Statistical Analysis of Homebuilt Aircraft Engines

[rwanttaja]

I did an analysis of the FAA registration database to develop some statistics on what kind of engines are installed in homebuilts. Here's an overall summary:
http://www.wanttaja.com/engine_breakdown.jpg
About 4000 homebuilts in the registry are listed as having "AMA/EXP" engines. As part of this analysis, compared the "AMA/EXP" engine aircraft with the NTSB accident database, which almost always includes the engine type. I found about 550 of the AMA/EXP aircraft in the accident database, and from them, extrapolated the percentage of AMA/EXP aircraft that have specific aircraft engines. I used those percentages to add the appropriate number of engines to each of the main analysis engine categories.

This aspect isn't precise... it produced slightly more aircraft with Rotorway engines than there are Rotorway helicopters registered... but I think proportionately, the numbers are pretty good.

Ron Wanttaja

[rwanttaja]

Here's a more detailed breakdown by specific engine types.

The first column is the engine, the second is how many homebuilts are registered with that engine type.

The third column is the result of my comparison of the FAA "AMA/EXP" engine registries to the NTSB accident database. About 550 of the ~4,000 "AMA/EXP" aircraft matched. The third column is how many "Hits" I got for that engine type.

The "Calculated from AMA Set" is an estimate of the total number of AMA/EXP aircraft that are that engine type. if 55 of my "hits" are a given engine type, I assumed that 10% of the AMA/EXP engines are that type (55 out of 550). So the "Calculated" value in Column 3 would be 10% x 4000. I then add the actual number registered (Column 2) to the number calculated (Column 4) to come up with an estimated total.

Mind you, the results in Column two are really the only reliable data. The rest is based on my projections. Note that there ARE no aircraft powered by Chevrolet liquid-cooled or KFM engines in the FAA database. These only came out when I cross-referenced the AMA/EXP engines to the NTSB accident database.



Engine

Registry

In AMA Compare Set

Calculated from AMA Set

Total


Aerosport

221

4

29

250


AeroVee

90

6

43

133


Ama/EXP

263

7

0

263


Cam 100

2

2

14

16


Chevrolet Liquid-Cooled

0

21

153

153


Continental

2900

34

247

3147


Corvair

40

7

51

91


Cubcrafters

30

1

7

37


Cuyuna

42

2

14

56


ECI

18

6

43

61


Ford

81

6

43

124


Franklin

106

9

65

171


GM

5

10

72

77


Great Plains

12

5

36

48


HAPI

5

7

51

56


Hirth

76

10

72

148


HKS

67

2

14

81


Honda

19

5

36

55


Jabiru

547

23

167

714


Kawasaki

18

0

0

18


KFM

0

18

131

131


LOM

3

3

21

24


Lycoming

11373

80

583

11956


Mattituck

225

3

21

246


Mazda

17

10

72

89


McCullough

207

3

21

228


No Entry

115

0

0

115


None

476

0

0

476


NSI

0

5

36

36


Oldsmobile

1

0

0

1


Onan

23

3

21

44


P&W Recip

41

0

0

41


P&W Turbine

91

0

0

91


Revmaster

145

2

14

159


Rotax All

4962

14

226

5188


Rotec

10

2

14

24


Rotorway

187

43

313

500


Solar

12

5

36

48


Stratus

0

1

7

7


Subaru

525

32

233

758


Superior

332

6

43

375


Suzuki

11

3

21

32


Toyota

0

1

7

7


UL Power

28

8

58

86


Vendeneyev

34

7

51

85


Viking

39

2

14

53


VW

562

49

357

919


Walther

67

10

72

139


Warner

28

0

0

28


Wright

7

0

0

7


Yamaha

14

5

36

50


Zenoah

29

1

7

36



Ron Wanttaja

[Frank Giger]

I'm trying to wrap my head around the data, and I think I might be over-thinking it.

Is this simply "what kind of engines are in homebuilts" and nothing more? Or with the inclusion of NTSB data is there some other data we can tease out?

Btw, if NTSB data is used, there could be more aircraft with a special type of engine installed than registered. One would only have to, through misfortune, make the NTSB records twice with the same aircraft.

[rwanttaja]

I'm trying to wrap my head around the data, and I think I might be over-thinking it.

Is this simply "what kind of engines are in homebuilts" and nothing more? Or with the inclusion of NTSB data is there some other data we can tease out?

Yes, this is basically trying to answer a question in another forum about how many auto-conversions are installed in homebuilts. It should be easy enough to extract from the FAA registration database. But about 4000 homebuilts have an entry that just says, "Amateur-Built/Experimental Engine."

While the FAA database includes a lot of VW and Subaru engines (and a smaller number of Corvairs), it really lists few of the V6/V8 sorts of conversions that seem to get all the press. As you can see above, the January 2016 FAA database shows *no* aircraft powered by Chevrolet engines (other than Corvairs). Yet they can be found in the NTSB database, so the assumption is many of the liquid-cooled Chevy engines are actually listed with the generic AMA/EXP tag.

