With the addition of NASA and the USAF as X-31 partners, things had become even more complicated. The X-31 team now consisted of contractors (Rockwell and MBB), governmental agencies (DARPA, NASA, and the German MOD), military services (the U.S. Navy, USAF, and the German Air Force), and flight-test agencies (Dryden, WTD-61, NAS Patuxent River, and the Air Force Flight Test Center). In order to manage all of this, the various partners established an official International Test Organization that attempted to promote cooperative
Into the Air: Initial Flight Testing
decision making with no single person as the “chief.” The logo that was devel-oped was an oval shape with no single entity at the “top.” The structure of this organization will be discussed in later sections.
The move to Dryden brought with it a change in the responsibility for aircraft maintenance from Rockwell to Dryden and for flight clearance authority from NAVAIR to Dryden.
It now took even more time for the aircraft to be inspected and the systems to be understood by Dryden personnel so that the flight clearance activity could be started. By April 1992, the aircraft was deemed ready for its first flight under the auspices of the ITO. On April 23, 1992, Karl Lang took to the air in Ship 2 for a systems-checkout flight that included flying qualities in all three reversionary modes.
The X-31 program was still very fragile. Funding was always a concern, and for this reason DARPA managers wanted to move forward with some deliberate speed. Col. Michael Francis, USAF, who had replaced Tack Nix as DARPA program manager, was concerned that “the flight safety community wanted to baby-step us,” which would lead to delay and complication.21 But NASA had sound reasons—including flight safety and data-reduction quality—for not rushing along. However, the financial reality was that delays could cause the program’s cancelation. There was the constant pull and tug of finding the right balance between “safe,” “perfect,” and “good enough” that plagues all flight-test programs, but perhaps affects research programs to a greater extent.22 In an attempt to address safety concerns and to develop some milestones for the post-stall envelope program, Francis held an offsite meeting in nearby Lancaster, CA, between May 19 and 20, 1992. This planning session reviewed the program goals and the current program status, prioritizing the following flight-test objectives for the post-stall envelope expansion:
• •
•
X-31 International Test Organization logo.
(NASA DFRC)
Demonstrate dynamic post-stall flight,
Assess unique EFM technologies (agility, the vehicle, human-machine interface), and
Prepare for the tactical evaluation.
Test planners discussed the maneuvers that needed to be demonstrated to com-plete the dynamic post-stall flight envelope clearance, acknowledging that the envelope had to be gradually expanded so that areas of concern were adequately characterized. They named these demonstration maneuvers “maneuver mile-stones,” and these consisted of the following:
Flying Beyond the Stall
• Deceleration to 70° angle of attack at 1 g;
Full deflection, 1-g rolls at 70°angle of attack (executing this maneu-ver contractually completed Phase III of the program);
Dynamic, level turn entry to post-stall from corner speed; and Turn-optimized/gravity-assisted post-stall maneuver to 180°heading change (this maneuver would subsequently receive the moniker, the
“Herbst turn”).
•
• •
Planners also discussed concerns over insufficient roll control at high angles of attack, along with the planned addition of strakes on the aft fuselage as a fix. The above maneuver milestones assumed that this roll-control problem would be fixed by the strake installation. There was some discussion of the methodology for conducting the follow-on tactical evaluations. Interestingly, it was agreed that, ideally, flying one X-31 against the other with only one of the aircraft using thrust vectoring would provide the best technical assessment of the program (though, as mentioned previously, this subsequently did not take place).23
Since Dryden and the Air Force Flight Test Center were now partners in the test effort, it was time to get their test pilots checked out in the airplane. Rogers Smith from Dryden flew his first checkout flight on June 4, 1992, followed by Lt. Col. Jim Wisneski, USAF from AFFTC, on June 9, 1992.
As the post-stall envelope expansion continued, testers evinced continuing concern that the effectiveness of the trailing-edge elevons was different than had been predicted before flight. This had been noted in the initial testing at Palmdale, as was discussed previously.24 This was a serious issue because most delta-wing aircraft (including the X-31) do not have separate ailerons (for roll) and elevators (for pitch). Rather, they use a single surface called an elevon (for elevator-aileron) that combines both functions. Operated differentially, an elevon furnishes roll control; operated symmetrically, it furnishes pitch control.
