It seems most of the pilot world is chalking this crash up to pilots relying too much on automation and not getting enough stick time. Perhaps the best in the business, John Nance, flatly blames "the philosophy behind automation uber alles that encourages us to use autoflight systems almost exclusively, keeping pilots around for emergencies only." ("When Pilots Aren't," ProPilot, March 2104.)

— James Albright

image

Updated:

2020-02-11

image

NTSB Docket, Exhibit 6-b, photo 5

I don't think this was an issue of over reliance on automation. Korean Air Lines (later branded Korean Air) and Asiana Airlines have a history of pilots mishandling visual approaches and not fully understanding the automation. See: Airlines of Korea. I think the problems can be boiled down to three areas:

  1. A culture that frowns upon crew resource management: the pilot flying the airplane was senior in rank and stature to both the instructor pilot and copilot who would have been reluctant to speak up. This isn't an Asian thing at all, the Japanese have fully embraced CRM. The Koreans and Russians have not.
  2. Pilots that don't have a background in flying visual approaches: there are several Asiana and Korean Air Lines mishaps where the pilots could not bring themselves to fly anything but an ILS and made very bad decisions on the way to mishaps because they simply couldn't keep the airplane on a glide path to the runway without an electronic glide slope.
  3. Pilots who don't understand the finer points of the automation in their aircraft. More about this below.

Taken from a historical context, this accident was just another in a long string of similar accidents from airlines based in Korea. There have been so many, in fact, that the name of one of the airlines has been subtly changed over the years. So why do I make a case study about this mishap, one that could have easily been prevented by an airline unwilling to fire a few pilots and institute some real changes? Because there is a lesson for us all: when you change equipment, operators, or flying environment, you need to be willing to relearn all that you knew before. You need to invite outside critique. In short, you should be very careful to ward off complacency. More about this: Complacency.

There is a lot in the NTSB report about emergency response that is vitally important but not a part of pilot procedures so I will leave that to you.

The NTSB accident report is exceptionally well written but I think they missed two critical points. First, the cultural aspect that made the PM and observer pilots reluctant to be more forceful in their CRM roles. Second, the poor understanding of how to visually fly an approach in a large aircraft, where energy management is an important factor.

1 — Accident report

2 — Narrative

3 — Analysis

4 — Cause

5 — Postscript

image

1

Accident report

  • Date: 6 July 2013
  • Time: 11:28
  • Type: Boeing 777-28EER
  • Operator: Asiana Airlines
  • Registration: HL7742
  • Fatalities: 0 of 16 crew, 3 of 291 passengers
  • Aircraft Fate: Destroyed
  • Phase: Landing
  • Airports: (Departure) Seoul-Incheon International Airport (ICN/RKSI), South Korea; San Francisco International Airport, CA (SFO/KSFO), USA

2

Narrative

image

NTSB AAR-14/01, figure 2.

What happened? The crew started the approach too high, made a few automation mistakes that caused them to get even higher, and then as they were plummeting down to briefly pass through the correct glide path (at a very high descent rate), they ended up with the auto throttles in a mode that would not correct for a speed that got too low to safely recover from.

We'll get into the "Why" in the Analysis section. But look at the date of hire of the three pilots to understand that even though the PF was being evaluated by the instructor pilot who was in the PM position, the PF was the senior pilot. Looking at other accidents from airlines from Korea, we see a trend where deference to the captain is a factor in poor CRM.

History of flight

  • On July 6, 2013, about 1128 Pacific daylight time, a Boeing 777-200ER, Korean registration HL7742, operating as Asiana Airlines flight 214, was on approach to runway 28L when it struck a seawall at San Francisco International Airport (SFO), San Francisco, California. Three of the 291 passengers were fatally injured; 40 passengers, 8 of the 12 flight attendants, and 1 of the 4 flight crewmembers received serious injuries. The other 248 passengers, 4 flight attendants, and 3 flight crewmembers received minor injuries or were not injured. The airplane was destroyed by impact forces and a postcrash fire. Flight 214 was a regularly scheduled international passenger flight from Incheon International Airport (ICN), Seoul, Korea, operating under the provisions of 14 Code of Federal Regulations (CFR) Part 129. Visual meteorological conditions (VMC) prevailed, and an instrument flight rules (IFR) flight plan was filed.
  • Two of the flight crewmembers, a trainee captain and an instructor pilot (IP), were the primary flight crew, and the other two flight crewmembers, a second captain and a first officer (FO), were relief pilots. The flight, which was the first of a scheduled 2-day trip with a scheduled layover in San Francisco, was an operating experience (OE) training flight for the trainee captain.
  • As shown in figure 2, the trainee captain occupied the left seat and was the pilot flying (PF) for the takeoff and landing. The IP, who was the pilot-in-command (PIC), occupied the right seat and was the pilot monitoring (PM) for the takeoff and landing. The relief captain and FO occupied seats in the cabin for takeoff and during the initial part of the flight. About 4 hours 15 minutes after takeoff, they came forward to the cockpit and assumed flight crew duties for about the next 5 hours 15 minutes of the flight, allowing the primary flight crew to rest in the cabin.
  • The PF stated in an interview that he returned to the cockpit about 0938. According to the relief captain, he told the PF that he had programmed the flight management computer (FMC) with the instrument landing system (ILS)/localizer (LOC) 28L approach and advised him of the likelihood that the flight would be held at high altitude and/or high speed by air traffic control (ATC) for longer than normal during the approach to SFO. The relief captain stated that the PM returned to the cockpit about 10 minutes after the PF’s return.
  • According to the cockpit voice recorder (CVR), at 0955:45, there was a transfer of aircraft control to the PF and the PM, and the relief captain and FO returned to the cabin. About 6 minutes later, the PF and the PM discussed expectations for receiving radar vectors for a visual approach. At 1042:28, the PF began an approach briefing. During his briefing, the PF referred to automatic terminal information service (ATIS) Juliet, which included the information that visual approaches to runways 28L and 28R were in progress, and the ILS glideslopes for these runways were out of service. The PF stated that he expected vectors for a visual approach to runway 28L, would use the LOC to maintain lateral path, and, after capturing the LOC, would use the automatic flight control system (AFCS) to manage the vertical profile. He said that the minimum descent altitude for the LOC approach was 460 ft mean sea level (msl) and that he would set a go-around altitude of 3,000 ft msl in case of a missed approach.
  • At 1047:28, the PF called for the descent checklist. The PM acknowledged the callout and began to run the checklist, which included verification that the reference landing speed (Vref)6 was 132 knots. At 1047:54, the PM stated, “checklist completed.” At 1112:33, the relief FO returned to the cockpit, occupied the center jumpseat, and acted as an observer during the approach and landing. As the flight proceeded towards SFO, it was cleared to descend to lower altitudes and vectored to intercept a straight-in approach to runway 28L.
  • At 1121:49, a controller at Northern California Terminal Radar Approach Control (NorCal) asked if the flight crew had the airport in sight. The PM replied, “okay runway in sight,” and the controller cleared the flight for a visual approach to runway 28L. According to flight data recorder (FDR) data, when the approach clearance was given, the airplane was descending through about 6,300 ft msl. The airplane’s airspeed was about 211 knots; it was configured with the flaps and landing gear up; the autothrottle (A/T) was in hold (HOLD) mode; and the autopilot flight director system (AFDS) was in flight level change speed (FLCH SPD) pitch mode and heading select (HDG SEL) roll mode.
  • At 1122:07, the PF stated, “I am intercepting localizer,” and the PM responded, “yes.” At 1122:11, the LOC push-button switch on the AFCS’s mode control panel (MCP) was pushed, arming the LOC mode, and the PM stated, “localizer armed,” to which the PF replied, “check, cleared visual approach.” At 1122:46, the PF stated, “next, three thousand one hundred” and then “cleared visual approach,” and the PM replied “check.” At 1122:48, the MCP-selected altitude changed to 3,100 ft. At 1122:52, LOC capture occurred, and the AFDS roll mode changed from HDG SEL to LOC, where it remained for the duration of the flight. At this time, the airplane was about 15.4 nautical miles (nm) from the runway threshold, descending through about 5,300 ft msl at an airspeed of about 210 knots

