Photo: Pinnacle Airlines 3701 wreckage, from NTSB.
This crew was nominated for a Darwin Award in 2004 and rightfully so. Their lack of professionalism deserves no less. But the incident also served to once again underline the fact many pilots do not understand the cause and remedy of an aerodynamic stall.
- Aerodynamics — Not everyone who drives a car understands what is going on between the gas pedal and the tires but every pilot should know the difference between the relative wind and the chordline of a wing. See Basic Aerodynamics / Lift for more about this.
- Stall Recovery — When a highly cambered wing is stalled there is only one way to get it out of the stall and years of simulator practice trying to hold every inch of altitude needs to be unlearned. See Basic Aerodynamics / Angle of Attack for more about AOA. See Abnormal Procedures / Stall Recovery for more about this.
Much has been made in this mishap about the possibility of both engines suffering a core lock, preventing a relight. A core lock occurs when an engine is shutdown at altitude and the cold temperature causes various parts of the engine to contract at different rates, preventing the compressor and turbine blades from spinning. It may or may not have been the case in this mishap but that misses the point. If there was a core lock, the pilots caused it by allowing their airspeed to get unacceptably low.
But there is one more factor that I think everyone has missed. When pilots who have never stalled a large airplane are trained exclusively in simulators, having an abnormal situation outside of the box can come as a shock. When in a panic, our brains often jump to an instinctual reaction (pull back on the yoke) and fail to think things through. The best way to deal with this is to desensitize oneself against the fear in a real airplane. More about this: Pilot Psychology / Panic.
This mishap does provide another lesson: if you are gliding an airplane to lower altitude in hopes of relighting the engines, keep an eye on the spool rotation. In a G450, for example, you cannot attempt a relight until 25,000 feet. It would be wise to keep both spools rotating at some speed during the descent.
What follows are quotes from the relevant regulatory documents, listed below, as well as my comments in blue.
- Date: 14 Oct 2004
- Time: 22:15
- Type: Canadair CL-600-2B19 Regional Jet CRJ-200LR
- Operator: Pinnacle Airlines
- Registration: N8396A
- Fatalities: 2 of 2 crew
- Aircraft Fate: Destroyed
- Phase: Approach
- Airports: (Departure) Little Rock National Airport AR (LIT/KLIT), United States of America; (Destination) Minneapolis-St. Paul International Airport MN (MSP/KMSP), United States of America
[NTSB AAR-07/01, ¶1.1.]
- Flight 3701 departed LIT about 2121. The flight plan indicated that the company-planned cruise altitude was 33,000 feet. About 5 seconds after takeoff, when the airplane was at an altitude of about 450 feet mean sea level (msl) (about 190 feet above ground level), the first of three separate pitch-up maneuvers during the ascent occurred when the flight crew moved the control column to 8° airplane nose up (ANU), causing the airplane's pitch angle to increase to 22° and resulting in a vertical load of 1.8 Gs. The rate of climb during this pitch-up maneuver was 3,000 feet per minute (fpm). Immediately afterward, the flight data recorder (FDR) recorded stickshaker and stickpusher activations, a full airplane-nose-down (AND) control column deflection, a decrease in pitch angle, and a drop in vertical load to 0.6 G.
- About 2125:55, when the airplane was at an altitude of about 14,000 feet, the flight crew engaged the autopilot. The air traffic control (ATC) transcript and FDR data showed that the flight crewmembers changed seats in the cockpit during this time, but the ATC transcript did not indicate the reason for the seat change. About 2127:15, when the airplane was at an altitude of about 15,000 feet, the flight crew disengaged the autopilot.
- About 2127:17, when the airplane was in level flight at an altitude of 15,000 feet, the second pitch-up maneuver began when the flight crew moved the control column to 3.8° ANU, causing the airplane's pitch angle to increase to 17° and resulting in a vertical load of 2.3 Gs. The rate of climb during this pitch-up maneuver reached 10,000 fpm briefly. Between about 2128:40 and about 2128:43, the flight crew made a left rudder input of 4.2°, a right rudder input of 6.0°, and a left rudder input of 0.4°, resulting in lateral loads of -0.16 G, 0.34 G, and -0.18 G, respectively. About 17 seconds later, the flight crew made a right rudder input of 7.7°. About 2132:40, when the airplane was in level flight at an altitude of 24,600 feet, the third pitch-up maneuver began when the flight crew moved the control column to 4° ANU, which increased the airplane's pitch angle to more than 10° and resulted in a vertical load of 1.87 Gs. The rate of climb during this pitch-up maneuver reached 9,000 fpm briefly.
