Blind following of flight directors yet again
 

B738, vicinity Bristol UK, 2019 - SKYbrary Aviation Safety

Pilots blindly following the flight director regardless of its veracity. The Children of the Magenta Line syndrome continues to defy common sense.

From the report it looks more like a rushed and impossible approach to me - especially with the reduced track miles. They were too close with too much energy to accept the shorter routing.

They then seem to have got into a right bugger's muddle with the 737 "automation" and were at least 10nm behind the aircraft - not a good place to be.

Whatever it was, it seems amazing that pilot recruitment these days concentrates on how many verbal reasoning and maths questions you can do in 10 minutes, and clever answers to HR questions, rather than being able to actually fly a plane with Situational Awareness.

I suspect that these qualities are the ones that prevent you from getting into such sticky situations which require the flying ability to get you out of.

No. That comes from experience. Flying experience prevents you getting into a bad situation in the first place, e.g. trying to approach within too few track miles.

Let's look at the article shall we..."It was noted that the 55 year-old Captain had been acting as PM for the flight until he took over control when the aircraft began to descend soon after the go around had been initiated. He had a total of 14,750 hours flying experience which included 10,770 hours on type."

Can't see a lack of experience here.

Yes, I walked into that one didn't I :)

However, the verbal reasoning and maths tests don't seem to have helped this Captain either.

I'd put my money on fatigue and commercial pressure for this sort of misjudgement.

Some of the old/bold treated the FD with some suspicion They used to tell a story about Ron Hardy, BMIs much respected chief pilot, who always did his checks with the most junior instructor in the fleet. As an engine failure at V1 = 1 knot happened and there seemed little input from him except a couple of tweaks to the trimmers, the instructor leant forward and said "follow the command bars captain", Ron pushed his seat back removed his hands and feet from the controls and watched. The sim ran along the runway rotated and climbed away. Ron turns to the instructor and says see what I mean?. These, points to shoulder, are command bars, they are advice bars

Moved from CRM to Tech Log at OP's request.

While we can all make mistakes with profiles, I recall at AN as a young chap there was a great emphasis on keeping the mental picture going throughout the initial pre-descent and descent proper. Folks tended to have the descent profile where they desired throughout with a bit of ROD tweaking as required while the altimeter wound its way down.

Flexible thought processes facilitated out of left field recoveries (eg air traffickers' needing to keep the aircraft high due to traffic conflicts and so on). For sure the boxes do a useful job but, unless the pilot knows what is what, or lets his/her guard down, it is easy to allow the gee whizz stuff first to seduce and then to leave one in flagrante. Not infrequently, say coming into SYD from the north for a straight in landing during busy periods, one could find oneself exploiting the steep side of the aeroplane's descent capability.

More importantly, one recognised early when one really needed an orbit or some extra track miles to run and that made things easier for everyone concerned.

The sudden cutting of track miles opening the door to the ensuing chaos was certainly a part of it. The blind, task-saturated FD following (in day VMC, at that) was another. Both were fully factors in their own right, and the importance of one does not diminish the other (as in "more like a..." comparison).

If they had a normal, calm approach then the whole event probably would not have happened, yes; but the underlying inability to break away from the FD would still be there, only hidden and waiting for some other triggering event.

It's disappointing but not surprising. The FD taboo is so pervasive that I remember one sim where we had a PRM breakaway, which is mandated to be hand-flown (at least for us) due to the time criticality. OK, we get the breakaway vector and I start flying it. Then some other radio-related stuff happens and the other pilot can't get to the FCP to set the heading bug and modes. I turn off the FD as it's a distraction making flying difficult (giving a full opposite turn indication, and the wrong vertical command as well), and the sim instructor scolds me for doing that and decreasing the available SA tools, or some similarly deranged take.

I'm not seeing experience either, until more details are specified. How many of those hours are flying in something other than a straight line or climbing turn, with the FD on?

The SA of the crew was already gone at 10,000’. Intentional non compliance with regard to airspeed? I doubt it, that AS exceedance points back to lack of SA. UAS, already there at 10,000 above the ground, they just didn’t know it. Blindly following the FD was a symptom of the problem.

