There’s a saying that a good landing starts with a good approach. The pilots of Asiana flight 214 didn’t manage their approach well and crashed into a seawall.
IMPACT
When Asiana flight 214 crashed at San Francisco International Airport (SFO) on July 6, 2013, many in the aviation community were incredulous. There were cries of “How could an airline crew with three pilots on the flight deck crash a perfectly good airplane on visual approach on a perfectly clear day with light winds?” Like many crashes that I dealt with while at NTSB, the causation of this accident was complex and involved the interaction of several elements of the system. Those elements collided on this day, just as Asiana 214 collided with the seawall surrounding the airport.
The Boeing 777 was approaching SFO runway 28L when it careened onto the airport at the completion of a 10 ½ hour flight from South Korea. The captain was completing his initial operating experience to wrap-up his PIC qualifications on the 777. His total flying time was around 9,700 hours, including 45 in the 777. He was transitioning from the Airbus 320 where he had been a check airman. With 12,000 total hours, including over 3,200 in the 777, the check pilot occupied the right seat. This was his first trip as an instructor. The third pilot was the relief first officer who was sitting on the jumpseat.
Setting the stage for the crash were fatigued pilots, which NTSB found likely degraded their performance. The two pilots were new in their respective roles. Combine these factors with an airline culture that didn’t promote manual flying and widespread misunderstandings of the limitations of the Boeing 777’s Automatic Flight Control System (AFCS), and you have a formula for disaster. Central to this crash was the notion of expectancy; the captain expected the airplane to do something it wasn’t designed to do. Specifically, he expected the autothrottle system to provide speed control for him, but unbeknownst to him, the system would not do so while in a certain autothrottle mode. Not only did this captain not understand this part of the autothrottle system – other, more experienced 777 pilots referred to this as the “FLCH Trap.”
This shows how HOLD would be annunciated on the Flight Mode Annunciator. Credit: NTSB
The 777 AFCS controls the autopilot and autothrottle. As with most highly automated autoflight systems, there are several modes and sub modes. “FLCH” is an abbreviation for the AFCS Flight Level Change mode. It is used for two things: To climb to a preselected altitude above you with climb thrust, and to descend to a preselected altitude below you with idle thrust. In the decent mode, once the throttles reach the idle stop for two seconds, they transition into HOLD mode. In this mode, the throttles are disengaged from the autothrottle servos and will not move until commanded by the AFCS or the pilot.
If engaged, the autothrottles will not allow speed to drop below the commanded speed. The autothrottle logic also has a nice speed protection feature: If the autothrottles are disconnected with the autothrottle disconnect switch on the side of the throttles, they will reengage (“wake up”) and apply thrust to protect speed if airspeed gets too slow. This autothrottle wake up feature was demonstrated to the captain during simulator training. Impressed with this protection system, the captain told investigators that he was “astonished” that the airplane would do this.
FLCH can be used several times during flight, which presents a paradox: If the AT are completely disconnected, they will wake up and apply thrust if airspeed gets slow. However, if they are still connected, but in the HOLD mode as a result of FLCH, or as a result of being manually overridden by the pilot, the autothrottles will not wake up if speed gets slow. Oddly, neither Asiana’s training nor Boeing’s manuals mentioned this situation, a factor NTSB found contributed to the crash. In fact, even the ground instructor who taught the captain’s class did not understand this nuance. Before the board meeting where we wrapped up this investigation, I asked the head of 777 training for a large US airline how well this autothrottle “failure to wake-up while in HOLD mode” was understood prior to this accident. He replied that it was not well understood at all. Shockingly, four years after this crash, where the FLCH trap became even widely discussed, I found myself on the jumpseat of a 777 of a United Air Lines flight. I asked the captain if the autothrottles would wake up if thrust was idle and in a HOLD mode. He got the answer wrong.
