Gulfstream 650 N652GD

• Gulfstream was performing field performance flight testing to (1) gather data to support type certification of the G650 under 14 CFR Part 25, “Airworthiness Standards for Transport Category Airplanes,” and (2) develop takeoff and landing speed schedules and distances for the G650 airplane flight manual.
• The takeoff performance flight tests were being conducted with an angle-of-attack (AOA) limiter function disabled. The AOA limiter function was intended to be the primary stall protection system for the G650 (instead of a traditional stick pusher) once the airplane was certificated. The development of the AOA limiter software was not completed, so Gulfstream intended that the stick shaker would provide test pilots with a tactile warning of an impending stall. In addition, the pitch limit indicator (PLI) on the primary flight display would provide the pilots with a visual indication of an impending stall. [The figure] shows a representative PLI (as depicted in Gulfstream's draft G650 aircraft operating manual).

Source: NTSB AAR-12/03, ¶1.1

According to test team members, [flight test engineer 1] FTE1 briefed the team members on the day before the accident. During this briefing, FTE1 indicated that the target pitch attitude for continued takeoff tests with the flaps set to 10° (flaps 10) would be reduced from 10° to 9°‚ (± 1°). FTE1 also indicated that they should discontinue a test if pitch reached 11° during the initial takeoff and then decrease pitch and add engine power. Gulfstream's principal engineer for airplane performance (who discussed the change in target pitch with FTE1 at an informal meeting in late March 2011) and [airplane performance group head engineer] APG1 stated that FTE1 made the change in target pitch to be consistent with the procedure for takeoff tests conducted with flaps set to 20° (flaps 20) and ensure that the AOA would remain below the range at which two previous uncommanded roll events (as discussed in section 1.3.2) had occurred.

Source: NTSB AAR-12/03, ¶1.1

The takeoff speed schedules to be used by the flight crew consisted of tabulated values for the decision speed (V₁), rotation speed (V_R), and takeoff safety speed (V₂) as a function of flap setting and airplane gross weight; liftoff speed (V_LOF) values were included in the speed schedules for one-engine-inoperative (OEI) continued takeoffs. The speed schedules were based on the free-air (out-of-ground-effect) stall speeds of the airplane, as determined by previous flight testing, and the Part 25 takeoff speed requirements.

Source: NTSB AAR-12/03, ¶1.1

The G650 program's takeoff performance guarantee target was 6,000 feet ± 8 percent at standard sea level conditions. Gulfstream indicated that achieving the target V₂ speeds was necessary to maintain the takeoff distance within the guaranteed target, or the operation of the airplane would be limited to longer runways. (For the G650, the runway length required for takeoff is minimized if the V₂ speed is minimized. Thus, there is a performance advantage to keeping the V₂ speed as close as possible to the minimum required.) The test team completed its first OEI continued takeoff test run (7A1), but the airspeed reached 145 knots at 35 feet and exceeded the target V₂ value (136 knots) by 9 knots. The OEI continued takeoff (with the same flap configuration) was being repeated during the accident test run to reduce V₂ to the target value for that run (135 knots).

