wanttaja
En-Route
May I point out: High Wing.
I'm gonna dig up a plot I did a while back on the subject of cruise speed and wing position.
Ron Wanttaja
Accident Rates for Common GA Aircraft
- Thread starter wanttaja
- Start date
I'm gonna dig up a plot I did a while back on the subject of cruise speed and wing position.
Ron Wanttaja
Tantalum
Final Approach
- Joined
- Feb 22, 2017
- Messages
- 9,250
- Display Name
Display name:
San_Diego_Pilot
I inherently prefer low wings, but have often read that high wings are "safer" during a crash since you are cocooned inside several sturdy structural elements as opposed to just sitting on top of them.. when I'm relegated to the 172 for scheduling and $$ reasons this makes me feel better
wanttaja
En-Route
I generated this plot a long time ago, and it makes me itch to take another look and update it. Anyway, it's a comparison of the fatality rate vs. published cruise speeds for a number of production and homebuilt aircraft. This particular chart was taken from a presentation I gave to some PA-28 folks a few years back.
Ron Wanttaja
Ron Wanttaja
denverpilot
Tied Down
I generated this plot a long time ago, and it makes me itch to take another look and update it. Anyway, it's a comparison of the fatality rate vs. published cruise speeds for a number of production and homebuilt aircraft. This particular chart was taken from a presentation I gave to some PA-28 folks a few years back.
Ron Wanttaja
Tease! Haha.
Now I want to know what that thing is that is a low wing that cruises at 80 knots that has a 35% fatal rate?!
LOL. Numbers are fun.
wanttaja
En-Route
denverpilot
Tied Down
wanttaja
En-Route
On the analysis so far, I've taken the number of accidents over the past ten years, and divided it by the number of each aircraft type registered on 1 January 2017. As I mentioned, this is kind of a bastard number...great for comparing accident rates between types (as long as there are sufficient examples of the type and the same process is used), but it really has no absolute meaning.
It was suggested that I look at Cirrus and compute the *per year* Fleet Accident Rate. In this case, one would take the total number of accidents that occurred in, say, 2010 and divide it by the number of registered examples of the type in that year.
Here's the results:
The spike in 2005-2006 might explain why some folks gained the impression of poor safety of the Cirrus line. The rate has been working its way down, since.
Ron Wanttaja
It was suggested that I look at Cirrus and compute the *per year* Fleet Accident Rate. In this case, one would take the total number of accidents that occurred in, say, 2010 and divide it by the number of registered examples of the type in that year.
Here's the results:
The spike in 2005-2006 might explain why some folks gained the impression of poor safety of the Cirrus line. The rate has been working its way down, since.
Ron Wanttaja
Checkout_my_Six
Touchdown! Greaser!
that chute training is making a big impact......
3 in the green
Line Up and Wait
Thanks for putting together!
With respect to the fatality rate, the inactivity of an aircraft shouldn't be a factor, since the fatality rate is a function of the accident count and you need to be flying (presumably...) to have an accident.
But otherwise yes, the accident rate itself would be skewed with many inactive airframes. That said, there are tons of inactive airframes of various types, so I'm wondering if that smooths that factor out a bit. Impossible to quantify...
Also curious about the DA40 rate. I was under the impression there have only been a few fatal accidents EVER for that airframe (at least two of which I recall being loss of control VFR-into-IMC types).
wanttaja
En-Route
Absolutely correct.
My assumption is that is smooths out for aircraft of approximately the same vintage. This was the problem, way back when I first tried to look at the Cirrus accident rate vs. the 172. Which is why I did a separate analysis for 172 accidents occurring to the 1990 and later models.
You're kind of describing how I got into this accident analysis gig. I got too suspicious of "common knowledge" and decided to find the answers myself.
When I open the date filter all the way on my NTSB database, I see eight fatal DA-40 accidents involving N-numbered aircraft.
ATL05FA034: Probable Cause, "The pilot's failure to follow IFR procedures and to maintain assigned altitude resulting in a collision with a transmission wire and trees." This is not a classic VFR into IMC accident, as the pilot was IFR qualified and operating on an instrument flight plan.
LAX07FA160: Probable Cause, "The pilot's failure to maintain sufficient altitude above the surface of water during an intentional buzzing maneuver." At night, yet.