- The ATC transcript showed that the captain requested a climb to 41,000 feet, which is the Canadair regional jet (CRJ) maximum operating altitude, about 2135:3611 and received clearance to climb to that altitude about 2136:13.12 The cockpit voice recorder (CVR) recording began about 2144:44 with the captain and the first officer discussing the climb to 41,000 feet. About 2148:44, the first officer stated, "man we can do it. Forty one it." About 2151:51, the first officer stated, "there's four one oh my man." About 2152:04, the CVR recorded the first officer laughing as he stated, "this is ... great." FDR data showed that, about 2152:08, the airplane was in level flight at 41,000 feet. FDR data also showed that the airplane climbed from 37,000 to 41,000 feet at an airspeed that decreased from 203 knots/0.63 Mach at the start of the climb to 163 knots/0.57 Mach as the airplane leveled off. The FDR data further showed that the autopilot vertical speed mode was engaged during the climb with a commanded vertical speed of 500 fpm and that the airplane's angle of attack (AOA) at 41,000 feet was initially 5.7°.
- About 2152:22, the CVR recorded the captain asking the first officer whether he wanted something to drink and then the first officer responding that he wanted a soda. CVR evidence indicated that the captain left his seat shortly afterward to get the drink.
- About 2153:28, the CVR recorded the captain stating, "look how high we are." About 2153:42, a controller at the Kansas City Air Route Traffic Control Center (ARTCC) asked the pilots whether they were flying a CRJ-200. The captain confirmed this information, and the controller stated, "I've never seen you guys up at forty one there." About 2153:51, the captain replied, "we don't have any passengers on board so we decided to have a little fun and come on up here." About 2153:59, the captain added, "this is actually our service ceiling."
- About 2154:07, the captain told the first officer, "we're losing here. We're gonna be ... coming down in a second here." About 3 seconds later, the captain stated, "this thing ain't gonna ... hold altitude. Is it?" The first officer responded, "it can't man. We ... (cruised/greased) up here but it won't stay." About 2154:19, the captain stated, "yeah that's funny we got up here it won't stay up here."
- About 2154:32, the captain contacted the controller and stated, "it looks like we're not even going to be able to stay up here ... look for maybe ... three nine oh or three seven." About 2154:36, the FDR recorded the activation of the stickshaker. FDR data showed that, at that point, the airplane's airspeed had decreased to 150 knots, and its AOA was about 7.5°.
- The FDR recorded activations of the stickshaker and the stickpusher three times between 2154:45 and 2154:54. FDR data showed that, after the second activation of the stickshaker and stickpusher, the No. 1 (left) and No. 2 (right) engines' N1 (fan speed) and fuel flow indications began decreasing. FDR data also showed that, at the time of the second stickpusher activation, the airplane's AOA had increased to 12° and that, after the stickpusher activated for the third time, the pitch angle decreased from 7° to -20°.
- About 2154:57, the FDR recorded the fifth activation of the stickshaker and the fourth activation of the stickpusher. Even with the stickpusher's activation, the motion of the airplane continued to increase its AOA to the maximum measurable value of 27°. The pitch angle increased to 29°, and the airplane entered an aerodynamic stall. Afterward, a left rolling motion began, which eventually reached 82° left wing down, the airplane's pitch angle decreased to -32°, and both engines flamed out. About 2155:06, the captain stated to the controller, "declaring emergency. Stand by." FDR data showed that, during the next 14 seconds, the flight crew made several control column, control wheel, and rudder inputs and recovered the airplane from the upset at an altitude of 34,000 feet. During the recovery, the CVR recorded a sound similar to decreasing engine rpm, and FDR data showed that the No. 1 and No. 2 engines' N1 indications continued to decrease and that the engines' fuel flow indications were at zero.
