-T / T / +T | Comment(s)

Monday, February 10, 2003

Icing Considered Cause Of Taiwan ATR Crash

The crash of a TransAsia Airways ATR-72 last December has led to a series of recommendations aimed at improving the training of turboprop pilots to handle severe icing. Although the Dec. 21 accident that sent the twin-turboprop spinning into the Taiwan Strait is still under investigation, the Aviation Safety Council (ASC) of Taiwan has called on all turboprop operators to review their training programs to ensure they "contain the necessary training for pilots to recognize and effectively respond to all levels of icing conditions." It also recommended that operators emphasize additional training in pilots' situational awareness of icing conditions.

The ASC recommendation was subsequently picked up by the BEA French investigating authority, which further recommended that Avions de Transport R�gional (ATR) "inform the operators of the facts known at this stage of the investigation and assist them in the implementation of the safety recommendation emitted by the ASC."

The BEA also recommended that the French Directorate General of Civil Aviation (DGAC) inform the certification authorities in those countries in which the ATR is being operated to "draw their attention" to the ASC recommendations.

ATR has now issued a notice to all operators discussing the ASC and BEA recommendations, noting that the ATR's Airplane Flight Manual (AFM) addresses the problem of severe icing conditions and the procedures with dealing with it. The manufacturer said that it "reiterates that AFM procedures shall be strictly applied" whenever the aircraft encounters those conditions.

Although the investigation is still continuing, the most likely cause being considered is the severe icing. During 20 minutes of cruise flying in icing conditions, the airplane's speed decreased from about 200 knots to 158 knots. Just three seconds after the captain uttered "severe icing," the airplane's stall protection system activated.

The radar track of its flight showed the airplane falling with the dead weight of a brick at a rate of more than 22,000 feet per minute. The circumstances are similar to those of other events involving various airplane types in cruise, hold, climb or even descent.

The pilots commented they were in severe icing 73 seconds before the abrupt end of the cockpit voice recording (CVR). They might have recovered with a power-on descent to warmer levels. Not to go for a rapid descent is to invite the classic stall, "roll upset" and spin. Once the airplane starts to enter a spin, recovery under the disorienting conditions at night - as in this case - would have been difficult.

John Dow, a recently retired Federal Aviation Administration (FAA) expert on in-flight icing, said he could wallpaper a room with digital flight data recorder (DFDR) tracings of icing incidents and accidents that look like this crash, based on the details at this early stage in the postmortem - the steady erosion of speed, the increase in nose-up trim by the autopilot, perhaps the final addition of power, and the wrenching roll and departure from controlled flight.

Dow described the nature of icing as one where the insidious trap is set when stall speed increases from ice buildup on the wing, then the trap is sprung on a surprised crew. With insufficient available thrust, the crew would be unable to accelerate while remaining in level flight and so would need to unload the airplane by applying nose-down elevator. This method of escape increases airspeed and reduces angle-of-attack away from the deadly stalling angle - where lateral control easily can be lost. This maneuver also gets them down towards warmer air, Dow said.

In the TransAsia crash, Dow said, "If the DFDR does not show us anything different, this case appears similar to literally scores of twin-turboprop icing events." "We may not be able to completely prevent stall from icing, but we can better train our pilots to recover," he said, adding that any training that emphasizes holding pitch and applying power "may be protracting recovery. Pilots also need to be mindful of inadequate margins and to be wary of pulling up too soon or too abruptly," he added. (See box, above)

Crews may need to be better attuned to the fact that time is of the essence. "One of my students, a former airline pilot, said when he got into icing he had to apply full power not to slow down or sink below the glide path in just 20 seconds. I believe it," Dow recalled. In air with a high liquid water content (LWC), Dow said, "there can be some very rapid accumulation of ice.

As we have also learned, thin and distributed rough ice also can result in very severe aerodynamic penalties."

The pilots of the TransAsia airplane, Capt. Pan Teh-chung and First Officer Liu Ching-hai, requested permission to descend from 18,000 feet to 16,000 feet just moments before contact was lost.

Dow noted that after a fatal EMB120 crash from icing near Detroit in 1997, a modified pilot reporting (PIREP) taxonomy was proposed "to better reflect that fact that thin ice may be associated with severe results." The proposed format was based on records of scores of events (see box above).

"This format represents an effort to get the pilots to listen to what the airplane is telling them," Dow said. He pointed out that the 60-knot speed loss in level flight experienced by the TransAsia accident airplane "placed it in Level 4, which may mean that the accident investigator fills out the PIREP. Level 3 should connote 'ATC please clear out the airplanes underneath me now because I am either on my way down or about to be'."

With appropriate awareness, procedures and recovery techniques, Dow said of the TransAsia crash, "There should be no reason to lose an airplane from that altitude."

(This report was taken from Air Safety Week, a sister PBI Media publication written and edited by David Evans.) >> Dow, e-mail jdowsr@earthlink.net <<

Post-Stall Recovery With Ice Contamination

The problem:

  • DFDR time histories of dozens of roll upset events show that pilots are maintaining or attempting to increase pitch attitude while applying power.
  • This has resulted in failure to recover from the stall and protracted roll oscillations with altitude loss in excess of 3,000 feet with several accidents resulting in altitude losses in excess of 10,000 feet.

Why?

  • Pilots are being trained to recover using this technique at first sign of stall, usually stick shaker in lieu of reducing AoA [angle of attack] by pitch attitude change or flap extension if ground contact is imminent or airplane pitch response is degraded.
  • Stall protection systems may not have sufficient margin to protect against contaminated stall so that the stick shaker does not fire or fires at stall onset, instead of activating in the approach to stall.
  • Insufficient thrust is available to accelerate airplane out of stall, so stall upset is protracted and total loss of control may occur. IMC can exacerbate the problem (Note: the TransAsia ATR-72 crash was at night in bad weather).
  • Pilots do not experience this type of event in training and are not trained to recover. Very few regional turboprop operators have simulators that can realistically present these events for pilot training.

Conclusion:

  • At first sign of a stall, which may be stick shaker, uncommanded roll, buffet or other aerodynamic cues, apply nose down pitch control and level the wings while advancing RPM and torque until sufficient increase in airspeed for type.
  • If unable to lower nose, extend flaps from the cruise configuration.
  • Recover, maintaining higher airspeed than at upset! Retract flaps if extended for the recovery. Source: Dow
A Taxonomy for Pilot Reporting in Icing Conditions
Effect on Aircraft
Speed
Power
Climb
Control
Vibration and/or buffet
Level 1 < 10 kt loss < 10% increase required No effect or < 10% loss No effect No effect
Level 2 10-19 kt loss 10% - 19% increase required 10% - 19% loss rate of climb No effect No effect
Level 3 20-39 kt loss 20% - 39% increase required = 20% loss of climb Unusually slow or sensitive response from control input Controls may have slight vibration
Level 4 = 40 kt loss Not able to maintain speed Not able to climb Little or no response to control input May have intense buffet or vibration
Source: FAA Task 1b icing plan terminology subcommittee