Asiana 777 crashes in San Francisco

[Phil Bunch]

Here's a graph of the apparent descent, published by the NY Times:

http://www.nytimes.com/interactive/2013/07/07/us/asiana214-uneven-descent.html?ref=us

[John Golin]

(My understanding is no - it will not 'reengage' .  Interestingly VNAV does have some altitude protection with AT disengaged - http://www.youtube.com/watch?v=IJKU19Rn0GA - not what I expected until I saw the clip ).


I've never understood why the AT, by design, can be 'on' and not active in a mode - and why HOLD exists (outside of takeoff for both).

Is the AT "on" or is it simply "armed"? To me, when the L/R AT switch is pushed up, the AT is "armed"  When the A/T button is pushed and the light is on, the AT is on.

Armed - yes... my question is probably poorly phrased.  

It goes to the entire logic of the AT - why does it need to be armed but not  Active in some active mode (SPD / THR etc)?  The autopilot is not like this - if it is on, you always have a roll and pitch mode (of some description).

Automatic Speed Mode Engagement

If the A/T is not engaged, it will automatically engage in the speed (SPD) mode when all these conditions are true:

*At least one A/T ARM switch is in the ARM position
*Radio altitude is more than 100 feet, or more than 400 feet after takeoff
*Engine thrust is below the engine thrust limit
*The autopilot or the flight director are not in a speed-through-elevator mode
*The airspeed is just above stall speed

Thanks Mr Watson!

In the above note, if it was either ON (and engaged in a mode) or OFF, what would we lose?  Yes, things would have to be done differently, but I don't understand what is the advantage of the complexity we have with an 'Armed'  (that couldn't be achieved some other way)?

I have the feeling someone is about to embarrass me with a really obvious reason...  probably to do with interaction with AP modes...?

 :?
 

[John H Watson]

Armed-but-not-engaged allows the pilots to control the thrust manually (as it probably should have been done in this situation). However, if a go-around is needed (think panic & reflex), a TOGA switch is pushed. This not only provides a visual (F/D) cue for manoeuvres, but also GA thrust.

The partial automation gives the pilot a free hand (for other tasks or even an extra hand on the control column for more precise control).

HOLD mode gives the pilot the ability to add extra thrust without pushing buttons/changing modes. e.g. in FLCH descent mode to decrease the descent rate. However, you need to engage the A/P or follow the F/D to maintain selected airspeed.

Hope this makes sense.

Rgds
JHW

[Jeroen D]

Lot of articles flying around the net on the experience of the captain.

When we were still living in the USA and I was still flying I used to participate in the FAA excellent safety program WINGS. Aimed at GA it's goal is to increase overall safety and reduce accidents.

I remember I went to one of their seminars. Topic was how to reduce fatal accidents in GA. The FAA has a lot of data pertaining on this. The most telling part of their analysis was that there is no correlation between hours flown and fatal accidents rate. Put differently, statistically speaking a new pilot with say 100 hours flying a Cessna 150 has the same statistical chance of being in a fatal accident as the captain of a Learjet with 10.000 hours.

The big differentiator was safety training and awareness. Pilots that engage in constant safey education/training are statistically speaking much less prone to end up in fatal accidents. The WINGS programs is exactly that, it constantly encourages pilots to study, read, be engaged in the broadest sense possible of enhancing your flying skills and your attitude. The latter probably being the most relevant.

There are other safety programs that do exactly the same. For instance Cirrus runs special (safety) programs for their owners. And statistically speaking those that participate crash 50% less then the pilots that don't participate.

This 777 is not part of GA, but I'm sure the same principle applies. Flying safe is first and foremost an attitude.

Not executing a go-around or to late is, unfortunately, something that happens a lot. In itself it is not a particular difficult manoevre. It is considered a "safety manoevre". It gets you out of potential trouble.

I keep reading that they 'requested a go-around'. Not sure what they mean, but you certainly don't request a go around from ATC. When you're not happy with your approach, you just push the TOGA button or firewall the throttles and when you're well established in the climb you tell ATC you're going around.

Nobody will question a pilot that goes around. There is no legal requirement to report anything. At least not with the authorities. I understand some carriers require their pilots to explain why they went around. Not sure if that is a good practice.

Of course, we still need to hear what the real course of this accident is.

Jeroen

[John Golin]

Armed-but-not-engaged allows the pilots to control the thrust manually (as it probably should have been done in this situation). However, if a go-around is needed (think panic & reflex), a TOGA switch is pushed. This not only provides a visual (F/D) cue for manoeuvres, but also GA thrust.

