The Downsides of Cockpit Automation - The Crash of Asiana 214
2 months 2 weeks ago - 2 months 2 weeks ago #17537 by Westcoast
You may remember the crash of Asiana 214 in 2013; it got a lot of press at the time. It was a B777 making an approach to Rwy. 28L at KSFO, which crashed into the seawall and cartwheeled down the runway. Remarkably, only three people were killed. As I remember from reading press reports at the time, the ILS glideslope was inoperative and the accident was blamed on a relatively inexperienced pilot who got himself above the glideslope and pitched down to get back down to it, without realizing that the A/C auto-throttle was disengaged. When he pulled up to land, the aircraft slowed well below Vap and the main gear impacted the seawall at the near end of the runway. As you are about to learn, this story is basically correct, but there was a lot more to it than that.So here is a YouTube video () which explains the background and sequence of errors that led to the crash. This is from a new source, MentourPilot.com, which I have not introduced previously in this forum, but I think you will find it to be of really excellent quality. It’s a bit long, but well worth your time.So, there are a lot of lessons to be learned from this story. Let me try to list some of them. First of all, automation can be your enemy, particularly if the flight crew is tired, already task saturated and not intimately familiar with the subtle interactions between the modes of the automated flight control system. Additionally, the tendency to try to use the automated flight control system to fly the aircraft, particularly in non-normal conditions, can further task saturate the flight crew and lead to disaster. This tendency can be enhanced by company policies (as in the Asiana case), or by the flight crew’s inexperience in hand flying the aircraft. Finally, the reluctance of the flight crew to call and execute a go-around, particularly if a stable approach has not been established early on, must be overcome. FAA/NTSB recommendations following this crash emphasized clarity in the manufacture’s manuals, increasing pilot training hand-flying the aircraft, and initiating a go-around early in the event of a non-stabilized approach.I have several friends who are, or were, active ATP qualified commercial airline pilots and, because this subject interests me, I have a tendency to pick their brains mercilessly. One recommendation I have heard, I don’t recall just where, is to fly flight segments in a repeating A, B, C sequence of A: fully automated, B: hand flying using the flight director only (non- coupled autopilot) and C: fully hand-flown. All of these probably using the autopilot during the long, usually uneventful, cruise segments of the flight. As a general rule, I am too lazy to follow this advice, but that has gotten me in trouble.Coincidently, over the few weeks just before I discovered this particular video, I had several simulator flights which demonstrated each of the principal problems which afflicted Asiana 214. These were a couple of Econ Mode/KMSP Hub flights, flown in the aircraft on which I am current, the ERJ-145LR. I make these flights under a particular self-imposed constraint to add to the realism. I fly in the present (current time), day or night and in the current weather. Under the Econ rules, one has to commit to the flight before it is flown, and your flight gets graded, at least with respect to variables like your fuel planning and your handling of the aircraft. If the flight does not go well, you can accept the outcome, but if your landing descent rate exceeds 1000 fpm, you will be “publicly” shamed by the dreaded “Ask Agent” on the SPA Flight Arrivals board. If you just quit the simulator before your screw-up becomes public, you will be penalized by being charged $5000 for failure to complete the flight. It’s not real money, of course, but still it provides a little more motivation to be on your toes and take the flight to its conclusion.A particular detail about the Feel There ERJ-145 is relevant here: It doesn’t have an auto-throttle (AT). The speed is controlled during climb/descent by pitch when using the FLC (flight level change) mode - sound familiar? Otherwise, there is no automated speed control. This is a clear deficiency when compared to aircraft like the B737 and it makes a stabilized approach a good deal harder to manage. This is exacerbated by the time it takes for turbojet engines to spool up or down. When you make any change in the flight parameters, pitch, flaps, spoilers, gear, it takes a fair amount of time to settle on a desired airspeed as the throttle lag interacts with the descent rate to put you in a tail chase, using the throttle and pitch to try to control the airspeed and descent rate – sound familiar again? Obviously, this is more problematic when an ILS glide slope is