”FDR data showed that once descent from FL100 commenced, the crew set the Bristol QNH of 1019, deployed the speed brakes and increased the ‘Selected Airspeed’ to 300 knots. This airspeed was selected despite the fact that the maximum permitted airspeed below this altitude in UK airspace is 250 KIAS.”

Discussed here when the report was first published in May 2020:

PPRuNe: Blind use of flight director could have led to crash

Slightly off topic, do any of you folks feel that the automation in modern jets is so complex as to be a hindrance, rather than a help? I fly a DC-9 myself; my automation is anything but complex, and I kind of like it that way. I see so many accident reports referencing different modes of automation, and pilots confused about which mode is selected. Granted, many of these accidents also occur with poorly trained crew, but training aside, has the industry gone too far with the bells and whistles?

I feel I have a firm grasp of the different modes available. I agree with what Vessbot said in post 10- when it becomes a distraction, turn it off.

What being children of magenta, following FDs blindly got to do with this incident? The pilot simply had no idea what to do with vertical displacement closer to runway. By disconnecting automation he wasn't going to get some revelation about what to do. Without any plan in mind he should have declined the offer. But if he wanted to give it try he had to create high ROD at slow speed and not by diving at high speed in clean configuration. Just pull the speed brake(initially), put gear down and configure all the way to Vapp and then descend at the resultant high ROD and would have made it. He went Karachi PIA way. Didn't they fly manually and still crash?

It seems you're addressing an earlier part of the event (the flurry of late and half-understood mode changes in the approach) than the thread title (the descending go-around).

Not sure about you, but that story doesn't do anything to impress me. Sounds more like a deliberate attempt to get one over on a junior instructor.

Most modern aircraft will fly away safely after an engine failure with no control inputs provided there's sufficient excess power and they're not grossly out of trim (used to be a sales technique on the A320 series).. Whether they'll clear the mountain in the distance is a totally different question.

The oldest commercial plane I have experience on was a 1970s design and the newest is a 2000s design. I don't think that automation has advanced to the point of being a hindrance. I recently had a look at some older aircraft's checklists, all of which were considerably longer and more complicated than those of my current steed. Less switching during turn-arounds and during taxiing is a welcome reduction in workload (which helps to offset the additional chores introduced by FMS programming and switching from paper to EFB plus the various CYA paperwork introduced over the last 30 years). Less switching during abnormal ops will leave more capacity for the PM to monitor and the commander to manage.

Transitioning to more advanced avionics designs will result in some "what's it doing now" moments but usually you get to know the possible tripwires rather quickly.

What might be an issue is the lack of experience with "fallback options", e.g. pilots not being used to fly conventional navigation SIDs without FMS guidance any more. That needs to be trained and pilots should be allowed and encouraged to reduce the level of automation used during normal operation to stay sharp. Company policies or individual captains prohibiting this are counterproductive.

When I started flying we were rather proud to be able to use the full operational envelope of the aircraft. Over the years a lot of its edges have been declared no-go in the interest of safety (the theoretical underpinning being that using the edges increase the chances of transgressing them and scraping metal.) While this apparently has improved safety, it has also eroded pilots' abilities to deal with those edges. I am not entirely sure that is good, but far from being able to offer a comprehensive alternative.

What a voice of reason! :D :ok: :)

The average pilot has de-skilled in aviating and transformed to a higher level of being in conducting himself. The latter pushed out the former, made possible through evolution of technology. The overall result is a less-accident prone commercial aviation by order of magnitude or multiples of.

N.b. holes below the waterline need to be plugged no matter the skill-set involved. The newly achieved increased safety level, combined with less exposure to risks, presents a steeper slope to complacency than before. The arousal-stress / performance curve explains.

This interesting Mike Riley blog post deals with the issue.

There was a gentle reminder from management the other day to say that if you are planning a raw data approach, please be mindful of the conditions and brief how you might fix it if it goes wrong. It also reminded us that other operators have recently banned flight without AP/FD unless on the last section of a non-precision approach.

I do tend to very briefly discuss a navigation headache at take-off, but only to the extent of "if all the clever stuff falls over shall we ask for radar vectors?".