Source: NTSB AAR-14/01, ¶1.1

3,100 feet is an intermediate altitude on the approach, just outside the FAF.

image

NTSB AAR-14/01, figure 3

  • At 1122:54, the PM stated, “let’s descend slowly to one thousand eight hundred feet, and it’s visual.” The PF replied, “yes yes sir I will set to one thousand eight hundred.” At 1123:02, the MCP-selected altitude was changed to 1,800 ft, which was the minimum altitude for the DUYET waypoint, located 5.4 nm from the runway and the final approach fix (FAF) for the ILS/LOC 28L approach. At 1123:05, the PM stated, “localizer capture.” The PF replied, “check, flaps one sir,” and the PM stated, “speed check flaps one set.” At 1123:11, the flap lever was moved to the flaps 1 position as the airplane was descending through about 4,900 ft msl at an airspeed of about 214 knots. At 1123:16, the PF stated, “speed one nine two set,” and the MCP-selected airspeed changed from 212 to 192 knots.
  • At 1123:17, when the airplane was about 14.1 nm from the runway, descending through about 4,800 ft msl at an airspeed of about 215 knots and a descent rate of about 900 ft per minute (fpm), the NorCal controller instructed the flight to reduce airspeed to 180 knots and to maintain that speed until 5 miles from the airport; the PM acknowledged the instruction. At 1123:31, the PF stated, “speed one eight zero,” and the PM replied, “check.” At 1123:33, the PF stated, “flaps five.” About 3 seconds later, the PM made an unintelligible comment, and the MCP-selected airspeed changed to 180 knots. At 1123:42, the PF stated, “flaps five sir,” and the PM replied, “speed check” and then “flaps five.” At 1123:45, the flap lever was moved to the flaps 5 position.
  • By 1123:50, the airplane’s descent rate had decreased to about 300 fpm, and the PM made an unintelligible comment. About 2 seconds later, the MCP-selected airspeed changed to 172 knots. At 1123:53, the PF stated, “yeah, I am descending now,” and the PM replied, “yeah.” At 1123:57, the AFDS pitch mode changed to vertical speed (V/S); the A/T mode changed to speed (SPD) mode; and the MCP-selected vertical speed was set to -900 fpm. At 1123:58, the PM called out the pitch mode change by stating, “V/S,” and the PF replied, “one thousand.” About 1 second later, the MCP-selected vertical speed changed to -1,000 fpm, and the PM stated, “check.” There was no communication between the flight crew for about the next 31 seconds.
  • At 1124:32, when the airplane was about 9.5 nm from the runway, descending through about 3,900 ft msl at an airspeed of about 185 knots and a descent rate of about 1,000 fpm, the relief FO who was observing (hereafter referred to as the observer) commented that the flight was to maintain 180 knots until 5 miles from the airport by stating, “to one eight zero five miles.” The PM responded, “ah ah ah one eight zero,” and the observer repeated, “one eight zero five miles.” At 1124:36, the PF stated, “huh?” The observer again repeated the speed to be maintained as instructed by the NorCal controller, and the PF stated, “okay one eight zero five miles.” There was no communication between the flight crew for about the next 12 seconds.
  • At 1124:51, when the airplane was about 8.5 nm from the runway, descending through about 3,500 ft msl at an airspeed of about 188 knots and a descent rate of about 1,000 fpm, the PF called for the landing gear to be extended. At 1124:53, the gear handle was moved to the down position, and the PM stated, “this seems a little high.” About 2 seconds later, the PF stated, “yeah,” and the PM repeated, “this should be a bit high.” About 3 seconds later, the PF stated, “do you mean it’s too high?” The PM made an unintelligible reply. At 1125:02, the PF stated, “I will descend more,” and at 1125:04, the MCP-selected vertical speed changed to -1,500 fpm. There was no communication between the flight crew for about the next 21 seconds.

Source: NTSB AAR-14/01, ¶1.1

At this point on the visual approach, the airplane was 400' above the profile given on the instrument approach.

  • At 1125:23, the PM made an unintelligible comment, followed about 6 seconds later by his comment, “ok.” At 1125:31, the PM stated, “one thousand,” and the MCP-selected vertical speed changed to -1,000 fpm. At this time, the airplane was about 6.3 nm from the runway, descending through about 2,600 ft msl at an airspeed of about 178 knots and a descent rate of about 1,500 fpm.

