - Foot Tunnel. ) M. Because of wind tunnel test section wall, floor, and ceiling interactions, LA - 8 hover testing re sults may not accurately reflect flight conditions for hover in ground effect or at altitude. (Reference Sec. X .) Futu r e planned LA - 8 hover testing in 12 - Foot will include mounting the airframe vertically (so the propeller flow is aligned with the normal a irflow throu gh the test section) to help resolve some of these questions.

Acknowledgments This research was funded by the NASA Aeronautics Research Mission Directorate (ARMD) Transformational Tools and Technologies (TTT) project. Testing support was provid ed by Earl Harris, Clinton Duncan, and Wes O’Neal. LA - 8 vehicle support was provided by David No rth, Gregory Howland, Steven Geu ther, and Robert McSwain. Photography support was provided by Lee Pollard.

References [1 ] Busan, R. C., Rothhaar, P. M., Croom, M. A., Murphy, P. C., Grafton, S. B., and O’Neal, A. W., “Enabling Advanced Wind - Tunnel Research Methods Using the NASA Langley 12 - Foot Low Speed Tunnel,” 14th AIAA Aviation Technology, Integration, and Operations Conference , AIAA Paper 2014 - 3000 , 2014.

[2] Murphy, P. C., and Landman, D., “Experiment Design for Complex VTOL Aircraft with Distributed Propulsion and Tilt Wing,” AIAA Atmospheric Flight Mechanics Conference , AIAA Paper 2015 - 0017 , 2015. https ://doi.org/10.2514/6.2015 - 0017 .

[3] Rothhaar, P. M., Murphy, P. C., Bacon, B. J., Gregory, I. M., Grauer, J. A., Busan, R. C., and Croom, M. A., “NASA Langle y Distributed Propulsion VTOL Tilt - Wing Aircraft Testing, Modeling, Simulation, Control, and Fli ght Test Development,” 14th AIAA Aviation Technology, Integration, and Operations Conference, AIAA Paper 2014 - 2999, 2014.

https://doi.org/10.2514/6.2014 - 2999 .

[4 ] North, D. D., Howland, G., and Busan, R . C., “Design and Fabrication of the LA - 8 Distributed Electric Propulsion VTOL Testbed,” AIAA SciTech 2021 Forum , 2021. To be published .

[ 5 ] McSwain, R. G., Geuther, S. C., Howland, G., Patterson, M. D., Whiteside, S. K., and North, D. D., “An Experimenta l Approach to a Rapid Propulsion and Aeronautics Concepts Testbed,” NASA TM – 2020 - 220437, 2020.

[ 6 ] Geuther, S. C., North, D. D., and Busan, R. C., “Investigation of a Tandem Tilt - wing VTOL Aircraft in the NASA Langley 12 - Foot Low - Speed Tunnel,” NASA TM – 20 20 – 5003178, 2020.

[ 7 ] Simmons, B. M., and Hatke, D. B., “Investigation of High Incidence Angle Propeller Aerodynamics for Subscale VTOL Aircraft ,” NASA TM, 2021. To be published.

[ 8 ] Simmons, B. M., and Murphy, P. C. “Wind Tunnel - Based Aerodynamic Model Identification for a Tilt - Wing, Distributed Electric Propulsion Aircraft,” AIAA SciTech 2021 Forum , 2021 To be published .

[ 9 ] Simmons, B. M., “System Identification for Propellers at High Incidence Angles,” AIAA SciTech 2021 Forum , 2021. To be published.

[ 10 ] Murphy, P. C., Simmons, B. M., Hatke, D. B., and Busan, R. C., “Rapid Aero Modeling for Urban Air Mobility Aircraft in Wind - Tunnel Tests,” AIAA SciTech 2021 Forum , 2021. To be published.

[ 11 ] Murphy, Patrick C., Buning, Pieter G., Simmons, Benjamin M., “Rapid Aero Modeling for Urban Air Mobility Aircraft in Computational Tests,” AIAA SciTech 2021 Forum , 2021 . To be published.

[ 12 ] Murphy, Patrick C. and Landman, Drew, “Experiment Design for Complex VTOL Aircraft with Distributed Propulsion and Tilt Wing,” AIAA Atmospheric Flight Mechanics Conference , AIAA Paper No. 2015 - 0017. , 2015.

https://doi.org/10.2514/6.2015 - 0017 .

[13] Landman, D., Simpson, J., Mariani, R., Ortiz, F., and Britcher, C., “Hybri d Design for Aircraft Wind - Tunnel Testing Using Response Surface Methodologies,” Journal of Aircraft , Vol. 44, No. 4, 2007, pp. 1214 – 1221.

https://doi.org/10.2514/1.25914 .

[14] Foster, J. V., and Hartman, D. , “High - Fidelity Multi - Rotor Unmanned Aircraft System (UAS) Simulation Development for Trajectory Prediction Under Off - Nominal Flight Dynamics,” 17th AIAA Aviation Technology, Integration, and Operations Conference , AIAA Paper 2017 - 3271, 2017. https://doi.org/10.2514/6.2017 - 3271 .

[1 5 ] Owens, D. Bruce, Brandon, Jay M., Croom, Mark A., Fremaux, C. Michael, Heim, Eugene H., and Vicroy, Dan D. “Overview of Dynamic Test Techniques for Flight Dynamics Research at NASA LaRC (Invited),” 25th AIAA Aerodynamic Measurement Technology & Ground Testing Conference , AIAA P aper 2006 - 3146, 2006 . https://doi.org/10.2514/6.2006 - 3146 .

[1 6 ] Montgomery, D. C., Design And Ana lysis of Experiments, 8th ed., John Wiley & Sons, Inc., Hoboken, NJ, 2013.