The partial automation gives the pilot a free hand (for other tasks or even an extra hand on the control column for more precise control).

HOLD mode gives the pilot the ability to add extra thrust without pushing buttons/changing modes. e.g. in FLCH descent mode to decrease the descent rate. However, you need to engage the A/P or follow the F/D to maintain selected airspeed.

Hope this makes sense.

Rgds
JHW

It does indeed, thanks  

Though I still wonder why couldn't you design a TOGA push alone without AT armed to go to TOGA? And we don't 'half fly' the autopilot - if you want to tweak the throttles, use the thumb switch to disengage, and when you want to give it back to the automation, reengage...

No confusion then

Meh... what would I know - I don't fly, and in PS1 leave the A/T on until 100ft anyway (unless Roddez silently turns it off when I'm looking out the window!  :evil:  )

[Jeroen D]

Just picked this up on aviation Harald:

The aircraft joined a 17nm final, the crew reported the runway in sight before being handed off to tower. The autopilot was disconnected at 1600 feet 82 seconds prior to impact, the aircraft descended through 1400 feet at 170 KIAS 73 seconds prior to impact, descended through 1000 feet at 149 KIAS 54 seconds, 500 feet at 134 KIAS 34 seconds, 200 feet at 118 KIAS 16 seconds prior to impact. At 125 feet and 112 KIAS the thrust levers were advanced and the engines began to spool up 8 seconds prior to impact, the aircraft reached a minimum speed of 103 KIAS 3 seconds prior to impact, the engines were accelerating through 50% engine power at that point, and accelerated to 106 knots. The vertical profile needs to be assessed first. There was debris from the sea wall thrown several hundred feet towards the runway, part of the tailcone is in the sea wall, a significant portion of the tail is ahead of the sea wall in the water.


So they were really slow!!
Jeroen

[martin]

Here's a reasonable preliminary analysis worth reading (link found on PPRuNe, and recommended by several pro's there).

Based on FlightAware radar data (caveats apply), and restricting itself to technical aspects (as opposed to "company culture", CRM issues, etc.)

Martin

Update:

1. Belated thanks to John Watson for supplying the relevant 777 manual quote and a good clarification!

2. Here is  (on PPRuNe) a more detailed description of the "FLCH trap" I mentioned earlier.

[John Golin]

Interesting (for me) comment on the Boeing throttle quadrant from PPRuNe...

"The throttles on all Boeing North airplanes have the same force requirements to move - about 2.5 lbs force at the knob. In fact, with the exception of the 787, they all use minor variations on the same friction device for throttle 'feel'. Naturally, the 747 requires twice the force to move all the throttles since it has twice as many throttles to move. The autothrottle servo has plenty of force margin to move the throttles, its the friction devices that slip - as designed - to allow manual override of the A/T. The throttle quadrant on the 747-8 is unchanged from the 747-400.

That's roughly half the force that was required on the pre-FADEC throttle cabled engines - typically about 4.5 lbs at the knob (sometimes as high as 6 lbs.)."

[Jeroen Hoppenbrouwers]

More via the AV Herald (from NTSB). Sigh.

On Jul 9th 2013 the NTSB reported in their third press conference based on pilot interviews, that at 500 feet AGL the PAPIs were showing three red one white and the pilot began to pull back on the yoke to reduce rate of descent assuming the autothrottles would maintain the speed set to 137 knots. A lateral deviation developed taking the attention of the crew. Descending through 200 feet all PAPIs were red and the speed had decayed into the red/black marked range, the crew realised the autothrottles were not maintaining the target speed, at that point the autothrottles started to move the levers forward.

This pretty much explains what happened. Not yet 100% why. I don't get why the levers started to move forward eventually. Somebody claims that it wasn't the A/T but the P/F that did this.

More NTSB findings:

There were three pilots in the cockpit, the captain under supervision was pilot flying occupying the left hand seat, the training captain was pilot monitoring occupying the right hand seat, the relief first officer was occupying the observer seat, the relief captain was in the cabin at the time of the landing. The captain under supervision, 9700 hours total flying experience, had flown 10 legs for a total of 35 hours on the Boeing 777-200 so far and was about half way through his supervision. The training captain was on his first flight as training captain, the two pilots had never flown together before.

Well . . .