Source: NTSB AAR-14/01, ¶1.1

At 6.3 nautical miles a three-degree glide path would have called for being 2,000 feet above the runway, so they were about 600 feet high.

  • The airplane crossed DUYET at 1125:46, while descending through about 2,250 ft msl at an airspeed of about 176 knots and a descent rate of about 1,100 fpm. When it reached DUYET, the airplane was about 450 ft above the 1,800-ft minimum altitude depicted on the approach chart. The airplane reached a point about 5 nm from the runway at 1125:55, while descending through about 2,085 ft msl at an airspeed of about 174 knots and a descent rate of about 1,000 fpm. When it reached the 5.0 nm point, the airplane was about 400 ft above the altitude for the desired glide path of 3°.
  • At 1125:56, the PM radioed the tower controller stating the flight’s position but did not receive an immediate response. About 5 seconds later, the PF called out “flaps twenty,” and the PM replied, “flaps five ahh” and then “flaps twenty.” At 1126:06, the flap lever was moved to the flaps 20 position as the airplane was descending through about 1,900 ft msl at an airspeed of about 175 knots and a descent rate of about 1,000 fpm, and the PF stated, “yeah.” At 1126:10, the MCP-selected speed changed to 152 knots. About 2 seconds later, the PF called out “flaps thirty,” and the PM replied, “speed check flaps thirty sir.” At this time, the airplane’s airspeed was about 174 knots, which was above the flaps 30 limit speed of 170 knots.
  • At 1126:25.7, the AFDS pitch mode changed to FLCH SPD, and the A/T mode changed to thrust (THR) mode. The AFCS responded to the mode change by starting to slow the airplane to the MCP-selected speed of 152 knots and initiating a climb toward the MCP-selected target altitude of 3,000 ft, as seen in FDR-recorded AFDS pitch commands, a slight increase in thrust lever angles, and a slight pitch up. At 1126:28.3, the PM stated “flaps thirty,” and at 1126:29.5, the PF made an unintelligible statement that included the word “sir.” Between these two remarks, at 1126:28.8, the flap lever was moved to the flaps 30 position, and simultaneously, the autopilot (A/P) was disconnected. At this time, the airplane was about 3.5 nm from the runway, descending through about 1,500 ft msl at an airspeed of about 169 knots and a descent rate of about 1,000 fpm.

Source: NTSB AAR-14/01, ¶1.1

It appears the MCP target altitude was changed to the published missed approach altitude of 3,000 feet some time after they had descended below it, which would have been appropriate. (This was not commented upon in the accident report until now, after the fact.) But it also appears the pilot selected FLCH in an effort to speed his descent not realizing the target altitude was above them, prompting the autothrottles to command an increase in thrust.

  • According to FDR data, the AFCS-initiated forward movement of the thrust levers that began when the A/T mode changed to THR was manually overridden, and the thrust levers were moved aft. At 1126:33, the thrust levers reached the idle position, and the A/T mode changed to HOLD. Immediately before the A/T mode change occurred, at 1126:32.5, the PM stated, “flight director,” and immediately after the change, at 1126:34.0, the PF replied, “check.”

Source: NTSB AAR-14/01, ¶1.1

The PF appeared to have pulled the throttles to idle without understanding why they had moved forward in the first place, or that they would then be in the HOLD mode. See: More about the auto throttles, below.

  • During a postaccident interview, the PF stated that he considered pressing the FLCH push-button to obtain a higher descent rate but could not recall whether he did so or not. When interviewed, none of the three flight crewmembers recalled seeing the changes to the A/T mode displayed on each primary flight display’s (PFD) flight mode annunciator (FMA) that resulted from the selection of FLCH.
  • At 1126:36, the PM stated, “speed,” and the PF replied, “target speed one three seven.” At 1126:38, the MCP-selected airspeed changed to 137 knots. At this time, the airplane was about 2.9 miles from the runway, descending through 1,300 ft msl at an airspeed of about 165 knots and a descent rate of about 1,000 fpm. By this point, the flight crew should have been able to clearly see the precision approach path indicator (PAPI) lights; the PAPI indication would have been four white lights, showing that the airplane was significantly above the PAPI glide path angle of 2.98°.

Source: NTSB AAR-14/01, ¶1.1

They were 350' high.

  • At 1126:40, the PF called out “flight director off,” and the PM replied, “okay.” According to quick access recorder (QAR) data, at 1126:43, the left (PF’s) flight director (F/D) switch was turned off, and the right (PM’s) F/D switch remained on. There were no further changes in the F/D switch positions for the duration of the recording. At 1126:44, the PM stated, “it’s high,” and over the next 8 seconds, the airplane’s descent rate increased from about 1,000 to 1,500 fpm. At 1126:52, the PM stated, “one thousand,” and the PF replied, “check.” At 1126:54.9, the airplane was about 2.1 nm from the runway when it descended through 1,000 ft radio altitude20 (RA) at an airspeed of about 151 knots with a descent rate of about 1,500 fpm. When it descended through 1,000 ft RA, the airplane was 243 ft above the altitude for a 3° glide path, and the PAPI indication was four white lights. Table 1 lists selected events during the last 1,000 ft of the approach.

Source: NTSB AAR-14/01, ¶1.1

image

NTSB AAR-14/01, Table 1.

  • At 1127:14.8, the airplane was about 1.3 nm from the runway when it descended through 500 ft RA at an airspeed of about 137 knots with a descent rate of about 1,200 fpm. As shown in figure 4, which depicts a profile view of the last 40 seconds of the flight, the PAPI indication was three white lights and one red light—a slightly above glide path indication—when the airplane descended through 500 ft RA. The thrust levers were at the idle position, and the engines’ N1 speeds were about 24%. At 1127:15.5, an electronic voice announced “five hundred,” and about 1 second later, the PF called out “landing checklist.”

Source: NTSB AAR-14/01, ¶1.1

image

NTSB AAR-14/01, figure 4.