Increased automation has lowered the accident rate - I know it's not popular but its true

Look back at the days when most piloting was done by hand.... scarey

Yes but at the same time, pilots flying exclusively with automation are the one crashing airplanes nowadays.
Decent hand flying skills and a good understanding of automation are both essential. Every pilot should feel comfortable to fly without FD and auto thrust at any phase of the flight and that can only be achieved by doing it a in regular basis during line operations. On
my last flight, the FO was really good at managing the automation and descent profile but disconnected the AP late ( around 600 feet, was over-controlling, followed aggressively the FD below 100 feet and ended with a sketchy landing.
I said it just before the last Indonesian 737 accident, and I say it again: Fatal accidents due to pilots being reluctant to hand fly will happen again and more frequently if nothing changes.

Please support with the list of accidents you have in mind. Indication how the conclusion is reached the flightdeck flew manually less than their regional average is appreciated.

One more idea. It is very likely that the US has the most hand-flying pilots in the world, as a group statistics. Why not have a list of N.American accidents and see how that goes.

Contrary to yours, my opinion is that people crash airplanes today not because they can't handle them, not because the tech breaks down down, but because the commander does not take the correct decisions. If true, manual flying is widely recognized as a task that increases mental saturation. The quote from S. Sullengberger's book about his gliding experience (see the blog post link by Alpine Flyer above) is revealing.

About the fact that today's pilots, again as a group, are less precise and less capable with their hand-flying - there's no argument i guess.

Hi FlightDetent,

I agree with you that the commander decision plays a big role in many accidents.
I should have been more precise and said: '' IMHO, those pilots flying exclusively with automation are the one....''
I can not prove it but I would put money on it that in all these accidents where the hand flying skills have been a factor, those pilots did not fly raw data in a regular basis. I mean when was the last time we heard an accident and the pilots were flying raw data?
I can only take my outfit as an example where the most serious incidents, hard landings, or the failure to upgrade as commander are most of the time pilots who never fly raw data in line. I only know one case where a pilot who basically flies raw data every sector who was called in the office for over banking. But again flying raw data every sector...looking for trouble. Same as flying with AP all the time is looking for trouble. There is a place and time for everything.

learn how to learn, and think critically.
 

FlightDetent, …’the fact that today's pilots, again as a group, are less precise and less capable with their hand-flying - there's no argument i guess.”

Yes, and …

“… they’re very bad at generalizing their skills. “We often can’t count on them if the environment differs, sometimes even in small ways, from the environment on which they are trained,”
… the rigid nature, prevents them from tackling problems in open-ended domains.
… need to learn how to learn, be allowed to be creativity, benefit from on-the-fly learning, and think critically.”

The above is copied from an article on Artificial Intelligence, and continues noting AI’s need to be robust, with:-

“Intelligence that, while not necessarily superhuman or self-improving, can be counted on to apply what it knows to a wide range of problems in a systematic and reliable way, synthesizing knowledge from a variety of sources such that it can reason flexibly and dynamically about the world, transferring what it learns in one context to another, in the way that we would expect of an ordinary adult.”
Those are key features missing from current deep learning systems. Deep neural networks can ingest large amounts of data and exploit huge computing resources to solve very narrow problems, such as detecting specific kinds of objects or playing complicated video games in specific conditions.”

Many similarities with views of modern pilot performance; this suggests that if we imagine, train, and expect people to behave like machines, then we will get machine-like behaviours, with all the limitations and consequences of our expectations.
Yet todays industry is increacinly machine like and expects machine like performance.

IMHO that only works out when you group most of the warning equipment added since the 1970s under "automation". (E)GPWS and wind shear warning systems as well as Alpha protection certainly have their share as well, plus more reliable systems have reduced the number of occasions superior skills are required.


As long as there is no certification requirement for AP/FD being available all the time regardless of hydraulics/electrics failure, pilots should be able to fly without it and IMHO that can only be achieved by regularly flying raw data under appropriate conditions. Flying raw data all the time is not the answer, it's being able to choose the correct level of automation for the circumstances. This is not a binary thing either. You can use the autopilot and still botch raw data navigation because you're no longer used to or never have had to make a plan which navaids to tune and intercept in which sequence and/or fail to intercept the next course because you're still struggling to track / trying to intercept the current.


