Motors manufacturers

Natti Notti

Filing Flight Plan
Joined
Dec 11, 2020
Messages
25
Display Name

Display name:
Natti
Good day to all!

I do Data Engineering for living and I was always curious to analyse NTSB accident dataset for my research on engines reliability. After I loaded whole database I came to notice their dataset is not that clean. For example I see many tens of "CONTINENTAL..." manufactures:
upload_2021-5-5_12-39-31.png

So my question to experienced pilots and mechanics: Besides obvious misspellings, is it fair to assume all these CONTIS are in fact one company? And is this the case for other engine manufacturers like LYCOMING? Or there are some exceptions? Thanks!
 
Last edited:
analyse NTSB accident dataset for my research on engines reliability.
Is it fair to assume all these CONTIS are in fact one company?
FYI: There's no single answer to either point you bring up. Were the majority of the "Continental" engines listed above originally produced by the company currently known as Continental Aerospace Technologies? Yes. However, once those engines left the factory that original engine conformity could have been changed by anyone. For example, the PPonk listed in items 40 and 42 of your chart would indicate those engines have been modified by a popular 3rd party STC. As to other variances it could have been simply how the name "Continental" was spelled/entered on the original accident form prior to the info entered in the database.

As to reliability, what are you trying to track? Overall Continental reliability, or factory new engines, or field overhauled engines, or STC modified engines, or...? In my experience, engine reliability is more subjective to an engines individual history vs a simple general overview. Are some engine models more reliable? Sure. Are some brand/type engine cylinders more reliable? Sure. So it may take a deeper review to get a truer picture. Plus the fact the NTSB data only gives you a partial dataset as not all engine failures make it into that database.
 
Last edited:
Good day to all!

I do Data Engineering for living and I was always curious to analyse NTSB accident dataset for my research on engines reliability. After I loaded whole database I came to notice their dataset is not that clean. For example I see many tens of "CONTINENTAL..." manufactures:
View attachment 96089

So my question to experienced pilots and mechanics: Besides obvious misspellings, is it fair to assume all these CONTIS are in fact one company? And is this the case for other engine manufacturers like LYCOMING? Or there are some exceptions? Thanks!

The entries on the NTSB database are based on what the investigator manually entered. Their spelling sometimes varies, and/or they use whatever abbreviation they prefer (e.g., not standardized). And, as Bell206 mentions, sometimes the engines are on Experimental Amateur-Built aircraft and have modifications which may or may not be reflected in the investigator's entry for the engine type.

Best bet, if you're looking to study reliability, is to limit yourself to aircraft with Standard airworthiness. Fewer modifications, and there are standards for care that most should undergo.

The easiest way do to this is to select only accidents where the "Homebuilt" or "Amateur-Built" flags are "No". The title of the flags varies depending on whether you download the database or are using the online search tool. Either way, the title is a misnomer...it is actually used to denote Special Airworthiness (of which Amateur-Built is just one element). But if you're looking for Standard airworthiness, that won't matter. "No" should mean "Standard."

Still, it's not all that reliable...any given year, I find 5-10 homebuilt accidents where the "Homebuilt" flag says "No".

A more reliable method is to cross-reference it with the FAA Aircraft Registration database. There's a "Certification" column; if it starts with a "1", it's a Standard-Category aircraft. Of course, if you're looking at past accidents, there's a likelihood that the N-Number may have been reassigned. And the format of the Make/Model entries on the NTSB database won't necessarily match that in the FAA database. But the most common types (Cessna, Piper, Beech, Mooney, Cirrus) probably will be about the same.

Ron Wanttaja
 
Great points Bell206 and Ron, thank you. My initial idea was to filter on all engine related accidents for certified aircraft and calculate average time to failure for each manufacturer on factory new or overhauled engine. For my purpose I consider overhauled engine having the same quality as a new one. Later when I loaded the data I found that field which indicates hours since overhaul (which I wanted to use) is null across all engines. So, I have to come up with some approximation based on airframe age or taking remainder dividing total engine time by 2000, which of course is not ideal either. Ultimately I wanted to get some picture on reliability of Rotax vs Lycoming and Contis. This interest stemmed from ATSB research https://www.atsb.gov.au/media/5769864/ar-2013-107-final-report.pdf circa 2013, where they ranked engine manufacturers based on number of engine failures. But I find their metric is bit flawed because they counted number of engine failures w/o taking into account engine times for each manufacturer.
 
I guess "English Proficient" does not include spelling.
FYI: you'll find information from the initial Form 6210.1 Report is used for the basic record data which is completed by the pilot or operator especially when no NTSB IIRC travels to the site.
 