The difference between the predicted and actual trailing-edge elevon deflection necessary to hold a trim angle of attack at increasingly high angles was as much as 10° greater than predicted.
This problem meant that there now was insufficient elevon control authority left to provide adequate roll control. Elevon deflections available for roll control were reduced by 50 percent. The nose-down aerodynamic safety margin was also reduced. How could the preflight prediction have been so out of synch with postflight reality? It was discovered that the X-31 wind tunnel model was connected to the wind tunnel’s mounting device by two attachment rods located along the aft fuselage of the model. These rods themselves contributed a nose-down pitching moment to the wind tunnel model. Since the actual aircraft did not have these rods, the airplane did not have the nose-down pitching influence of the rods. Langley engineers quickly provided a fix to this
Into the Air: Initial Flight Testing
problem. They designed two aft strakes that were 4 feet long and attached to the aft fuselage between the trailing edge of the elevons and the nozzle area.
These strakes were tested on the existing 19-percent scale model of the X-31 in the NASA Langley 30- by 60-foot wind tunnel, which validated that the strakes indeed returned the pitching moments to the required values.25 The strakes were fabricated out of plywood sandwiched between two metal sheets and then covered with fiberglass. The initial flight with the strakes attached
X-31 without aft fuselage strakes, 1992. (NASA DFRC)
X-31 with aft fuselage strakes. (NASA DFRC)
Flying Beyond the Stall
was flown on September 10, 1992, with Navy test pilot Al Groves. Flight-test results confirmed the new tunnel predictions, and testing was now able to safely continue expanding the post-stall envelope to 70°.
The X-31 achieved its maximum design angle of attack of 70° during a flight by Groves on September 18, 1992. Groves decelerated and stabilized the aircraft at 70° angle of attack for approximately 40 seconds. This was sub-sequently repeated by Karl-Heinz Lang and Rogers Smith on the same day.
Additional aircraft controllability was demonstrated on September 22, 1992, when Fred Knox performed 30° bank-to-bank rolling maneuvers at 70° angle of attack.26 The goal now was to demonstrate “agility” at high angles of attack, which is the ability to energetically and rapidly maneuver from one flight condition to another. This was accomplished on November 6, 1992, during a flight by Fred Knox in which he performed 360° rolls about the velocity vector while the airplane was flying at 70° angle of attack.27 With this particular accomplishment, the X-31 had truly entered a regime of flight where no other conventional aircraft had been.
During the expansion to high angles of attack, the pilots noticed several
“lurches,” or dramatic changes in roll rate with a constant roll command on the stick. Engineers experimented with fixing this by placing “grit strips” (strips with a sandpaper-like surface to activate airflow in the boundary layer around the forebody) on the radome and on the noseboom. They suspected that vortex shedding off the radome or noseboom was causing these lurches and hoped to control it with the grit strips. The results were mixed. In some cases, no lurches were noted while in others, the change in roll rate was substantial.
The next area to explore was the so-called “dynamic entry.” Researchers realized that in actual combat, fighter pilots would enter the high-AOA regime in an aggressive manner, likely with relatively high rates of g-onset. The high-AOA testing of the X-31 to date had been with relatively slow decelerations to the desired angle of attack. The same approach was used in rolling the aircraft, whereby the airplane would be slowed to achieve the target angle of attack and would then be rolled through the desired roll angle. As noted, Fred Knox reached 360° of roll at 70° angle of attack. Now it was time to achieve post-stall flight with a dynamic entry more representative of what would be encountered in an actual combat-maneuvering entry.
The method of achieving a dynamic entry was to roll the airplane inverted and then abruptly apply full-aft stick to attain the desired angle of attack as set on the AOA limiter. Naturally, if an angle of attack in excess of 30° was desired, the post-stall switch on the control stick had to be actuated. The first attempt at a dynamic entry was flown by the Air Force member of the team, Jim Wisneski, on November 25, 1992. This first entry was to attain an approximately 2-g post-stall entry and was followed by some other maneuvers.
