Air India Crash

[Slam62]

If it was a software error or a procedural error, that would have huge repercussions.

I wonder if they have issued any sort of directive to mitigate the risks.

Either way a software error would do Boeing a lot of harm.

 

[adam cool dude]

If it was a software error or a procedural error, that would have huge repercussions.

I wonder if they have issued any sort of directive to mitigate the risks.

Either way a software error would do Boeing a lot of harm.

As long as the bean counters are in charge at Boeing nothing will change.

 

[Disco_P]

Youtube chose to recommend this video to me today

I assumed it had been picked up by the algorithm due to the recent crash but the last comment was 7 years ago, the third last comment is

 

[Penfold101]

Youtube chose to recommend this video to me today

I assumed it had been picked up by the algorithm due to the recent crash but the last comment was 7 years ago, the third last comment is

Classic YouTube comment section.

 

[Alibaba99]

An alternative is that the tanks had condensation inside, causing either corrosion or icing but this was on the ground.

Following the BA belly flop into Heathrow GE (who admitted they'd had several instances of fuel icing at altitude) / RR / Airbus and Boeing spent a lot of time and money researching fuel icing as it turned out that fuel could freeze at temperatures and conditions previously thought to be safe. Whilst it is always possible that other conditions were missed, it would be surprising.

 

[200sols]

It was 42c on the tarmac? Hard to believe this would be frozen fuel and again the likelihood of that taking down both engines at once?

 

[NickK]

It was 42c on the tarmac? Hard to believe this would be frozen fuel and again the likelihood of that taking down both engines at once?

An odd one could be a fire but not an explosion. The result is that the fire reduces the volume of gasses in the fuel tank, which then creates a partial vacuum. However that requires the nitrogen to fail and the vents to be closed or blocked (perhaps as part of something that went bang).

I could hallucinate possible scenarios all day, but best wait for the official findings.

 

[Alibaba99]

It was 42c on the tarmac? Hard to believe this would be frozen fuel and again the likelihood of that taking down both engines at once?

Look up BA Flight 38. That was a twin engine flame out due to icing in the fuel / oil heat exchangers. Icing is a physical phenomenon that can affect both engines simultaneously as the conditions are consistent around the aircraft and the systems are mirrored for each engine.

 

[200sols]

Look up BA Flight 38. That was a twin engine flame out due to icing in the fuel / oil heat exchangers. Icing is a physical phenomenon that can affect both engines simultaneously as the conditions are consistent around the aircraft and the systems are mirrored for each engine.

Thing is Ba38 was landing, hadn't been sat in 40+c heat.

 

[adam cool dude]

It was 42c on the tarmac? Hard to believe this would be frozen fuel and again the likelihood of that taking down both engines at once?

It isn't temperature that is a problem but dew point of the air. Humid air is a big problem. The light aircraft I fly can get carb ice even at 30+ degrees. It is part of our checklist on the downwind leg of a circuit to pull out the carb heat to make sure we have maximum power available if we have to initiate a go around. Same with descending altitude at a low rpm to stop the carb icing up. It is also part of our "FREDA" checks whilst in cruise to make sure carb is operating at best performance.

Things can get very cold with moist air. Especially in the venturi of a carb then add the cooling properties of fuel.

 

[200sols]

It isn't temperature that is a problem but due point of the air. Humid air is a big problem. The light aircraft I fly can get carb ice even at 30+ degrees. It is part of our checklist on the downwind leg of a circuit to pull out the carb heat to make sure we have maximum power available if we have to initiate a go around. Same with descending altitude at a low rpm to stop the carb icing up. It is also part of our "FREDA" checks whilst in cruise to make sure carb is operating at best performance.

Things can get very cold with moist air. Especially in the venturi of a carb then add the cooling properties of fuel.

You're talking about in flight though not on the ground.

 

[adam cool dude]

You're talking about in flight though not on the ground.