Put bluntly, the captain of Asiana 214 mismanaged the approach. He started out high and through a series of manipulations of the AFCS, he placed the system in FLCH. However, because the altitude set in the altitude window of mode control panel was above them, the autothrottles advanced to climb thrust – not idle as he expected. Not understanding this action, he manually pulled the throttles to idle and held them long enough for them to placed in HOLD. Now descending at idle thrust, the aircraft was passing through 500 feet with a descent rate of 1,200 fpm. The aircraft was also descending through PAPI glade path. (The glideslope was out of service). At this point, he did what would have worked on the Airbus, the aircraft that he was just transitioning from, and would have worked on the 777 had he not unintentionally placed the throttles in a HOLD mode - he pulled back on the control column to get back on the proper glidepath. Because the throttles were in HOLD, they would not increase thrust as the captain expected.
Estimated aircraft position at impact with seawall. Credit: NTSB
In concentrating on the below glidepath situation, believing the autothrottles would take care of speed, NTSB concluded that the captain focused on getting the aircraft back on a proper glidepath and discontinued monitoring airspeed. NTSB analysis was that the captain did not monitor airspeed for at least 24 seconds, and the check pilot didn’t monitor speed for at least 17 seconds. According to NTSB: “Human factors research has demonstrated that system operators often become complacent about monitoring highly reliable automated systems when they develop a high degree of trust in those systems and when manual tasks compete with automated tasks for operator attention.” The captain had developed trust that the autothrottle would take care of speed, so as he focused exclusively on getting back on the proper glidepath he dropped his scan of airspeed. “Thus, the flight crew’s inadequate monitoring of airspeed and thrust indications appears to fit this pattern involving automation reliance,” stated NTSB.
As the captain pulled back on the control column to recapture the glide path, airspeed continued to drop. Vref for this approach was 132 kts. Including a 5 kt additive, the approach was to be flown at 137 kts. Airspeed remained below approach target speed for 28 seconds, ultimately reaching 110 kts before the check pilot reacted. At 90 feet agl the check pilot yelled “speed,” added full thrust and pulled the control column full aft. Unfortunately, it was too little, too late. The aircraft careened into the seawall and cartwheeled across the runway and burst into flames. Of the three fatalities, NTSB found that two of those were not wearing seat belts and likely would have survived the crash had seat belts been worn. Remarkably, of the 303 occupants, 255 received either no or minor injuries.
One of NTSB findings was “If the autothrottle automatic engagement function (‘wakeup’), or a system with similar functionality, had been available during the final approach, it would likely have activated and increased power about 20 seconds before impact, which may have prevented the accident.” Despite that finding, I was outvoted in a 3-1 vote for a safety recommendation for Boeing to redesign the autothrottle wake up logic. I’m pleased to say that a few years after the Asiana crash, Boeing did just what I pushed for – they modified the 777 autothrottle logic so they now will wake up even if autothrottles are in HOLD.
The investigation found that Asiana had a culture that promoted heavy use of flight path automation with little emphasis on manual flying. An Asiana contract simulator instructor told investigators that manual flying was a “big scare for everybody,” and he believed that pilots avoided flying manually because of concern that they might do something wrong. The chief pilot told NTSB that turning off the autopilot at eight miles from the airport at 2,800 ft would not be recommended. Asiana provided NTSB with data that showed that nearly 20 percent of Asiana’s 777 landings were auto-lands. I compare this to my airline days where the only time we did an Autoland was on Cat 2 and 3 landings.
NTSB stated concluded that “by encouraging flight crews to manually fly the airplane before the last 1,000 ft of the approach, Asiana Airlines would improve its pilots’ abilities to cope with maneuvering changes commonly experienced at major airports and would allow them to be more proficient in establishing stabilized approaches under demanding conditions…” The ensuing NTSB recommendation to Asiana was sensible and to the point: “Modify your automation policy to provide for more manual flight, both in training and in line operations, to improve pilot proficiency.”
As you think about this crash, ask yourself how well do you know the hidden corners of your aircraft flight path automation system? Do you have an overreliance on automation? How effectively are you at flight path monitoring? Are your fatigue mitigation strategies effective?
—Robert Sumwalt is executive director for the Boeing Center for Aviation and Aerospace Safety at Embry-Riddle Aeronautical University. He was a member of the NTSB from 2006-21, including being chairman from 2017-21. Before that he managed a corporate flight department for a Fortune 500 company, and previously was a pilot for US Airways and Piedmont Airlines.
