MH17 shot down. And ATR 72 crashing in South Asia. Now Algerian Air flight disappearing.

What the hell is happening? It's a bad week for sure, and a bad year as well, for civil aviation safety.

I'm left wondering if there are any reasons for that, or if it's just terrible luck.

Sheer tough luck and unfortunate circumstances in my view. It's happened before that accidents have come in two's or three's within days then nothing for many many months.

Quoting rolfen (Thread starter): South Asia.

Not to be pedantic, but Taiwan is nowhere near South Asia. Here in the East, Taiwan is usually considered either East Asia or Northeast Asia. South Asia typically refers to the subcontinent, e.g. India, Pakistan, Sri Lanka, Bangladesh, and Nepal (and Bhutan?).

What kind of reasons are you looking for? As far as it appears (yet to be confirmed), you can blame this week on the Russians (MH17) and bad weather (GE 222 and AH 5017). Not sure what kind of other explanation you are looking for.

Cheers
Coal

And we wonder why the FAA banned US carriers from flying into a war zone.

Seems like caution needs to become more the rule than the exception. If you need to fly around wars, you should do it, even if it means a few more pounds of fuel or a few million in lost revenue.

Why would you assume there's a reason at all?

There will be a certain amount of aircraft accidents in a year. Nowhere in the laws of statistics does it say that they have to happen at regular intervals.

But then, there will always be the:



And the:

Poisson distribution.


David

Well, the last few years haven't had too many noticeable incidents, right? I guess the lack of crashes is starting to catch up.

Quoting Coal (Reply 2): bad weather (GE 222 and AH 5017).

Sorry, bad weather didn't bring down the TransAsia ATR and it's too soon to know about AH5017. Planes fly thru or around bad weather all the time. Just like in other "bad weather" crashes I think we will find out that someone's poor judgement caused the crash.

Nevertheless, commercial airplane disasters do happen from time to time. It just so happens that we've had 3 unrelated disasters in the same week.

Quoting Coal (Reply 2): What kind of reasons are you looking for?

I'm not looking for anything in particular, just sharing my thoughts. If there is any reason for more accidents, it could be an increase in traffic, for example.

Quoting Coal (Reply 2): Not to be pedantic, but Taiwan is nowhere near South Asia.

Thanks for the correction. I wasn't sure where it happened so I though South Asia was a safe bet   it still is the Southern part of Asia, right?

Chemtrails! It's all the fault of those darned chemtrails!

Seriously now, this is life! Shit happens. The only constants is life are death and taxes. Everything else is sheer luck (good or bad).

I have read stories where incredibly good luck saved someone but in the same incident, incredibly bad luck killed another.

Do I turn left or right? Or should I go straight on or turn around?
I might die right after typing this or live on to a ripe old age.

No one who, what, where, when and why before hand. That's life!

Is not the old saw that plane crashes, like celebrity deaths, happen in threes?

[Edited 2014-07-24 07:55:44]

Quoting PHX787 (Reply 6): Well, the last few years haven't had too many noticeable incidents, right? I guess the lack of crashes is starting to catch up.

Yeah, i think so too. Law of averages catching up. Nonetheless pretty sad week for aviation.  

Quoting vvbkumar (Reply 11):

If I may:

Chaos Theory does a better job than Classical Statistics.

The law of averages works when you are throwing dice, since axiomatically the dice can only fall in a fixed number of combinations. You can then compute an 'expected' relative frequency for a number to appear. The law of large numbers gives us an anticipated distribution that we can assign probability intervals.

Given that many airline accidents are either "Swiss Cheese" oriented or the result of unpredictable and complex behavior, I would go with Chaos Theory when dealing with systems of high complexity. Otherwise I could set up a machine learning program and train it with thousands of variables to predict the safety of each flight! (Hey, there's a Kickstarter idea!).

Finally, the concept of "We are due for one" is also a common error, known as "The Gambler's Fallacy". Untold sums have been lost at betting tables because of this - random outcomes are not serially correlated!

Disclaimer: I am not an aviation safety professional, and glad to be corrected.
Pet Peeve: A lifetime of watching the misuse of Statistics in the world of Finance.

Cheers!

The likelihood of a crash, all things being equal, is just as likely after 10 terrible crashes as it is after 5 years with no mishaps. There is no "catching up" and we aren't any safer now for while now that a few have crashed.