But...how many of the 4000 are of each type? Are they all auto conversions? Or are folks registering their Continentals and Lycomings as AMA/EXP for their own reasons (such as their not coming up on an AD search....).

So I took those 4000 and compared them to my NTSB accident database. The NTSB almost always lists the engine. So, for about 550 of those 4000, I found accident reports (I did have to be careful it wasn't another plane with the same N-Number). I worked out how many of the 550 were Lycomings, Continentals, etc. As you'll see above, while the FAA database shows no Chevy-powered homebuilts, there were 21 just in that 550-aircraft sample. So, obviously, the number of auto-engine homebuilts were being under-reported, and my analysis was an attempt to extract the true fleet sized.


Btw, if NTSB data is used, there could be more aircraft with a special type of engine installed than registered. One would only have to, through misfortune, make the NTSB records twice with the same aircraft.

Oh, yes, found several of those. Some planes had more than one accident, with a different engine each time.

Ron Wanttaja

[Anymouse]

Here's a quick data point. My engine is a Superior XP-360. I just looked up my info on the FAA site. They have it listed as AMA/EXPR.

I also looked up the Rotorway I owned. It shows as "unknown."

[rwanttaja]

Btw, if NTSB data is used, there could be more aircraft with a special type of engine installed than registered. One would only have to, through misfortune, make the NTSB records twice with the same aircraft.
The other aspect of this is *how* airplanes end up in the NTSB database. If Engine AAA has low reliability, it'll have a disproportionate number of accidents. This may distort my prediction of total aircraft with that engine, producing a larger number of predicted planes with the AAA engine. I see my table's calculated number as more of a max.

The best message, I think, is from the pie chart. Other than VWs and Subarus, the percent of the Homebuilt Fleet operating auto engines is really quite low. Even with a process that might be biased towards them.

Ron Wanttaja

[Frank Giger]

The statistical analyst of my past cringes at the sussing of engine reliability by type and incident/accident from the rather dubious databases available - though I know you've done it in the past.

I'm actually not that surprised at what you dug out when I got to thinking about it.

I've come to put homebuilts into two categories in my head - Light and Heavy.

"Light" as in either within LSA criteria or very close to it. At those weights and wing areas, Rotax and VW are more than enough to do the job pretty well. My little Nieuport, for example, or a Pietenpol for another, are as much "flying machine" as airplane in the common way of thinking of it.

That they make up just a third of what's out there also fits.

"Heavy" as in "going places." RV's, KR2's, all the way up to twin engine JAG's, put little VW's out of the mix. What one needs is a robust engine that is preferably air-cooled, and no matter how much we try to come up with a good alternative, the Lycombings and Continentals fit the bill the best. Once one factors in all the modifications and possible complexities of converting a large automotive engine for aircraft use, we tend to default to the safest and surest route for power plants.

I'm not saying the Corvair folks haven't dialed in that conversion - they have - but it's the exception, and it's a pretty tight community with the Corvair College being one of the coolest things going, but it's the exception rather than the rule.

We all like to trump up our calls to innovation, but in the end 99% of us go with what has something of a proven track record.

[rwanttaja]

The statistical analyst of my past cringes at the sussing of engine reliability by type and incident/accident from the rather dubious databases available - though I know you've done it in the past.

I hears ya, bro. The data is crippled, no question, but it's the only information we've got to work on. This is why I spend a lot of time explaining my process, so folks understand how much assumptions and side-calculations are involved.

I try...despite the flaws in the data...because either people ask outright questions, or make outrageous claims. "There are 1500 examples of this engine flying, of COURSE it's reliable!" somebody might say...but if the FAA database shows only three, and even my loosest approximations reveal only a dozen or so more, what does that mean?

I'm actually kind of surprised at how few inroads the Chevy-Ford-GM etc. liquid-cooled engines have made, in homebuilt aviation. What's more, there's some about of roil. I looked in my 2007 aircraft registration database and found five examples of one given engine type. There were six examples using the same engine in my 2015 database... all but one were different aircraft.


I've come to put homebuilts into two categories in my head - Light and Heavy.

"Light" as in either within LSA criteria or very close to it. At those weights and wing areas, Rotax and VW are more than enough to do the job pretty well. My little Nieuport, for example, or a Pietenpol for another, are as much "flying machine" as airplane in the common way of thinking of it.

That they make up just a third of what's out there also fits.
I think it's probably more, once you factor in all the Continental A65s through O-200s, and the Lyc O-235.

My next analysis project, I guess. Retiring in a couple of months. Fortunately, I apparently already have a hobby.....

Ron Wanttaja

[martymayes]

My next analysis project, I guess. Retiring in a couple of months. Fortunately, I apparently already have a hobby.....