  • At 1127:16.8, an electronic voice announced “minimums, minimums.” At 1127:17.5, the PM stated, “landing checklist complete cleared to land.” At 1127:19.8, the PM stated, “on glide path sir,” as the PAPI indication changed to two white and two red lights, showing the airplane’s on-glide path position. At this time, the airplane was descending through about 400 ft RA at an airspeed of about 134 knots (where it remained for about 4 seconds) and a descent rate of about 1,100 fpm, which was above the descent rate of about 700 fpm needed to remain on glide path. At 1127:21.2, the PF responded, “check.” There was no communication between the flight crew for about the next 11 seconds.
  • At 1127:23.3, the airplane’s airspeed dropped below 132 knots (Vref). At 1127:24.1, the airplane was about 1 nm from the runway at 331 ft RA, the airspeed was about 130 knots, the descent rate was about 1,000 fpm, and the PAPI indication changed to one white light and three red lights, showing the airplane’s slightly below-glide path position. Over the next 5 seconds, the airplane’s pitch attitude increased from about 2° to 4° nose up, where it remained for about 3 seconds. At 1127:31.0, the airplane was about 0.7 nm from the runway at 219 ft RA, the airspeed was about 122 knots, the descent rate was about 900 fpm, and the PAPI indication changed to four red lights, showing the airplane’s significantly below-glide path position. Over the next 5 seconds, the airplane’s pitch attitude increased to about 7° nose up, where it remained for about 3 seconds before continuing to increase.
  • During a postaccident interview, the observer stated that he saw two white lights and two red lights on the PAPI as the flight descended through 500 ft RA. He recalled that beginning sometime after the airplane descended below 500 ft, he could no longer see the runway or the PAPI lights through the windscreen. The PM stated that at 500 ft RA, the airplane was slightly low, and he saw one white light and three red lights on the PAPI. The PF stated that he began to see one white light and three red lights on the PAPI at 500 ft RA and recalled thinking that if he allowed the PAPI indication to go to four red lights, he would fail his flight and would be embarrassed. The PF pitched the airplane up to avoid going low. He stated that about that time, he “saw some light and was in blindness for a second.” The light prompted him to stop looking outside and instead look down at the instrument panel. He stated that the blindness was only momentary, and he could see the airspeed tape and noticed that the airspeed was low. The PM stated that he did not see any bright light. The PM further stated that about 200 ft RA, the airspeed was about 120 knots; he saw four red lights on the PAPI and thought perhaps the A/T was not working.

Source: NTSB AAR-14/01, ¶1.1

The aircraft was very slow with a high descent rate, a very poor energy state. With the engines at or near idle, the spool up time would have been high, making the problem of adding energy become worse.

  • At 1127:32.3, an electronic voice announced “two hundred.” At 1127:33.6, the PM stated, “it’s low,” and the PF replied, “yeah.” At 1127:36.0, one of the flight crewmembers made an unintelligible comment. At 1127:39.3, the quadruple chime master caution alert sounded. When the alert sounded, the airplane was about 0.45 nm from the runway at 124 ft RA, the airspeed was about 114 knots, and the descent rate was about 600 fpm. At 1127:41.6, an electronic voice announced “one hundred.” At 1127:42.8, the PM stated, “speed.” Less than a second later, both thrust levers were advanced by the PM.24 At 1127:44.7, the A/T mode changed from HOLD to THR. At 1127:46.4, the CVR recorded the stick shaker activating, and the lowest airspeed during the approach of about 103 knots was recorded by the FDR at 1127:46.9. At this time, the airplane was about 0.35 nm from the runway at 39 ft RA, the descent rate was about 700 fpm, the N1 speeds for both engines were increasing through about 50%, and the pitch attitude reached about 12° nose up. The airspeed then began to increase. At 1127:47.8, the PM called out, “go around,” and at 1127:48.6, the airspeed was about 105 knots, and the stick shaker stopped. The initial impact with the seawall occurred at 1127:50. At that time, the N1 speeds for both engines were increasing through about 92%, and the airspeed was about 106 knots.
  • Video from airport surveillance cameras showed that following the initial impact, the tail of the airplane separated, the airplane slid along the runway, and the rear of the fuselage lifted up, tilting the airplane into about a 30° nose-down angle. The airplane pivoted counterclockwise about 330° before impacting a second time and coming to rest off the left side of the runway, about 2,400 ft from the initial seawall impact point. The airplane came to a stop about 1128:06.26.

Source: NTSB AAR-14/01, ¶1.1

Personnel information

The PF was hired by Asiana Airlines on March 2, 1994, as a cadet pilot with no previous flight experience, and he received ab initio (from the beginning) flight training at a flight school in Florida from 1994 to 1996. He began FO training on the 737 in 1996 and served as a 737 FO and a 747-400 FO before upgrading to 737 captain on December 15, 2005. He transitioned to A320 captain on October 22, 2007. The PF began transition training to 777 captain on March 25, 2013.

After serving as a pilot in the Republic of Korea Air Force, the PM was hired by Asiana Airlines on February 1, 1996. He was initially qualified as a 767 FO and upgraded to 767 captain on March 21, 2001. He transitioned to 777 captain on January 16, 2008. He underwent 777 IP training in May and June 2013 and became qualified as an IP on June 12, 2013. The accident flight was his first time acting as an IP.

After serving as a pilot in the Republic of Korea Air Force, the observer was hired by Asiana Airlines on December 12, 2007. He was initially qualified as an A320 FO and transitioned to the 777 on March 3, 2012. Asiana records showed that the observer flew seven trips that transited SFO as a 777 FO from May 29, 2012, to April 10, 2013, and made no landings at SFO during those trips. The observer’s most recent flight into SFO before the accident was on May 10, 2013.

Source: NTSB AAR-14/01, ¶1.5


3

Analysis

The auto throttles play a role in this accident. More correctly, the pilots' understanding of the auto throttles play a role. But the pilots were not alone in their misunderstanding. See: Dissenting Opinions for more about this.

The other obvious issue is the failure to call for a go around when the approach was so obviously unstable.