Cabin crew did better. Lot of things went horribly wrong, such a slides inflating inside the cabin and trapping two cabin crew members which then were cut free by pilots while they tried to extinhuish fire aboard etc. etc. and two cabin crew members were actually ejected at impact and survived being bowled out onto the runway at 100 kts.

[Hardy Heinlin]

I don't get why the levers started to move forward eventually.

Wouldn't this agree with that?

So lets say 137kts is set in the MCP, AT switches on MCP are armed, and AT is disengaged via the TQ switch, will the AT kick in at any point as the speed trend approaches 137kts, assuming the levers are at idle, without any intervention?

From the 777 Maintenance Manual

Automatic Speed Mode Engagement

If the A/T is not engaged, it will automatically engage in the speed (SPD) mode when all these conditions are true:

*At least one A/T ARM switch is in the ARM position
*Radio altitude is more than 100 feet, or more than 400 feet after takeoff
*Engine thrust is below the engine thrust limit
*The autopilot or the flight director are not in a speed-through-elevator mode
*The airspeed is just above stall speed

Without knowing what mode the FD was in, it would be hard to say if the A/T would have engaged.

[Jeroen Hoppenbrouwers]

I mean, the A/T wasn't doing much good, probably not designed to do anything. Why would it then suddenly spring to life well below stall speed? It is much more likely that the throttles (were) moved forward by the P/F.

The "777 FLCH trap" would explain why the A/T wasn't doing anything. Unless my memory fails here, I think on the 747-400 the A/T keeps working if you are in FLCH and then disengage the A/P -- it engages in SPD mode, right? The P/F came off the 747-400, likely with many, many hours.

On the -400 it does not retard for the flare, which is the common surprise people have that use the A/T while flying the yoke manually.


Hoppie

[evaamo]

From the 777 FCOM, in the flight controls section (Pitch Envelope Protection & Stall Protection):

"The autothrottle can support stall protection if armed and not activated. If speed decreases to near stick shaker activation, the autothrottle automatically activates in the appropriate mode (SPD or THR REF) and advances thrust to maintain minimum maneuvering speed"

I spoke to a friend of mine who flies the 300 variant, and he told me something I also validated in the FCOM:  the EEC enters an "approach idle" mode when certain criteria is met... one of these being using flaps 25 or 30 (which was the case). This prevents "under spooling" the engines in case of a GA or if an engine goes out.

Fact of the matter is: they were flying an unstable approach and should have performed a go-around before reaching 500' AGL.

cheers
-Enrique

[Jeroen Hoppenbrouwers]

Just for the sake of completeness: the 777 FLCH trap is described as follows.

When flying with both A/P and A/T engaged, and the AFDS working in a speed-on-elevator mode such as FLCH, for a descent the A/T will sit at idle and then HOLD as you expect. If you then disengage the A/P but leave the A/T engaged, the AFDS will keep the A/T in the speed-on-elevator mode -- i.e., it does not activate the throttle servo if your speed drops under the target MCP speed. It simply expects its other half (the now disengaged A/P or the real pilot) to fix the speed by pitch.

It is quite possible there is another protection far lower down the speed tape, just above stall speed, but either that didn't work or it wasn't in time. I can imagine that if you pull up hard enough to get the PAPIs back to two red, two white, you bleed off speed so rapidly that by the time the A/T comes back up, you face spool-up times longer than time to impact.


Hoppie

[Jeroen Hoppenbrouwers]

Question for the pros from an amateur.

If I would have had to fly the 28L as directed (from high and hot) during WorldFlight, I would probably have selected Vertical Speed instead of FLCH. I would have set the speed bug on the flaps schedule and manipulate the V/S wheel to have the speed move slowly but steadily down towards the desired speed. This would not have protected me against running into the bricks either low or high, but it would give the A/T a fair chance of picking up when the speed reached the MCP speed bug. When the PAPI would be captured I'd adjust the V/S to 800 and then prepare for manual takeover.

Would this technique be reasonable, or open all kinds of cans of worms?


Hoppie

[evaamo]

Hoppie,

I agree with your post regarding the FLCH trap... now it would seem that if the A/T was on HOLD mode, the "protections" mentioned in my last post would have been triggered by being near (or below) Vmca, the low altitude, the flap settings and pilot input via the yoke (there's even a tail strike protection, both during takeoff and landing).

-Enrique

[evaamo]

As a follow up to my last message, here's the next interesting bit (also found on the 777 FCOM):

"below 100 feet radio altitude on approach, the autothrottle will not automatically activate"

-Enrique

[martin]

Moi,

all kinds of cans of worms?

