DA40NG

[nrenno]

Karl,

Thank you for the response and additional information, including your impressive aviation background. I am just a pilot, with engineering/science background, trying to understand the general aviation market to make informed decisions. I would refer to the issues that you mentioned as powerplant failures. Anything that cause the engines to malfunction needs to be taken into account in the statistics. An engine overhaul is quite different from an engine replacement.

The low lifetime of some components must be at minimum due to lack of confidence in their reliability. A powerplant with fewer low lifeline components is likely more reliable than the others.

[TimS]

nrenno,

Actually low life limits is a cost saving measure. EASA and the FAA have raised failure rate standards to the point that almost no engine or new design, even by Lycoming or Continental would pass. Just consider the failure rate of vacuum pumps for example.

By having lower limits and gradually increasing the limits, the manufacturer has a reduced certification costs and barrier. For example, the clutch change from 300 hours to 600 hours was not a change in parts or manufacturing. There are many other examples, such the life limit increases for Cirrus, Diamond DA20....

There is plenty of documentation available that the FAA and EASA have historically for the past thirty/forty years had an attitude that any change is a new risk and therefore should not be certified. Look how long it took the FAA to come to the conclusion that AOA enhances safety. Or that the Beech Starship or Premier is over engineered by a factor of four or more....

Lastly, there is a real cultural aversion to change among pilots. How many have screaming for a modern auto engine? But then they all make noise about the new failure modes, and forget about all the existing failure modes for the venerable Lycoming or Continental when making a comparison.

Tim

[nrenno]

Tim,

I did not know that low life limits are a cost saving measure for the manufacturer. This is interesting, but it also means that the low life limits are due to lack of knowledge (referred to as unknowns unknowns in my field). That is, the data for estimating the reliability of these parts with low life limits does not exist yet.

I love modern equipment, but I also try to understand the data available to make decisions based on it. The fact that an airplane is modern and has low fuel consumption is not enough for me. I love my IO-360 DA40 because the data shows that it is safe and reliable. In addition, it is fun to fly and has great visibility.

I would not buy a Cirrus, because the data tells me that it is not a safe airplane. The fact that it is fast, modern and looks cool is not enough for me.

[nrenno]

Scott,

I agree that we also need data to understand the failure rate (or the probability of failure) of the Lycoming engines. At this point, I am skeptical about all numbers mentioned. Without data, as you suggested, we need to use the accident record for making informed decisions.

--Nilton

[TimS]

Tim,
I would not buy a Cirrus, because the data tells me that it is not a safe airplane. The fact that it is fast, modern and looks cool is not enough for me.

Look closer and see if the problem is the pilots or the equipment.
At that point, your opinion about aircraft in general will take a nose dive.
As for unknown unknowns; yeah this applies. Based on a few discussions with engineers who have been through certification of parts, planes and STCs; it really comes down to a few fundamental choices (this is how it was explained to me):
1. Over engineer the part way beyond requirements. This increases manufacturing costs forever due to more material or more expensive material, and generally makes the parts heavy.
2. Test the system and parts to death a few million times. This has very significant costs.
3. Use computer and other model techniques. Then have a replacement of the part just at the maximum amount before you break the magical MTBF rate of the FAA group you are dealing with. Most of these models require base assumptions, which error on the conservative side, which over time gets compounded. So you build in some buffer, release to the field with a lower limit; the idea is to gather enough data to adjust the models and reduce the compounded conservative numbers; allowing for an increase in life limits.

Tim

[nrenno]

Thank you, Tim!

I have been learning a lot with this discussion. I agree that we should not forget to take into account pilot errors, as unfortunately highlighted by the recent accidents with the Icon A5.

—Nilton

[Colin]

Data is very difficult to collect properly and when you do, I think you are looking at a pretty small data set, which makes it dangerous to draw real conclusions.

I loved the book The Killing Zone but right after I finished it my brother read it, looked at the statistics, and said that *he* was pretty sure that time flown in the last 90 days was more of a factor than total time.

Cirrus planes seemed very dangerous to me, but their pilot association stepped up and changed the syllabus suggestions to Cirrus for the training. Or worked with them in collaboration. The training is now entirely different and the accident rate plummeted. What sort of conclusions can one draw from the entire data set? Certainly ones that are incorrect for someone willing to go through the entire factory training program.

One of the ways I saw that is probably annoying to anyone with even a little statistics knowledge. I saw the Nall Report and the vast number of fatal accidents that were VFR pilot into IMC. I got my instrument rating and concluded that the VFR portion of the pie chart no longer applied to me. (This is not quite as bad as carrying a bomb onto an airliner (since the chances of there being TWO bombs is tiny), but I think it is probably similarly wrong.)

[Keith M]

This is not quite as bad as carrying a bomb onto an airliner (since the chances of there being TWO bombs is tiny), but I think it is probably similarly wrong.

That should set off the alarms at the NSA

[Rich]

Ah, yes. If one is select in choosing data sets you can find anything you choose. Cirrus did show its fatal accident drop precipitously upon the improved training regimen. But it was temporary and the count crept back up in subsequent years. Fatal accidents found in the NTSB DB in the years 2012-2017, in order: 10,8,3,5,7,8. Stall-spins and weather are the primary recurring themes.

For Diamonds (all models): 1,2,2,0,2,0

But if you just choose the years 2012-2014, Cirrus improved greatly, while Diamond is actually getting worse.

On the other hand, based on this database, looking at the entire history (call it 15 years) the two marques were selling airplanes, Diamond averaged about 1.5 fatal accidents per year, Cirrus about 8.2, 2014 being an aberration. Overall 16% of Diamond accidents were fatal, for Cirrus: 36%.

[TimS]

Ah, yes. If one is select in choosing data sets you can find anything you choose. Cirrus did show its fatal accident drop precipitously upon the improved training regimen. But it was temporary and the count crept back up in subsequent years. Fatal accidents found in the NTSB DB in the years 2012-2017, in order: 10,8,3,5,7,8. Stall-spins and weather are the primary recurring themes.

For Diamonds (all models): 1,2,2,0,2,0

But if you just choose the years 2012-2014, Cirrus improved greatly, while Diamond is actually getting worse.

On the other hand, based on this database, looking at the entire history (call it 15 years) the two marques were selling airplanes, Diamond averaged about 1.5 fatal accidents per year, Cirrus about 8.2, 2014 being an aberration. Overall 16% of Diamond accidents were fatal, for Cirrus: 36%.

And during that period how many planes did Cirrus sell versus Diamond? And then compare the differences in which markets the planes were predominately sold.

Tim