Source: NTSB AAR-12/03, ¶1.1

• Test run 7A2 began at 0933:00. According to the on-board video recording, at that time, the PIC advanced the thrust levers for takeoff. Recorded flight data showed that, between 0933:36 and 0933:37, when the airspeed was about 105 knots, the SIC moved the right thrust lever to the idle position. About that time, the CVR recorded the SIC stating “chop” to confirm this action.
• At 0933:46, the CVR recorded the SIC stating, “standby, rotate.” About 1 second later, when the airspeed was about 127 knots, the video recording showed the PIC pulling on the control column for rotation. At 0933:50, the pitch attitude and AOA reached about 10°, and then the PLI appeared. About that time, cockpit displays showed that the airplane's wings were about level and that the slip indicator was displaced slightly to the left. At 0933:50.4, the airplane's pitch and AOA exceeded 11°. Immediately afterward, the CVR recorded the PIC stating, “[unintelligible] going on,” and the video recording showed that the bank angle was increasing to the right and that the PIC was making a slight left wheel input.24 The video recording ended at this point.
• Between 0933:52 and 0933:53, the CVR recorded the SIC and the PIC repeating, “whoa.” Recorded flight data showed that the stick shaker activated at 0933:52.2 for 0.6 second (with the pitch at 12.7° and the AOA at 12.4°) and at 0933:53.5 for 6.4 seconds (with the pitch at 11.8° and the AOA at 12.2°). At 0933:53.6, the CVR recorded the electronic annunciation “bank angle”; recorded flight data showed that the bank angle at that time was about 16.2°. The PIC then stated, “power, power, power,” and the SIC responded, “power's up”; flight data showed that the right thrust lever had been advanced all of the way forward about that time. At 0933:58.5, the CVR recorded the electronic annunciation “bank angle,” which was 30.5° at that time. The last communication recorded on the CVR (which was unintelligible) was at 0934:05, and the CVR recording ended at 0934:10.
• The National Transportation Safety Board's (NTSB) aircraft performance study for this accident found that, when the airplane's AOA reached 11.2° during the accident takeoff, the AOA exceeded the stall AOA for the combination of flap setting, height above the ground, Mach number, and roll angle present at the time, resulting in a loss of roll control. This finding was based, in part, on the results of Gulfstream's simulation residual analysis, which indicated that, at 0933:50.5, as pitch angle and AOA were increasing through 11.2°, large aerodynamic rolling and yawing moments to the right were acting on the airplane. These aerodynamic moments were indicators of flow separation on the right outboard wing and an asymmetric stall of the airplane. Before the accident, Gulfstream estimated that the in-ground-effect stall AOA would be 13.1° and set the AOA threshold for the activation of the stick shaker stall warning at 12.3°.
• Recorded flight data and ground scars and markings on the runway and airport property indicated that, shortly after rotation, the airplane's right wing contacted the runway. The airplane subsequently yawed to the right, departed the right side of the runway, traveled along about 3,000 feet of airport property, and came to rest about 8,404 feet from the runway 21 threshold and 1,949 feet to the right of the runway centerline.

Source: NTSB AAR-12/03, ¶1.1

On February 11, 2011, during a meeting to discuss issues from Roswell I field performance testing, the Gulfstream senior vice president of programs, engineering, and test and the Gulfstream vice president of the G650 program (also referred to as the G650 program manager) were informed that the recorded V₂ speeds were high. Specifically, the V₂ values at 35 feet were consistently higher than the target V₂ values planned for Roswell II testing. As a result, the field length needed for takeoff would be longer than the program's takeoff performance guarantee (6,000 feet ± 8 percent). Gulfstream personnel from the flight sciences, flight test, and flight operations departments believed that changes to the takeoff technique could improve these results.

Source: NTSB AAR-12/03, ¶1.1

• Two days later, FTE1 led a 1-day takeoff technique development testing effort in Birmingham, Alabama (flight 111), during which time the test team assigned to that flight experimented with different takeoff rotation techniques and rotation speeds to try to eliminate the V₂ overshoots, which would reduce the field length needed for takeoff. A total of seven simulated OEI continued takeoff test runs were performed using 20° of flaps and a target pitch attitude of 9°. The G650 project test pilot was the flying pilot, and the accident SIC was the monitoring pilot. Two flight test engineers (including FTE1) were also part of the on-board test team. (No airplane performance engineers were present for the test.)
• Changes to the takeoff technique included (1) adding 2 knots to the VR speed schedule for a given thrust-to-weight ratio while keeping V_LOF and V₂ the same and (2) increasing the pitch angle beyond the target pitch angle as soon as the airplane lifted off (instead of holding the target pitch angle until 35 feet). In addition, changes were made to the control column input used to initiate rotation. As the testing progressed, the abruptness and magnitude of this input increased. According to the on-board video recording, FTE1 asked the G650 project test pilot whether he could convince FAA certification officials that the rotation technique being explored was a “normal technique.” The project test pilot responded that the technique would have to be modified “slightly.”

