Report Pinpoints Factors Leading to YF-22 Crash
Aviation Week and Space Technology / November 9, 1992
Michael A. Dornheim / Los Angeles
The Lockheed/Boeing/General Dynamics YF-22A advanced tactical fighter prototype that crashed early this year was operating in a condition that was very prone to pilot-induced oscillation (PIO), according to Air Force accident investigators.
The airplane was performing a planned go-around using afterburner with thrust vectoring activated when it entered a pitch oscillation at 175 kt. and an altitude of 40 ft. After a severe series of pitch oscillations, the YF-22 crashed onto the runway and slid to a stop in flames (AW&ST May 4, p. 20).
Gen. Ronald W. Yates, commander of the Air Force Materiel Command, delayed release of the public "collateral board" investigation report to Oct. 22 to add segments from the confidential safety mishap report, including an engineering evaluation headed by David J. Moorhouse, currently the chief engineer for the F-15 STOL/MTD test program.
The report concluded there were no aircraft malfunctions and that it performed as designed. The investigation was aided by extensive telemetry data that were recorded during the accident.
In brief, the report stated the YF-22A was operating in a regime that was susceptible to PIO. The PIO was stimulated by retracting the landing gear during a pulse of full forward stick, which increased aircraft sensitivity. This, along with a nose-up bias from trim and software, started the PIO as Lockheed test pilot Thomas A. Morgenfeld briefly tried to fly a smooth climb-out.
Control surface actuators become rate limited, introducing further control log and worsening the oscillation. Morgenfeld was not aware he was in a PIO and thought the aircraft had malfunctioned. At 40 ft., he did not have many options and bellied the aircraft onto the Edwards AFB concrete Runway 22. The aircraft slid 8,000 ft. before stopping 185 ft. left of centerline.
The PIO-sensitive flight condition was "thrust vectoring on" at low altitudes and the lower airspeeds. The main contributor was not the thrust vectoring itself, but the flight control system gains that are in effect when thrust vectoring is activated. The aircraft would still be PIO-sensitive if these gains were used and the vectoring nozzles were disconnected.
At low airspeeds, YF-22A stick deflection basically commands a pitch rate, with zero deflection ideally giving a constant body attitude. From 200-260 kt. calibrated airspeed (KCAS), the pitch command blends to a g-command. The airspeed in the accident varied between 175 and 215 KCAS, so the aircraft was largely in the pitch command mode.
The accident investigators applied the "Ralph Smith PIO prediction criteria" to data from the accident go-around and four prior go-arounds, all using thrust vectoring. Smith is a consultant in Tehachapi, Calif. His criteria are not universally accepted but "they've been validated a lot," Moorhouse said.
Analysis of the early part of the accident PIO showed the pilot stick input was lagging pitch rate by 0.15 sec, the stabilator lagging the stick by 0.05 sec., and the pitch rate lagging the stabilator by 0.35 sec., for a total of 0.55 sec. lag in the pilot control loop. The period of oscillation was twice that -- 1.1 sec -- meaning that the lag was 180 deg. out of phase, making the aircraft "extremely susceptible to a PIO," the report found.
Analysis of the four previous go-arounds using military thrust gave similar results, predicting moderate PIO when the pilot attempts to closely track pitch angle.
From the time the gear was retracted in the accident, the stabilator was moving at its software rate limit of 60 deg./sec. and the vectoring nozzles were closely tracking the stabilator, with 0-0.2-sec. lag. Because the surfaces were rate-limited, they introduced a further 0.3-sec. lag in the pilot control loop, making the aircraft more difficult to control.
The Smith criteria assume the pilot tries to control load factor instead of pitch or pitch rate when in a PIO. "It's controversial, but Smith assumes the pilot switches to what he feels instead of what he sees," Moorhouse said. The accident data showed a 1.2-sec. delay in the control loop between pilot and load factor response while the oscillation was occurring with a 1.85-sec. period, for a phase lag of 234 deg. "Given the magnitude of the control inputs [full up/down deflections], this suggests a well-developed PIO in which it would be impossible for the pilot to recover unless he got out of the control loop," the report concluded.
The YF-22A had made a smooth go-around with thrust vectoring at military power 2 min. before the accident, and the report investigated the differences between them:
Three sets of pitch rate-stick deflection curves are possible. In order of increasing sensitivity, they are connected with landing gear down, vectoring off ("power approach"); gear up, vectoring off; and gear up, vectoring on. Placing the gear handle down automatically disables vectoring.