- About 2155:14, the controller told the pilots to descend and maintain an altitude of 24,000 feet; about 5 seconds later, the captain acknowledged the assigned altitude. About 2155:20, the FDR stopped recording because normal a.c. power to the airplane was lost. (The CVR had a different source of power and continued to record.) The last reliable N2 (core speed) recorded by the FDR before it stopped operating was 46 percent for the No. 1 engine and 51 percent for the No. 2 engine.
- About 2155:23, one pilot stated to the other, "we don't have any engines," and, about 10 seconds later, the captain stated, "double engine failure." About 2156:42, the flight crew began performing the double engine failure checklist, which required pilots to maintain 240 knots until they were ready to initiate the double engine failure procedure. The checklist indicated that, if the airplane were at or below 21,000 feet and above 13,000 feet, pilots should relight the engines using the windmill restart procedure, which required an airspeed of at least 300 knots. The procedure indicated that an altitude loss of 5,000 feet could be expected when accelerating from 240 to 300 knots.
- The FDR resumed operation about 2159:16. FDR data showed that the auxiliary power unit (APU) was supplying electrical power to the airplane, both engines' N1 indications continued to decrease, and both engines' N2 indications were at zero. FDR data also showed that the airplane's altitude was 29,200 feet and that its airspeed was 178 knots.
- About 2200:38, the captain told the first officer to increase the airspeed to above 300 knots, and the first officer acknowledged this instruction. FDR data showed that the airplane pitched down to -4.4° and accelerated to an airspeed of 200 knots but that, during the next 25 seconds, the airplane pitched up to 0° while its airspeed remained at 200 knots. About 1 minute later, the captain again told the first officer to increase the airspeed to 300 knots. FDR data showed that the airplane pitched down to -7.5° and accelerated to an airspeed of 236 knots (the maximum airspeed achieved during the windmill restart attempt) but that, during the next 22 seconds, the airspeed decreased to 200 knots.
- About 2201:51, the captain stated, "we're not getting any N two at all. So we're gonna have ... to go to ... thirteen thousand feet. We're going to use the APU bleed air [restart] procedure." Shortly afterward, the captain resumed the double engine failure checklist, which indicated that pilots were to maintain between 170 and 190 knots until they were ready to initiate the APU bleed air restart procedure.
- About 2203:09, the controller asked the flight crew about the nature of the emergency. The captain responded, "we had an engine failure up there ... so we're gonna descend down now to start our other engine." About 2203:30, the captain stated, "we're descending down to thirteen thousand to start this other engine," and the controller replied, "understand controlled flight on a single engine right now." FDR data showed that, during the next several minutes, four APU-assisted engine restarts were attempted, but the N2 speed for both engines remained at zero throughout the restarts. About 2206:40, the controller asked the flight crewmembers whether they wanted to land; the captain replied, "just stand by right now we're gonna start this other engine and see ... if everything's okay." About 2206:54, the controller informed the flight crew that JEF was up ahead, and the captain acknowledged this information.
- About 2208:17, the CVR recorded the captain stating, "switch." About 2209:02, the captain instructed the first officer to tell the controller that they needed "to get direct to [an] airport neither engine's started right now." The first officer informed the controller for the first time of the double engine failure, and the controller then asked the pilots if they wanted to go direct to JEF. The captain stated, "any airport and closest airport," and the first officer told the controller, "closest ... airport. We're descending fifteen hundred feet per minute we have ... nine thousand five hundred feet left."
- Between about 2210:21 and about 2211:20, the controller provided information about the winds, the approach frequency, and the localizer frequency for an instrument landing system (ILS) landing to runway 30 at JEF. About 2212:24, the first officer asked the controller where to look for the airport, and the controller provided position, distance, and heading information. About 1 minute later, the controller provided additional location information for JEF. About 2213:37, the captain asked the first officer whether the airplane was aligned with the runway, and the first officer notified the controller that he did not see the runway. The controller provided further directional information, and the first officer told the controller that he thought he had the approach end of the runway in sight. The controller received no further transmissions from the flight crew.