WPR09FA146: Probable Cause, "The pilot's failure to maintain proper altitude and glidepath while executing a night instrument approach." Again, not VFR-into-IMC, the pilot was instrument qualified on an instrument flight plan.
ERA11FA085: Probable Cause, "The pilot’s controlled flight into terrain likely due to spatial disorientation, while flying over sparsely lighted terrain at night." VFR conditions (pilot reported the airport in sight), pilot had instrument rating. Not VFR-into-IMC, just plain disorientation...flying a lot of circles at night.
ERA13FA083: Probable Cause, "The noninstrument-rated pilot’s improper decision to depart in dark, night marginal visual flight rules conditions, which resulted in his spatial disorientation and subsequent loss of airplane control." VFR into IMC.
ERA13WA319: N-numbered aircraft in Peru. Hit the mountains at night.
CEN14WA502: N-Numbered aircraft flying from Iceland to Greenland.
WPR16FA144: No probable cause issued, yet. VMC, daytime, hit almost horizontally.
So there's one case, possibly two, of VFR into IMC conditions, out of the eight accidents. 12.5% to 25%; That's a bit high, in my experience. My Bonanza A/B36 accidents shows 3.3%, Cessna 210s are 4.1%, and Cirruses are 4.8%. Still, only eight accidents...not much to base an estimate on.
Ron Wanttaja
Shepherd
Final Approach
Thanks for all your effort.
Again, it kind of goes back to usage rates. C-172's can see upwards of 600hours/year because they are frequently used as rentals. I don't see J3's racking up the amount.
However, I do see a lot of value in your fatality rate. In other words, if you do get in an accident how likely are you going to kill someone.
It was suggested that I look at Cirrus and compute the *per year* Fleet Accident Rate. In this case, one would take the total number of accidents that occurred in, say, 2010 and divide it by the number of registered examples of the type in that year.
Here's the results:
Any chance you can overlay the fatality rate on that graph? Thanks.
wanttaja
En-Route
Yes, but what's the AVERAGE annual flight time for C-172s vs. J-3s? We don't know. Most of us figure the 172s are higher...but how much higher?
The FAA's annual usage report generates an estimate of flight hours...but not by type. There's a value for fixed wing piston-powered single-engined aircraft of up to three seats, and a value for fixed wing piston-powered single-engined aircraft of four or more seats. It's no help when you're trying to compare particular GA types that fall into the same category.
BTW, the FAA's estimate in 2014 was about 87 hours a year for the four-seaters and 69 for the two-seaters.
Certainly, once I remembered how to plot to the secondary axis. On this plot, the Cirrus Yearly Accident Rate is plotted in blue, and the scale is the left vertical axis. Each year's Cirrus Fatality Rate is plotted in red, and scaled on the right vertical axis.
One of my pet peeves is magazine reporters who take the statistics from the previous year and bemoan how GA safety has declined, or get all pumped up if the numbers are better. The folly of that approach should be pretty obvious from the above. Accident rates vary year-to-year; there's a lot of luck and a whole spitload of Murphy involved. You have to take the long view, and examine the statistics over a broader interval.
Ron Wanttaja
Last edited:
Also, who wants to fly a cub in crappy weather, or for a long trip?
I don't understand the logic in your methodology of calculating the fatality rate. Using the ratio to the number of accidents in the type makes it completely incomparable from model to model in my mind. For example, if there were only one accident, and it were fatal, the model would have a fatality rate of 100% using your method. That would not in any way be useful to compare against a model that had 10,000 accidents with 10% fatality rate (1,000 fatalities)
Checkout_my_Six
Touchdown! Greaser!
Instead of calling it a fatality rate....it's more of a survivability rate for comparison purposes.
Also, I'd also consider "normalizing" the data....by dividing each of the numbers by the total. That will produce a ratio or fraction of the total for each category.
Also, I'd also consider "normalizing" the data....by dividing each of the numbers by the total. That will produce a ratio or fraction of the total for each category.
I still don't see how it can be used to compare model to model, unless you intend to have an accident on purpose.
Checkout_my_Six
Touchdown! Greaser!
how else could you determine there were accidents without deaths?
btw....this data is only for those that "reported".....there are many that don't report, especially if there were no injuries.
A separate accident rate and fatality rate, but with the fatality rate being number of accidents with a fatality per registered aircraft rather than per accident.