You can have frozen carb on the ground too. Especially if the plane hasn't had time to cool down and is in constant rotation. Carb heat checks are part of my pre flight checks too and several times I have had icing on a beautiful hot day.

 

[esmozz]

Look up BA Flight 38. That was a twin engine flame out due to icing in the fuel / oil heat exchangers. Icing is a physical phenomenon that can affect both engines simultaneously as the conditions are consistent around the aircraft and the systems are mirrored for each engine.

A quick check of the AAIB report in that event suggests that the engines failed 7 seconds apart. I think there's a distinct possibility that both engines failed (or rolled back) at the same time in the Air India crash.

Also, guys, carb icing on piston engined light aircraft has nothing to do with what happened in the Air India crash,

 

[ChrisD.]

Our BU engineering teams sometimes take days, if not weeks to provide a proper RCA for an outage. There are hundreds of log files with thousands of lines in each. For one appliance. And there might be 6+ appliances involved. Timestamps from different services which produce their own log files have to be systematically pieced together to provide a clear picture. Then in most cases the underlying code has to be checked for that particular product version to see where a fault may have occurred, or how some condition somewhere may have triggered a bunch of events. Then once that is done labs are built to reconstruct the failure scenario to see if it can be repeated.

I have zero idea about how an aeroplane logs data, but it's not difficult for me to realise that despite the relatively short flight path, it takes an awful lot of time just to debug the logs. If they can get them - the FDR might have to be flown somewhere for them to piece the data together. Then they have to trawl through the wreckage, flight history, maintenance history, things at the airport - fuel deliveries etc. Then all of the staff work history, training etc. I can't even begin to comprehend the level of effort it takes to produce a preliminary report.

 

[squerble]

Look up BA Flight 38. That was a twin engine flame out due to icing in the fuel / oil heat exchangers. Icing is a physical phenomenon that can affect both engines simultaneously as the conditions are consistent around the aircraft and the systems are mirrored for each engine.

Yea, on a plane that had just spent a few hours in -60ºC air, not on a plane that had been sat on the tarmac in 42ºC heat, having just been filled with fuel from fuel tanks in the ground at not far off that temperature.

Are people seriously suggesting icing here?!


EDIT: It also wasn't a flameout, the engines just rolled back to around flight idle due to restricted (but not completely cut-off) fuel flow.

 

[Mercenary Keyboard Warrior]

Our BU engineering teams sometimes take days, if not weeks to provide a proper RCA for an outage. There are hundreds of log files with thousands of lines in each. For one appliance. And there might be 6+ appliances involved. Timestamps from different services which produce their own log files have to be systematically pieced together to provide a clear picture. Then in most cases the underlying code has to be checked for that particular product version to see where a fault may have occurred, or how some condition somewhere may have triggered a bunch of events. Then once that is done labs are built to reconstruct the failure scenario to see if it can be repeated.

I have zero idea about how an aeroplane logs data, but it's not difficult for me to realise that despite the relatively short flight path, it takes an awful lot of time just to debug the logs. If they can get them - the FDR might have to be flown somewhere for them to piece the data together. Then they have to trawl through the wreckage, flight history, maintenance history, things at the airport - fuel deliveries etc. Then all of the staff work history, training etc. I can't even begin to comprehend the level of effort it takes to produce a preliminary report.

Apparently recent aircraft black boxes record readings from hundreds of sensors. Stuff like the 747 was a handful.

 

[NickK]

Our BU engineering teams sometimes take days, if not weeks to provide a proper RCA for an outage. There are hundreds of log files with thousands of lines in each. For one appliance. And there might be 6+ appliances involved. Timestamps from different services which produce their own log files have to be systematically pieced together to provide a clear picture. Then in most cases the underlying code has to be checked for that particular product version to see where a fault may have occurred, or how some condition somewhere may have triggered a bunch of events. Then once that is done labs are built to reconstruct the failure scenario to see if it can be repeated.