I will add a small disclaimer, I'm talking about all things being equal. Crashes can affect each other sometimes, for example, we probably won't have a shoot down over Ukraine for a while. But as far as MH370 being related to MH17 or the ATR or this crash, nope

Quoting rolfen (Thread starter): I'm left wondering if there are any reasons for that, or if it's just terrible luck.

Because random does not equal even/uniform.

[Edited 2014-07-24 11:28:57]

The United States is still experiencing a period of incredible safety. 0 passenger deaths by a US 121 operator since 2009.

Quoting 32andBelow (Reply 15): The United States is still experiencing a period of incredible safety. 0 passenger deaths by a US 121 operator since 2009.

The thing is, airline safety is much more than just counting bodies. Even a near miss where no one gets hurt is a very serious safety incident.

Quoting comorin (Reply 12):

I learn something new everyday on this forum 

Quoting airbazar (Reply 16): The thing is, airline safety is much more than just counting bodies. Even a near miss where no one gets hurt is a very serious safety incident.

It can be used as a measure against other time periods in domestic aviation. A controller may mess up and clear two planes too close, but systems such as TCAS/and soon ADS-B will help provide a safety net in these situations. The sheer fact that there are 0 passenger deaths in hundreds of million flights is astonishing for any form of transportation.

Aviation is going to have to get safer, and for reasons of perception rather than reality. The stratospheric rise in traffic over the last five years *will* by definition lead to an increase in the number of accidents, if the accident rate remains the same. It will therefore reach a point where the public perceive air travel is becoming "dangerous" (where of course it isn't; there's just more of it). To maintain public confidence therefore, air travel will need to become progressively safer.

Bad things tend to come in threes for some reason.

Quoting Coal (Reply 2): What kind of reasons are you looking for?

Quoting Coal (Reply 2): Not sure what kind of other explanation you are looking for.

Quoting francoflier (Reply 4): Why would you assume there's a reason at all?

      . Does there have to be a reason.

Quoting rolfen (Reply 8): I'm not looking for anything in particular, just sharing my thoughts. If there is any reason for more accidents, it could be an increase in traffic, for example.

Increase in traffic? Can you please elaborate.

Quoting flightsimboy (Reply 21): Increase in traffic? Can you please elaborate

One accident per million flights = more accidents if there are more flights.

Quoting Stitch (Reply 10): Is not the old saw that plane crashes, like celebrity deaths, happen in threes?

Quoting jolau1701 (Reply 20): Bad things tend to come in threes for some reason.

It's an old wives tale and totally not supported by the facts.

There have been seven fatal commercial crashes this year, two fatal military aircraft carrying large groups of civilians.

(I'm not counting the military aircraft shot down in conflict areas)

Quoting usdcaguy (Reply 3):

Actually, Taiwan is Southeast Asia indeed. It's sure not Northern Asia, if anything its East Asia, but Wikipedia says it's Southeast Asia.

Not to be pedantic......

Cheers.

Simple probability. With all the weeks without crashes, it was only a matter of time before there was a week with them. You didn't think to yourself "no crashes this week, that is weird" even though statistics show airplanes eventually crash. Sometimes only one, sometimes a few, sometimes many. It happens until designs and travel can be made 99.999999% safe with new technologies.

This reminds me of the fall 2001 (also from swiss and also a non-av prospective);
first sep. 11, just a few weeks later there was a gun rampage in a State parliament, then the swissair grounding, also in october there was frontal crash of two trucks in the Gotthard Tunnel (longest tunnel of the alps), then the Crossair Crash near ZRH, then the A310 crash in NY...
There was one mayor (bad) event a week... I remember thinking what's next?

I think there are just times where such big events occur more frequent. With more and more people living here and also the number of aircraft rising, more events will happen. But we have to wait for the long term statistics to see if there are really more fatalities per people flown or not.
But lets hope it won't continue this way!

Quoting 32andBelow (Reply 15): The United States is still experiencing a period of incredible safety. 0 passenger deaths by a US 121 operator since 2009.

Dude, I hope you threw a gallon of salt over your shoulder, rubbed your lucky rabbits foot, and knocked on every square milimeter of the Spruce Goose after saying that.

