American Eagle Flight 4184

A delayed return to Chicago

American Eagle flight 4184 was the second leg of the crew's day, returning from Indianapolis to Chicago O'Hare on 31 October 1994. Simmons Airlines operated the scheduled passenger service with N401AM, an Avions de Transport Regional model 72-212. The high-wing twin turboprop carried 64 passengers, two pilots and two flight attendants. The first officer was flying the aircraft. All 68 people aboard died when it crashed near Roselawn, Indiana.

The flight was scheduled to leave its gate at 14:10 and reach Chicago at 15:15. It left the gate at 14:14, but deteriorating weather at O'Hare led controllers to change the arrival traffic flow. Flight 4184 remained on the ground for 42 minutes before receiving its instrument flight clearance. The Indianapolis ground controller warned that the crew could expect some airborne holding. Takeoff clearance followed at 14:55:20.

Before departure, the crew received a combined flight release and weather package prepared by the company. It omitted AIRMET advisories and forecasts of light to moderate turbulence or icing along the intended route. Dispatchers could include AIRMETs at their discretion, and pilots could consult them at the departure station. Investigators could not determine whether these pilots had obtained that information separately.

The autopilot was engaged during the climb through 1,800 feet. The crew subsequently received clearance to climb to 16,000 feet. At 15:13, the aircraft began descending toward 10,000 feet, and the flight recorder showed airframe deicing operating during the descent. O'Hare's arrival controllers were preparing to hold inbound aircraft.

Thirty-nine minutes in the hold

At 15:18:07, the crew received clearance to the LUCIT intersection, with instructions to hold southeast on the airway. The expected onward clearance was initially 15:30, then moved to 15:45. The captain arranged ten nautical mile holding legs and confirmed right turns. At 15:24:39, he reported entering the hold. The crew also notified the company of the delay through the aircraft's messaging system.

The first circuit was flown at about 175 knots indicated with the flaps retracted. Airframe deicing had been switched off, and propeller speed was reduced to 77 percent. During the hold, the captain commented on the aircraft's nose-high attitude in turns. At about 15:33, the crew extended the flaps to 15 degrees. The recorded angle of attack decreased as the flaps moved.

The onward clearance time was extended again, to 16:00. At 15:41:07, a caution tone sounded, followed by activation of Level III airframe deicing. Propeller speed increased to 86 percent. A pilot remarked that ice was visible at about 15:48:41, and the first officer reported ice again at 15:55:42. The pilots remained in the hold.

The NTSB's analysis found two small drag increases consistent with ice accretion, one around the time the flaps were extended and another about six minutes before the upset. It concluded that the aircraft probably encountered large supercooled droplets intermittently. Ice formed on the upper wing behind the deice boots and ahead of the ailerons. The hold lasted about 39 minutes before control was lost.

Flap retraction and the first roll

At 15:56:16, the controller cleared flight 4184 to descend to 8,000 feet. He also advised that clearance onward would take about another ten minutes. The descent began at 15:56:51 from 10,000 feet. Engine power was reduced to flight idle, propeller speed remained at 86 percent and the autopilot stayed engaged in vertical speed and heading select modes.

At 15:57:21, the aircraft was descending in a right bank at 186 knots indicated when its flap overspeed warning sounded. The pilots retracted the flaps. To maintain the selected vertical speed as the flaps came up, the autopilot increased the pitch attitude. The angle of attack rose through approximately 5 degrees. Over the ice-contaminated right wing, airflow began separating near the aileron.

This altered the aerodynamic force acting around the aileron hinge. The force reversed and drove the ailerons rapidly toward a right roll. At 15:57:33, as the aircraft passed through about 9,130 feet, the recorded aileron deflection reached 13.43 degrees right wing down. The autopilot disconnected because the deflection rate exceeded its permitted control rate. The aircraft rolled rapidly to the right.

The NTSB found that the angle of attack at the onset of the aileron movement was below the stall warning threshold. As the angle of attack then decreased, the ailerons returned toward neutral and the roll stopped at about 77 degrees right bank. The pilots applied corrective control inputs, and the aircraft began rolling back toward level.

A second reversal and the steep descent

During the initial recovery, nose-up elevator inputs increased the angle of attack again. At 15:57:38, the aircraft was rolling left through about 59 degrees right bank when the ailerons again deflected rapidly toward a right roll. The second hinge moment reversal initiated another rapid rotation. In the NTSB's reconstruction, the aircraft rolled approximately one and a quarter turns over the next nine seconds.

Nose-up elevator and the existing pitch trim kept the angle of attack above the value at which airflow separated near the aileron. The ailerons remained directed toward a right roll and oscillated. When the nose-up elevator input decreased, the angle of attack fell, airflow reattached and aileron control returned. By then, the aircraft was descending through approximately 6,000 feet at about 24,000 feet per minute.

The pilots attempted to pull out of the dive. Pitch reached about 73 degrees nose down before beginning to recover, while airspeed continued rising. Near 1,700 feet, further elevator input increased the aerodynamic loading. A loud crunching sound was recorded. The NTSB concluded that excessive loads separated the outer sections of both wings and the horizontal tail close to the ground.