Asiana pilots' statements about A/T function

  • The PF stated that the 777 A/T system would always maintain the selected airspeed as long as the A/T was on. He said that if a pilot overrode the thrust levers manually, the A/T would resume controlling airspeed when the pilot released the thrust levers. He stated that it was irrelevant whether he had pushed the FLCH button immediately before disconnecting the A/P during the accident approach because he was in manual flight and the A/T was always working. He thought the A/T should have automatically advanced the thrust levers upon reaching the MCP-selected airspeed during the accident approach, and he did not understand why that did not occur. Furthermore, he thought the A/T system should have automatically transitioned to TO/GA when the airplane reached minimum airspeed. In that respect, he believed that the 777 A/T functionality was similar to alpha floor protection on the Airbus A320/321. Asked how confident he was in his understanding of the 777 AFCS, he said he had followed the Asiana training program but was not confident in his understanding, and he thought he needed to study more.
  • An Asiana 777 OE instructor captain who had conducted an OE instructor check flight with the PM was asked to predict how the 777 A/T system would behave if the FLCH pushbutton was selected while descending through 2,000 ft msl in V/S mode with the A/P on and the MCP altitude set to 3,000 ft msl. He said that the thrust levers would advance, and the airplane would try to climb. When asked what mode the A/T would be in if the pilot manually pulled the thrust levers back, he said that the A/T would remain in SPD mode, and the thrust levers would advance after the pilot released them. Three additional Asiana pilots were asked what would happen in this same scenario if a pilot pulled the thrust levers back manually and then disconnected the A/P. Another 777 OE instructor captain who had conducted an OE training flight with the PM said he did not know what thrust mode the A/T would be in after this sequence of actions. Two 777 OE instructor captains who had conducted OE training flights with the PF said that the A/T would transition to HOLD mode and remain there until a different A/P pitch mode was selected by the pilot.
  • The Asiana 777 chief pilot was asked if there were any conditions in which the 777 A/T system would not provide low speed protection, and he said it would not provide low speed protection in HOLD mode. He said this information was not included in Asiana’s 777 pilot training curriculum but was contained in the 777 FCOM, and he thought Asiana 777 pilots were aware of it. When the same question was posed to three OE training captains (two of whom had conducted OE training flights with the PM and one of whom had conducted an OE flight with the PF), they said that the A/T system did not provide low speed protection in HOLD mode; however, one said he had only learned that after the accident. A fourth OE instructor was asked the same question and said that the A/T system would not provide low speed protection if the A/P and A/T both failed. A contract simulator instructor (who had been providing 777 simulator training to Asiana pilots since 2006) was asked if he thought most pilots understood that the 777 A/T system did not provide low speed protection in HOLD mode, and he said he was not aware of that himself until Boeing provided an update to the 777 FCOM in 2012.

