Better awareness of latent and subtle human-performance factors — as well as the human-machine interface and contextual issues minutes before Asiana Airlines Flight 214 struck a sea wall at San Francisco International Airport (SFO) — can benefit the world’s air transport safety professionals, say two of the U.S. accident investigators. This means not just knowing how things went wrong but also what alternative preparations and courses of action would have enabled an uneventful landing.
Behind-the-scenes logic and reasoning used by the U.S. National Transportation Safety Board (NTSB) during this investigation also may aid the work of others in averting repetition of this scenario by, for example, revealing unrecognized risks within existing practices, said the NTSB’s William Bramble, senior human performance investigator, and Roger Cox, a captain and senior air safety investigator. They were co-chairs of the investigation’s joint operations and human performance working group.
Bramble and Cox elaborated in a presentation and paper on several of the final report’s key accident-prevention lessons in November at Flight Safety Foundation’s 67th annual International Air Safety Summit in Abu Dhabi, United Arab Emirates. Previous AeroSafety World articles have discussed the factual information, principal findings and safety recommendations regarding the Boeing 777 crash on July 6, 2013 (ASW, 2/14, 9/14, 10/14 and 11/14).
Two captains were on the flight deck of Flight 214. The captain in the role of pilot monitoring (PM) was pilot-in-command and an instructor pilot providing training — one of a series of such flights providing initial operating experience on the aircraft type — to the trainee captain, who had the role of pilot flying (PF) for the approach and occupied the left seat. A relief first officer was seated on the jump seat as an observer.
The flight crew conducted a visual approach to Runway 28L in visual meteorological conditions and light winds. After impacting the sea wall, the tail separated; the airplane slid down the runway, shed parts and pivoted about 330 degrees when it became airborne momentarily before coming to a stop off the left side of this runway. Casualties included three passenger fatalities and serious injuries to 40 passengers, eight flight attendants and one pilot, the final report said.
The report said the probable cause was “the flight crew’s mismanagement of the airplane’s descent during the visual approach, the [pilot flying’s] unintended deactivation of automatic airspeed control, the flight crew’s inadequate monitoring of airspeed, and the flight crew’s delayed execution of a go-around after they became aware that the airplane was below acceptable glide path and airspeed tolerances.” In the interest of disseminating lessons learned, this statement was structured to identify four separate, human-performance deficiencies in chronological order, Bramble said.
Investigators also considered a fifth, implied, deficiency — the flight crew’s non-detection of the PF’s error — to be “important for understanding the causal logic of the report,” he said (Figure 1). “Why did this mismanagement occur?” he asked.
Cox said, “The accident flight was the first one the pilot flying had attempted visually in the 777 without a glideslope, outside the simulator, and he felt stressed about [‘exactly controlling the descent profile and the lateral profile’]. The NTSB found the pilot’s lack of exposure to straight-in, speed-restricted, visual approaches affected his ability to manage the vertical profile of the accident approach. … Asiana has since added ‘straight-in visuals’ to their syllabus.”
Bramble noted that a number of issues identified in the report’s findings — for example, flight crew role confusion, increased workload, expectancy, automation reliance, surprise and nonstandard communication — are “all common human factors issues that have been observed in other accidents.” Interviews with the flight crew also revealed significant influences on individual performance such as personal apprehension about safely conducting a visual approach and misunderstandings of aircraft systems that were unspoken and unresolved.
“[The PF] told us that he was afraid if he went from three reds to four reds [on the precision approach path indicator (PAPI)], he was going to fail his OE [operating experience] flight, [or this] would be reported to his supervisors, and so he was very concerned about not having that happen,” Bramble said.
As a result of the series of flight crew actions and omissions that ASW previously has described, “the airplane was both high and fast when the speed restriction [air traffic control (ATC) instruction to maintain 180 kt until 5 nm (9 km) from the runway threshold] was lifted at 5 nm,” he said. “This subsequently delayed and compressed various flight crew tasks. … Although flaps 20 could have been selected 11 or more nm [20 km] from the runway before the gear was even deployed, the pilot flying did not call for this setting until the airplane was about 4.5 nm [8.3 km] from the runway. This in turn delayed the selection of landing flaps and lengthened the time for deceleration.
“Although the PF had recently completed the simulator portion of 777 transition training, he had not recently practiced the cognitive tasks associated with maintaining an appropriate vertical path during a lengthy, straight-in, high-energy approach without precise vertical guidance. The cognitive components of such skills can atrophy from lack of practice because when you have precision guidance, it relieves you of the need to perform some of that cognitive work.”
The PF’s unintended deactivation of automatic airspeed control resulted from the complex series of interactions between the pilot and the autoflight system detailed in one of the previous ASW articles (10/14). “Realizing he was quite high and close to the runway, the pilot flying selected FLCH [flight level change mode], thinking it would minimize thrust and allow him to descend and decelerate more rapidly. However, he apparently forgot that less than a minute earlier, he had selected a missed-approach altitude of 3,000 ft,” Bramble said.
“[When] the pilot flying realized that the airplane was not behaving as he intended, [he] very quickly disconnected the autopilot, leaving only the flight director, which was subsequently turned off on his side by the [other] captain. … The pilot flying incorrectly believed that the autoflight system would resume controlling speed after his manual throttle override and … that the autothrottle automatic-engagement feature [A/T wakeup] would ensure a minimum safe operating speed. Although the autothrottle had such a feature, it was not available in hold mode. … In this case, the pilot flying — and other Asiana pilots — were found to have gaps in their understanding of relevant aspects of autothrottle functioning.”