Source: NTSB AAR-12/03, ¶1.1

The takeoff performance presented in this section is predicated on employing the following techniques during rotation and climb out following an engine failure. After reaching the target rotation speed, a rapid and aggressive column pull shall be applied. Adjust pitch following liftoff to achieve V₂ at the 35 foot height.

The takeoff performance presented in this section is predicated on employing the following techniques during rotation and climb out following an engine failure. After reaching the target rotation speed, a column pull force of approximately 75 pounds should be applied in an aggressive manner (less than 1 second). The following table presents recommended pitch attitudes (q) that should be targeted from liftoff to the 35 foot height. These pitch attitudes should allow the aircraft to accelerate with one engine inoperative to approximately V₂ at the 35 foot height. Airspeed control is critical during a continued takeoff with an engine failure, therefore, the pitch attitude should be adjusted as required to capture and maintain V₂. Pitch attitudes during a normal takeoff (all engines operating) may be increased to obtain the desired climb out speed and flight path, but should be limited to a maximum of 20 degrees.

Source: G450 Aircraft Operating Manual, §12-01-10, ¶1

The test team found that an abrupt column pull force of about 70 to 75 pounds was the most successful in reducing the magnitude of the V₂ overshoot. (The maximum column pull force permitted by FAA regulations was 75 pounds.) The test team also found that, if the flying pilot rotated rapidly (at peak pitch rates between 6.1° and 8.5° per second) to the 9° target pitch attitude and then exceeded 9° shortly afterward, V₂ overshoots (and V₂ + 10 knot overshoots for all-engines-operating [AEO] takeoffs) could be reduced to within a few knots of the target speeds. The takeoff rotation technique that produced the best results during flight 111 resulted in a V₂ speed that was still about 3 knots high. According to Gulfstream's GVI Field Performance Certification Flight Test Plan (revision A, dated October 2010), the required tolerance for the target V₂ speed was ± 2 knots.

Source: NTSB AAR-12/03, ¶1.1

The two previous uncommanded roll events during field performance testing at ROW occurred on November 16, 2010 (during flight 88), and on March 14, 2011 (during flight 132).

• The flight 88 uncommanded roll event occurred during minimum unstick speed (V_MU) development testing at a flap setting of 20° and a pitch target of 9° to 10°. The flying pilot was the accident PIC, and the G650 project test pilot was the monitoring pilot. The PIC had participated in, but had not performed, previous V_MU tests in the G650. The uncommanded roll event (8° right wing down) occurred immediately after liftoff during the PIC's initial V_MU test run. The PIC expressed surprise at the rotation rates obtained, which led to an overshoot of the target pitch attitude by 3°. The flight crew recovered the airplane (when the monitoring pilot pushed the control column forward to lower the nose and AOA), continued to climb out, and landed without further incident. The airplane did not contact the ground during the roll event. Testing was not stopped after the flight to investigate this matter. Instead, the test was repeated immediately afterward, and the PIC performed the maneuver successfully.
• The flight 132 uncommanded roll event occurred during the second test run for an OEI continued takeoff (right engine reduced to idle) at a flap setting of 20° and a target pitch attitude of 9°. The flying pilot for flight 132 was the SIC of the accident flight, and the monitoring pilot was a company senior test pilot assigned to field performance testing. FTE1 was also part of the on-board test team. The test card for the incident run stated, “rotate at V_R using 70 lb pull until rotation begins, reduce force to gradually capture 9°.” During the test run, the accident SIC pulled with 65 pounds of force and held sufficient force on the column to allow the pitch angle to reach 12° about 0.5 second after liftoff; about 1 second later, the airplane rolled 8° to the right. The flight crew recovered the airplane (when the monitoring pilot pushed the control column forward to lower the nose and AOA) and continued the takeoff without the airplane contacting the ground.
• According to the senior test pilot, he and FTE1 met informally after flight 132, and FTE1 expressed concern that the airplane had stalled. FTE1 noted that “we were at” an AOA of 11.5° during the event but that an in-ground-effect stall was not predicted to occur until at least an AOA of 13°. Because the AOA during the event had remained 1.5° below the predicted in-ground-effect stall AOA, the senior test pilot and FTE1 did not attribute the event to a stall but instead to a “lateral-directional disturbance” (that is, a roll event due to sideslip) that was aggravated by the unavailability of the yaw damper, which had been deactivated because of a temporary in-flight restriction resulting from a previous event. The senior test pilot suggested that, to prevent an uncommanded roll from recurring, takeoff testing should be discontinued until the yaw damper was back in service; FTE1 agreed with this suggestion. An analysis of flight 132 data performed by Gulfstream and the NTSB after the accident indicated that a stall occurred at an AOA of 11° and that the sideslip angle was -3° during the onset of the stall.