This interest stemmed from ATSB research https://www.atsb.gov.au/media/5769864/ar-2013-107-final-report.pdf circa 2013, where they ranked engine manufacturers based on number of engine failures. But I find their metric is bit flawed because they counted number of engine failures w/o taking into account engine times for each manufacturer.

"Rotax powered aircraft were the next most common with 87 (one in 36), followed by aircraft with Lycoming (58 – one in 35) and Continental (28 – one in 35) engines." So, no real difference between those.

The primary flaw in there metric would be the small sample sizes. Further, I suspect that total engine hours is probably not going to be statistically significant. The factors that are more likely to be significant would be time since maintenance and frequency of use. But there is no way (that I am aware of) to get that kind of data.
 
I guess "English Proficient" does not include spelling.

FAA. Yep.
Believe we're referring to the NTSB accident database, not the FAA. FAA registration database just says, "CONT MOTOR" as the make for Standard category aircraft with Continental engines. For homebuilts, one can probably blame both the DAR and, probably, the owners themselves.

In any case, through much of its history, the NTSB investigators were probably not the people who physically typed the data into the system.

Ron Wanttaja
 
"Rotax powered aircraft were the next most common with 87 (one in 36), followed by aircraft with Lycoming (58 – one in 35) and Continental (28 – one in 35) engines." So, no real difference between those.

The primary flaw in there metric would be the small sample sizes. Further, I suspect that total engine hours is probably not going to be statistically significant. The factors that are more likely to be significant would be time since maintenance and frequency of use. But there is no way (that I am aware of) to get that kind of data.
The referenced file is also from the Australian board, not the US one. May be some methodology differences.

Ron Wanttaja
 
Great points Bell206 and Ron, thank you. My initial idea was to filter on all engine related accidents for certified aircraft and calculate average time to failure for each manufacturer on factory new or overhauled engine. For my purpose I consider overhauled engine having the same quality as a new one. Later when I loaded the data I found that field which indicates hours since overhaul (which I wanted to use) is null across all engines.

According to my last downloaded database, this is listed as a "new field." So one would hope it would gradually get populated over time, but only for new accidents. And, of course, it depends on the NTSB investigator logging the information.

I ran a search of the NTSB database, looking for the phase "since major" in the preliminary or factual narratives. Got 963 hits. So SOME data is there.

Ron Wanttaja
 
I am noticing aircraft with conversions are designated as a separate aircraft make in NTSB. I wonder if this is correct designation in GA and if it is proper leave make as a parent (CESSNA in this case) and indicate the fact of conversion in the model of the aircraft (ex: Wren 460)? Here is example for Cessnas:

upload_2021-5-9_23-40-12.png
 
I am noticing aircraft with conversions are designated as a separate aircraft make in NTSB. I wonder if this is correct designation in GA and if it is proper leave make as a parent (CESSNA in this case) and indicate the fact of conversion in the model of the aircraft (ex: Wren 460)?
Again, like the engine data, you cannot assume aircraft make/model designations are consistent. Yes, if it says "Reims-Cessna" you can assume it's a French-manufactured Cessna, but if it DOESN'T say "Reims-Cessna," you cannot assume it isn't. I'm not sure of the current processes, but through much of the span of the NTSB, the make/model entries were made manually on paper forms, and subsequently keyed into digital systems by ordinary clerks.

If the make/model just says, "Cessna 172" instead of something like "Cessna 172S", you CANNOT assume it is a 1956-1959 model (the 172A came out in 1960).

To make it even more fun, realize that some accidents are listed as a "Textron Aviation 172" rather than a Cessna 172.

Like I said before, look at the narratives and cross-reference to the FAA database.

Ron Wanttaja
 
The FAA database doesn't actually have the TEXT for the manufacturer or model in the individual plane entries in the raw data, they just have a unique numeric code which you have to look up in a separate database. Much less chance to get things wrong that way.
 
@Natti Notti

What @wanttaja is alluding too is the data in the NTSB field can point you in a direction. But you then have to read the actual reports to confirm what you find.

So use the NTSB database to filter down to standard aircraft, then pull each report looking for the data you want.

Tim

Sent from my HD1907 using Tapatalk
 
I wonder if this is correct designation in GA and if it is proper leave make as a parent (CESSNA in this case) and indicate the fact of conversion in the model of the aircraft (ex: Wren 460)?
To add to the above, the "official" designation of an aircraft model or variant of that model, can be found on the TCDS or as required by an STC or other approval. Any other model designation outside those instances is whatever the person making that record decides the designation to be. For example, the Cessna/Air Repair, Inc. line item is what someone entered to reflect that this particular Cessna 305 (L-19) falls under the current TC holder Air Power and not the previous holder Cessna. Same with Cessna Soloy as this references the turbine conversion but I do not believe the Soloy STC requires a change to the data plate or manufacturer. So the model list above can be taken more as a notation than any official model designation.
 
