Lion Air crash: Investigators say plane was 'not airworthy' | Hacker News Reader

Lion Air crash: Investigators say plane was 'not airworthy'

bbc.com

54 pointsm_eiman45 comments

For a time, I instructed airline pilots in simulators. The aircraft I taught had a similar mechanism with a "cutout switch" like this Boeing. Often, even when they knew it was coming, the pilots would fail to immediately shut down the system when given a related emergency situation, similar to this one. After learning more about this crash, I immediately thought of the tens of times I've seen this happen in the simulator. Sometimes, I'd pause it and ask the pilots if they've realized their error, other times, I'd simply let them crash (poor instructing on my part). In a previous post, this opinion was called racism/profiling, however, I maintain that extreme incompetence exists in many SE Asian airlines. The particular airline for which I instructed (in the US) had many low-ability pilots, and the reaction of these Lion Air pilots matches theirs exactly. We need to increase automation and, little by little, continue to remove the pilot from the equation until we have pilotless/single-pilot commercial aircraft. Humans simply aren't meant to "become the machine", which is essentially what we're asking airline pilots to do.

Trying to "remove the pilot from the equation" is exactly what caused this crash, by giving primacy to this automatic trim adjustment system over the pilot's control inputs. We need to design systems that are more respectful of the pilot's input (in particular, that respond to the pilot's obvious intent rather than requiring the pilot to use some unusual cutout switch to achieve the most basic aeroplane functionality of: push stick forward => pitch down and vice versa).

I'd consider that even more dangerous. Stick pushers actually activate before the plane stalls. These large aircraft are designed in a way that a full stall could be catastrophic. Stick shaker events (the thing that happens before the stick pusher) happen quite frequently due to pilot inattention. In this case, an accident occurred, however, more frequently, they save lives.

A traditional stick pusher is a much better design: the pilot has clear visibility into what's happening, they can see that the plane is pitching down because the stick is being pushed forward. Pilots do make mistakes but so do automated systems; the way forward is to build systems that collaborate with the pilot rather than cutting the pilot out.

If the aircraft were being piloted manually, this crash wouldn't have happened. The increasing imposition of automation and removing the pilot from the equation was the cause of this crash. Your recommendation is to take this further, rapidly?

First of all he said little by little so calling it rapid is a bit misleading. Second this was not ultimately a hardware/software failure but the failure of the overall supervisor of the system(= pilot). We can't (yet) make hardwares that are immune to failure but we have redundancy and procedures to deal with these failures. It seems like in this case the procedures were not properly executed which lead to the tragedy. To also address @lmm's point: the leading cause for aircraft accidents are human error and not hardware/software failure so more automation = less accidents. Edit: after reading some more on the subject it seems like boeing share some of the blame for not including the new system(or system changes) in their training program which is again a human error.

Is pilot error more or less of a cause of fatal aircraft incidents than automation gone awry? What does the data say?

The FAA thinks that cockpit automation leads to complacency. It's a question on the commercial pilot written test.

There's a talk on youtube (Children of Magenta, https://www.youtube.com/watch?v=pN41LvuSz10) given to an audience of pilots, it's a complicated issue.

> We need to increase automation and, little by little, continue to remove the pilot from the equation until we have pilotless/single-pilot commercial aircraft.

there seems to be a growing consensus in the self-driving car debate that incremental automation is actually pretty dangerous, or at least much less safe than either fully manual or fully autonomous control, to the point where some consider it irresponsible to sell L3 autonomous vehicles at all. is there a particular reason why planes should be different?

I have never understood what problem self-driving cars is trying to solve. I can, on the other hand, see that it could be more desirable to have self flying planes. But to even think that self flying = issue free and non crashable planes is beyond naive. Personally I believe that the best is to let humans do what they do best, and let computers do what they are good at. Without trying to force one into the other's shoes. (Computers has shoes, don't they?)

Self driving cars solve two problems: 1. driving is boring, 2. driving can lead to accidents if you are not attentive at the right moment.

That's why the get into the simulator, so they can learn good habit. This has nothing to do with pilots, but with humans and losing track of time when trying to troubleshoot something.

I recommend folks read "Black Box Thinking: Why Most People Never Learn from Their Mistakes"

Preliminary report: https://www.flightradar24.com/blog/wp-content/uploads/2018/1...

As far as I can tell, one of the angle of attack sensors failed on the previous flight, the aircraft started to continuously trim nose down, the pilots trimmed it back up again a few times before switching off the automatic trim motors, declaring PAN-PAN, and continuing the flight. They reported the fault, the engineers "fixed" the wrong thing, and on the fatal flight it started playing up again. Except this time, the pilots didn't switch off the automatic trim motors, but just kept re-trimming nose up to counter the continuous automatic trim nose down. Eventually they got overwhelmed and crashed.