Source: NTSB AAR-12/03, ¶1.3.2

• After the test run, the test team noted that the airspeed at 35 feet exceeded the target V₂ speed and discussed how the takeoff technique might be modified during the next test run to reduce the V₂ overshoot. The PIC indicated that the maneuver could be repeated with a shorter pause at the target pitch value. The test team did not discuss that the airplane had reached the 9° target pitch about 3 seconds before liftoff or that the airplane had lifted off at an airspeed that was 1 knot below the target V₂ speed. Also, as with previous test runs, no team member questioned the safety of the takeoff technique or considered whether the test procedures or takeoff speed schedules needed to be reevaluated. The NTSB concludes that the test team's focus on achieving the V₂ speeds for the flight tests and the lack of guidance specifying precisely when the pitch angle target and pitch limit applied during the test maneuver contributed to the team's decision to exceed the initial pitch target and the pitch angle at which a takeoff test was to be discontinued.
• During the accident takeoff (test run 7A2), there was no pause at the 9° pitch target, and the pitch rate slowed only as the airplane pitched through 9° about 1 second before liftoff. The airplane then stalled during liftoff at a pitch angle and an AOA of about 11.2°. The NTSB concludes that the airplane stalled at an AOA that was below the in-ground-effect stall AOA predicted by Gulfstream (13.1°) and the AOA threshold for the activation of the stick shaker stall warning (12.3°). The test team members' exceedance of the stall AOA likely resulted, in part, from their confidence in, and reliance on, the PLI and stick shaker system, which they did not realize had been set too high to account for the actual in-ground-effect stall AOA.
• Also during the accident takeoff, an uncommanded roll to the right began about 2 seconds before liftoff. The PIC made a left control wheel input of about 12°, which reduced the roll rate from 1.5° to about 0.9° per second. Liftoff occurred at a bank angle of 2.6°, and the roll rate increased immediately afterward to about 5° per second. The PIC stated “[unintelligible] going on” and doubled the amount of control wheel and rudder correction he was using (to counteract the increasing roll and yaw), but these actions did not stop the roll.
• About 1.8 seconds after liftoff, with pitch tracking just under the PLI and the bank angle exceeding 10°, the SIC began to exclaim “whoa” multiple times. About 2 seconds after liftoff, pitch increased above the PLI, the stick shaker activated, and the PIC responded by pushing the control column forward and adding full left control wheel and full left rudder to counteract the roll. (The stick shaker activated at the programmed AOA, but a stall on the right outboard wing had already occurred.) About 2.3 seconds after liftoff, the right wingtip struck and began to drag along the ground at a bank angle of 13.4°. Pitch then decreased below the PLI, and stick shaker activation stopped. The PIC made a brief aft control column input and then held the column slightly aft while continuing to apply full left wheel and full left rudder. Even though pitch had decreased below the PLI, the airplane remained stalled, and the right wingtip continued to drag along the ground as the airplane veered off the runway.
• About 3.3 seconds after liftoff, one or both pilots advanced the right thrust lever, the PIC called out repeatedly for power, and the SIC confirmed, “power's up.” However, the airplane rolled farther to the right, pivoting on its right wingtip. The stick shaker activated again and continued to activate for the next 6.4 seconds (which was essentially the remainder of the flight). About 4.3 seconds after liftoff, at a bank angle of 18° and with pitch increasing to 13.8°, the PIC pulled back abruptly on the control column. The pitch and AOA responded accordingly, eventually increasing to peak values of 14.8° and 22.7°, respectively. The PIC maintained this substantial aft column input for several seconds while continuing to maintain full left control wheel and rudder. The bank angle increased to a maximum of 32° before the airplane began to roll out of the bank, the pitch angle decreased, and the fuselage impacted the ground. The NTSB concludes that a stall on the right outboard wing produced a right rolling moment that the flight crew was not able to control, which led to the right wingtip contacting the runway and the airplane departing the runway from the right side.

