Hacker News Reader

Comment by marcosdumay | Hacker News Reader

Aviation has an well known acceptable risk level at around 10^-9 for each issue. That's the number that leads government intervention, pilot procedure designing, aircraft designing and everything else. It's expected to lead to a less than 10^-6 chance of accidents per flight. (Somebody calculated the B783 Max odds on 4*10^-6 yesterday, what is a crazy high level.)

That number has been higher in the past, and is moving into 10^-10 per issue with 10^-7 overall risk right now, with large airplanes in scheduled flight very near that level.

View on HN

> Somebody calculated the B783 Max odds on 4*10^-6 yesterday, what is a crazy high level.

I think that may have been me (1/250000), but that was based on a couple of generous assumptions - two crashes across 4 flights/day on 350 planes for an average of 365 days. Unfortunately I think a more reasonable flights/day number is 3 or lower - a lot of Max 8s are on longer routes - and the flight day average is almost certainly lower than 365 (which assumes linear deliveries for the past two years, with no days for maintenance).

Also, I believe the statistic of interest here is incidents per operating hour, not per flight?

Given that the Max crashes both happened shortly after takeoff, I'd be more interested in a per-flight statistic. But I'm not sure how the industry does it.

The industry/NTSB/FAA generally cares more about flight hours, but I agree with you. Both of these accidents happened shortly after takeoff. I am surprised that no accident has happened during landing.

Is the same sensor used the same way in landing?

I would imagine so. As long as the pilot wants to control the rudder, that same sensor would be used. What I would like to know is if the MAX has a different setup. I think it'd be idiotic of Boeing to use the same rudder control configuration as the old 737 considering the problems that they had as demonstrated by this post.

All of accidents per hour, accidents per operation, and fatality per passenger are relevant.

When something happens 2 times, even a simplistic error analysis is going to show at least a 3x range in probably (2+/-1 incidents is 1 to 3)

A few years ago I calculated with a friend the accident rate of several cohorts of Brazilian aviation (well, the part of aviation that operates on aerodromes, so no agriculture or sports). Our worst cohort was for private planes, private pilots, on small airports, on fast growing cities. It was just a little bit under 4*10^-6 accidents (all, not only fatal) per airport operation (that is, twice your number).

I was really scared by your number. I wouldn't ever jump into a random private plane, and seems that people refusing to jump into a B733 Max (looks like I wrote it wrong earlier) are doing a reasonable thing.

Do those people jump into cars?

For long distance trips where they have an option? I'd guess that many don't.

Still that 4e-6 per trip number is worse than the city driving statistics I've found. And that 8e-6 per trip (4 per airport operation) is around the motorcycle values.

That's also assuming there's no correlated issue causing all the crashes.

Is that 10^-6 figure is for any sort of accident, or just catastrophic ones?

It is called "accident" is when there is severe damage to the airplane or fatality. In aviation, the other are called "incident" when there is some damage, or merely "occurrence". But that last one is overloaded with way too many meanings.

Reply on news.ycombinator.com