Source: NTSB AAR-14/01, ¶1.17.2.2

Accident sequence

  • The airplane intercepted the final approach course for runway 28L about 14.1 nm from the runway threshold and 4,800 ft msl, which was close to the desired glide path and set up the flight crew for a straight-in visual approach. However, after the flight crew accepted the ATC instruction to maintain 180 knots to 5 nm from the runway, the crew made inputs to the AFCS that reduced the airplane’s descent rate and caused it to diverge well above the desired 3° glide path.
  • The PF decreased the MCP-selected speed from 210 knots to 180 knots and commanded flaps 5. While these selections at this time were appropriate, leaving the AFDS in FLCH SPD pitch mode without using the speedbrake to ensure an adequate rate of descent was not effective because the A/P was controlling airspeed with elevator inputs and the reduction in selected speed caused the airplane to pitch up and the descent rate to decrease. As a result, the airplane started to drift above the desired glide path. In addition, the PF did not appear to promptly recognize that the airplane was drifting above the desired glide path after he had selected a lower airspeed in FLCH SPD.
  • When the airplane was about 11.5 nm from the runway, descending through 4,300 ft msl, the PF changed the A/P pitch mode from FLCH SPD to V/S and adjusted the vertical speed to -1,000 fpm; there was a corresponding automatic change in the A/T mode from HOLD to SPD. The airplane’s vertical speed began to increase toward the target value; however, a descent rate of 1,000 fpm was not high enough to maintain, let alone recapture, the desired glide path, so the airplane continued to drift above it.
  • When the airplane was about 8.5 miles from the runway threshold, descending through 3,400 ft msl at an airspeed of 188 knots, the PF commanded landing gear down. Gear extension added drag and made it easier to decelerate. Although he did not do so, the PF could also have commanded flaps 20 at this time, adding additional drag, because the airspeed was 7 knots below the flaps 20 limit speed of 195 knots. The airplane was about 900 ft above the desired glide path, and a higher flap setting would have allowed the flight crew to descend the airplane at a steeper angle while maintaining the assigned airspeed and would have saved time later when the flight crew needed to establish a landing configuration and decelerate to final approach speed.
  • When the airplane was about 6.3 nm from the runway at about 2,600 ft msl and an airspeed of 178 knots, the selected vertical speed was changed back to -1,000 fpm. By examining the altitude and distance to the runway, both of which were displayed on the instrument panel, and applying the well known rule of thumb that a 3° glide path requires about 300 ft of altitude loss per nm, the pilots could have quickly estimated that they were still several hundred feet high. Changing the vertical speed back to -1,000 fpm at that time was, therefore, inappropriate because the airplane was still well above the desired glide path, and the flight crew needed to continue descending the airplane at more than 1,000 fpm to return to the desired glide path. The crew’s action indicated a lack of awareness of the airplane’s position relative to the desired glide path and of cues in the cockpit that could have alerted them to this. As a result of this lack of awareness and their early reversion to a descent rate of 1,000 fpm, the airplane remained high.
  • When the airplane reached the 5 nm point, it was about 400 ft above the desired glide path, descending through 2,085 ft msl, at an airspeed of about 174 knots. As a result, the PF needed to slow the airplane below the flaps 30 limit speed of 170 knots, while also losing the excess altitude. When test pilots from Boeing and the FAA attempted to fly a stabilized approach in a flight simulator beginning from this starting point, they found it difficult to achieve a stabilized approach by 500 ft agl. In fact, they found it impossible to do so without exceeding maximum descent rates published in Asiana’s FOM. However, when the starting point was changed to an on-desired-glide path altitude (1,650 ft msl) with the same airspeed, the test pilots were consistently able to achieve a stabilized approach by 500 ft agl, and they did so without exceeding Asiana’s maximum descent rates. Clearly, the airplane’s excess altitude increased the difficulty of achieving a stabilized approach. The NTSB concludes that the flight crew mismanaged the airplane’s vertical profile during the initial approach, which resulted in the airplane being well above the desired glide path when it reached the 5 nm point, and this increased the difficulty of achieving a stabilized approach.
  • The flight crew’s difficulty in managing the airplane’s vertical path continued as the approach progressed. The PF was aware of the need to lose excess altitude and attempted to do so using FLCH SPD pitch mode, which initiated a sequence of events that had the unintended consequence of deactivating automatic airspeed control. The changes on the FMA that occurred during the airplane’s descent from 1,850 to 1,330 ft msl, when it was between about 4.5 and 3.0 nm from the runway threshold, are shown in figure 14 and discussed below.
  • After passing the DUYET waypoint, the PF called out and set the MCP altitude to 3,000 ft, the planned altitude in the event of a go-around. After passing the 5 nm point, the PF commanded flaps 20; the PM moved the flap lever to the flaps 20 position; and the PF set the MCP speed to 152 knots (panel A of figure 14), the target speed for flaps 20.93 The PF then commanded flaps 30, but the PM delayed setting them to the flaps 30 position because the airspeed had not yet slowed to below the flaps 30 limit speed.
  • About 13 seconds after the PF commanded flaps 30 and 3 seconds before the PM set flaps 30, the AFDS pitch mode changed from V/S to FLCH SPD. By design, the A/T mode simultaneously changed from SPD to THR. Because the MCP-selected speed of 152 knots was less than the airplane’s current airspeed of 172 knots (and in FLCH SPD, the AFDS controls speed via elevator pitch commands), the AFDS responded to this mode change by increasing pitch to slow the airplane. Also, the F/D command bars displayed an increase in target pitch. Because the MCP-selected altitude of 3,000 ft was above the airplane’s current altitude of about 1,550 ft msl (and in THR mode, the A/T applies thrust appropriate to attain the selected altitude), the thrust levers began to advance to initiate a climb. About three seconds later, the PF disconnected the A/P and manually countered the forward motion of the thrust levers, moving them to the idle position. He then decreased the pitch by pushing the control column forward. As a result of the PF’s manual override of the A/T, the A/T mode switched to HOLD, a mode in which the A/T would not move the thrust levers and was not controlling thrust or airspeed.
  • The process of decelerating and configuring for landing would have been much easier, the pace of activities slower, and the workload less if it was not necessary for the PF and PM to both lose excess altitude and slow down after reaching the 5 nm point. Because they had to do both, they encountered a period of increased workload when they were engaged in parallel tasks for an extended period of time during a critical portion of the approach. This, in addition to the PF’s failure to make a callout when he selected FLCH SPD pitch mode on the MCP, reduced their ability to effectively crosscheck each other’s actions and monitor changes in the status of the autoflight system. The NTSB concludes that the flight crew’s mismanagement of the airplane’s vertical profile during the initial approach led to a period of increased workload that reduced the PM’s awareness of the PF’s actions around the time of the unintended deactivation of automatic airspeed control.
  • After setting the target approach speed of 137 knots in the MCP, the PF commanded, “flight director off.” It was Asiana’s informal practice to fly visual approaches with the PF’s F/D turned off and the PM’s F/D turned on, but only after turning both F/D switches off first. In postaccident interviews, the PF and PM stated that the PM followed this practice. However, QAR data indicated that the PM turned off the PF’s F/D switch but did not turn off his F/D switch. If both F/Ds had been in the off position at the same time, this would have cleared the AFDS modes and caused the A/T to transition to the SPD mode and resume automatic control of airspeed. However, the AFDS remained active and continued to display the FLCH SPD guidance that the PF had commanded. The PF’s and PM’s FMAs continued to display A/T HOLD mode, LOC roll mode, and FLCH SPD pitch mode. The PM’s F/D command bars were providing guidance to stay on the localizer and pitch to maintain the commanded 137-knot speed. The PF’s F/D command bars were removed.
  • Although, at 500 ft, the airplane met some of Asiana’s stabilized approach criteria as listed in the 777 POM (including being on target airspeed, in the landing configuration, and on the correct flightpath), it failed to satisfy other criteria. It was descending at greater than 1,000 fpm, and the thrust setting was not appropriate (it should have been about 56% N1 speed). Because the approach was not stabilized at 500 ft agl, the flight crew should have conducted a go-around. Either the pilots did not notice that these parameters exceeded stabilized approach criteria or they believed that the deviations were minor and could easily be corrected. In either case, the crew’s decision to press ahead was not unusual, as industry statistics indicate about 97% of unstable approaches are continued to landing.
  • In a postaccident interview, the PM recalled that he first noticed the airspeed was low (120 knots) about 200 ft and thought at that time that the A/T was not working. FDR data are consistent with his recollection that the airspeed was 120 knots at 200 ft. The observer also recalled noticing that airspeed was low and seeing that the red hatch marks at the bottom of the airspeed tape (the barber pole) were moving up around 200 ft. He estimated that this was 5 or 10 seconds before impact (the airplane reached 200 ft about 18 seconds before impact). The observer did not recall a specific airspeed value at 200 ft.
  • About 8 seconds after the PAPI indication changed to four red and 7 seconds after the PM recalled seeing the airspeed at 120 knots, the quadruple chime caution aural alert sounded, and the master caution light illuminated. The airplane was about 11 seconds from impact at an altitude of 124 ft and a speed of 114 knots. According to the performance study, after this point, the airplane likely no longer had performance capability to accomplish a go-around.