A plane shot down by a SAM is not a accident, just like 9/11 was no accident and the same as we will find out MH370 was no accident either.

Quoting lxlucien (Reply 26): then the A310 crash in NY...

Assume you're referring to the AA crash after takeoff from JFK on November 12, 2001. It was an A300, not A310.

Quoting comorin (Reply 12): I would go with Chaos Theory when dealing with systems of high complexity. Otherwise I could set up a machine learning program and train it with thousands of variables to predict the safety of each flight

The difficulty here is nothing to do with chaos theory.

The problem with applying machine learning to flight safety data is getting enough decent training data. There are a large number of flights, but almost all of them pass off without incident.

On top of this, human beings are already working very hard to spot the patterns in aircraft safety data, and we probably do a fairly good job of it, and therefore change the patterns as we find them.

To give a purely hypothetical example, say a particular engine has a design flaw, such that it explodes if aircraft suddenly goes from very humid air to very dry air. Given say, a million flights in which a thousand resulted in crashes because of this defect, a machine learning algorithm would definitely be capable of spotting the pattern "Aircraft with this kind of engine, flying through areas where this kind of weather often happens are less safe".

But of course, that data would never be collected from the real world. Humans will figure out the design flaw very quickly, perhaps after the first explosion, and fix the design flaw so it doesn't happen. So the pattern will get squashed before it is sufficiently strong for the machine-learning algorithm to detect. So, the machine learning algorithm would see, perhaps, one or two examples of the engine exploding in this kind of weather, and then thousands of examples of the engine working fine in the same situation, and it would do the algorithmic equivalent of shrugging it's shoulders and saying "Those crashes were probably outliers, there's nothing interesting here".

There probably are a few patterns that machine learning would find, but they'd most likely be things we know already: long flights are less safe than short flights, flights through bad weather are less safe, flights on old aircraft are less safe, flights operating in countries that have less rigorous safety standards are less safe, flights with inexperienced pilots are less safe, etc. etc.

Disclaimer: I know nothing about aviation, and I didn't finish my PhD in machine learning (but I picked up a few things along the way)

Quoting Viscount724 (Reply 29): Assume you're referring to the AA crash after takeoff from JFK on November 12, 2001. It was an A300, not A310.

Yes you're right, sorry. Was a bit too early to write  

Quoting lxlucien (Reply 26): This reminds me of the fall 2001 (also from swiss and also a non-av prospective);

And there was also a SAS MD-87 accident in Milan in early october 2001. These were horrible months and I remember I unintentionally made a Swissair flight attendant break down crying in the MD-11 gallery HKG-ZRH when I expressed my sympathy in nov 2001.

Quoting comorin (Reply 12): Chaos Theory does a better job than Classical Statistics.

... yet you go on arguing with classical statistics rather than chaos theory.   Just saying. Otherwise you're correct of course.

Quoting B747400ERF (Reply 25): 99.999999%

I wonder if the good old travel technology called "walking" is that safe..

Quoting seahawk (Reply 28): A plane shot down by a SAM is not a accident, just like 9/11 was no accident and the same as we will find out MH370 was no accident either.

Do you know something we don'f? They don't even know where the plane is or if it is still together. Maybe we will see it appear seconds before it collides with a building or some terrorist organization is using it to prepare for a new round of attacks.

Quoting s5daw (Reply 34):

A statistic printed in an issue of the Swiss Alpine Club magazine said that alpine hiking (in itself more dangerous than plain "walking") has an accident rate of 1 per 7143 hours.  


David

Quoting rolfen (Reply 8): I wasn't sure where it happened so I though South Asia was a safe bet

Wow, what a ridiculous statement......why would South Asia be a "safe bet" for this kind of thing?

Well since them MH17 was shot down it shouldn't be counted as a real airplane accident, the airplane itself was flying safely....


For the rest it's just statistics.... this year 2 accidents in a week, next year one in January one in December for example.... just statistics....

Quoting Gr8Circle (Reply 37): why would South Asia be a "safe bet" for this kind of thing?

The OP was talking about calling Taiwan a South Asian country, as opposed to North Asia.

As an aside, the geography geek in me is thinking South East Asia.

North East Asia could be Japan, Kamchatka, Korea (the area, not the countries specifically).

I'd draw the line right though Shanghai, if we're talking about the North/South Asian divide.