I have zero idea about how an aeroplane logs data, but it's not difficult for me to realise that despite the relatively short flight path, it takes an awful lot of time just to debug the logs. If they can get them - the FDR might have to be flown somewhere for them to piece the data together. Then they have to trawl through the wreckage, flight history, maintenance history, things at the airport - fuel deliveries etc. Then all of the staff work history, training etc. I can't even begin to comprehend the level of effort it takes to produce a preliminary report.

Historically it can take a large amount of time. However (and I'm going to use the term) AI in a form specifically for incident analysis of large volumes of logs can reduce this considerably. Furthermore it's possible for the AI to work with audio and all the signals to model.

I've owned an enterprise cloud platform for 11 AWS orgs (that's more AWS orgs than AWS publicly allow for a company, essentially think of that as a group of 6 sub-companies with two AWS clouds each for non AWS techies). You can now guess the complexity of tracing incidents from lift and shift data centre migrations into that! Now when you have incidents, with over 1600 servers for one application tracing issues can only be done by automated means - enter data science and at the end AI. It branches out beyond the application logs, and takes data from the infrastructure and other non-application 'events'. TL;DR is that it's possible to bring non-traditional silo'd information for processing with automated processing that data science now provides.

So I suspect there's a large push to use automated forensics by insurance companies and operators to reduce the costs of air investigations.

 

[esmozz]

Apparently recent aircraft black boxes record readings from hundreds of sensors. Stuff like the 747 was a handful.

Modern black box recorders certainly record a lot more data than they use to and are designed with survivability in mind, only being accessed after an incident. However all modern aircraft have quick access recorders (QARs) as well, which I think record even more parameters. These are less robust and basically record onto SD cards. Often they upload to the mobile phone network and are routinely analysed shortly after the flight has landed. Despite not being designed with survivability in mind, nonetheless they may well survive and be able to be offer a lot more data to investigators.

I'm pretty sure the engines also record and store their own data independently of other aircraft recorder devices which again, may be recoverable. The FMC's as well. I'm sure there are others.

The point being, as mentioned above - there is a vast amount of data to wade through in a air accident investigation.

 

[Penfold101]

Modern black box recorders certainly record a lot more data than they use to and are designed with survivability in mind, only being accessed after an incident. However all modern aircraft have quick access recorders (QARs) as well, which I think record even more parameters. These are less robust and basically record onto SD cards. Often they upload to the mobile phone network and are routinely analysed shortly after the flight has landed. Despite not being designed with survivability in mind, nonetheless they may well survive and be able to be offer a lot more data to investigators.

I'm pretty sure the engines also record and store their own data independently of other aircraft recorder devices which again, may be recoverable. The FMC's as well. I'm sure there are others.

The point being, as mentioned above - there is a vast amount of data to wade through in a air accident investigation.

This is my concern - assuming uncommanded engine shutdown, I suspect we’ll be able to see that the signal is sent to both controllers to shut fuel off simultaneously and probably which box has done it, but seeing WHY it’s sent that command it’s going to be the difficult part.

 

[Mercenary Keyboard Warrior]

Modern black box recorders certainly record a lot more data than they use to and are designed with survivability in mind, only being accessed after an incident. However all modern aircraft have quick access recorders (QARs) as well, which I think record even more parameters. These are less robust and basically record onto SD cards. Often they upload to the mobile phone network and are routinely analysed shortly after the flight has landed. Despite not being designed with survivability in mind, nonetheless they may well survive and be able to be offer a lot more data to investigators.

I'm pretty sure the engines also record and store their own data independently of other aircraft recorder devices which again, may be recoverable. The FMC's as well. I'm sure there are others.

The point being, as mentioned above - there is a vast amount of data to wade through in a air accident investigation.

Indeed with just about everything being processor controlled its basically trivial and in terms of the cost of aircraft "free" to add data recording at that point.
Plus if your going to want to check eg engine performance you would ideally just have data for engine performance which is easy to access compared to the black box.

I saw an engineer say that the latest black boxes store data from over 1500 sensors. Up from 4, a few decades ago.