Pinnacle Airlines Flight 3701

A repaired aircraft needed in Minneapolis

Pinnacle Airlines Flight 3701 left Little Rock National Airport at about 21:21 on 14 October 2004. The Bombardier CL-600-2B19, a CRJ-200 registered N8396A, was being moved to Minneapolis-St. Paul for its next assignment. Only the captain and first officer were aboard. Pinnacle operated as Northwest Airlink, and this positioning flight was conducted under Part 91 with an instrument flight rules flight plan.

Earlier that day, a different crew had rejected a takeoff in the aircraft after a warning about a right engine bleed air duct. The passengers and crew transferred to another aircraft. Mechanics sent from Memphis found chafing damage on a sensing loop in the engine pylon, replaced the loop and ran the engine for about 30 minutes. N8396A was then released for service. Its takeoff weight for the positioning flight was 39,336 pounds, below the 53,000-pound maximum, and its centre of gravity was within limits.

The company flight plan specified a cruise altitude of 33,000 feet. Both engines were operating normally before the later upset. NTSB found the aircraft properly certified, equipped and maintained, and found weather, fatigue and hypoxia were not factors in the accident.

Aggressive manoeuvres during the climb

The first unusual pitch manoeuvre came about five seconds after takeoff, roughly 190 feet above the ground. The pilots moved the control column aft, producing a 22-degree nose-up pitch and a load of 1.8 G. The stickshaker and stickpusher activated, followed by a full nose-down control column movement. These systems warned of an approaching stall and pushed the control columns forward to reduce the angle of attack.

Two further pitch-up manoeuvres followed at about 15,000 and 24,600 feet. The first briefly produced a climb rate of 10,000 feet per minute and a load of 2.3 G. The crew also made alternating rudder inputs. NTSB found the manoeuvres had no operational purpose and likely took the aircraft beyond its certificated flight envelope.

The pilots changed seats during the ascent. Investigators established their positions by correlating recorded voices and radio transmissions with microphone key data. The captain occupied the right seat and the first officer the left through much of the flight, including the upset. At about 21:35:36, the captain requested 41,000 feet, the CRJ's maximum operating altitude. The controller cleared the climb at about 21:36:13.

Reaching 41,000 feet with too little speed

During the final climb from 37,000 to 41,000 feet, the autopilot used vertical speed mode with a commanded climb rate of 500 feet per minute. Airspeed fell from 203 knots, or Mach 0.63, to 163 knots, or Mach 0.57, as the aircraft levelled at about 21:52:08. The required climb schedule was 250 knots until Mach 0.70 became the controlling speed.

At the lower speed, the aircraft entered a region where increased drag exceeded the thrust available to accelerate it. NTSB found that selecting vertical speed mode allowed the climb to continue at the expense of airspeed and brought the aircraft to 41,000 feet in a critically low energy state. The flight crew had failed to understand how airspeed affected the aircraft's performance at that altitude.

The cockpit recording captured both pilots expressing excitement about reaching 41,000 feet. The captain left his seat to fetch a drink. When a controller remarked on the unusually high altitude, the captain explained that they had no passengers and had chosen to go there for enjoyment. Within minutes, the pilots recognized that the aircraft was losing performance. At about 21:54:32, the captain requested a lower altitude.

The stall and the loss of both engines

While the controller coordinated the descent, airspeed continued to deteriorate. At about 21:54:36, the stickshaker activated at approximately 150 knots. Further stickshaker and stickpusher activations followed. Each time the stickpusher moved the columns forward, the pilots pulled aft again. Their increasing aft inputs opposed the system's attempt to reduce the angle of attack.

At about 21:54:57, another stickshaker and stickpusher activation occurred. The aircraft pitched up to 29 degrees and entered an aerodynamic stall. It then rolled as far as 82 degrees left wing down and pitched as far as 32 degrees nose down. Airflow into both engine inlets was disrupted, producing engine stalls and a complete loss of power. Both engines flamed out.

At 21:55:06, the captain declared an emergency and asked the controller to stand by. The crew recovered control near 34,000 feet. Fuel flow indications had fallen to zero, and engine speeds continued to decrease. At 21:55:20, loss of normal electrical power interrupted the flight data recorder. The cockpit voice recorder continued operating.

NTSB found the stickshaker and stickpusher had functioned normally. A software error placed the displayed low speed cue about 10 knots too low. The crew had been taught to respond to the stickshaker regardless of the displayed cue, but their control inputs intensified the stall.

The cores stopped during the descent

The pilots recognized the loss of both engines shortly after the upset. They began the double engine failure checklist at about 21:56:42. Its initial memory items included establishing a target airspeed of 240 knots. That speed was important because airflow through the engines kept their cores rotating and preserved the possibility of restarting them.