Source: NTSB AAR-12/03, ¶2.2

As stated in section 1.1, ground effect refers to changes in the airflow over the airplane resulting from the proximity of the airplane to the ground. Ground effect results in increased lift and reduced drag at a given AOA as well as a reduction in the stall AOA; thus, the stall AOA is lower for airplanes in ground effect compared with the stall AOA for airplanes in free air (out of ground effect). Ground effect decreases as the distance from the ground increases and is generally negligible above a height equivalent to the wing span of the airplane (which is about 100 feet for the G650). [The figure] depicts the changes in the airplane's lift and stall AOA due to ground effect.

Source: NTSB AAR-12/03, ¶2.3

• During an October 7, 2010, meeting of the Gulfstream flight test safety review board (SRB), an estimate of the reduction, or decrement, from the free-air stall AOA to the in-ground-effect stall AOA was presented as 2°. This 2° decrement (which was previously provided to Gulfstream's flight test engineering department by the company's flight sciences department) was based on G650 low-speed wind tunnel testing. A 2° decrement was also used during the GIV and other Gulfstream programs. After the accident, a G650 aerodynamicist indicated that the decrement was a generally accepted and agreed-on value that could not be further refined during flight tests because of the expectation that the airplane would always be operated below the stall AOA near the ground.
• During a March 24, 2011, meeting to discuss Roswell II takeoff performance testing, FTE1 indicated that he had revised the decrement from the free-air to in-ground-effect stall AOA to about 1.6°.
• In addition to the revised decrement for the in-ground-effect stall AOA, the stick shaker activation threshold had been changed (starting with flight 125 on March 7, 2011) from 85 to 90 percent of normalized AOA, which reduced the margin for stall protection. The Gulfstream chief flight test engineer stated that he and FTE1 made this change to allow predicted takeoff speeds to be achieved without stick shaker activations that would invalidate tests.
• For the accident flight, the free-air stall AOA was 14.7°, and the 1.6° decrement for the in-ground-effect stall AOA resulted in a predicted in-ground-effect stall AOA of 13.1°. Thus, the stick shaker AOA set to 90 percent of normalized AOA (equivalent to an actual AOA of 12.3°) provided a 0.8° margin to the in-ground-effect stall AOA assumed at the time. However, the stick shaker (and the PLI) did not provide any warning before the actual stall on the accident flight, which occurred at an AOA of about 11.2°. The NTSB's aircraft performance study for this accident found that the flight test data that Gulfstream had collected during previous G650 field performance takeoffs, particularly the data from the flight 88 and flight 132 uncommanded roll events, were sufficient to quantify the changes in aerodynamic lift and the actual reduction in the stall AOA because of ground effect. Thus, Gulfstream should have been able to accurately predict the G650 in-ground-effect stall AOA before the accident flight.

