Overview of Experiment ? Flight Conditions – M = 0.56, 25000 feet (Subsonic condition) – M = 0.86, 36000 feet (Transonic condition) ? Nose-To-Tail (N2T) Distances – 20, 55, 110 and 190 feet ? Nomenclature – X-direction (longitudinal) – Y-direction (lateral) – Z-direction (vertical) The subsonic flight condition (M=0.56, h=25000 feet) was selected to match pre-existing data from vortex-effect prediction codes. These codes needed to be validated to determine their utility on future applications of AFF. Since a possible future application of AFF is for transport airplane, flight data were also acquired at M=0.86 and h =36000 feet. This transonic flight condition is representative for that class of vehicle. The vortex effects were also mapped at different longitudinal distances behind the leader airplane. These Nose-To-Tail (N2T) distances were monitored by the control room and maintained by the pilots through periodic radio calls. Only the results from the subsonic condition, 55’ N2T will be presented here. 55’ is equal to the length of the F/A-18. The reference axis system was as shown above. It should be noted that although Z is positive down, this presentation will refer to positions above the lead airplane (or high) as positive and positions below the lead airplane (or low) as negative. Page 1 Autonomous Formation Flight Program NAS4-00041 TO-104 Page 2 Autonomous Formation Flight Program NAS4-00041 TO-104 Lift and Drag Analysis Flight Test Database Engine Data Air Data INS Data In-Flight Thrust Model F G , F RAM , FE DRAG Wind Axis Accelerations A XW , A YW , A ZW Air Data Computations Gross Weight, V inf , P o D est. =T Trail. - J Lead Performance Model D = cos(D est ) F G –F RAM –FE DRAG -F EX C L , C D , C Di = C D –C D0 Vortex Effect = Vortex – Baseline %'C D , %'C Di , %'WFT Predicted Performance (outside vortex influence) C L , C D , C D0 F EX =GW*A XW Performance data was determined using classical techniques. A force balance along the flight path was used to determine drag while a force balance perpendicular to that was used to determine lift: D = cos(Dest) FG – FRAM – FEDRAG-(GW*AXW); L = sin(Dest) FG + (GW*NXW). Three primary data reduction areas feed the performance mode; 1) Air Data, 2) IFT, and 3) Accelerations. The Air Data model computes gross weight (GW) using empty weight and the remaining total fuel accounting for crew weight. It also includes a calculation of an estimated alpha, Dest, which is based on the trailing aircraft’s pitch angle and the lead aircraft’s flight path angle (Dest = qtrail-Jlead). This was required because the trailing aircraft’s alpha probes are unusable during formation flight due to localized upwash influences of the lead aircraft. Because the lead aircraft flew at steady-state conditions (constant speed and altitude), the flight path angle, Jlead, was always close to zero. The engine manufacturer’s IFT model was used to calculate thrust on the F404-GE-400 engines installed in the trailing F/A-18 Aircraft. The next chart describes the measurements used to run this model. The model calculated gross thrust (FG), ram drag (FRAM) and engine throttle dependent drag, (FEDRAG). Gross thrust is the primary force the engine produces out the tail pipe, FRAM represents the force loss due to the momentum of air, W1, entering the inlet, and FEDRAG accounts for the external drag forces associated with the engine nozzle and inlet spillage flow. The INS was used to obtain vehicle acceleration data. This data was corrected for rotation effects due to not being mounted exactly on the center of gravity. It then was translated into the flight path (wind axis) coordinate system. Axial acceleration was used to compute vehicle excess thrust: FEX = GW*AXW The performance model used the information from the three paths described above to obtain lift, drag and respective coefficients. To obtain drag reduction values, data obtained during formation flight (vortex) was compared to baseline (non-vortex) points completed in a back-to-back fashion. Some formation flight test points did not include a slide-out maneuver to obtain baseline conditions. For these few points, baseline data were estimated based on data trends in drag related to gross weight. A simple prediction model was used to calculate baseline lift and drag values to evaluate the reasonableness of the baseline data. Test Point Procedure ? Pilot Procedure – Acquire and hold position within the influence of the vortex for 30 seconds of stable data – Engage auto-throttle velocity-hold and maintain position for 20 seconds of stable data – Laterally slide out of position (away