We have a policy that requires all FD/AP and navaid switching to be done by the PM during hand-flying. This can be justified statistically as the chance of the other pilot and the autopilot going belly-up at the same time is minuscule. Should you end up at the wrong end of statistics it might nevertheless be helpful if you can do it. So why not train this and other stuff deemed to dangerous for line ops on the sim with the understanding that it is to be used in case of emergency only. Negative training? Overtraining? I don't think so. I don't think sim session coddling (one malfunction at a time only, etc.) makes for better pilots than having very hard sim sessions where you walk away with a "phew, we just made it" but are pretty sure that you will be able to hand (almost) anything life might throw at you while airborne.


Gliding experience won't help but flying more visual approaches will give you a better "eye" to judge whether you're high or low.

It would be impossible to gather this statistic. But enough anecdotes from personal experience of watching the vast majority of flights over the years, plus discussions with pilots of culture at airlines all over the world, after a while paints a picture of how things are overall. And it doesn't take more than "their regional average," it's the regional average itself that is a problem.

One more idea. It is very likely that the US has the most hand-flying pilots in the world, as a group statistics. Why not have a list of N.American accidents and see how that goes.

I disagree with your model of "the commander does not take the correct decisions." A "decision" implies that the person looks at the alternatives and consciously takes the path of one of them instead of the other... decided to do A vs. B. This picture does not mesh in any sensible way with people having made the "decisions" to attempt an idle-thrust goaround, or slow the plane 30+ knots below approach speed, or watch the airplane roll up into a steeper and steeper bank without doing anything about it, or watch it go further and further below the runway without doing anything about it. To say that the pilot incorrectly decided to do those things seems.... like a farcical description.

No, I would argue that none of those actions were based on "decisions" under any meaningful sense of the word. A was not consciously chosen over B. A (for Automation) was the default, and esablished as such over thousands of hours of consistent operation. For any alternative B to happen, this default has to be overcome, which is extremely difficult when it takes the comfortable mental place of the final backstop of flight control manipulation.

Yes it will and watching carefully if the runway/touchdown-target point is stationary & remaining in one place or "walking up the windshield or down"... ;-)
PJ2

That's why I challenged pinteam as the predicament to his statement was such statisctic. Irony lost in trasnlation, sorry for that.

You're reading too much into it, I am not that versed. Facing the simple choice of saying "making wrong decisions" I elected play smart with "not taking the good decisions". At the time of writing it looked reasonable as I tried to convey observed empirical reality that handflying reduces the brainresource to make decisions (in the true meaning you explain agreeably).

Agreed! That is exactly why I said NOT TAKING the good decisions.

Sorry, now that I noticed you partly try to disagree with the opposite of what I wrote, deciphering what may be relevant in this last paragraph is not compatible with my timezone a.t.m. For the record I never discussed the choice of engaging the AP or not, but the other choices and desicions that need to be solved regardless of automation state, and specifically which of the basic chess-board laouyts (AP=on vs. hand-flown) leaves more spare processing power.

First thought, your opinion: If Asiana@SFO was hypothetically flown remotely by a remote pilot UAV-style, would the captain at chair 0A / 0B had better chance to a) notice b) evaluate correctly c) take preventive action against the speed-loss (combined)? Compared to himself flying manually at the time.

@pinteam Neither am I in disagreement with the concepts.
The point was trying to make that hand-flying skills accidents do not happen anymore in our line of work, ... strangely your last part of the sentece says the same. :confused:

At my company, part of our sim session involves raw data flying. I've been told that the sim instructors quickly identify the people who regularly turn off the FD on the line.

What about reducing the the automation accidents?

:-) "Neither am I in disagreement with the concepts"
Sorry, just pointing out that picking the correct arguments to support each side makes the debate more enjoyable. Pinteam saying more handflying would reduce the number of hand-flying accidents just look at them - I do not see that many at all. There are valid, just, and noble reasons for the case, only this is not one of them.

Is hand-flying being discouraged these days? Absolutly sadly yes. And it seems to bring benefits previously unattainable, but not for the pilot (ego) indeed. Still saddens me.

With the brush already at hand, the other day a couple of guys claimed they need and we all should increase the amount of handflying once back on the line after COVID. Pushing the idea it cannot wait for later months as it its a critical, job-defining skill (it is, okay) and most specifically so that you are ready and up to speed if the black-swan that requires properly polished hand flying comes
.
on the very first day upon returning!?