Most definitely I'm not a pro (I sometimes wonder even about "amateur"), but FWIW: after ploughing through 1360+ posts (the dole lets you do that...) in the "Asiana Crash" thread of PPRuNe, the following consensus regarding the "FLCH trap" seems to emerge:

¤ The "trap" notwithstanding, FLCH is still useful (and popular),
¤ because it provides the fastest way to lose altitude
¤ while still offering automatic overspeed protection.
(V/S, by contrast, may require too much "fiddling" (quote) with setting the rate and/or deploying speedbrakes, in order to keep the speed in check.)

But to avoid the "trap", FLCH is a no-no (per personal rule or company policy) for any or all of the following:
¤ on approach
¤ below 3000 ft
¤ with the MCP altitude set to 0000 (or runway altitude)

If anyone wishes to dig deeper: To spare you the ploughing, here is a list of links to posts from the PPRuNe thread which I found helpful for understanding the issue:
(Disclaimer: subjective selection, FWIW and YMMV apply; in chronological order; no guarantee of completeness or correctness; posts ploughed through up to #1372, 10th Jul 2013, 06:44 GMT)
* asterisks denote posts I personally (!) found especially illuminating

¤ Link01  
¤ Link02  
¤ Link03  
¤ Link04  
¤ Link05*
¤ Link06  
¤ Link07*  
¤ Link08  
¤ Link09  
¤ Link10*
_ but see also Link11 and Link12  
¤ Link13   (picture!)
   
Plus two links recommended by PPRuNe posters:
¤ Cpt. Chesley "Sully" Sullenberger (of Hudson fame) comments (video)
¤ "Wired" article
(yes, that "Wired", but the article has been recommended).

Note that this is all about the 777 (Link 13 above demonstrates that differences do exist to other types).

Now, how would it all look on a 744?

HTH,
Martin

[Phil Bunch]

Not sure this link adds new info, but the NY Times is reporting that the autothrottles were armed:

https://www.nytimes.com/2013/07/10/us/inquiry-suggests-chance-that-mechanical-failure-had-role-in-crash.html?pagewanted=print

Here's an excerpt:
-----------------------------------------------------

Investigators in the cockpit of the wreckage found the auto-throttle switches set to the "armed" position, meaning that the auto-throttle could have been engaged, depending on various other settings, she said. The disclosure is far from conclusive, but raises the clear possibility that there was a mechanical failure or that the crew misunderstood the automated system it was using.

The chairwoman, Deborah A. P. Hersman, also said that interviews of the three pilots who were in the cockpit at the time of impact showed that the speed indicator on the flat-panel displays in the cockpit had drifted down into a crosshatched area, meaning that the instruments were saying that the plane was moving too slowly.

At the dual controls, the pilot flying the plane was undergoing initial training as he upgraded from a smaller plane, and was supervised by a veteran pilot who was new as an instructor, Ms. Hersman said. The instructor told investigators that between 500 feet and 200 feet in altitude, the crew was also correcting from a "lateral deviation," meaning that the plane was too far to the right or left (she did not specify which) and realized they were too low.

At 200 feet, the instructor pilot told investigators in an interview, he noticed they were too slow. "He recognized that the auto-throttles were not maintaining speed," and began preparing the airplane to go around for another try. But it was too late.

[Jeroen D]

Not everybody is happy on the 'transparency' the NTSB provides:

http://sanfrancisco.cbslocal.com/2013/07/09/air-line-pilots-union-criticizes-ntsb-handling-of-sfo-crash-probe/


Jeroen

[martin1006]

In my believe it must have been a pilot error. How is it possible when you land with a 4 mans crew in the cockpit without noticing the pitch of the plane is to high and the speed is to low?

Unbelievable!

As far as I know the captain gave the flying pilot command to pitch up and at that moment they found out the trust levers where still on idle and the speed already near stall speed.

It means they probably wanted to use auto throttle at landing and they where not aware of the functions of the meaning of the ARM position in a Triple.

I know at Asiana they always land on auto. (in most Asian countries this is the case)

This is the same lack the Turkisch crew had on Amsterdam. Always landing on Auto and not aware of all the functions in a 737.

Resuming, for me the only possible cause of this crash is a crew depending to much on an automatic function and not monitoring the things what they should monitor.

If you land any airplane by hand and you have your hands on the controls and your eye on your PFD and the runway a crash like this is simply not possible.