Last edited:
I should also point out that the FAA (registration) database usually just lists the engine that came in the plane originally. I've got an STC'd engine change to an IO-550, but the FAA still says:
LYCOMING GO-435C&D SER. Which isn't even too specific (mine was a -C2).
 
The FAA database doesn't actually have the TEXT for the manufacturer or model in the individual plane entries in the raw data, they just have a unique numeric code which you have to look up in a separate database. Much less chance to get things wrong that way.
Yes, the FAA database is a bunch of related files with numeric codes for engine look-up, and alphanumeric ones for make/model. Each file provides additional information, such as the horsepower or number of seats.

Almost all the Continental engines are listed as "CONT MOTOR", with two exceptions. The first is the "IOX370 SERIES", listed as 205 horsepower and four-cycle. The second is the "XIO-370 SERIE" listed as a 180 HP reciprocating engine. That's not a typo, BTW, the listing doesn't have the "S" at the end. And I don't know why the first would have the code "8" for four-cycle, and the second with a "1" code for recip....

I don't see a currently-registered aircraft with the first engine, but there's one "XIO-370 SERIE" installed in an RV-6. There is a deregistered aircraft with the first engine...listed as a KR2.

Ron Wanttaja
 
As this quickly run into a wall, my next step will be to download FAA master registration database and match aircraft type in it based on N-number. I will report how far I will go with it. Thanks all!
 
As this quickly run into a wall, my next step will be to download FAA master registration database and match aircraft type in it based on N-number. I will report how far I will go with it. Thanks all!
Here's the Access query I use:
upload_2021-5-10_10-44-0.png
Also, keep in mind that the older the accident you look at, the more likely the N-Number has been reassigned. Best verify it's at least the same make.

The FAA registration database also includes the list of de-registered aircraft; these might be an aid as well.

If you're getting the impression that analyzing the NTSB and FAA databases takes a lot of manual effort...you're right!

The FAA Registration database download also includes a PDF file that explains each column. You've probably already found it, but for the other folks out there, the equivalent file for the NTSB database is embedded in the Access file itself...look for a table called "eADMSPUB_DataDictionary."

Ron Wanttaja
 
I did database stuff for the DoD. We arranged so that it was nearly impossible to put a wrong entry in the important fields. The NTSB database (which I've manipulated in Oracle) is a hot mess.
 
I did database stuff for the DoD. We arranged so that it was nearly impossible to put a wrong entry in the important fields. The NTSB database (which I've manipulated in Oracle) is a hot mess.

I disagree. The NTSB database is organized so no analysis can be performed in order to protect the participants who attempt to make money in the world of aviation.

Tim (said sarcastically)
 
Thanks all, yes I am aware of the data model, I used it to get myself up to speed with NTSB data. The issue with NTSB dataset is data quality. I've spent a week writing cleanup and standardization code to bring engine and aircraft data (Piper only for now) into common standard. It is tedious work and not feasible for one person for the whole dataset. I will think to expose this data for crowd editing at some point and/or use other more reliable data sources like FAA (which as other say also a mess :) ) Interestingly the `occurences` table in NTSB post 2008 dataset is completely empty. I myself is a certified Oracle DBA, but use GCP Big Query to analyze NTSB dataset since Cloud technologies are easier to use these days. Agree with TIm, I found interesting that some important fields are empty or sparsely populated, ex:
eng_time_overhaul field in engines table is only populated for 4995 engines out of 85090 making impossible any analysis around reliability. But I was able to gain some insights around in-flight breakups of the PA-46 which always interested me.
 
Last edited:
Thanks all, yes I am aware of the data model, I used it to get myself up to speed with NTSB data. The issue with NTSB dataset is data quality. I've spent a week writing cleanup and standardization code to bring engine and aircraft data (Piper only for now) into common standard. It is tedious work and not feasible for one person for the whole dataset. I will think to expose this data for crowd editing at some point and/or use other more reliable data sources like FAA (which as other say also a mess :) ) Interestingly the `occurenses` table in NTSB post 2008 dataset is completely empty. I myself is a certified Oracle DBA, but use GCP Big Query to analyze NTSB dataset.

If you are going that far. You are probably better off using an AI engine to read the raw NTSB reports and build the model directly from there and skip the whole polluted database.

Tim
 
Back
Top