The crew failed to achieve and maintain 240 knots. Before the aircraft descended through 28,000 feet, both cores had stopped. When the flight data recorder resumed at about 21:59:16, the auxiliary power unit was supplying electricity. The aircraft was at 29,200 feet and 178 knots, and both engine core speed indications were zero.

The checklist called for a windmill restart between 21,000 and 13,000 feet, with at least 300 knots to drive the cores fast enough for a relight. The captain twice instructed the first officer to increase speed. The resulting nose-down inputs were too small and too brief. The aircraft reached only 236 knots before slowing again.

NTSB found the captain failed to ensure the required speed was reached or take control and accelerate the aircraft himself. The manuals described 240 knots as a target but did not explain that maintaining core rotation was essential to a successful restart.

Repeated restart attempts and delayed disclosure

After the windmill attempt failed, the crew descended towards 13,000 feet to try restarting with bleed air from the auxiliary power unit. Four APU-assisted attempts followed between about 22:07:04 and 22:12:07, two on each engine. Both core speed indications remained at zero. None of the attempts restored engine power.

Controllers still lacked a clear description of the emergency. At about 22:03:09, the captain had referred to an engine failure and a descent to start the other engine. The controller replied that he understood the aircraft was in controlled flight on one engine. The captain allowed that understanding to stand. At about 22:06:40, another controller asked whether the crew wanted to land. The captain asked him to wait while they tried a restart.

The crew returned to their normal seating positions near the end of the descent. At 22:09:06, the first officer informed the controller that neither engine was running and requested the nearest airport. Almost 14 minutes had elapsed since the pilots recognized the double engine failure.

NTSB found the captain intentionally gave controllers an inaccurate description of the situation. The delayed disclosure reduced the assistance available for identifying a reachable airport and preparing for an emergency landing.

Landing options disappeared as altitude fell

Investigators calculated the aircraft's glide capability using a reference speed of 170 knots and including drag from the deployed emergency generator and flight spoilers. At 30,000 feet, six airports were within the calculated best glide range, including Jefferson City. At 20,000 feet, five remained within range. By 10,000 feet, Lee C. Fine Memorial Airport was the only airport within range, but the aircraft had already overflown it.

Jefferson City was just outside the calculated best glide range at 10,000 feet. It lay essentially ahead of the aircraft, while returning to Lee C. Fine would have required a 160-degree turn that consumed altitude, speed and distance. NTSB found the controller correctly directed the aircraft towards Jefferson City from that position. The pilots had descended from about 35,000 to 10,000 feet before disclosing the full emergency.

The controller provided wind, approach and localizer information for runway 30. After descending out of cloud near 5,000 feet, the crew sought further directions to the airport. At about 22:14:02, the first officer reported seeing the runway. The captain soon realized they could not reach it and asked about a road. Radar showed a left turn towards a straight, illuminated section of highway.

The crew kept the landing gear retracted. Ground proximity alerts sounded in the final seconds. At about 22:15:06, the cockpit recorder captured an impact sound and stopped shortly afterwards. The aircraft struck trees and the ground in a residential area about 2.5 miles south of the airport. Both pilots died, and impact forces and fire destroyed the aircraft. No one on the ground was injured.

Core lock, training gaps and the final findings

Engine examinations and tests of the starting system found no mechanical failure that explained the lack of core rotation. Investigators examined earlier CF34 engine restart failures during production acceptance testing. General Electric and Bombardier had attributed those failures to contact at a high pressure turbine air seal, a condition they called core lock.

NTSB concluded that the abrupt high power, high altitude flameouts caused severe thermal changes in the accident engines. Seal teeth contacted stationary material as clearances changed. Once the cores stopped, binding prevented them from turning during the windmill and APU-assisted restart attempts. Cooling subsequently restored the clearances, leaving no physical evidence of the temporary interference inside the engines.

The final report, adopted on 9 January 2007, identified unprofessional behaviour, departures from procedures and poor airmanship, inadequate training, delayed emergency landing preparations, and improper checklist management in the probable causes. Core lock and inadequate explanation of the need to maintain core rotation in flight manuals were contributing factors.

Pinnacle had not practised high altitude climbs or double engine failures in simulator training before the accident. Afterwards, it added high altitude demonstrations and a double engine failure scenario, and revised its checklist to emphasize minimum speeds. NTSB recommended improved high altitude training, simulator practice in high altitude stall recovery, clearer double engine failure guidance and greater oversight of nonrevenue flights. Recommendations issued in November 2006 also called for engine restart testing and explicit core lock guidance in flight manuals.