It's a tragic accident, and blame can be spread all round. It's good that the previous flight were able to work it out, and terrible that the fault wasn't properly fixed, and the subsequent flight didn't work it out. Also, that the aircraft wasn't resistant to this particular fault.

Sad.

One pilot was able to figure out the problem, find the solution and land safely. The next pilot did not and hundreds died.

Maybe I am confused, but it doesn't look from what I read in articles that the crash was caused by the problems that made the plane not "airworthy". If that's the case, it certainly needs to be flagged. But I don't get why that makes the headline.

When I was in Indonesia, about 8 years ago, Lion Air had the ominous badly translated slogan “We make people fly”. They had a bad track record at that time, too.

Edit: Apparently they still do, https://mobile.lionair.co.id/

That doesn't sound badly translated, does it? Lots of corporate slogans are that kind of vague phrase. They mean 'fly' as in 'soar', 'achieve', that kind of thing, as well as literally meaning they will fly you to your destination. Could have come from an English company, to my ears, and doesn't sound ominous. I think you're reading too much into it.

One very negative moment for this slogan was the moment one of their pilots was very audibly high on drugs on the PA before a flight.

Is ‘flying’ code for being on drugs? Most people don’t know about drug culture and wouldn’t make these kind of connections. I think it’s just an generic marketing phrase.

'Fly' (the English word) is an euphemism in Indonesian for being high. It's just unfortunate that they picked that phrase.

I like the way the news media tries to point at just one party to blame for any incident, and argues about which party should be to blame, when, like any incident, it was probably caused by tens of coinciding factors, and 'blame' really ought to be apportioned to all parties involved.

This is not the news media trying to point at just one party, it is the reporting of Boeing doing so in order to deflect the criticism that has been directed towards it. US pilot's unions are divided over whether they think Boeing made a serious error in not disclosing the existence of the new system in question, but the fact that many pilots apparently think it was an error counts for something - in addition to actually flying aircraft, professional pilots are our safety watchdogs.

From the article:

> The findings by the National Transport Safety Committee (KNKT) suggest that Lion Air put the plane back into service despite it having had problems on earlier flights.

Seems this implies there are at least two parties involved: airline and aircraft manufacturer.

Let's also not forget the pilot. They are supposed to read the manual so they can respond to a badly behaving auto pilot / guidance system. There is always a pilot override. I guess you can bundle that in to airline blame though (lack of training or bad org structure). But really, this is a new plane and Boeing probably don't have adequate software testing in place for the newly written / ported software.

The FAA issued a mandatory airworthiness directive as a result of this crash. Boeing didn't include training for a new anti-stall automation regime in their type cert update class for existing 737-rated pilots moving to the 737 MAX. There was apparently a malfunctioning angle of attack sensor, which caused a software data conflict which caused the anti-stall assist (which the pilot wouldn't have known existed) to repeatedly trim the aircraft nose-down.

There is also the possibility that a separate pilot assist system read a possible near-stall and enforced difficulty in pulling back on the stick (to prevent what it saw as a potential stall). So the pilot couldn't pull out of the dive that the MCAS (trim changer) had caused.

The blame for this incident does not lie on the pilots failing to RTFM. Boeing changed several systems without telling anyone. Now the plane's been type rated, so software changes to fix it are significantly difficult.

Imagine Boeing as a guy saying "Screw those idiots who operate our multimillion dollar safety-critical product; they're too stupid to know how to work it!" This is a closer perspective to the truth than the parent's comment.

Mentour pilot overview: https://www.youtube.com/watch?v=zfQW0upkVus FAA Emergency Airworthiness Directive: http://rgl.faa.gov/Regulatory_and_Guidance_Library/rgad.nsf/...

> Boeing didn't include training for a new anti-stall automation regime in their type cert update class for existing 737-rated pilots moving to the 737 MAX. There was apparently a malfunctioning angle of attack sensor, which caused a software data conflict which caused the anti-stall assist (which the pilot wouldn't have known existed) to repeatedly trim the aircraft nose-down.

If this is accurate, it's beyond criminal. If they're putting pilots in planes with automated systems without any knowledge of those systems, let alone how to operate/override them, someone needs to be behind bars and is responsible for the deaths of those people.

No, you need certification to be able to fly eact plane type, which equal to 6 months retraining. For this incident Boeing "forgot" to tell companies do not do what you usually do, as it doing what you do actually worsen the situation. Boeing released emergency notice after that incident, and data from black box showed that the pilot is doing precisely what they are used to do.

Plane it self is legal to fly at that time.

> do not do what you usually do, as it [...] worsen the situation

This sounds like the worst UI evolution. Could they not have come up with something that's more natural? I don't know anything about controls and procedures for commercial aircraft, but it is so complex that nothing is natural? Like listening to Formula1 team radio where the engineer tells the driver to set some cryptic mode in the steering wheel menu.