- About 2214:02, the first officer told the captain that he had the runway in sight. The captain questioned the first officer about the location of the runway and then stated, about 2214:17, "we're not gonna make this." About 2214:38, the captain stated, "is there a road? We're not gonna make this runway." Radar data showed that the airplane then turned left and headed toward a straight and lit section of highway. About 2214:46, the captain stated, "let's keep the gear up ... I don't want to go into houses here." About 2214:53, the final radar return was received when the airplane was about 0.58 nautical mile southeast of the crash site and at an altitude of 930 feet.
- About 2214:54, 2214:58, and 2215:00, the CVR recorded the enhanced ground proximity warning system (GPWS) alerts "too low gear," "too low terrain," and "pull up," respectively. About 2215:03, the CVR recorded the captain stating, "we're gonna hit houses," and, about 2 seconds later, the enhanced GPWS alert "pull up." About 2215:06, the CVR recorded a sound similar to an impact and stopped recording about 1 second afterward.
[NTSB AAR-07/01, ¶1.27.1] Examination of the engines at the accident site found that neither engine exhibited classic rotational damage or ingestion evidence. The engines were disassembled and inspected at GE's manufacturing facility in Lynn, Massachusetts. The engine core rotors were rotated, and neither core was found seized. Teardown inspections found no mechanical failures or evidence of seizure in either engine. A materials investigation of the high pressure turbine seal hardware found no abnormal rotational marks.
[NTSB AAR-07/01, ¶1.18.2]
- During the investigation of this accident, the Safety Board learned that GE CF34-1 and CF34-3 engines had a history of failing to rotate during in-flight restart attempts on airplanes undergoing production acceptance flight testing at Bombardier. The manufacturers referred to this condition as "core lock." Bombardier first identified this problem in 1983 during Challenger certification tests, and GE attributed the problem to interference contact at an air seal in the high pressure turbine.
- The CF34 high pressure turbine air seals are designed to control cooling and balance airflow. The seals include teeth on the rotating components that grind operating grooves into abradable surfaces on the stationary components. The efficiency of these seals significantly affects engine performance, so the seals are designed to operate with minimal clearances.
- Bombardier added a procedure that screened for core lock to the production acceptance flight tests for its airplanes powered by CF34-1 and CF34-3 engines. At the time of the accident, this screening procedure was as follows:
Engines that are found to be core locked are reworked using an in-flight "grind-in" procedure that was designed to remove seal material at the interference location. Engines that undergo grind-in rework are then rescreened for core lock. The grind-in procedure is as follows:
- Climb to 31,000 feet.
- Retard the test engine throttle to idle and stabilize for 5 minutes.
- Shut down the test engine.
- Descend at 190 knots.
- Slow the aircraft until N2 is reduced to 0 percent.
- At 8 1/2 minutes from shutdown, push over to 320 knots.
- If N2 is 0 rpm at 21,000 feet, the engine is declared to be core locked.
As testimony during the Safety Board's June 2005 public hearing on the Pinnacle Airlines accident indicated, neither Bombardier nor GE considered core lock to be a safety-of-flight issue. The manufacturers claimed that engines that passed the screening procedure, with or without grind-in rework, would not core lock as long as the 240-knot airspeed was maintained.
Bombardier's core lock screening procedure requires a cool-down period before engine shutdown to stabilize internal temperatures and clearances. However, this procedure does not produce the more severe thermal distress associated with the high power, high altitude flameouts that were experienced during the accident flight. As stated in the Safety Board's November 20, 2006, safety recommendation letter to the FAA, the successful demonstration of Bombardier's flight test procedure might not ensure that an engine will not experience core lock if the core is allowed to stop rotating after a high power, high altitude flameout. In its letter, the Board noted that the No. 1 accident engine had successfully passed the screening procedure during initial production acceptance testing. The Board further stated that the successful demonstration of Bombardier's flight test procedure might not ensure that slowing the airplane to an airspeed of 170 to 190 knots is sufficient to maintain core rotation during an attempted APU-assisted restart.