Do you wear seatbelts even though you don't intend to have an accident on purpose? Do you do so because accidents where people wear seatbelts have a lower fatality rate than those where people don't wear seatbelts?
You're missing my point. Unless you know you will have an accident, you need to factor the odds of having that accident into the odds of a fatality.
As I said in my example, if a model has only one accident and it's fatal, its not comparable to a model that has a higher probability of an accident, but lower fatality relative to accident rate. If my odds of an accident are high, but fatality rate as calculated here are low, it does not mean the fatality rate is actually low because there are so many more accidents.
True, but Ron has already included both and has filtered out the areas where the sample size is too small. Since both are included, it's mainly a matter of mental math to get the fatalities per registered model figure that you're looking for.
wanttaja
En-Route
Yes, which is why I require a certain minimum number of accidents before I publicly discuss results for that aircraft. Take the Icon A5 for instance...its fatality rate (so far this year) is 50%, but as there have been only two accidents, it's meaningless. And as I discussed in a previous post, I didn't split the Diamond results by model since the total number of accidents per model is pretty low. I noted the DA-40 results separately in response to a specific question.
It's easy to compute a fleet fatality rate, but, again, the results are affected by the number of inactive aircraft. In addition, since the fatality rate is generally low, you again end up with small sample sizes and the resulting inaccuracies. If you look at the table, you'll see about 1/3rd of the columns don't have enough fatal accidents to meet my 50-accident-minimum criteria.
If there was an *accurate* estimate of the number of active aircraft and average flight hours *per aircraft*, I'd do everything by flight hours. But there isn't....
Ron Wanttaja
wanttaja
En-Route
Yes, certainly true. A homebuilt had an engine failure on takeoff at one of my EAA Chapter events, wiping off the landing gear. The Chapter members had it tucked back into a hangar with the doors closed before the cops arrived. "What hump?"
I've also known homebuilders who have played fast and loose with the NTSB 830 definition of an accident.
However, a homebuilder is used to assembling, disassembling, and transporting aircraft pieces, and have some standing to be able to assess the degree of damage. Not quite as true with production aircraft. Sure, it can happen, but most folks need to hire someone to transport an aircraft from an off-airport landing. An accident on a private field might be hushed up more easily, but the majority of activity is going to take place pretty publicly.
And...even if one assumes that, say, 10% of reportable accidents are not reported, that's probably not very aircraft-type related. No reason to assume it happens more to, say, Bonanzas than Cirruses. So the *relative* rates aren't affected.
Ron Wanttaja
Checkout_my_Six
Touchdown! Greaser!
yup....totally agree. A Cirrus chute is a lot harder to hide....so it would get reported vs. a Bonanza bean field landing out in the middle of middle America....where folks just fly over.
I understand the difficulties you're facing, but I still don't think the fatality rate you've calculated can be used to compare model to model. I think it's misleading.
Checkout_my_Six
Touchdown! Greaser!
Look.....think of it as an "estimate"....not a statistic. The trends in the data are representative....in a qualitative sense....not quantitative though.
wanttaja
En-Route
Which is why I always describe my processes with the original posting, and subsequently answer questions about them: I want the reader to have enough info to decide whether or not my results are something they can believe. I may or may not accept criticism, but always appreciate the feedback.
Ron Wanttaja
And I'm not criticizing. Your numbers are what they are. I'm just pointing out my observation about that particular number. I don't think it can be used for comparison.
MIFlyer
Pattern Altitude
Ron, I know this is an old thread, but any info on PA-32 and PA-23?
tspear
En-Route
- Joined
- Dec 10, 2010
- Messages
- 3,587
- Display Name
Display name:
Timothy
Why not look at the data another way. Look at the hours per plane when it has crashed.
This would make it easier to determine which plane is safer based on the number of hours between accidents.
It will not give you a metric on the rate for the fleet, but will give a comparison between models.
Tim
This would make it easier to determine which plane is safer based on the number of hours between accidents.
It will not give you a metric on the rate for the fleet, but will give a comparison between models.
Tim
wanttaja
En-Route
Haven't run those numbers. My primary interest is homebuilts, so I get into the production-airplane world only when I need an example for comparison.
Ron Wanttaja
James331
Ejection Handle Pulled
- Joined
- Apr 18, 2014
- Messages
- 20,309
- Display Name
Display name:
James331
Y'all crack me up, maybe make a study of what scent of deodorant they were wearing, about as relivent.