Source: NTSB AAR-12/03, ¶2.3

• A takeoff technique needs to be well defined for the technique to be “consistently executed in service by crews of average skill.”
• However, on the basis of comments from Gulfstream flight test pilots and the concerns expressed by FTE1, the takeoff technique that was initially developed during flight 111 and attempted during flight 132 would not likely have been certifiable. Pilots would not have been able to consistently execute this technique during service without “exceptional piloting skill,” as required by sections 25.143 and 25.105, and by flight crews “of average skill,” as required by section 25.101. Also, the takeoff technique developed during flight 111 that came the closest to achieving the target V₂ speeds involved an overshoot of about 3 knots (which was above the required ± 2-knot tolerance for the target V₂ speed), and the technique had not been successfully demonstrated on any takeoff performance flight test after flight 111. The NTSB concludes that, before the accident flight, Gulfstream had sufficient information from previous flight tests to determine that the target V₂ speeds could not be achieved with a certifiable takeoff rotation technique and that the V₂ speeds needed to be increased.

Source: NTSB AAR-12/03, ¶2.4.1

The National Transportation Safety Board determines that the probable cause of this accident was an aerodynamic stall and subsequent uncommanded roll during a one-engine-inoperative takeoff flight test, which were the result of (1) Gulfstream's failure to properly develop and validate takeoff speeds for the flight tests and recognize and correct the takeoff safety speed (V₂) error during previous G650 flight tests, (2) the G650 flight test team's persistent and increasingly aggressive attempts to achieve V₂ speeds that were erroneously low, and (3) Gulfstream's inadequate investigation of previous G650 uncommanded roll events, which indicated that the company's estimated stall angle of attack while the airplane was in ground effect was too high. Contributing to the accident was Gulfstream's failure to effectively manage the G650 flight test program by pursuing an aggressive program schedule without ensuring that the roles and responsibilities of team members had been appropriately defined and implemented, engineering processes had received sufficient technical planning and oversight, potential hazards had been fully identified, and appropriate risk controls had been implemented and were functioning as intended.

Source: NTSB AAR-12/03, ¶4.2

• After the accident, Gulfstream revised its takeoff airspeed development and testing methods. According to Gulfstream, airspeeds are now generated using a desktop computer simulation that represents the dynamics of the maneuver, the aerodynamics of the airplane in and out of ground effect, and the “control effectiveness” of the airplane. Gulfstream indicated that the desktop computer simulation was developed to more precisely model the takeoff maneuver and predict V₂ speeds that ensure, among other things, (1) an achievable and repeatable initial pitch attitude at rotation and (2) a suitable margin between the operating AOA and the stall AOA during in-ground-effect operations and the climb to the obstacle clearance height. The revised takeoff airspeed development and testing methods were used to determine updated G650 takeoff speeds for the accident flight test conditions, as shown in table 4. As indicated in the table, the updated V₂ speed was 15 knots (11 percent) faster than the V₂ speed provided to the accident test team.
• In addition, Gulfstream indicated that it implemented actions (besides increased takeoff speeds) to ensure the safe operation of the G650 in achieving the 6,000-foot takeoff performance guarantee. First, Gulfstream modified the takeoff technique to reduce the necessary column pull forces so that a pilot could reliably attain the initial pitch attitude within 3 to 4 seconds. Second, Gulfstream changed the stall warning system (for both flight test and production airplanes) so that the in-ground-effect stall AOA would be continuously computed in the flight control computer using the height above the ground and Mach number. Gulfstream believed that this change would provide pilots with increased situational awareness and would help ensure a timely reaction if an over-rotation were to occur. Last, the maximum takeoff thrust was increased by 5 percent to minimize the performance penalties associated with higher takeoff speeds.

Source: NTSB AAR-12/03, ¶2.4.2