from lead a/c), engage altitude-hold and stabilize outside of vortex for 20 seconds ? Technique provides direct comparison of performance data in and out of vortex ? Use of auto-pilot and auto-throttle significantly improved maneuver and data quality Each test point was conducted in the same way. Once both aircraft were on condition, the trail aircraft maintained its position behind the lead aircraft for 30 sec. During this time, the pilot of the trail aircraft was controlling every aspect of his aircraft, including throttles. Because of the transient nature of the vortex effects, especially with significant wing overlap, the pilot’s throttle movements were, in some cases, coarse and over-corrective. This problem was exacerbated when combined with a significant longitudinal distance like 190’ N2T, because maintaining longitudinal separation became especially difficult when the pilots did not have a good visual (close) reference. After 30 sec of stable data, the pilot engaged the auto-throttle (ATC) velocity hold and held position for another 20 sec. More often than not, the ATC would have to be set a few times before the N2T closure rate was small enough to call stable. After 20 sec of stable, ATC-engaged data, the control room gave the call for ‘slide out’, at which time the pilot of the trail aircraft maneuvered laterally out of position to the right, engaged altitude-hold, and stabilized for another 20 sec outside of the vortex. The control room then gave a ‘test point complete’ call at the appropriate time. Following a video of an example test point, an explanation as to why the test point procedure was set up in this way will be given. Page 3 Autonomous Formation Flight Program NAS4-00041 TO-104 Flight Test Point Matrix 4' 6' 8' 10' 12' -50% -25% 0% 25% 50% (% Wingtip Separation) Real-time feedback in cockpit using ILS needles N2T (X) position monitored through control room calls (no direct feedback in cockpit) 55 ft N2T Page 4 Autonomous Formation Flight Program NAS4-00041 TO-104 -75% -50% -25% 0% 25% 50% Lateral Separation, Y, % wingspan 50% 25% 0% -25% -50% Vertical Se paration, Z, % w i ngspan Condition 2 Mach 0.86 36,000 ft Condition 1 Mach 0.56 25,000 ft To fully map the vortex, a grid of test points, or test matrix, was created. Several factors constrained this matrix, including: Limited test flights available The guidance (needle) display was limited to 60 target files. To maximize the resolution of the vortex mapping in the most efficient manner, the matrix was based on 1/8 of an F/A-18 wingspan, or just under 5 feet. To fully map the vortex, a grid of test points, or test matrix, was created. Because flight test time was limited, the number of matrix points had to be kept to a minimum without sacrificing the resolution of the vortex mapping. In addition, the guidance (needle) display used by the trailing pilot to fly each test point was limited to 60 target files. Designed within these boundaries, the test matrix was based on 1/8 of an F/A-18 wingspan (bF/A-18=37.5 ft), or about 4.7 ft. A grid of equally-spaced points in the Y- and Z- axis was then set up using this parameter. An example of such a grid is shown above. Autonomous Formation Flight Test Results Summary Of Phase 0 -February 2001 Summary Of Phase 1 -August 2001 Phase 0 Control Experiment #1 Steady-State Tracking AFF Flight 715 - February 21, 2001 2-Minute Tracking Task High Performance Gainset 20 15 Relative Lateral Position Error (ft) Rela tive V e rtical P o sition Error (f t) 0 10 5 -20 -15 -10 -5 -20 20151050-5-10-15 Extremely Accurate Position Outside The Vortex Demonstrated (Winter 2001). (Experiment: Dial In –75/-75 ft Translations) Page 6 Autonomous Formation Flight Program NAS4-00041 TO-104 20 15 Relative Lateral Position Error (ft) Rela tive V e rtical P o sition Error (f t) 0 10 5 -20 -15 -10 -5 -20 20151050-5-10-15 AFF Flight 714 - February 21, 2001 2-Minute Tracking Task Integral Gainset The additional feedback of the integral of the position error in the INTEGRAL gainset was very successful at eliminating any steady-state offsets in position error. An undesired side-effect of the integrator was larger overshoots for this gainset than for the others. However, performance and stability were still well within the acceptable region for these gains. The HIGH PERFORMANCE gainset exhibited extremely good disturbance rejection capability. Position errors during steady-state tracking with these gains were approximately 1 foot both laterally and vertically. Drag Change Contour Plot Contour plots: ? Provides a true perspective of the vortex's influence on vehicle