A Statistical Analysis of Cirrus Accidents

wanttaja

En-Route
Joined
Jun 7, 2008
Messages
4,779
Location
Seattle
Display Name

Display name:
Ron Wanttaja
As most of you are aware, I do a lot of Homebuilt accident statistical analyses, using the NTSB accident database. As part of that, I occasionally use the same processes to look at particular production-type aircraft.

The ongoing discussions re: Cirrus accidents got me curious. So I fired up my last-downloaded version of the NTSB database (December 2015) and ran my usual process against Cirrus SR20 and SR22 aircraft.

Let's compare them to two other groups: Those of Cessna 210 accidents (for 1998-2007), and for a combined set of Glasair and Lancair homebuilt aircraft (aka HPHB). All analyses include only accidents that occurred within the US...no foreign accidents are included.
This includes only accidents that occurred in the US, international accidents aren't included.
This includes only accidents that occurred in the US, international accidents aren't included.

The number of accidents were reasonably close:
Cirrus: 231
Cessna 210: 370
HPHB: 247

The second line on my results addresses fatal accidents, and computes the percentage of accidents that result in any fatalities.
Cirrus: 34.2%
Cessna 210: 23.2%
HPHB: 42.5%

Cirrus' 34.2% is about twice that of the overall GA fleet. But remember, the lethality of an an accident depends on how fast the airplane is going at impact...and a high-performance aircraft like the Cirrus, Cessna 210, and the high-powered homebuilts will have higher fatality rates.

Yes, the Cirrus rate is about 50% higher than the Cessna 210. But my previous analyses tend to show that high-wing aircraft have a lower fatality rate, probably due to better protection of the cabin.
Cirrus' 34.2% is about twice that of the overall GA fleet. But remember, the lethality of an an accident depends on how fast the airplane is going at impact...and a high-performance aircraft will have higher fatality rates:

Of the three groups, the pilots involved in Cirrus accidents have the lowest median total time:
Cirrus: 727 hours
Cessna 210: 1468 hours
HPHB: 1700 hours

This factors into the biggest category for my analysis process: "Pilot Miscontrol." This category addresses stick-and-rudder errors that lead to accidents.
Cirrus: 51.1% of accidents
Cessna 210: 35.1%
HPHB: 37.9%

The fact that the Cirrus pilots have lower total time probably is the main driver here. It's interesting to note that Pilot Miscontrol accounts for 55% of Cessna 172 accidents...even when training accidents are excluded. And, again, the pilot experience level is lower, about 500 hours total time, median.

When I looked at the in-depth data, one of the first things that struck me is how less often Cirrus engines quit running. The percentages below are the number of accidents where the engine quit for any reason. This includes mechanical failure as well pilot-induced issues
Cirrus: 13.0%
Cessna 210: 31.6%
HPHB: 30.4%

This factors into the survivability of engine failures, as well. 16.7% of Cirrus engine failures lead to at least one fatality or serious injury, vs. 39.3% for the Cessna 210 and 41.3% for the High Performance Homebuilts. Having the CAPS is probably a big factor, here.

The three groups see roughly the same percentage of engine mechanical issues. Other mechanical issues come up less for the Cirrus...their percentage is about a third of those of the Cessna 210 (and that doesn't include landing gear issues).

Continued VFR into IFR conditions occur at about the same rate for the three classes. 4.8% for Cirrus, 4.3% for Cessna 210, and 4.0% for the high-performance homebuilts (which, you'll recall, have the most experienced pilots).

One spot where the Cirrus' rate is much higher is accidents stemming from pilot disorientation:
Cirrus: 4.3%
Cessna 210: 0.5%
HPHB: 1.2%

Again, the Cirrus pilots have the least total time of the three groups.

Here's a summary of some of the main factors:

cirrus.jpg

Ron Wanttaja
 
Interesting, Ron.

Do I understand correctly that you are using recent data for Cirrus and homebuilt, but older 1998-2007 data for Cessna 210? That should be fine, assuming NTSB didn't change its reporting between the earlier and recent periods, but I'm just wondering why you made that choice.