Congratulations on retirement Ron! If you get bored there is always airplane building......

[rwanttaja]

Congratulations on retirement Ron! If you get bored there is always airplane building......
Currently contemplating the "Fly Baby Grande". Lengthened cockpit, widened fuselage, deeper instrument panel (current panel is too shallow to install most conventional electronics). A few other mods that have occurred to me, over the years.

The thought is to build the fuselage, then transfer the engine, wings, tail, and gear from my current Fly Baby. Actually, transferring most of the engine...leaving behind the generator, so I don't have to install a transponder or ADS-B. I figure I can thus build a new airplane for less than ADS-B would cost.

Other than that, I'll probably be writing quite a bit.

Ron Wanttaja

[Auburntsts]

Here's a quick data point. My engine is a Superior XP-360. I just looked up my info on the FAA site. They have it listed as AMA/EXPR.



Mine too. I have a BPE IO-540-X which is really a Lycoming IO-540-C4B5 with an uncertificated fuel injection system and a BPE data plate. In the database it's listed as AMA/EXPR and model UNKWN.

[martymayes]

Mine too. I have a BPE IO-540-X which is really a Lycoming IO-540-C4B5 with an uncertificated fuel injection system and a BPE data plate. In the database it's listed as AMA/EXPR and model UNKWN.

So is the problem with the FAA database fields? Are builders disclosing the nature of their experimental/unknown engine and the registry doesn't have the ability to enter that information so it defaults to AMA/EXPR and model UNKWN ??

[rwanttaja]

So is the problem with the FAA database fields? Are builders disclosing the nature of their experimental/unknown engine and the registry doesn't have the ability to enter that information so it defaults to AMA/EXPR and model UNKWN ??
The FAA database designates both engine make and model with a field that runs from 0 to 99999 (0 is none, and 99999 is the code for AMA/EXP). The 100,000 spaces are nowhere near fully occupied.

There might be a desire by some FAA types to not use up slots in the table with non-production-type engines. But there are a lot of production engines included as well.

I suspect it's a combination of laziness by the FAA/DARs and builders' desire to avoid AD traceability. If they issue an AD for a Continental O-200 and yours is officially "AMA/EXP", the FAA doesn't know it applies to you.

Ron Wanttaja

[Auburntsts]

So is the problem with the FAA database fields? Are builders disclosing the nature of their experimental/unknown engine and the registry doesn't have the ability to enter that information so it defaults to AMA/EXPR and model UNKWN ??

Well I can only speak for myself but my 8130 identified my motor as a Barrett Precision Engine IO-540-X and my AWC inspection was performed by a FSDO inspector who looked at it closely so nothing was hidden from him-- got a 40 hr Phase1 to prove it.

i can only assume Ron is on the right track with his answer.

[Anymouse]

So is the problem with the FAA database fields? Are builders disclosing the nature of their experimental/unknown engine and the registry doesn't have the ability to enter that information so it defaults to AMA/EXPR and model UNKWN ??

My Tango has been flying for just over 11 years. Quite frankly, I can't remember how I filled out the form, but I'm pretty sure I listed the engine as a Superior model.


I can only assume Ron is on the right track with his answer.

Me too!

[Frank Giger]

In the FAA's defense, we don't know how clunky their software is. This is an organization that still has forms with Ye Olde black carbon paper between sheets (!), and I can imagine the data entry stuff could well be the same one from 15 years ago in a DOS box.

Misplace the binder that has the ancient sheet of engine codes in it and "EXP OTHER" will do in a pinch.

:)

Frank "Data Entry Monkey" Giger

[SamP]

Hi Ron and Frank,
Forgive me as I try to understand the dataset. As a very rough reliability factor, would you divide the "In compare AMA set" column by the "Registry" column?

If only there was a way to dive just a little deeper. I know some people in the VW community are adamant against shrink fit prop hubs. Comparing the reliability of the Aerovee to the GP might be a rough way of seeing if those fears are founded.

[rwanttaja]

Hi Ron and Frank,
Forgive me as I try to understand the dataset. As a very rough reliability factor, would you divide the "In compare AMA set" column by the "Registry" column?
No, the "In AMA Compare Set" column in the number of aircraft registered with "Ama/EXP" engines I was able to determine had a particular engine type. Nominally, one would divide the number of accidents involving that engine type by the "Total" column. The problem is, I don't really like doing that when the fleet size is partially based on an estimate. I do it in categories (e.g., Certified engines vs. auto engines, etc.) but hate to point fingers at any particular engine based on deriving the fleet size.


If only there was a way to dive just a little deeper. I know some people in the VW community are adamant against shrink fit prop hubs. Comparing the reliability of the Aerovee to the GP might be a rough way of seeing if those fears are founded.

Ultimately, that would be my goal, but specifics about the engine are not always provided. If the NTSB says "VW", there are a lot of possible sub-types.

Ron Wanttaja