performance Factors: ? Number of test points ? Data smoothing – bicubic spline ? Extrapolation – missing data points ' C D , percent M=0.56, 25,000 ft altitude, 55ft nose-to-tail This particular contour plot (Mach 0.56, 25,000ft) contains 92 test points. Page 7 Autonomous Formation Flight Program NAS4-00041 TO-104 Actual Flight Test Results Validate “Drag Bucket” Theory Phase 1 Control Requirement Act Phase 1 Control Requirement Phase 0 Control Results Drag Reduction % Drag Reduction % Lateral Offset ('Y feet) V e r t i c a l O f f s e t ( V e r t i c a l O f f s e t ( '' Z f e e t ) Z f e e t ) +'Y -'Z X f Z f Y f -'X Page 8 Autonomous Formation Flight Program NAS4-00041 TO-104 Page 9 Autonomous Formation Flight Program NAS4-00041 TO-104 -25% -20% -15% -10% -5% 0% 5% 10% 15% -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 Wing-Tip Separation (Y Relative Position), % Wing-Span (C D-FF - C D-BL )/C D-BL Percent change in drag versus position at M=0.56, 25,000ft, 55’ N2T Vertical Separation, % Wing-Span Z=0.37 Z=0.25 Z=0.13 Z=0.06 Z=0.0 Z=-0.06 Z=-0.13 Z=-0.25 Z=-0.37 'CD +'Y -'Z X f Z f Y f -'X Contour Plot of Multiple Data Points The change in lift and drag coefficient were evaluated for each maneuver by comparing drag while in the vortex to baseline values. In general very small changes (variations of less than 2%) in calculated lift coefficient were found for all conditions as predicted. For the 55 feet N2T condition shown,up to 19% drag reduction was calculated with peak values between level and -13% vertical position and a lateral position of 10-20% wingtip overlap. Overall the data indicates a large region of significant gains. The data is not symmetric about the the peek position and shows increased sensitivity as the trailing aircraft moved inboard of the peak position as opposed to outboard of this position. In fact, drag increases were measured at some high wing overlap positions, verifying the importance of proper station-keeping to obtain the best results. Data quality varied for each test point with the outboard data tending to have better quality than the inboard data, primarily due to the pilots ability to maintain stable conditions. Some inboard test points were very difficult to fly due to the lead aircraft’s vortex impacting the tail or fuselage. Fortunately the region of best drag benefits was fairly stable and good data quality was obtained on most points. Atmospheric conditions also affected the data on some test points due to turbulence and vertical winds. The back-to-back comparison of vortex and baseline data helped to minimize these effects. Total Drag 'C D , percent 'C D , percent Condition 2 Mach 0.86 36,000 ft Condition 1 Mach 0.56 25,000 ft Page 10 Autonomous Formation Flight Program NAS4-00041 TO-104 55 ft N2T 110 ft N2T 190 ft N2T Compare / Contrast ? Max drag reduction 30-40% greater at Condition 1 than Condition 2 ? Optimal location is relatively constant at all flight conditions and longitudinal positions – Optimum Position ? -0.20 < Y < -0.10 ? -0.10 < Z < 0.0 ? Drag increases calculated with high wing overlap ? 190 ft: Broadly distributed cross-section, poorly-defined peak ? Demonstrated that only a relatively low-precision for Y-Z positioning is required to stay within beneficial region ? More complete picture can be developed with more exhaustive study Page 11 Autonomous Formation Flight Program NAS4-00041 TO-104 Condition 2 Mach 0.86 36,000 ft Condition 1 Mach 0.56 25,000 ft Discussion of Induced Drag ? Induced drag, C Di = C D -C Do –C Do obtained from wind tunnel data ? Theory predicts C Di will not change with flight condition – Simplest form, C Di = C L 2 / (SeAR) – AFF data indicates slightly greater improvement at higher / faster flight condition with larger regions of benefit – Difference between flight conditions is small percentage of overall reduction -- May be a function of data quality, technique, or noise? ? Optimum position – Similar to C D results ? Hoerner predicted C Di reduction of 40% – AFF results show improvement between 40 - 50% at both flight conditions and all longitudinal positions Page 12 Autonomous Formation Flight Program NAS4-00041 TO-104 Induced Drag Effects More Significant at Higher Mach But Remarkably Consistent as N2T Length Increases Condition 2 Mach 0.86 36,000 ft 'C D , percent Condition 1 Mach 0.56 25,000 ft 'C D , percent Page 13 Autonomous Formation Flight Program NAS4-00041 TO-104 55 ft N2T 110 ft N2T 190 ft N2T Discussion of Fuel Flow ? Results sensitive to time slice chosen – Throttle movement (lessened using Auto-Throttle) ? Complete throttle cycles used – Atmospheric