Especially interesting is your result that engine quit running is a less frequent cause for the recent Cirrus accidents, as compared to the Cessna 210s. I wonder if age of the fleet could be a reason for that, or maybe the prevalence of fuel totalizers in the more recent models.
 
I'm always so impressed with your stats work Ron. Thanks for doing it.
 
You don't have a graph representing what gets reported in the media for each and how it is interpenetrated by pilots :)
 
Thanks for this posting, and all the time and effort to go through and present the data.

A couple of things jump out on first perusal. Mechanical causes are lower for the Cirrus, and that seems what one would expect from a fleet that consists of much newer airplanes in aggregate (although I am wondering what "builder error" actually occured for the Cirrus? Is this the early factory airplane fatal aileron jamming incident?). Second thing in looking over the bar graphs they verify the pilot experience/human factors difference. Would seem the most dangerous thing in a Cirrus is inexperienced, low time pilots with the money to buy the plane.

Pretty well as you described it in your commentary Ron.
 
Would seem the most dangerous thing in a Cirrus is inexperienced, low time pilots with the money to buy the plane.

I think to a certain extent, you hit the nail on the head.
While I defend the use of the chute, I do believe the Cirrus is this generations' Dr. Killer.

I think one is best benefited to start in a trainer and transition up to a Cirrus versus, just buying it and learning to fly it.

I had a few OH S**T! moments in a 172 and TB9 (in hindsight they were little scares but at the time...)
Had I had those first moments in a plane closing in on twice the speed... there might not be hindsight.

That I believe is one problem.

Another problem which I can't grasp is the ones that are clearly in over their head and refuse to use the chute.
This one pizzes me off to no end.

 
You don't have a graph representing what gets reported in the media for each and how it is interpenetrated by pilots :)

Short of a wing falling off prior, every chute deployment in a Cirrus is going to be interpreted as one that would have been easily survivable in the hands of us non-Cirrus PoAers. Glide it to a pinpoint off-field touchdown in a turnip patch, not even a mud splatter on that shiny gel coat, no messy chute to re-pack, hero shot in the local papers with the law enforcement official who comes to investigate, and Mom would be proud (buy five copies for her). :D
 
What's really needed are rates. For example, accidents per 100,000 hours, instead of just numbers of accidents. It could change the conclusions a lot.

Too bad it's hard to estimate the numbers of hours flown for a particular fleet.
 
Do I understand correctly that you are using recent data for Cirrus and homebuilt, but older 1998-2007 data for Cessna 210? That should be fine, assuming NTSB didn't change its reporting between the earlier and recent periods, but I'm just wondering why you made that choice.
Based on the availability of previous analyses.

My main focus is homebuilt accidents, which is why the Glasair/Lancair data runs the full 17 years. About eight years ago, I wanted to compare homebuilt accidents to an equivalent set of production-type aircraft. I needed a mix of simple and complex aircraft, so I chose Cessna 172s (excluding training accidents) and 210s. I used an identical analysis process... extract the data on engines, number of injuries, etc. from the NTSB database, then read the narratives to determine the specific cause of each accident.

I've added a bit to the 172s since then, but haven't touched the 210s. Hence, my data set for 210s only runs to 2007. Note that I have the NTSB database for the entire period; what I'm missing is the accident-by-accident assessment I described above.

Ron Wanttaja
 
What's really needed are rates. For example, accidents per 100,000 hours, instead of just numbers of accidents. It could change the conclusions a lot.

Too bad it's hard to estimate the numbers of hours flown for a particular fleet.
The FAA makes these kinds of predictions, but not to the individual aircraft type level.

I've actually computed this for a number of types, based on the age of the aircraft at the time of accident and the total number of hours. It's coarse, of course, and there are a lot of problems with it. I get about 254 hrs/year for Cirrus, and 133 hours/year for the 210s. Hard to really tell with the homebuilts, since the "Model Year" in the FAA registry may not really reflect when the first flight occurred.

Ron Wanttaja
 
I think to a certain extent, you hit the nail on the head.
While I defend the use of the chute, I do believe the Cirrus is this generations' Dr. Killer.