Into the Air: Initial Flight Testing
X-31 in high-AOA flight, 1994. (NASA DFRC)
Flying Beyond the Stall
Then a rude and unanticipated surprise occurred. As reported by Rockwell flight-test engineer David Rodrigues,
The first maneuver was performed at 35,000 feet / 0.4 Mach and consisted of a full aft pitch input from inverted flight with maximum afterburner set and the angle of attack limiter set at 45°
(a split S maneuver). This maneuver was repeated for additional data. The angle of attack limiter was then set to 60° angle of attack for the next split S. During the next maneuver, PST was inadver-tently disengaged, limiting the angle of attack to 30°. A recovery was made and the aircraft set up to repeat the split S to 60°. This maneuver was again at 35,000 feet / 0.4 Mach with afterburner set. The aircraft stabilized momentarily at about 60° degrees but was yawing to the right and continued to departure from control with increasing angle of attack and yaw rate. [Emphasis added.]
The pilot initiated recovery with forward stick as a recovery call was issued from control. The aircraft immediately responded to the forward stick and with angle of attack reduction, the yaw rate damped to zero, completing recovery of the aircraft to controlled flight. No failures were noted in the aircraft and no engine or other system anomalies were noted during the departure.28
The departure came as a shock to mission planners, for the X-31 was designed to avoid doing just that! Fortunately, the aircraft recovered after 320° of turn due to the excellent nose-down pitch authority provided by the thrust-vectoring system.29 Analysis showed that the departure was probably caused by a large, unexpected yawing moment that was generated from the forebody of the airplane, a problem encountered on earlier programs such as the Northrop F-5E Tiger II development effort and the early F-15E Strike Eagle program (the former because of forebody shape and the latter because of an asymmetrically offset flight-test pitot noseboom). Wind tunnel tests at NASA Langley showed that the X-31’s nose configuration could have very nonlinear and unstable yawing-moment characteristics due to the influence of asymmetric vortices coming off the nose. These vortices could impart asym-metric drag, effectively “pulling” the nose to one side or the other and trigger-ing a departure from controlled flight. This testtrigger-ing also showed that roundtrigger-ing of the nosecone on the airplane and the addition of 20-inch strakes along the nose would provide adequate directional characteristics above 50° angle of attack.30 Grit strips were also added to the nose, as had been done previously to help make the vortices more uniform and predictable in behavior.
Into the Air: Initial Flight Testing
Postflight analysis of the incident from a flight control perspective resulted in changes to the flight control software to prevent sideslip buildup as well as to increase the thrust-vectoring control power.31 Technicians increased thrust-vector vane travel from 27° to 34°, allowing a greater amount of thrust deflection.32 The 27° limit prevented the vanes from hitting each other during operation.33 (It should be noted that changes to the flight control software are not unusual in flight-test programs, and, indeed, the X-31 had 32 software releases over the course of the program; such is the nature of research aircraft programs in the electronic flight control era.)34 On February 9, 1993, Jim Wisneski again took the X-31 into the elevated-entry regime with a buildup of split-S maneuvers to 50°, 55°, 60°, 65°, and ultimately 70° angles of attack. No problems were encountered on any of the elevated-g tests, even at the higher AOA marks.35
Harvey Schellenger, a Rockwell engineer who joined the program early on as it moved from Columbus to Los Angeles and later became chief engineer and then acting program manager of the EFM program, recalled to the author that solving these high-AOA vortex-shedding issues constituted an “interesting story”:
The low speed wind tunnel model had a nose with a small radius.
The aircraft also had a small radius, but after correcting for scale, the wind tunnel model radius was larger (on the order of a large marble at full scale). We saw no indication of large yaw asymmetries in the wind tunnel tests, but the aircraft had a yawing departure in flight. Back into the Langley [30- by 60-foot] tunnel with a bit of clay to sharpen the model nose, and the asymmetry showed up. We rounded the noses of the aircraft to match the original w-t [wind tunnel] model—plus a little, to about a golf ball. We also added strakes as sized in the tunnel but I think that just rounding would have been enough. The approach was to hit it with a big ham-mer and make sure it would not return, so it was rounding with strakes and back in the air. Aircraft 1 was rock steady from then on, but A/C 2 still showed a little “nervousness” in yaw at high AoA.
