Originally Posted by
Stranger
Perhaps. But surely that's an exercise in semantics. And/or one could argue that beyond a certain point the noise overwhelms the signal so there is no logic in going further.
No, this is definitely not an exercise in semantics by any stretch.
Except, I seriously doubt the figures for these estimates are all that reliable, more like educated guess. Not just in aviation obviously. Which brings me back to noise and signal...
There is no guesswork of any sort involved. I think most people here would probably go insane before being able to figure out just how to quantify the risk levels. There were many days where I was certain taking a bullet to head would have been less painful than going through the ARP safety assessment exercises.
The image below is probably the simplest of the safety assessments that are conducted.
....and it took me several years to learn the secrets to being able to perform that sort of work without going nuts - a secret that I'll share here.
Looking at the nuclear industry (to me less controversial -or not-), they had fault trees and that sort of crap leading to a supposed probability of something like a major accident in probably something like 10,000 years. In actuality they have had for or five since the thing started. Probably with none of the scenarios anticipated in the fault trees...
In aviation fault trees are used for troubleshooting procedures.
But of course coming back to aviation, the saving grace is that we have so many polanes and so many flights that even with very few accidents, we end up having a pretty good data base.
At least for accidents. Probably not for single component failure rates...
You'd think so, but not really, good quality prognostic data is not all that common. The recent development of IVHM (Integrated Vehicle Heath Management) for new platforms is changing that and single component failure rate prediction will get better as the years progress.