At full 0.26-in. nose-down stick deflection, a pitch rate of -2.6 deg./sec. is commanded in the power approach mode, while -10.5 deg./sec.-- four times greater -- is commanded with gear up and vectoring on, when at 205 KCAS. Similarly, at full 0.52-in. aft stick deflection, the power approach mode commands a 5.25-deg./sec. pitch rate, while gear up and vectoring on gives 17.1 deg./sec.
Stick force for the YF-22A's sidestick controller is 31 lb. full aft and 16 lb. full forward. It has a ± 0.03-in. deadband. The pitch rate curves are nonlinear and intermediate stick force gradients vary.
When other factors are included, analysis shows the YF-22A became 23-155% more sensitive when the gear handle was raised in the accident go-around, depending upon commanded pitch rate.
However, the control surfaces did not twitch instantly to follow the new gains because of anti-transient logic, called a "sump," located between the flight control laws and the surfaces. The sump adds an opposing bias to the control surface inputs equal to the amount of the transient due to the change in control laws. This bias washes out in 4 sec.
With the stick pressed full forward when the gear handle changed the control laws, the sump added an airplane nose-up bias initially worth about 8 deg./sec. pitch rate to prevent the stabilator from twitching further nose down. The aircraft also had some nose-up trim. When the pilot released the stick 0.2 sec. later, the 6-7-deg./sec. sump bias made the aircraft unexpectedly rotate nose up. That bias is worth more than full aft stick in power approach mode. The pilot countered with 0.3-sec. full down stick with the high-gain control laws freshly in effect, the control surfaces became rate-limited, and the PIO was underway as the pilot briefly tried to maintain a shallow climb-out.
"You can't predict when PIO will happen," Moorhouse said, explaining why PIO had not shown up before. "You can predict that an aircraft is susceptible to PIO, that at some time circumstances will make it happen."
Moorhouse said that the Smith criteria and other PIO measures are in Mil-Std-1797 handling specifications, but that "it's typical to not check for PIO during design," particularly if the aircraft is designed to have good Level 1 handling qualities.
The YF-22A thrust-vectoring system is intended mainly to enhance control at high AOA and to improve "g" capability at high supersonic speeds, both of which were tested at high altitude. "We did not do a lot of thrust-vectoring work on the simulator at sea level," Gerald T. Joyce, General Dynamics lead engineer for F-22A control laws development, told investigators.
The fixed-base YF-22A simulator was used to evaluate PIO under the accident conditions. When starting at 14 deg. to capture a 10-deg. attitude with the vectoring-on control laws, "the initial pitch down ... was very rapid and ... startling to the pilot," the report said. "Attempts to aggressively capture the desired pitch attitude always resulted in a large amplitude, undamped pitch oscillation with a period of 1.4 sec. Again, [control surface] rate limiting was apparently largely responsible."
But when the pilot was less aggressive, "the response was much different. The pitch down rate was smooth, but not abrupt, and capture easily performed." And when in a large PIO, "the oscillations died [rapidly] if the pilot let go of the stick."
The oscillations started about 1 sec. after the gear handle was raised, and the aircraft hit the ground 7.5 sec. later. Morgenfeld told investigators he did not understand what the problem was. "[I lit] the afterburner, gear up ... I thought I felt maybe a little bit of a bob," he said.
"The first time I knew something really was wrong, I felt a very strong nose-over pitch. [I was] looking at a lot of runway and the airplane had never done anything like that before. It surprised me, it really shocked me .... I thought something had broken and I didn't see any [warning] lights .... I'll undo the only thing that was different between this pass and the other, I snatched it back out of burner, come back to low [thrust]."
"[I started] thinking about coming back on up with the power now to mil and continue the go-around and get this thing away from the ground. And at about that time it made another big nose-down plunge ... and then at the next strong nose-down movement I thought, boy, that's it, something's drastically wrong here .... I just tried to get the nose up so I didn't do the lawn dart trick in the runway." The aircraft hit the ground at about 205 KCAS.
