Hacker News Reader

Comment by YeBanKo | Hacker News Reader

Maybe the reason why Lufthansa does not allow visual approach at SFO at night, could be after the Air Canada near miss.

> The NTSB determined the probable cause was the Air Canada flight crew's confusion of the runway with the parallel taxiway, with contributing causes including the crew's failure to use the instrument landing system (ILS), as well as pilot fatigue.

FAA changed the rules for SFO and made visual approaches forbidden at night "when an adjacent parallel runway is closed" [2]. Maybe Lufthansa plays it safe and requires ILS for all long haul night landings.

[1] https://en.wikipedia.org/wiki/Air_Canada_Flight_759

[2] https://www.flightglobal.com/faa-changes-san-francisco-landi...

View on HN

There's also this crash that happened while ILS was broken: https://en.wikipedia.org/wiki/Asiana_Airlines_Flight_214

This was ultimately pilot error but also due to the fact that the pilots were not as accustomed to doing visual landings.

Pilots are not going to be accustomized of visual landings in case of emergency, if you have a corporate policy that forbids them in controlled setting!

This one is bizzare, not only European pilots on average have less experience than US ones, but they are not allowed to gain experience by corporate policy.

(This one coming from a country where aviation is a big mess)

You do not train with passengers on board, period.

That is simply not how things work. Every airline trains with passengers on board.

What about the rookie copilot with 50 hours on type?

We don't test software systems in production either. Except we do.

We should be testing in non-production environments, but non-prod is not the same as production, so it is inevitable that you will be testing a system change when you deploy to prod. If your non-prod environment is similar enough that there are NO system changes when you deploy to prod ... then your non-prod env is actually prod, because that would mean that your non-prod network routing, auth, and database are the same as prod, because changing any one of those would be a system change.

A pilot's 'flight hours' is regarded as a key measure of their experience.

If a pilot only ever trained visual approaches in sim, and not the real world accounting for the overwhelming majority of their flight experience, that number would be meaningless.

"Experience building" by flight hours (and in properly managed organizations, regular "manual" flying) is not the same as "training", at least in local vocabulary as used by my flight instructors and other people I know in aviation.

Thus somewhat mixed reading of what I wrote earlier. Was obvious to me, from responses wasn't obvious to others.

It can absolutely make sense to "train" with passengers on board.

Automation complacency or simply being out of practice can have significant repercussions in emergencies.

Consider a task where humans have a marginally higher error rate than the autopilot. Obviously, per task, the safe thing to do is to let the autopilot fly.

But now consider a more difficult scenario where the autopilot just bails out, e.g. due to ILS being inoperable, wind shear, any kind of non-permissible sensor discrepancy on board etc. – how would you like be on a plane flown by a pilot that has never done a manual landing in the past few months or even years due to company policy?

Of course, a night approach at a foreign airport might not be the best opportunity to practice a manual landing (and it looks like the planned arrival time was during daylight hours!), but as a general policy, allowing some manual flying seems like a very good safety practice, at least until autopilots can fly and land in 100% of all scenarios.

Some of it might be terminology issue.

For me, a training flight is a flight that is written with appropriate "exercise" number into logbook and is part of the official training program. These days most airlines do those on simulators, and they will include "rarer" operations like rarely used approaches (2xNDB instead of ILS, how to fly DME arc approach, etc), instrument failures, and so on.

You do not do such flights with passengers, because you're intentionally lowering the safety level.

In addition to that, there's a recommendation to randomly, given safe conditions, do regular manual operations. Mind you, those SOPs probably explicitly say "do not experiment after long haul flight".

I might have been a bit more cross because people have died (including in ways that impacted me) because of organizations that mixed "training" and "passenger" flights.

You wrote: <<This one coming from a country where aviation is a big mess>>

Can you provide more details?

Russia - a history of crashes caused by rapid growth and weird aircraft park, followed by the war where the whole industry is now out of international law.

What does Russia have to do with anything?

Visual landings are only disallowed at night. Get your facts straight.

What if they have to do one in emergency scenario, and they only ever did this in daylight?

>> Pilots are not going to be accustomized of visual landings in case of emergency

> Visual landings are only disallowed at night

I don't agree with the rest of what he said... but he actually has a valid point here. Emergencies happen at night too, and sometimes the nature of the emergency will make an instrument approach impossible.

The visual approach isn't an issue, it's the requirements to maintain visual separation. Under emergency conditions, you'd expect the separation to be handled by ATC.

> Under emergency conditions, you'd expect the separation to be handled by ATC.

ATC is never responsible for general traffic separation in visual meteorological conditions in the United States.

ATC is responsible for separating IFR airplanes from other IFR airplanes. When flying below 18000ft MSL in VMC, those IFR aircraft are still responsible for looking out the window to see and avoid VFR aircraft who may not be talking to ATC (and to be clear, the VFR aircraft is jointly responsible).

Just to drive this home: it is still 100% legal to fly around in a Piper Cub at 9500ft MSL with no transponder or radio in most of the country. It is legal to fly around VFR in a jet at 17500ft squawking 1200 not talking to ATC almost everywhere in the country. Primary radar coverage is going away, not getting better. You have to look out the window.

Asiana 214 crashed (in broad daylight and perfect visibility) because Asiana pilots were overused to landing on the autopilot, which was off on this landing; also because they ignored the "sink rate" warning going off for a minute.

Not because landing at SFO is particularly dangerous.

https://en.wikipedia.org/wiki/Asiana_Airlines_Flight_214#Cra...

This is a bit simplified. See https://youtu.be/Toy92_qriZU?feature=shared

Reply on news.ycombinator.com