Because of this the pilots started calling it the “evil twin”. We were about to give up on A/C 2 and concentrate on just A/C 1 ops when I noticed that we had somehow not rounded 2’s nose as much as 1’s. How we missed this I can’t explain, but we did. And the strakes didn’t cover it up – not completely. We’re talking pretty small differ-ences here. From way back in the memory bank: the original radius was about 0.1", the final was ≅ 0.7", and the evil twin was initially
≅ 0.4. Changing A/C 2 to match A/C 1 completely eliminated the
“nervousness” and we had a two plane program again.36
Flying Beyond the Stall
Illustrating the criticality of nose shaping, Patrick “Pat” Stoliker, an engineer on the X-31 at the time and now deputy director of Dryden, recalled, “The two aircraft had different grit strip configurations (Ship 2 needing a longer strip). Ship 1 had 20-inch strakes while 47-inch strakes were used on Ship 2.”37
As testing progressed, changes in the air data–sensing system were made as well. Since the pitot-static-sensing function (as well as the angle-of-attack and angle-of-sideslip vanes) was expected to be unreliable at the high angles of attack anticipated for X-31 operations, the INU was used to provide estimated angle-of-attack, angle-of-sideslip, and dynamic pressure figures. Postflight data analysis showed large errors in angle of attack, angle of sideslip, and airspeed due to wind shifts. While this did not degrade flying qualities, it did make the monitoring of flight control system performance difficult, and it caused several maneuver aborts through erratic data input. Consequently, testers decided to modify the noseboom to provide usable data. With the original boom, the AOA vane was usable to 70° angle of attack, but the sideslip vane had large oscillations at 60° angle of attack. Indicated airspeed was always at the minimum of 48 knots when the airplane was above 60° angle of attack. The design change resulted in the installation of the Kiel airspeed probe at a 10°
nose-down attitude. Additionally, technicians mounted the sideslip vane on a wedge to provide it with a 20° nose-down attitude. Therefore, at 70° angle of attack, the sideslip vane remained below 60° and pilots no longer saw the sideslip oscillations.
The new noseboom configuration provided good angle of attack, angle of sideslip, and airspeed throughout the X-31 envelope, including post-stall flight to 70° and –5° angles of attack.38 At the time, there was not a heated version of the Kiel probe. The test team made the decision not to wait to have a heated ver-sion manufactured; therefore, in changing the noseboom configuration, technicians installed the new “canted” noseboom knowing that it was not heated. This was not considered to be problematic because the X-31 only flew in clear-sky day testing. This later would have dire consequences, highlighting yet again that in flight testing and flight research, one cannot be too careful in preflight mission planning and instrumenting research airplanes.
Since MBB engineers had the responsibil-ity for flight control law development, and because their technique (which differed from the American approach) relied largely on Kiel probe canted to compensate for
high angles of attack. (NASA DFRC)
Kiel probe “collar” to “collect” pitot pressure at high AOA. (NASA DFRC)
Into the Air: Initial Flight Testing
mathematical prediction of the aircraft’s response, it was important as envelope expansion progressed for the German test contingent to understand the values of the aircraft’s aerodynamic parameters. Aerodynamic parameters are called
“coefficients” and are determined by measuring various forces and moments produced by the aircraft and then normalizing them by dividing by dynamic pressure, a reference area (usually wing area), and, in the case of moments, a reference length. Such calculations naturally depend on accurate measurements of the forces and moments acting on the airplane. As might be expected, the accurate and repeatable measurement of these forces and moments in the post-stall arena proved to be difficult due to the very dynamic and nonlinear nature of aerodynamic flow conditions above stall speed. Several different tech-niques were employed to establish these coefficient parameters. Researchers tried both closed- and open-loop time-history matching, as well as maximum likelihood estimators.39
The problems in aerodynamic parameter identification arose from the fact that the airplane was unstable, making it difficult to merely integrate the mathematical state equations for the aircraft. Additionally, aircraft parameters and control parameters are highly correlated, so individual parameters often cannot be estimated independently. There was a large amount of noise due to vortices shedding from the aircraft nose forebody that contaminated the data further, complicating data acquisition and reliable analysis. Finally, the aircraft motion was not often sufficiently excited for parameter identification because
The problems in aerodynamic parameter identification arose from the fact that the airplane was unstable, making it difficult to merely integrate the mathematical state equations for the aircraft. Additionally, aircraft parameters and control parameters are highly correlated, so individual parameters often cannot be estimated independently. There was a large amount of noise due to vortices shedding from the aircraft nose forebody that contaminated the data further, complicating data acquisition and reliable analysis. Finally, the aircraft motion was not often sufficiently excited for parameter identification because