- ATS cross-bleed start.
- Ascend to 31,000 feet.
- Repeat core lock screening procedure but descend at an airspeed of about 240 knots to establish 4 percent N2.
- Maintain 4 percent N2 for at least 8 1/2 minutes.
- Confirm that no core lock exists by repeating screening procedure.
[NTSB AAR-07/01, ¶2.2.1]
- The flight plan for the accident flight indicated that the planned cruise altitude was 33,000 feet. However, the ATC and CVR transcripts showed the flight crew's desire to climb to an altitude of 41,000 feet, which is the CRJ maximum operating altitude. During postaccident interviews, Pinnacle Airlines pilots stated that some pilots had expressed curiosity about operating the airplane at 41,000 feet and that a "[flight level] 410 club" existed at the airline.
- Pinnacle Airlines required that a climb to 41,000 feet be conducted at an airspeed of 250 knots (0.70 Mach).98 However, FDR data showed that the flight crew conducted the climb from 37,000 to 41,000 feet at an airspeed that decreased from 203 knots (0.63 Mach) at the start of the climb to 163 knots (0.57 Mach) as the airplane leveled off at 41,000 feet. FDR data also showed that the flight crew conducted the climb from 37,000 to 41,000 feet with the autopilot vertical speed mode engaged and a commanded vertical speed of 500 fpm. The airplane could not sustain the required airspeed while climbing at 500 fpm, which resulted in the 40-knot loss of airspeed, and, once level at 41,000 feet, the airplane was operating in a "region of reversed command" in which available thrust was not sufficient to increase airspeed.
Given the background of both pilots, it is unlikely they were familiar with flight characteristics in the region of reversed command or how to return to the region of normal command. More about this: Basic Aerodynamics / Region of Reversed Command.
[NTSB AAR-07/01, ¶2.2.2]
- Even though the pilots recognized that the airplane's performance was deteriorating, neither pilot appeared to respond to this situation with urgency. The pilots contacted the controller about 2154:32 and requested a lower altitude; however, during the time that the controller was coordinating the descent, the airplane's airspeed further deteriorated to Mach 0.53 (150 knots), and the stickshaker activated for the first of five times. Afterward, the stickshaker activated four times, and the stickpusher activated four times, pushing the control column forward automatically. The flight crew responded to the stickpusher each time by pulling back on the control column. These control column inputs caused the airplane's pitch angle to increase to a maximum ANU value of 29° about 2154:59, and then the airplane entered an aerodynamic stall.
The pilots may have been oblivious to the danger because they did not understand the relationship of indicated airspeed to altitude when it comes to stall speed. The wing's stall angle of attack does not change, but the indicated airspeed at which the wing stalls goes up with altitude.
- According to Pinnacle Airlines, its flight crewmembers were taught that the top of the low speed cue indicated the airspeed at which the stickshaker would initially activate. However, the flight crews were also trained to respond to stickshaker activation, regardless of whether the airspeed was above the low speed cue, because the stickshaker is the primary warning to pilots of an impending stall. In addition, when an airplane's speed is about 10 knots from the top of the low speed cue, especially when operating at an altitude of 41,000 feet, pilots should realize that the airplane is dangerously close to an impending stall. When the stickshaker activated during the accident flight, the pilots should have immediately attempted to accelerate the airplane to a safe airspeed by descending; sufficient reserve power was not available from the engines while the airplane was at 41,000 feet. Thus, the accident pilots should have responded appropriately to the stickshaker and used the low speed cue only as a secondary indication showing that the airspeed was dangerously slow.
- Even though the flight crew was eventually able to recover the airplane from the upset event, the flight control inputs made during the upset were opposite of what was required to recover the airplane. FDR data showed that the flight crew moved the control column aft after the first stickpusher activation and that the crew moved the control column aft with increasing magnitude after the next three activations of the stickpusher.
- A reason that might explain why the pilots made these flight control inputs, which exacerbated the upset event, was that Pinnacle Airlines' stall recognition and recovery simulator training focused on recovery with a minimum loss of altitude (which is common throughout the aviation industry). As a result, the pilots were trained to apply power to restore the energy state of the airplane and to maintain altitude.