not available, as the pilot has to control both airspeed and descent rate independently through the use of the elevators and the throttle. This is the same thing that is required when hand-flying a descent profile.Most of the time while returning to KMSP, ATC tells me to expect an ILS approach to either Rwy. 12L/R or Rwy. 30L/R, both equipped with ILS approaches. However, for the past several weeks, the winds at KMSP after sunset, have been from the southwest (200 deg.), so ATC has told me to expect a visual approach to Rwy. 22 (approach plate below). Since I don’t practice hand-flown approaches all that much, I didn’t really want to make a VFR approach, particularly after dark. While Rwy. 22 doesn’t flat out require that I make a VFR approach, as there is a LOC approach, it means I will not have a glide slope (GS) available, just like Asiana 214. And since I have neither a GS nor an AT, I was in almost exactly the same fix as Asiana 214: I wanted to fly an automated approach, but had neither GS nor AT available (in Asiana’s case, the absence of AT was particular to the FLC mode), but otherwise, the circumstances were the same.The first flight where I encountered this situation, I naively thought there must be some way to use the AP to control roll (the LOC) and hand fly the pitch. I can’t explain why I didn’t immediately realize this, because a little thought soon convinced me that the AP is all or nothing. So, I was going to need some AP mode for the pitch axis, if I was going to use the AP coupled to the localizer. I knew that couldn’t use FLC, as that would not allow me to control my descent rate, but I could use the Vs mode (rate of descent). It wasn’t until I reached the FAF and tried to start my descent that I found I couldn’t enter the Vs mode with the AP engaged in either the LOC or APP modes. At that point I disengaged the AP and hand flew the rest of the approach. The results were not pretty, as by that time I was high and hot. I finally got her down, not quite on the runway center line, but well down the runway from the touchdown zone. Somehow, I managed to float so far down the runway that I bled off my excess speed and managed to touchdown with a modest Vs and was not penalized by John’s Econ police. But it was a close call. I started the approach without understanding the coupling of autopilot modes and never really established a stable approach. I should have executed a go around as soon as I discovered I couldn’t get into the Vs vertical mode. The only thing I did better than the Asiana crew is that I went to manual flight mode before a crash became inevitable.By the second time I found myself making an LOC approach to Rwy. 22 in the dark, I had figured out that the reason I hadn’t been able to get into the Vs mode was not that I was in the LOC (or APP) mode, but because my altitude setting in the MCP window was set to 2600’, the altitude where the descent path intersects the glide path. So, I reasoned, this time I would set the altitude window to the altitude of the MAP (1420’). Then, using the table of ground speed versus descent rate for a 3-deg. glide slope provided on the approach plate, I could set Vs for my computed Vap and use the AP to fly me down to the MAP and hand fly it from there on in. Correct in theory, but not all that useful in practice. For this particular approach, the elapsed time from the FAF to the threshold at my 140 Kts. approach speed is 2:19. but the MAP is just half way down, so from the beginning of the final descent to the MAP at Vap is about 1:10, hardly worth the trouble of using the AP. To add to that, trying to stabilize my airspeed at Vap for the final descent without an AT was more-or-less a fool’s errand, as I was trying to chase the airspeed and descent rate during that very busy one minute. I would have been better off by far, just hand-flying the whole approach (in VFR) conditions, even at night. Alternatively, I could have calculated the cross-wind component for 12R and, depending on the results, requested the ILS approach. Of course, since I hadn’t thought out these options beforehand, I wasn’t prepared to make the right choice in the event. Think about it, maybe you shouldn’t make every approach an ILS approach. Fly it by hand from the IAF just to be sure you can do it if you have to. MikeP.S. An additional vertical mode available in the B737, and probably similarly equipped aircraft, is called a constant rate descent. It allows you to set a range radius on the Nav display, expand that outward to touch the runway threshold and select the mode. That will set the descent rate to bring you to the flare point just at the runway threshold. Between that and the AT, you won’t need a glide slope.
Last edit: 2 months 2 weeks ago by Westcoast. Reason: TRying to add spacing. Didn't work.
Please Log in to join the conversation.
Time to create page: 0.073 seconds