I think one is best benefited to start in a trainer and transition up to a Cirrus versus, just buying it and learning to fly it.

I had a few OH S**T! moments in a 172 and TB9 (in hindsight they were little scares but at the time...)
Had I had those first moments in a plane closing in on twice the speed... there might not be hindsight.

That I believe is one problem.

Another problem which I can't grasp is the ones that are clearly in over their head and refuse to use the chute.
This one pizzes me off to no end.
Certainly agree with you, there.

I had to read all 231 accident reports for this analysis, which is really a depressing task. "On March 29, 2013, about 1045 central daylight time, a Cirrus SR22T airplane, N1967N, was substantially damaged after impact with terrain (frozen lake) near the Chandler Field Airport (AXN), Alexandria, Minnesota. "

Sigh. Then, on some of them, I'd read the next sentence: "The private pilot and one passenger sustained minor injuries, and two passengers were not injured...." and I knew I was going to read of another CAPS activation. Really made things a bit brighter.

I read cases where relatively inexperienced pilots managed to successfully force-land without using the CAPS. Good job, guys.

But three of the CAPS handles were pulled by ATPs. Obviously, their experience told them the best chance of survival was a 'chute pull (one iced up in a spin, another with a night engine failure... and a third case with two ATPs aboard and a daytime engine failure. All were probably happy with their choices.

I haven't quantitized it, but I got the impression that folks were MUCH more reluctant to use the CAPS in the early days. I may do some cross-correlation of fleet size vs. chute pulls per year and see if it bears that out.

Ron Wanttaja
 
Hmmmm.... year vs. successful CAPS deployments, from my data. This includes only US cases.
2001 0
2002 1
2003 0
2004 2
2005 1
2006 2
2007 2
2008 0
2009 2
2010 3
2011 3
2012 3
2013 3
2014 7
My list differs from the COPA one; they list some accidents that don't seem to be on my downloaded database.

Ron Wanttaja
 
...although I am wondering what "builder error" actually occured for the Cirrus? Is this the early factory airplane fatal aileron jamming incident?

I actually didn't count that case because the airplane was in Experimental category. Here are the "Builder Error" cases:

MIA05IA087 Inadequate welding procedure by the manufacturer resulting in a poor weld of the nose landing gear and collapse of the nose landing gear during a porpoised landing.

MIA05LA143 The manufacturer's defective wheel brake assembly design and a leaking wheel brake, resulting in an overheated brake assembly and a wheel fire during an aborted takeoff.

NYC07IA092 The airplane manufacturer's inadequate rudder-aileron interconnect rigging information. Contributing to the incident was the design of the rudder-aileron interconnect system.

SEA07IA201 The failure of the left turbocharger as a result of a production defect in the unit's compressor due to a machining process change by the manufacturer that was made without formal documentation or substantiation.

Ron Wanttaja
 
Great stuff. The issue with Cirrus is less pilot overall experience and more time in type. Here is some data;

Cirrus SR22 sales by year

sr22
1999
2000
2001 124
2002 292
2003 355
2004 459
2005 475
2006 565
2007 588
2008 427
2009 238
2010 222
2011 207
2012 169
2013 244
2014 277
Ok, now the guessing begins. I am assuming that it takes 2+ years to get out of the time-in-type danger zone of 200 to 300 hours in type. So that says 100% are low for 2001 and 2002. For 2003 I also leave it at 100% since sales occur throughout the year. For 2004, I use the formula (low t-in-t %)=100*(2003+2002)/(2001+2002+2003) i.e. prior two years of sales over sales total at start of year. The numbers are:

1999
2000
2001 0
2002 0
2003 0
2004 84
2005 66
2006 55
2007 46
2008 40
2009 31
2010 19
2011 12
2012 11
2013 9
2014 9

What we see is that thought 2009, most SR22 pilots are low time in type. This fits the fatal data:

Cirrus accident rates vs NTSB.png
 
OK, now that that data upload is out of the way here are some more thoughts.

I don't have it since I didn't do the work but I once saw a nice graph of landing kinetic energy (stall speed squared times max gross) vs. percentage of accidents that are fatals. You could do a nice linear fit. There was one outlier which was Diamond which was off the curve fit in the good direction by a meaningful amount. In other words, the percentage of accidents that are fatal goes up linearly with landing kinetic energy.