upsets ? Changes in the lead airplane fuel flow data during a test point were used to adjust trailing airplane data ? Turbulent day - demonstrated improved C D in the vortex, but increased fuel flow ? No data gathered at 20’ nose-to-tail separation – Auto-throttle technique deemed too risky at this position ? Fuel flow data tracks drag data well – Comparison of drag/fuel flow results adds confidence to overall data set ? Max measured reduction, over 18% from baseline Page 14 Autonomous Formation Flight Program NAS4-00041 TO-104 Effects of Longitudinal Separation ? Several longitudinal stations flown consecutively at the perceived “optimal” location – Y = -0.19 b – Z = -0.06 b ? Best benefits found at approximately X = 4.2 b, or 100 feet nose-to- tail – Close to predictions using horseshoe vortex model* – Afterwards, benefits slowly diminish, but are still significant * Ray, R., et. al, Flight Test Techniques Used to Evaluate Performance Benefits During Formation Flight AIAA-2002-4492 Page 15 Autonomous Formation Flight Program NAS4-00041 TO-104 Page 16 Autonomous Formation Flight Program NAS4-00041 TO-104 *All data obtained at Y = -0.19b, Z = -0.06b Performance as a Function of Longitudinal Position* Data was gathered to evaluate the effect of longitudinal spacing on drag and fuel-flow benefits. The “optimum” lateral and vertical position (Y=-19%, Z=-6%) was flow at a series of trailing positions to obtain this data. Additional data from the matrix of data points flow previously was also used to help fill in the chart. The results show the best benefits were found at a distance of 2.7 wingspans aft (270%). This is close to the prediction shown by Blake, W., and Dieter Multhopp (AIAA-98-4343, August 1998), which uses the horseshoe vortex model for a generic wing shape. Further aft the benefits begin to diminish, but are still significant. It was also found that the vortex position was not as stable further aft and seemed to wander around somewhat, based on pilot comments. The scatter in the data seems to confirm this. The open symbols represents a test point of very poor quality. The reduction in induced drag, forward of the maximum value, is predicted by the simple model. These results show a large region of benefit as a function of downstream separation. This is good news for designers developing autonomous controllers where the engine must control longitudinal spacing. Rapid throttle movements to hold precise position could use excessive amounts of fuel. A low gain controller on the engine throttle should improve its ability to minimize fuel usage. Chase (Control) Single-seat F/A-18A Lead Two-seat TF/A-18A Trail Single-seat F/A-18A Last Experiment Performed On Program X-Country Cruise Mission Formation Experiment Summarize everything, then ask if there are any questions... Page 17 Autonomous Formation Flight Program NAS4-00041 TO-104 Page 18 Autonomous Formation Flight Program NAS4-00041 TO-104 0 100 200 300 400 500 600 700 800 900 1000 9:00:00 9:15:00 9:30:00 9:45:00 10:00:00 10:15:00 10:30:00 10:45:00 Time Fuel Difference, lbm Formation Climb Estimated Duration Speed Brake Was Out Formation Cruise Formation Descent 14 % Fuel Savings TM Data Out of Range Lead - Trail Chase- Trail Mission Profile 40K/0.85 41K/0.84 Formation Flight 15K L o s t ~ 1 0 0 l b s 14% Fuel Savings Cruise Mission Demonstration ? Summary of cruise demonstration data – Simulated mission profile with independent chase of similar configuration – Estimated 110 N mi of range improvement if formation cruise continued A simulated F/A-18 cruise mission was performed to demonstrate the potential benefits of flying in the optimum formation position during extended periods. An independent chase aircraft was also deployed during this flight to obtain fuel burn data of a F/A-18 of similar configuration and weight. Both the trailing and chase aircraft were single-seaters while the lead aircraft was a two-seater. The independent chase had no data acquisition system installed. Fuel tank readings for all aircraft were recorded periodically during the mission as plotted above with the telemetry data from the two formation aircraft. The results show significant fuel savings were recorded for the trail aircraft despite problems with the mission. Telemetry data was lost during the mission (as expected) because the aircraft flew out of telemetry range to accomplish the mission. When TM was reestablished it was discovered the speed brake was partially deployed on the trailing aircraft. That was corrected for the remainder of the mission, however it is estimated approximately 100 lbs of fuel