Flight 4184 struck a wet soybean field partially inverted, with its nose down and left wing low. The report gives the accident time as about 15:59 CST and the site elevation as about 675 feet. The aircraft was destroyed, and the impact was unsurvivable.

Large droplets beyond the certification envelope

The NTSB estimated that the holding area contained a mixture of supercooled cloud droplets and drizzle or rain drops. Some droplets were estimated to be as large as 2,000 microns in diameter. The ambient temperature at 10,000 feet was about minus 3 degrees Celsius, with the freezing level between 7,000 and 8,000 feet. Weather radar, upper-air observations, satellite imagery and pilot reports supported the analysis.

The ATR 72's icing certification used the conditions specified in Part 25, Appendix C. Those envelopes described droplet size, liquid water content and temperature, with a maximum mean effective droplet diameter of 50 microns. Freezing drizzle and freezing rain fell outside those limits. The accident aircraft encountered droplet sizes and water content exceeding the certification envelope.

Its pneumatic boots removed ice by repeatedly inflating and deflating chambers on the leading edges. The wing boots covered about seven percent of the upper wing chord. Ice accumulating farther aft lay beyond their working area. The ATR 72's ailerons were unpowered, connected mechanically to the cockpit control wheels, with hydraulic spoilers supplementing roll control.

In the accident conditions, the ridge of ice disturbed airflow near the right aileron as the angle of attack rose. The resulting hinge moment reversal could force the control surface and cockpit wheel to move suddenly. The crew's training had prepared them to expect ice-related speed loss and a stall warning. It had not prepared them for this abrupt aileron behavior at an angle of attack below the stall warning threshold.

Reproducing the ice and the control response

In December 1994, ATR conducted flight tests at Edwards Air Force Base using an Air Force tanker to create an icing cloud. An instrumented Learjet measured droplet size, liquid water content and temperature. The tests examined both the established certification conditions and larger supercooled droplets at temperatures near freezing.

Within Appendix C conditions, ice accumulated in the active boot area and was shed during normal cycling. In simulated freezing drizzle outside that envelope, ice accumulated both within and behind the boots. A jagged ridge developed near the aft edge of the boot on the upper wing. It appeared with both retracted flaps and flaps set to 15 degrees.

ATR subsequently tested artificial ice shapes derived from the tanker results. With shapes associated with ice accretion at flaps 15, flight at flaps zero produced autopilot disconnection, uncommanded aileron deflection and an initial roll consistent with the accident data. Sharp edges on the artificial ice lowered the angle of attack at which reversal occurred and increased the control forces.

The NTSB found that tests with flaps extended produced a smaller drag increase than tests with flaps retracted. Together with the flight recorder evidence, this supported its conclusion that the accident ice ridge developed mainly after flap extension. The physical evidence and the test program supported a loss of lateral control caused by ice-induced airflow separation.

The amended findings on warnings and oversight

The NTSB adopted its final report on 9 July 1996 and amended the findings and probable cause on 13 September 2002. The amended probable cause identified loss of control following an unexpected aileron hinge moment reversal after ice formed behind the deice boots during exposure to supercooled droplets outside the certification envelope. The aircraft was susceptible to that loss of control, and the crew could not recover.

The contributing factors included inadequate DGAC oversight of the ATR 42 and 72 and delayed transfer of airworthiness information to the FAA. The NTSB also identified deficiencies in FAA icing certification and operational requirements, published icing information and continuing airworthiness oversight. It cited ATR's inadequate response to continued ATR 42 icing roll upsets, which should have prompted further research and updated manuals and training.

The revised findings said ATR's earlier information had inadequately warned operators about the adverse control characteristics and recovery techniques. The 1992 All Weather Operations brochure inadequately communicated the catastrophic potential of operations in freezing rain. The crew lacked specific warning that freezing rain could produce sudden autopilot disconnection, aileron reversal and rapid roll excursions.

The report also examined cockpit discipline. In its revised finding, the NTSB said a sterile cockpit would probably have reduced distractions and could have improved awareness of the conditions. The accident manuals did not prohibit holding with flaps at 15 degrees in icing, and the crew had received no published explanation of the associated adverse ice accumulation.

Wider boots and safety recommendations

ATR developed extended chord deice boots for the wing outboard of the engines, including the area ahead of the ailerons. A second series of tanker tests found no ice accretion behind the new boots in simulated conditions consistent with those estimated at Roselawn. In early 1995, ATR published an icing procedures brochure covering recognition cues, aircraft configuration, lateral trim and autopilot use.

The NTSB recognized that wider boots, improved procedures, training and pilot awareness had reduced the risk. It remained concerned about aircraft controllability across the full range of naturally occurring droplet sizes, water content, temperature, configuration and exposure duration, because the tests had covered a limited set of conditions.

Its recommendations called for expanded icing certification criteria, tests addressing aileron hinge moment reversal, clearer operational limitations and better ways for crews to identify hazardous icing. It recommended that dispatchers provide pertinent AIRMETs and Center Weather Advisories, that weather broadcasts include relevant icing information, and that training address freezing drizzle, freezing rain and recovery from unusual flight events. The safety program addressed the aircraft, its operating guidance and the flow of weather and airworthiness information.