Stunning logic, really. Either you are trained for it from the SIM or you suck and the airline will be rightfully torn into single rivets at court. Practicing in the first and second week won't help much for that event, would it?

Check Airman We fly the same type and surely your sim checkrides (thinking some extended programme to return after a break) would include the G+Y, which is hand-flown by necessity. For us it is OVC 500 X-W 15 at checkride, OVC 300 X-W 25 during training. Handling the latter well is a pre-requisite to entering the check-phase, where you still need to land it straight and proper, usual tolerances apply.

If on the line solo, you passed this about a fortnight ago. Another 10 day of LIFUS/LOE inbetween to get comfortable and the line-check where you will be asked to do a hand flown approach or disconnect early.

What type of BLACK SWAN do they have in mind you would not be ready for YET? Need help with this ...

IMHO the pilot ego still can (and should) be satisfied by hand-flying under the right conditions. There is no need to prove it during minimal weather (although many airlines still allow hand-flown circling approaches at published weather minima while hand-flying is considered much too dangerous for Cat 2 approaches...), with malfunctions, or with cumulus granitus around. If done under the right conditions it will both satisfy the ego and keep you sharp for the (statistically improbable) moment something fails that takes the autopilot and maybe even FD with it (which AFAIK can happen in all but the very newest flying toys). Restricting raw data flying to the sim won't do as there's simply not enough time given all the other stuff that has been added over the years to be ticked off.

The operating manual of my current (Brazilian) airplane recommends all-time use of FD and use of AP from after TO to DH. There's an added caution that a high level of automation induces the crew to drop out of the loop and develop excessive confidence in automatic systems and a recommendation to occasionally fly manual approaches to keep proficiency. Interestingly we have to fly Cat IIIa by hand using the HGS (which is a bit more of an exercise than watching an autoland but somehow more satisfying, but I'd never dare to do it would I let the AP fly every other ILS down to minimums).

FlightDetent

The reason I support raw data on day 1 is because if we wait for "a few sectors" to get comfortable, I think it's pretty easy to imagine a situation where some people never reach that day. My airline does have training programme set up for those on an extended leave. I'm uncertain what exactly it entails though.

You only have to look at the videos filmed from the flight deck on YouTube to realise that there's a big difference between what you might consider acceptable and what seems to be par for the course at other operators. Landing outside the touchdown zone; hands kept on thrust levers after V1, ignoring GPWS pull-up warnings and describing it as an "approach with the bells and whistles going off".

I occasionally dig out approach videos if I'm going somewhere new, since it's a handy way to get a quick look at what might happen. Unfortunately the end result is often that yet another airline ends up on my own no-fly list.

Airbus FBW; No.
B737 Classic; Yes.

Not so much the complexity, but how well designed and how well integrated the automation is. Some automatics are very good, others are not very well designed or integrated, and in those latter aircraft, it can be more of a hindrance.

The pilots have to understand the automatics to fly safely - it is not just a case of engaging the autopilot and auto-thrust - they must understand exactly how the autos are working. (I can still hear some pilots shouting that the Airbus FBW was fighting against them - no it wasn't).

For me, converting onto the Airbus FBW, I found the automatics to be brilliant. Its automation is so well designed and integrated - including the thrust lever gate switching design - that it assists the pilot extremely well. After the Airbus I found the B737 Classic's automation to be quite patchy. For example, the Boeing auto-thrust was confusing to me - sometimes it was active, other times it wasn't, but I hadn't switched it off ?? - so I found it a lot easier to hand fly the Boeing in certain situations.

Reading about the A350 a while ago, I seem to remember Airbus saying that training now begins with fully manual only for the first couple of days in the SIM, with the automatics being introduced gradually after the basics have been assimilated?

On the OP question, I think that it is not modern automatics per se but Flight Directors that are causing our manual flying to rust. Some fall into the trap of just looking at two crossed lines and following them like some sort of video game. Without any flight directors, we would be scanning the PFD properly - pitch, bank, speed, V/S, altitude, heading - and completely see what was going on - even when the autos were engaged.

I think if we could retain all the automatics but just get rid of the flight directors, then our flying would improve.

Without the FD's you wouldn't know if the automatics are working correctly. That's the reason in an Airbus doing a NPA without vertical guidance the FD is kept on. The bird just floating without a reference will make no sense.