I wonder how much of a competitive advantage it would be to make easier/safer to fly commercial aircraft. It would be sad to hear that these are the results of great effort of that goal.

But isn't moving "STAB TRIM" to CUT OUT a memory item even in the 737 NG in the case of a trim runaway?

Yes. The newfangled plane had a new way for trim runaway to occur, but the action to take in the event of trim runaway shouldn't be (AFAICT) any different. They managed it on the previous flight.

Not necessarily. Lion Air could be responsible for the problems on earlier flights too (through poor pilot training or maintenance procedures).

But did these problems caused the crash?

Yes they did. The previous pilot figured out a workaround in time, not this one.

One of the most heavily regulated industries on the planet, and all this automation still caused a devastating crash.

How are we supposed to trust autonomous cars when the smartest engineers on the planet can't automate the condition of not diving a plane into the ground.

Exactly this. People are blaming the pilot, Lion Air, SE Asia, but ultimately a Boeing aircraft decided that it was very important to pitch down and dive into the sea. How many backup sensors and alternate systems have to fail/pass before this course of action is taken? Does the angle of attack sensor not check what a backup sensor is saying, and then what the altimeter is saying, airspeed sensor, and on and on. If a single sensor failing can make this happen then the sensor shouldn't exist.

Yes, it's truly a case of inadequate engineering IMO. If an automated system designed to prevent crashes (stalls, as I understand it) merely changes the failure mode into something harder to detect and correct manually, it's a poorly designed system.

A sibling comment mentioned that the sensors fail constantly. Seems that it might be possible to operate such a system with a significant number of critical inputs in a degraded state which could lead to unknown effects IMO.

In many ways, automated driving is much more complex than successfully piloting an aircraft in regulated airspace.

The previous flight had PIC stick shaker activation for the entire flight. PIC gave control to SIC. PIC realized the trim would stop moving with the STAB TRIM switches to CUTOUT. Flew the rest of the flight with manual trim and no autopilot, stopping at FL280 for RVSM.

Sensors fail daily on hundreds of planes. Virtually everything down to radios is able to be switched off at the circuit breaker level.

The responsibility is on the pilot to identify and switch off anything malfunctioning.

> How are we supposed to trust autonomous cars when the smartest engineers on the planet can't automate the condition of not diving a plane into the ground.

Easy: air travel is still safer than car travel by orders of magnitude. Treating an incident like this one is an avoidable tragedy with lessons to learn is important, and thankfully this is changing for car crashes slowly [1], but a single failure of an autonomous system does not mean that it's not at all viable, because the alternative, purely human driving, is clearly a lot worse.

1: https://www.vox.com/2015/7/20/8995151/crash-not-accident

This happened in Indonesia. SE Asian domestic airlines have abysmal safety records, and most of them are on blacklists that prevents them from flying anywhere else for that reason.

>"In our opinion, the plane was no longer airworthy and it should not have continued," he said. The committee's report itself, though, does not spell out that conclusion.

Well that's kind of a useless conclusion without the report having anything to back it up. It's really easy to pat yourself on the back and say "yup, wasn't air/road/sea-worthy" in hindsight after there's been an accident.

>Pilots tried to correct this by pointing the nose higher, until the system pushed it down again. This happened more than 20 times. It is unclear why the pilots did not employ procedures to disable the automated system.

This is the million dollar question. Why didn't the pilots turn off the system and fly in "legacy mode" like they would if they were flying the prior revision of the 737.

>This is the million dollar question. Why didn't the pilots turn off the system and fly in "legacy mode" like they would if they were flying the prior revision of the 737.

As a pilot, if my plane is doing something which I don't expect it to - my first thought isn't to disable auto-trim, but to consider why it's doing that. A previous crew did have the issue, and did report the problem. As a pilot, I trust that it has been fixed, so no, it wouldn't be the first thing on my mind.

From what I understand from reading PPrune forums, pulling the circuit breaker for trim irregularities is the de facto standard operating procedure on the 737, and has been for over a decade+.

Hopefully they find the CVR and that gives us more insight into what the crew was thinking. It is odd that the pilots struggled with trim controls for ~11 minutes and didn't think to disable it.

The media has focused on the new MCAS system on the 737-Max, but I think that is somewhat irrelevant, considering that the procedures for addressing trim irregularities non-Max models are the same: pull the trim circuit breaker.

The full quote just said there’s less ambiguity with the title:

“Nurcahyo Utomo, aviation head at the National Transport Safety Committee, said it was "too early to conclude" whether the anti-stall system had contributed to the crash. He added that the plane experienced similar problems during its previous flight from Denpasar in Bali to Jakarta. "In our opinion, the plane was no longer airworthy and it should not have continued," he said. The committee's report itself, though, does not spell out that conclusion.”

Reply on news.ycombinator.com