Source: NTSB AAR-14/01, ¶2.2

More about the auto throttles

image

NTSB AAR-14/01, figure 9

  • The A/T is armed by moving the left and right engine A/T arm switches on the MCP [Mode Control Panel] to the ARM position. Once armed, the A/T will engage when certain AFDS pitch modes (including FLCH SPD) are selected or the MCP A/T engage button is pushed. The A/T may be overridden manually at any time by taking control of the thrust levers and placing them at a desired position.
  • Mode summary
    • Thrust reference (THR REF) — Thrust set to the reference thrust limit displayed on EICAS.
    • Speed (SPD) — Thrust applied to maintain target airspeed set using the MCP or FMC.
    • Thrust (THR) — Thrust applied to maintain the climb/descent rate required by AFDS pitch mode.
    • Idle (IDLE) — Occurs when A/T controls the thrust levers to the aft stop.
    • Hold (HOLD) —Occurs when A/T removes power from the servo motors. In this mode, A/T will not move the thrust levers.
  • Airplane speed can be controlled by the AFDS [Autopilot Flight Director System] or the A/T (Autothrottle).
    • When the AFDS is controlling speed, this is informally called “speed-on-elevator,” as the speed is controlled by modifying the pitch of the airplane through elevator movement.
    • When the A/T is controlling speed, this is informally called “speed-on-throttle,” as the speed is controlled by movement of the thrust levers. The A/T controls speed only when it is in SPD mode.

Source: NTSB AAR-14/01, ¶1.6.3

"Speed on elevator" is typically during a climb when the thrust is set at an upper limit or during a descent when the thrust is set to idle. "Speed on throttle" is typically at times other than a climb or descent, such as in cruise or on an approach.

  • In FLCH SPD mode, the rate at which the thrust levers move depends on the difference between the current altitude and the target altitude. The A/T is limited by the thrust limit at the forward range of thrust lever travel and by idle at the back range of travel. During a FLCH descent, HOLD mode will engage when the thrust levers reach the aft stop or if the pilot manually overrides the A/T. During a FLCH climb, HOLD mode will engage only if the pilot manually overrides the A/T. When the HOLD mode engages, the annunciation for the A/T mode will change from “THR” to “HOLD,” and the annunciation will be surrounded by a green box for 10 seconds. The A/T will remain in HOLD mode until one of the following conditions is met:
    1. The airplane reaches the MCP target altitude.
    2. The pilot engages a new AFDS pitch mode or new A/T mode.
    3. The A/T arm switches are turned off.
    4. The thrust is manually commanded to increase past the thrust limit.
    5. The A/P is disconnected, and both F/D switches are turned off.

Source: NTSB AAR-14/01, ¶1.6.3

Note that in many accidents where automation is a factor, the investigators never fail to point out that the mode in question is clearly given to the pilot. An important factor in this case was that the pilots failed to realize the autothrottles were in HOLD mode when it was clearly annunciated at the top of their pilot flight displays:

image

NTSB AAR-14/0, figure 9

I'm not so sure. There is a lot of information on that display and the "HOLD" is tucked away on top with other things that compete for attention. Of course the pilot should have spent some time looking at the airspeed indicator and the approaching amber band and barber pole. There is a trend vector forecasting what is to come. But there was an even better predictor of what was to come in this Boeing: the throttles which move even when being manipulated by the autothrottle system. I don't believe the pilot's hands were on the throttles during the approach, robbing him of valuable information. Had he been in the practice of having a hand on the throttles during the approach and landing, he would have noticed the fact they were at idle. That should have cued him about his impending low energy state.

Go-Around Decision-Making

As stated earlier, the pilots did not initiate a go-around until reaching an altitude of 90 ft, which was about 7 seconds before the impact. This delay prevented a successful go-around because the airplane was not able to develop the power necessary to climb out in the time available before impact. As shown during testing, a go-around that was initiated 11 seconds before the impact (at an altitude of 124 ft) likely would not have resulted in the ground impact. In fact, the go-around should have been initiated well before that point: when the airplane reached 500 ft and the approach was unstabilized. The previous section discussed the role of insufficient monitoring in the flight crewmembers’ failure to execute a go-around between 500 and 200 ft.

Source: NTSB AAR-14/01, ¶2.3.4


4

Cause

The report's list of causes is good, but incomplete.

The report does mention that the PF was the senior pilot and does show that the PM and observer paid deference to this fact. It also notes these less-senior pilots could have been more forceful in their corrective callouts. But the report does not address why these pilots behaved this way. See: An Expat's Opinion for an idea about this.

The report makes mention in the analysis section but does not follow through in the causal section the fact the pilots did not understand the need to put the airplane on a proper glide path and, more importantly, the correct way to manage their energy state to simultaneously increase their descent rate while decreasing their airspeed. Using FLCH was the wrong answer. The correct answer was to increase their drag earlier.

  • The flight crew mismanaged the airplane’s vertical profile during the initial approach, which resulted in the airplane being well above the desired glide path when it reached the 5 nautical mile point, and this increased the difficulty of achieving a stabilized approach.
  • The flight crew’s mismanagement of the airplane’s vertical profile during the initial approach led to a period of increased workload that reduced the pilot monitoring’s awareness of the pilot flying’s actions around the time of the unintended deactivation of automatic airspeed control.
  • About 200 ft, one or more flight crewmembers became aware of the low airspeed and low path conditions, but the flight crew did not initiate a go-around until the airplane was below 100 ft, at which point the airplane did not have the performance capability to accomplish a go-around.
  • The flight crew was experiencing fatigue, which likely degraded their performance during the approach.
  • Nonstandard communication and coordination between the pilot flying and the pilot monitoring when making selections on the mode control panel to control the autopilot flight director system (AFDS) and autothrottle (A/T) likely resulted, at least in part, from role confusion and subsequently degraded their awareness of the AFDS and A/T modes.
  • Insufficient flight crew monitoring of airspeed indications during the approach likely resulted from expectancy, increased workload, fatigue, and automation reliance.
  • The delayed initiation of a go-around by the pilot flying and the pilot monitoring after they became aware of the airplane’s low path and airspeed likely resulted from a combination of surprise, nonstandard communication, and role confusion.
  • As a result of complexities in the 777 automatic flight control system and inadequacies in related training and documentation, the pilot flying had an inaccurate understanding of how the autopilot flight director system and autothrottle interacted to control airspeed, which led to his inadvertent deactivation of automatic airspeed control.

Source: NTSB AAR-14/01, ¶3.1

The National Transportation Safety Board determines that the probable cause of this accident 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. Contributing to the accident were (1) the complexities of the autothrottle and autopilot flight director systems that were inadequately described in Boeing’s documentation and Asiana’s pilot training, which increased the likelihood of mode error; (2) the flight crew’s nonstandard communication and coordination regarding the use of the autothrottle and autopilot flight director systems; (3) the pilot flying’s inadequate training on the planning and executing of visual approaches; (4) the pilot monitoring/instructor pilot’s inadequate supervision of the pilot flying; and (5) flight crew fatigue, which likely degraded their performance.