I'd wager 95% of all these crashes were due to pilot error/ bad decisions, other 5% were a mix of other factors, type not being one, ain't the airplane that kills anyone anymore than the gun, it's the dude himself that ultimately is the instrument of his own demise.
Add to that anyone reading into all this static, if you're more than 20lbs overweight slap yourself, cuz its not the flying machine that stats say is going to be your undoing, it's that gut of yours.
I'd wager 95% of all these crashes were due to pilot error/ bad decisions, other 5% were a mix of other factors, type not being one, ain't the airplane that kills anyone anymore than the gun, it's the dude himself that ultimately is the instrument of his own demise.
Add to that anyone reading into all this static, if you're more than 20lbs overweight slap yourself, cuz its not the flying machine that stats say is going to be your undoing, it's that gut of yours.
gsengle
Pattern Altitude
- Joined
- May 9, 2016
- Messages
- 2,099
- Display Name
Display name:
Gsengle
There is a difference in accident rate that can absolutely be ascribed to the good or bad qualities of an airframe. Is it most of it? Maybe not. But we can try and tease it out.
There is also the reality that different types have different pilot groups in experience and attitude, in percentages holding instrument ratings and in type of flying. This is also a significant factor we can try to tease out.
Yep the pilot is always the biggest factor. But for any given pilot, we do want to know the relative safety of the type.
Sent from my iPhone using Tapatalk
James331
Ejection Handle Pulled
- Joined
- Apr 18, 2014
- Messages
- 20,309
- Display Name
Display name:
James331
Again, it's still not the plane, best argument....maybe... is it's the demographic who's attracted to said plane.
gsengle
Pattern Altitude
- Joined
- May 9, 2016
- Messages
- 2,099
- Display Name
Display name:
Gsengle
It's all three. The pilot/demographic is probably #1, the usage profile is probably just as important - trainers don't kill as many people as high performance turbocharged traveling singles or seaplanes for example. But all else being equal, the airplane design itself is a factor in the accident profiles. The part 23 certification rules tries to keep the differences minimal, but they are there, and get bigger the higher performance the plane I think.
You are who you are - and you can improve your training and attitude, and you can be risk averse in your mission. But those kinda are what they are. But you can pick an aircraft.
Sent from my iPhone using Tapatalk
MIFlyer
Pattern Altitude
I do hear though about different airframes being more docile and forgiving of pilot error than others..
think: gentle stall breaks, low propensity to spin, low stall speed, reliable gear, reliable engines, lack or presence of CG sensitivities, etc.
I'm just a low time guy, but in addition to working on my proficiency and judgement, it does seem like you can make airframe choices to stack the deck for or against you.
think: gentle stall breaks, low propensity to spin, low stall speed, reliable gear, reliable engines, lack or presence of CG sensitivities, etc.
I'm just a low time guy, but in addition to working on my proficiency and judgement, it does seem like you can make airframe choices to stack the deck for or against you.
wanttaja
En-Route
I use a "Control Group" of production airplanes to compare to homebuilts...it consists of about 3,000 Cessna 172 and 210 accidents. Roughly 66% are pilot error or bad decisions. Homebuilts are less (53%), but a lot of that is due to more-experienced pilots.
I haven't spent much time comparing the causes of production-aircraft accidents, but there are obvious cases where flaws in the airplane contribute to accidents. About 1.3% of Bonanza accidents are due to problems with the landing gear or brakes, vs. 12.5% of the Cessna 210 accidents. Only 0.5% of Cessna 172 accidents stem from fuel starvation (e.g., having fuel but not the pilot doesn't feed it to the engine), vs. 4.6% for the Cessna 210s and 9.2% of the Bonanzas.
When one in eight Cessna 210 accidents can be attributed to the landing gear system, I think that's a good indicator as to how the aircraft themselves affect their accident rates.
Ron Wanttaja
deonb
Pattern Altitude
- Joined
- Aug 17, 2015
- Messages
- 2,266
- Display Name
Display name:
deonb
Comparing a subject against itself is not valid for statistical analysis.
One in eight Boeing 777's fatal accidents can be attributed to hitting a seawall on landing. Does that teach us anything about how a 777 compares to other aircraft?