The pilot disorientation number got me thinking. Two things stand out about the SR22 compared to other planes I have flown such as the 172, 182 and Columbia. The SR22 is difficult to trim precisely. The plane is fast in pitch and the trim control is electric and also fast. Additionally all Cirrus aircraft have excellent autopilots. On a long flight in actual conditions, the SR22 really discourages disconnecting the AP and hand flying in IMC.
 

When you look at the breakout of flying by type (personal vs.training etc.) what becomes apparent is that personal flying is much more dangerous than the GA average. Mission profile also affects safety. In trying to look at SR22 data I went looking for the closest similar mission plane that had an adequate fleet size. I decided on the A36. The C210 is also a good choice except I suspect the fleet is older which may push it more towards VFR vs. IFR flight. I wanted to use Flightaware numbers to normalize for flight hours and that requires different models to have similar IFR/VFR ratios. Here is some data:

Flightaware counts

SR22 BE35 BE36

7/31/13 18:00 41 15 29
1/27/12 8:34 15 3 8
1/28/13 15:17 30 7 15

2014 SR22 fatals US 3
2014 SR22 CAPS US 7
2014 BE36 fatals US 3

2013 SR22 fatals US 1
2013 SR22 CAPS US 5
2013 BE36 fatals US 7

2012 SR22 fatals US 8
2012 SR22 CAPS US 4
2012 BE36 fatals US 6

Here is an observation I think is interesting. If you add up the fatals for the BE36 and SR22 over 2012, 2013 and 2014 and divide by the total of three different Flightaware counts you get a fatality metric assuming (IFR hours)/(total fight hours) is the same for the two aircraft. The Flightaware totals are BE36=137 and SR22=229. The BE36 had 16 fatals over that period. The SR22 had 12 plus 16 CAPS pulls. The metric then is 0.052 for SR22 and 0.12 for the BE36. Interestingly, if you treat CAPS events as fatals then you get 28/229=0.12 i.e. the same as the BE36.


 
Another problem which I can't grasp is the ones that are clearly in over their head and refuse to use the chute.
This one pizzes me off to no end.

I've looked at this accident in detail and AOPA misses at least one point. Here are some thoughts.

1) The weather looked workable about 2 hours before the flight so I won't fault the pilot there. However, it was low enough that he should have had a planned out. At 200 hours this sounds like a low time issue together with poor training. Also, he should have been following the weather in flight. Admittedly it was a short flight but he waits till the last minute to call the airport. More importantly there is no indication he was listening to ATIS of the airports up ahead during the trip. He clearly got surprised by the weather. That's really inexcusable.

2) Not in the video but I believe the pilot had limited training in the plane. It was a club plane and I think he had been through a basic check out and that was it. There's more to flying any plane than just V speeds and how to land.

3) The pilot overflying the airport is a huge clue. The plane has a 10" moving map on the MFD. I think the pilot was focused on the AI already and just trying to maintain control. Otherwise he would have been better aware of where he was. The plane had an Stec 55X AP. People on pilot forums talk all the time about automation leading to poor hand flying skills. I see more deaths because the pilot didn't know how to use what was in the plane. When I did my transition training to the SR22, the instructor asked about VFR into IMC. I was a VFR pilot at the time. I answered "Turn around." He snapped back and said "No!" I was shocked. He then explained that I had no business flying in the clouds. I was to hit HEADING SYNC on the HSI and then HEADING and ALTITUDE HOLD on the AP. Next I was to use the heading bug and let George turn the plane around. Had this pilot used his AP and let George fly it he would have probably lived. I read through a bunch of NASA reports on TAA issues. I was amazed. One guy busted class B because a message came up on his moving map that covered up the map. He didn't know how to clear the message. Time and again there were issues with pilots not knowing how to operate the equipment in their plane.

4) One reason, besides higher time in type, that the Cirrus fatal rate has become excellent is that pilots have started using CAPS. The attitude that good pilots don't use CAPS has been deadly. This is another case that I chalk up to training. Pilots who attend COPA CPPP events get CAPS use drilled into them. As you mentioned, had this pilot pulls the red handle he would have lived.