Source: NTSB AAR-14/01, ¶3.2

Dissenting (in part) opinion

Member Sumwalt makes an excellent point here and the implication on Boeing manual writers is clear. I agree with his sentiment but I don't think it excuses the pilots for failing to fly the airplane correctly.

Member Robert L. Sumwalt filed the following concurring (in part) and dissenting (in part) statement on July 1, 2014.

I believe setting the stage for the crash was expectancy; the pilot flying expected the airplane to do something that 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 HOLD mode.

This pilot was not the only one who didn’t understand this nuance of the automation, and I am convinced that misunderstandings of this system are actually fairly common. For example, somewhere during the pilot flying’s ground school training, a ground instructor supposedly mentioned to him—perhaps in passing—that on three occasions, he had experienced a similar situation where the autothrottle system did not maintain speed. To show that the ground instructor himself did not understand this situation, he described it as an “anomaly.” However, it’s not an “anomaly” at all—it’s the way the system was designed. This represents data point number one in a wider misunderstanding of the system logic.

Data point number two comes in the form of an FAA test pilot. In 2010, this pilot was flying a Boeing 787 on a test flight when he was alarmed to see that the autothrottle did not maintain speed while in FLCH and HOLD. The autothrottle logic on the 787 is essentially the same as the 777, and this pilot had 500 hours in the 777. Why would an experienced FAA test pilot not understand this system logic if it was that clear? He raised his concern throughout the FAA certification chain. Officials with EASA also became concerned. As a result, Boeing added additional and clearer guidance in the Boeing 787 manuals regarding this system logic. But for the Boeing 777—an airplane with the same basic system—nothing was done.

Data points three and onward are less well documented but are no less relevant or important. Before the Board meeting, I asked the head of 777 training for a very large airline how well this autothrottle “failure to wake-up while in HOLD mode” was understood prior to this accident. His answer: not well understood at all.


5

Postscript

An ex-Pat's opinion

I agree with just about everything this ex pat instructor has to say except: "You just can't change 3000 years of culture." The Japanese have similar baggage when it comes to learning and never challenging authority. And yet they don't have the same issues. I grew up in the "this side of the cockpit is mine, that side is ours" era and have witnessed a sea change in attitudes in the U.S. military and civilian aviation communities. Old dogs can learn new tricks.

Here are some extracts from a U.S. pilot sent to Korea to help correct things, at their request. He spent five years in Korea with both airlines. He found them to be identical, "the only difference was the color of the uniforms and airplanes."

  • We ex pat instructors were forced upon them after the amount of fatal accidents (most of the them totally avoidable) over a decade began to be noticed by the outside world. They were basically given an ultimatum by the FAA, Transport Canada, and the EU to totally rebuild and rethink their training program or face being banned from the skies all over the world. They hired Boeing and Airbus to staff the training centers. KAL has one center and Asiana has another.
  • This solution has only been partially successful but still faces ingrained resistance from the Koreans. I lost track of the number of highly qualified instructors I worked with who were fired because they tried to enforce "normal" standards of performance. By normal standards, I would include being able to master basic tasks like successfully shoot a visual approach with 10 kt crosswind and the weather CAVOK. I am not kidding when I tell you that requiring them to shoot a visual approach struck fear in their hearts ... with good reason. Like this Asiana crew, it didn't compute that you needed to be a 1000' AGL at 3 miles and your sink rate should be 600-800 Ft/Min.
  • The Koreans are very very bright and smart so I was puzzled by their inability to fly an airplane well. They would show up on Day 1 of training (an hour before the scheduled briefing time, in a 3-piece suit, and shined shoes) with the entire contents of the FCOM and Flight Manual totally memorized. But, putting that information to actual use was many times impossible. Crosswind landings are also an unsolvable puzzle for most of them. I never did figure it out completely, but I think I did uncover a few clues. Here is my best guess.
  • First off, their educational system emphasizes ROTE memorization from the first day of school as little kids. As you know, that is the lowest form of learning and they act like robots.
  • They are also taught to NEVER challenge authority and in spite of the flight training heavily emphasizing CRM/CLR, it still exists either on the surface or very subtly. You just can't change 3000 years of culture.
  • The other thing that I think plays an important role is the fact that there is virtually NO civil aircraft flying in Korea. It's actually illegal to own a Cessna-152 and just go learn to fly. Ultra-lights and Powered Hang Gliders are Ok. I guess they don't trust the people to not start WW III by flying 35 miles north of Inchon into North Korea. But, they don't get the kids who grew up flying (and thinking for themselves) and hanging around airports. They do recruit some kids from college and send then to the US or Australia and get them their tickets. Generally, I had better experience with them than with the ex-Military pilots. This was a surprise to me as I spent years as a Naval Aviator flying fighters after getting my private in light airplanes. I would get experienced F-4, F-5, F-15, and F-16 pilots who were actually terrible pilots if they had to hand fly the airplane. What a shock!
  • Finally, I'll get off my box and talk about the total flight hours they claim. Actually, this is a worldwide problem involving automation and the auto-flight concept. Take one of these new first officers that got his ratings in the US or Australia and came to KAL or Asiana with 225 flight hours. After takeoff, in accordance with their SOP, he calls for the autopilot to be engaged at 250' after takeoff. How much actual flight time is that? Hardly one minute. Then he might fly for hours on the autopilot and finally disengage it (MAYBE?) below 800' after the gear was down, flaps extended and on airspeed (autothrottle). Then he might bring it in to land. Again, how much real "flight time" or real experience did he get. Minutes! Of course, on the 777 or 747, it's the same only they get more inflated logbooks. So, when I hear that a 10,000 hour Korean captain was vectored in for a 17-mile final and cleared for a visual approach in CAVOK weather, it raises the hair on the back of my neck.

References

(Source material)

Nance, John, When Pilots Aren't, ProPilot, March 2014.

NTSB Aircraft Accident Report, AAR-14/01, Descent Below Visual Glide Path and Impact With Seawall, Asiana Airlines Flight 214, Boeing 777-200ER, HL7742, San Francisco, California, July 6, 2013