[NTSB AAR-07/01, ¶3.1]
- The pilots' aggressive pitch-up and yaw maneuvers during the ascent and their decision to operate the airplane at its maximum operating altitude (41,000 feet) were made for personal and not operational reasons.
- The flight crew's inappropriate use of the vertical speed mode during the climb was a misuse of automation that allowed the airplane to reach 41,000 feet in a critically low energy state.
- The improper airspeed during the climb demonstrated that the pilots did not understand how airspeed affects airplane performance and did not realize the importance of conducting the climb according to the published climb capability charts.
[NTSB AAR-07/01, ¶2.2.3]
- The double engine failure checklist indicated that, between the altitudes of 21,000 and 13,000 feet, the windmill restart procedure should be used to relight the engines. This procedure required that the pitch attitude of the airplane be reduced to and maintained at -8° to accelerate the airplane to an airspeed of 300 knots or greater. However, the pitch inputs made by the flight crew were not of sufficient magnitude and were not sustained. As a result, the crew did not achieve the 300-knot or greater airspeed required for the procedure; FDR data showed that the highest airspeed attained during the restart attempt was 236 knots and that the engines' N2 (core rotation) indications remained at zero. The Safety Board concludes that the captain did not take the necessary steps to ensure that the first officer achieved the 300-knot or greater airspeed required for the windmill engine restart procedure and then did not demonstrate command authority by taking control of the airplane and accelerating it to at least 300 knots. The Safety Board further concludes that the first officer's limited experience in the airplane might have contributed to the failed windmill restart attempt because he might have been reluctant to command the degree of nose-down attitude that was required to increase the airplane's airspeed to 300 knots.
- Because the flight crewmembers were unsuccessful in their attempt to relight the engines using the windmill procedure, they elected to descend the airplane to an altitude of 13,000 feet so that they could attempt to restart the engines with the APU. Once the airplane descended to an altitude of 13,000 feet, the flight crew attempted four APU-assisted engine restarts (two attempts per engine) during a 5-minute period (2207:04 to 2212:07) and between the altitudes of about 12,900 and about 5,000 feet. FDR data showed that the N2 indications for both engines remained at zero during the restart attempts. The Safety Board concludes that, despite their four APU-assisted engine restart attempts, the pilots were unable to restart the engines because their cores had locked. Without core rotation, recovery from the double engine failure was not possible.
[NTSB AAR-07/01, ¶3.1]
- The upset event exposed both engines to inlet airflow disruption conditions that led to engine stalls and a complete loss of engine power.
- The pilots' lack of exposure to high altitude stall recovery techniques contributed to their inappropriate flight control inputs during the upset event.
- The captain did not take the necessary steps to ensure that the first officer achieved the 300-knot or greater airspeed required for the windmill engine restart procedure and then did not demonstrate command authority by taking control of the airplane and accelerating it to at least 300 knots.
, ¶3.2] The National Transportation Safety Board determines that the probable causes of this accident were (1) the pilots' unprofessional behavior, deviation from standard operating procedures, and poor airmanship, which resulted in an in-flight emergency from which they were unable to recover, in part because of the pilots' inadequate training; (2) the pilots' failure to prepare for an emergency landing in a timely manner, including communicating with air traffic controllers immediately after the emergency about the loss of both engines and the availability of landing sites; and (3) the pilots' improper management of the double engine failure checklist, which allowed the engine cores to stop rotating and resulted in the core lock engine condition. Contributing to this accident were (1) the core lock engine condition, which prevented at least one engine from being restarted, and (2) the airplane flight manuals that did not communicate to pilots the importance of maintaining a minimum airspeed to keep the engine cores rotating.
Basic Aerodynamics / Low Speed Flight
Abnormal Procedures & Techniques / Aerodynamic Stall
Pilot Psychology / Complacency
Pilot Psychology / Panic
NTSB Aircraft Accident Report, AAR-0y/01, Crash of Pinnacle Airlines Flight 3701 Bombardier CL-600-2B19, N8396A, Jefferson City, Missouri October 14, 2004