French investigators are advising that simulator training should incorporate shock effects after a pilot's startled reflex reaction to an overspeed alarm sent an Air France Airbus A340-300 into a rapid climb, unnoticed by the crew.
The A340 reached climb rates of 5,700ft/min (30m/s), powering from its assigned transatlantic cruise level of 35,000ft (10,700m) to a maximum altitude of 38,000ft as the first officer - the non-flying pilot - reacted to the sudden alarm with a nose-up input to the side-stick.
French investigation agency Bureau d'Enquetes et d'Analyses found the aircraft, operating Caracas-Paris last 22 July, drifted into a turbulent region of convective weather because the weather radar setting had "not been optimal".
The A340 (F-GLZU) had been cruising at M0.83 northeast of Guadeloupe but a wind gust triggered an overspeed warning as the aircraft's airspeed rose to Mach 0.87. Reflexively, the first officer disconnected the autopilot using the side-stick button and pulled back on the stick for 6s.
"The [pilot] indicated that he did not remember these actions," says the BEA, which points out that it reconstructed the incident through flight-data recorder information because the cockpit-voice recorder was overwritten.
Although the overspeed alarm stopped after only a few seconds, it masked a separate warning that the autopilot had disconnected. The crew also could not remember hearing an altitude warning, which signals deviation of 200ft from an assigned level, as the aircraft - climbing at 1,950ft/min - passed through 35,200ft. The BEA says the effect of surprise on human memory "probably contributed" to the crew's failure to recall the altitude warning.
Unaware the autopilot was not activated and that the A340 was climbing, the crew selected various Mach settings. The pitch increased and, as the aircraft's Mach speed fell, the crew selected M0.93.
The aircraft soared to a climb rate of 5,700ft/min and exceeded 38,000ft before the captain realised, with surprise, how high the A340 was flying. Its airspeed had fallen to M0.66 and was displayed as 226kt, 19kt below the minimum selectable speed.
Only as the pilots queried how to return the aircraft to its 35,000ft cruise level did they begin to realise the autopilot was not active, nearly 90s after its disconnection.
The BEA says the crew missed several visual cues pointing to the A340's attitude and position, including the nose-up pitch of 12˚, the high climb rate, the excessive altitude, and instrument panel signals that the autopilot was off.
Without the cockpit-voice information, however, the inquiry was unable to explain the lack of monitoring of these parameters.
However, it states that long-haul flights are routinely calm and that an unexpected alarm can generate a "startle effect", a temporary period of stress which sparks instinctive reactions and motor responses.
These basic reflexes might bring the wrong response to an alarm and be "difficult to correct" under pressured circumstances. The inquiry highlights that the crew engaged in low-priority actions - such as visual observation of the weather, and making a passenger announcement - to the "detriment" of monitoring crucial aircraft parameters.
The A340's captain had logged more than 3,000h on type, while the first officer had over 2,400h.
"Recent studies show that the effect of surprise is the subject of little, or no, simulator training," says the BEA. It is recommending to the European Aviation Safety Agency that surprise effects be introduced to training scenarios to help pilots react to them and work under stress.
Investigators have also advised that aircraft autopilot disconnection alarms be brought into line with current certification requirements, after pointing out that the A340's disconnection alert, designed to an earlier standard, only sounded for 1.5s.