It's time we focus on knowing how to use the automation in the plane. In this case understanding the AP and CAPS would have saved lives.
 
It's time we focus on knowing how to use the automation in the plane. In this case understanding the AP and CAPS would have saved lives.

I wouldn't get in a Cirrus if I didn't know how to use the AP.
Sounds silly but that spring loaded side stick gets tiring if you are having to work it.
 
Thanks for the research,your figures always tell a story.
 
I wouldn't get in a Cirrus if I didn't know how to use the AP.
Sounds silly but that spring loaded side stick gets tiring if you are having to work it.

How much of that do you think is the trim? I have often wished for the trim wheel on a 172. Also, the fast pitch is great on a VFR day but I find the Columbia better for IFR where I want to trim out and be stable.
 
How much of that do you think is the trim? I have often wished for the trim wheel on a 172. Also, the fast pitch is great on a VFR day but I find the Columbia better for IFR where I want to trim out and be stable.

I hate that elec trim.
I wish it has a wheel for sure
 
I wouldn't get in a Cirrus if I didn't know how to use the AP.
Sounds silly but that spring loaded side stick gets tiring if you are having to work it.

I've read this comment on several different forums and it concerned me a little because I was planning to learn in (and ultimately purchase) a Cirrus.
However, now that I've got almost 12 hours hand flying a Cirrus I can't begin to understand why anyone would feel this way. Just on Friday I flew for almost 3 hours completely by hand and didn't have even the slightest bit of fatigue (13 touch and goes during the training) . The stick just sits neutral in my hand the same way any other stick would. Now granted if you badly trimmed the airplane and had to hold the stick back for extended periods of time fighting the trim I'm sure it would get a little tiring but that's no different than any other plane. I did get a little finger fatigue pushing the button on the radios so much on the touch and goes (uncontrolled airport) but I suspect that's not unique.
The trim is so exceptionally easy to use that I will trim the plane at all the different phases of flight for near hands off stability. When climbing out I tend to bump a bit of trim to hold the climb and then when leveling out on the turn to downwind I'll bump the trim to level. (thumb hat switch for elevator/aileron trim for those not familiar)
I only did an hour in a C172 before I started in the Cirrus, but IMHO the Cirrus was even more comfortable to fly due to the ergonomics of the throttle and easy access to trim. Obviously it's a subjective thing and I don't have much experience to compare, but I can confirm that it doesn't take any effort at all to hand fly. ;)
 
I hate that elec trim.
I wish it has a wheel for sure

I guess it's all I know, but I really like the elec trim. I used the wheel in the C172 and really didn't like it because of the reach. I felt i couldn't trim on landings and take offs as safely/easily.
 
At worst, I thought pitch trim in the Cirrus was a tad fast.

But I adapted to that pretty quickly, just making quick "punches" at the trim to make tiny corrections.

Did not find it a big deal at all.
 
At worst, I thought pitch trim in the Cirrus was a tad fast.

But I adapted to that pretty quickly, just making quick "punches" at the trim to make tiny corrections.

Did not find it a big deal at all.

I don't like it.
Roll trim as well. way too touchy.
 
At worst, I thought pitch trim in the Cirrus was a tad fast.

But I adapted to that pretty quickly, just making quick "punches" at the trim to make tiny corrections.

Did not find it a big deal at all.

For me it isn't tiring to hand fly a Cirrus but I do seem to have trouble getting that last little tweak of trim that will let me take my hands off the controls for more than a few seconds. I could get a 172 or 182 to where I could release the controls and hold altitude for awhile while I did other things.
 
For me it isn't tiring to hand fly a Cirrus but I do seem to have trouble getting that last little tweak of trim that will let me take my hands off the controls for more than a few seconds.

Yes, a little touch on the trim goes a long way. Like many have said, I wish it was less touchy.

One workaround is to "let George do it." Use the autopilot to get in straight and level flight. Then disengage the autopilot and hand fly -- the trim will often be exactly what you want.
 
Back
Top