savings were not realized due to this problem. Also, the pilot in formation began to realize the guidance needles were not reading accurately as the aircraft flew further away from Edwards AFB. He continued the mission by flying by the “seat of his pants”, using the experience he had gained during the flight test program. Even with these problems, a 640lbm(14%) fuel savings was realized over the chase aircraft and over 700lbm of saving was measured compared to the lead aircraft over the duration of the formation. Independent checks of the fuel required to fill-up each aircraft verified these readings to within 50lbs. AFF Roadmap Page 19 Autonomous Formation Flight Program NAS4-00041 TO-104 9Complete Fall ’01, Summer ‘02 Summer ’03 The objectives of the initial phase (Phase 0) of AFF research were to reduce the programmatic risks for achieving automated drag reduction. Four potential risk areas were identified, the primary of which was guidance, navigation and control. Information was sought regarding the feasibility of using the Global Positioning System (GPS) for formation navigation and the achievable control precision of an outer-loop autopilot. In addition, flight data was collected to be used to validate and improve the project’s design tools. Other areas of concern were aerodynamics, flight systems and flight operations. Measurements of the effects of the vortex were collected during piloted drag reduction tests to compare with simulation vortex models. Much of the systems integration effort required for the autonomous drag reduction tests was accomplished and tested during the Phase 0 program. In addition, experience was gained by the entire project that will aid in the flight test planning of future experiments. Roadmap Charts –1 Performance Seeking Control Page 20 Autonomous Formation Flight Program NAS4-00041 TO-104 “Drag Bucket” Hypothesis Is An Ideal Application For Performance Seeking Control Page 21 Autonomous Formation Flight Program NAS4-00041 TO-104 Drag Reduction (I 0) 100 200 300 400 500 600 700 800 900 1000 0.00 0.50 1.00 1.50 2.00 2.50 3.00 0 +280 +240 +200 +160 +120 +080 +040 +020 +010 +005 +000 -005 -010 -020 -040 -080 -120 -160 -200 -240 -280 'Y 'Z ? Hypothesis validated by CFD results, linear theory ? Incremental force and moment coefficients function of -lateral separation -vertical separation -relative bank -lead A/C lift coefficient ? Peak drag reduction near “sweet spot” ?400 in laterally ?level flight ?same altitude Dunn, Ken 5/21/99 Internal Boeing Report Peak-seeking control uses nonlinear and adaptive techniques to estimate and optimize a performance measure. C D form / C D + ' Y -'Z X f Z Y f - ' X Axis Reference System f Formation Flight Control ? Peak-Seeking control adjusts control commands adaptively to minimize drag and reduce throttle usage. Guidance Control Estimation F/A-18 Lead Vortex Guidance Control Estimation F/A-18 Trail ?Vortex is unknown nonlinear coupling. ?Theory is still limited. ?Key design trade off is ‘persistent excitation’ vs. tracking error. ? Adaptive leader commands allow cooperation. Page 22 Autonomous Formation Flight Program NAS4-00041 TO-104 Roadmap Charts –2 Autonomous Aerial Refueling Concepts Page 23 Autonomous Formation Flight Program NAS4-00041 TO-104 Leader-Follower Guidance Can Be Directly Applied To Aerial Refueling Tasks Concept Analogous To Piloted Formation Flight. GPS/IMU Located Onboard Leader (tanker) and follower. Boom dynamics & position translated onboard leader and transmitted to follower Follower Aircraft Tracks Position of Refueling Boom And Tanker ? Guidance – leverage AFF guidance SW / HW. – fuse differential GPS with optical/radar based sensors. ? Control – leverage AFF controls method. – improvements. ? Simulation – leverage AFF vortex modeling method. – develop batch simulation software for closed-loop system performance analysis and evaluation. Page 24 Autonomous Formation Flight Program NAS4-00041 TO-104 Page 25 Autonomous Formation Flight Program NAS4-00041 TO-104 Conclusions ? Dramatic Demonstration of Relative Positioning/Control ? Accurate In-Flight Wake Field Vortex Mapping ? Maximum Reductions – Drag, up to 22%, calculated at Condition 1 for longitudinal separations of X=3.0 and 4.4 – Total fuel flow, over 18% ? Optimum Positioning – X = 4.2 b (nose-to-nose) – -0.20 < Y < -0.10 b – -0.13 < Z < 0.0 b – ‘Sweet Spot’ is larger than predicted (18 ft 2 ) ? Region with > 25% C Di or >10% C D reduction – Optimum Position matches location of max rolling moment – Favorable effects degrade gradually with increased N2T distances after peaking at 3 span lengths aft