During the investigation several days later, Morgenfeld said he "was extremely surprised to see that I had used so much forward stick." However, "almost by definition, a pilot is not aware he is in a PIO," Moorhouse said. He also noted that sidestick controllers tend to be used more in a full-displacement pulsed manner than large-throw center sticks do.
The airplane flight manual and operating limits had no restriction on the use of thrust vectoring, but the flight test card called for it to be off for landing. The two lead flight test engineers said this card instruction was a carryover from earlier testing and was to protect against a failure of the vectoring system at low altitude. Since the vectoring system had been reliable, they said they would have eliminated the card item if requested.
However, flight control engineers said vectoring should be off at low altitudes for "flight control reasons," but they did not issue an operating limit "because they had no known or suspected reason to do so," Not all individuals in the YF-22A program were aware of the card instruction, and whether it was being obeyed.
USAF Report Details April YF-22A Crash
Aviation Week and Space Technology / November 16, 1992
Michael A. Dornheim / Los Angeles
Air Force photos of last April's advanced tactical fighter prototype crash show the aircraft to be relatively intact despite a 1 1/2-hr. fire and sliding 8,000 ft. down the Edwards AFB concrete Runway 22.
The Lockheed/Boeing/General Dynamics YF-22A made a gear-up hard landing after 8 sec. of pilot-induced oscillation (PIO) during a flyby at 40 ft., according to the Air Force public accident report (AW&ST Nov. 9, p. 53).
Lockheed pilot Tom Morgenfeld pyrotechnically blew off the canopy after the normal mechanism did not work, and then climbed out of the cockpit. The canopy can be seen by the rear of the aircraft, surrounded by firefighting foam.
The aircraft had about 6,000 lb. of fuel on board when it crashed, more than desired for landing, and a reason for igniting the afterburner during the flyby was to consume fuel before landing. The fire crew arrived at the accident site within several minutes of the crash, found a fire at the left rear of the aircraft, and extinguished the external flames immediately.
However, an internal fire continued. Firefighters had difficulty entering the engine bay because the lower access panels were on the ground and a cutting tool could not penetrate the titanium skin. An hour after the crash, an internal explosion rocked the YF-22A, injuring a firefighter atop the aircraft. The fire was officially declared out 2 hr. after the crash.
Shredded plies of the composite structure can be seen at the rear. The belly appeared relatively undamaged, but there was extensive bending and cracking of internal structure. The vectoring nozzles are in a down position.
Fixes being considered to make the YF-22A less PIO-susceptible include altering the flight control gains, making the switch between gain schedules gradual instead of instantaneous, and preventing the stabilator from becoming rate-limited.
Editorial: Promote Risk Levels in Flight Tests
Aviation Week and Space Technology / November 23, 1992
What will the U.S. Air Force learn from the crash of the YF-22A advanced tactical fighter prototype? The answer is several lessons that are poles apart. The service could have been more careful, validated every corner of the envelope and taken a slow, cautious approach to minimize all risks (AW&ST Nov. 9, p. 53).
But that ignores the fact that the ATF demonstration/validation program was an amazingly productive program that warranted the risks taken. The crash was an unfortunate hiccup that capped an overall stellar effort.
This argument has more merit but is harder to support. One cannot prove what the ATF program would have lost had it been shorn of its aggressive spirit.
Conversely, the C-17 transport program shows what a minimum-risk philosophy does. The first operational C-17s delivered to the Air Force next spring probably will be prohibited from aerial refueling with KC-135 tankers because excessive caution has slowed fine tuning of the C-17 system to fix handling problems in some regimes.
The Air Force requires validation of all flight control software changes. This sounds reasonable, but it means changes are limited to six-month intervals. A software fix missing a deadline may languish for a year. And, if the fix does not work, another half-year cycle can pass. At that rate, the C-17 could be in development another five years.
That is the damage a low-risk doctrine inflicts when it becomes the canon of zealots. The C-17 is an aerodynamically stable airplane with perfectly acceptable and proven mechanical backup flight controls -- it can tolerate some mistakes. It should be treated the same as the F-16 and others were -- with software patches quickly devised to permit rapid flight control development.
The goal of test programs must be reasonable risk, which means the Defense Dept. has to stop promising perfection to Congress and re-educate lawmakers to the realities of risk in development.
The successful ATF moderate risk approach should be the model, but with more reasonable compensation for the contractors. The bogged-down C-17 program is a prime example of what happens otherwise.