To put the PF’s gap in perspective, Bramble cited a 2012 study in which 60 percent of participating pilots said they did not “feel comfortable, fully comfortable, in their understanding of the autoflight system until they’d been flying the line for about three months. They also estimated that about 42 percent of their learning of the autoflight system came from experience on the line as opposed to classroom instruction. Such gaps can lead to automation surprise, particularly in dynamic and nonroutine situations.”
Non-detection of the PF’s mode confusion by the other pilots — despite the 10-second flight mode annunciation — proved to be significant in the causal sequence. “None of the flight crew noticed it, and none of the flight crew noticed the PF’s selection of FLCH or the autothrottle’s transition to HOLD [mode,] nor were they aware that airspeed was no longer being automatically controlled,” he said.
“How did this occur? … The instructional nature of the flight led to some blurring of PF and PM roles, with the PM initiating more actions and the PF behaving more deferentially than normal. This disrupted SOPs [standard operating procedures] involving mode selections and callouts. In addition, the flight crew was experiencing elevated workload due to their earlier mismanagement of the vertical flight path, and the pilot monitoring was preoccupied with changing the flap configuration at the time that FLCH was selected. … A callout could have drawn attention to the pilot flying’s selection of FLCH — but it was not performed.”
As to the inadequate monitoring of airspeed, and the lack of corrective action by the flight crew until the airplane descended to 58 ft above San Francisco Bay, the NTSB assessed the primary factor as the mistaken beliefs about automation. “Why did it take the crew so long to detect decaying airspeed and correct it?” Bramble asked. “The pilot monitoring did not detect low airspeed for about 17 seconds, and the pilot flying for at least 24 seconds. … Decades of human factors research indicate that monitoring of automated parameters decreases as workload increases on manual tasks. This pervasive phenomenon is known as automation overreliance, and it is difficult to eliminate through practice and training. … The crew believed the autothrottle was controlling airspeed, and the pilot flying thought A/T always ensured a minimum airspeed. This influenced their allocation of attention. Expectancy drives visual scanning in a selective-attention context.”
Delayed attention to the airspeed partly was an indication of a need to accomplish many tasks in a compressed timeframe, the NTSB found, and a reflection of how busy the pilots became. Fatigue — specifically, flying during the window of circadian low (i.e., when people adjusted to the departure-airport time zone were sleeping) and the PF’s acute sleep restriction just prior to the flight — also was acknowledged as a foundational influence on the crash circumstances.
“By the time airspeed dropped below [approach speed, VREF], workload on manual flying tasks had increased. For example, the crew had to prompt ATC for their landing clearance around 600 to 700 ft. Immediately thereafter, the pilot monitoring was busy performing the before-landing checklist. Just at 500 ft, they were getting the two whites and two reds on the PAPI, and the pilot flying was trying to arrest his rapid descent rate and remain on the glide path. So a lot was going on right around this time period,” Bramble said.
The PF’s manual handling technique also interfered with airplane state awareness. “The pilot flying did not ever trim the airplane after he disconnected the autopilot, depriving him of a salient tactile cue as airspeed dropped below VREF. He was already carrying 15 or 20 lb [33 or 44 kg] of back pressure on the column at approach speed,” he said. “So when [back-pressure force] went from 15 to 20, 22, 23 [lb, it was] not as noticeable as from zero to 5 or 10 [lb; 11 or 22 kg]. And it’s possible that this resulted from his transition [to the 777] from the Airbus [A320,] where he did not manually trim.”
Using pilots’ recollections of their thought processes, the investigators considered a few explanations for the flight crew’s irrecoverably late decision to conduct a go-around. They included surprise, non-standard communication, and that the PF and PM “experienced role confusion, in that each believed the other was responsible for initiating the go-around … possibly due to inconsistencies in company policies about who had the authority to initiate a go-around,” he said.
Cox suggested that aviation professionals studying the lessons of this accident also would do well to keep in mind specific challenges of conducting stabilized visual approaches to SFO’s closely spaced Runways 28L and 28R, adequate flight path cues that could have been used by these pilots, underlying crew training issues and the airline’s automation policy, considered by the NTSB to be out of step with the most recent best practices adopted by some international air carriers.
Alternative vertical navigation (VNAV) guidance included the position of the landing runway in the windscreen (nominally halfway between the top and bottom edges of the windscreen [Figure 2]); Boeing’s published pitch attitudes and power settings for each phase of flight in this 777; timely selection of VNAV mode; the PAPI; the green altitude range arc; and the airplane’s glideslope-like computed vertical path indication on the navigation display. “[For example,] when the arc is beyond the crossing fix, you can see you are high,” he said. “[The vertical path] indication was available during the accident flight, and the crew could have used it — even coupled to the autopilot, if desired — to fly the approach.”
Cox noted that some Asiana Airlines training personnel confirmed to the NTSB their awareness of exceptions to the expected conditions for automatic autothrottle engagement and wake-up. Moreover, one ground instructor interviewed said that the accident PF had been among students warned about the significance. The instructor also argued that the Boeing flight crew operations manual should contain a more explicit warning about the exceptions. “He called it an ‘anomaly,’ but Asiana didn’t do anything about it. They didn’t change their manuals. They didn’t call Boeing and say, ‘What’s up with this?’ It just was allowed to continue,” Cox said. “Of course, the exception is exactly what happened.”