Mooney Accident Analysis

[jmbaute]

I wanted to learn a new data analysis tool (Power BI) so I decided to build a simple dashboard based on the NTSB database. I thought it might be interesting to the Mooney community so I am sharing the link with you. This is the same data you'll find in lots of other places, but hopefully this visualization makes it easy to slice & dice. I'm working on some other similar ones that include the GA fleet at large and one that is focused on Van's RV aircraft (GA is here and Van's is here if you're curious).

A couple notes:
- The data source is the XML file on the NTSB site. They update the data once per day. At this time I'm not sure if the link I'm posting refreshes against that data or if I have to trigger that refresh on my side.
- The report is fairly interactive, for example click on the model donut to filter the heat map and grid. Same thing on the heat map, click to filter by phase of flight. Click again to clear your filter.
- Right clicking various objects gives you a few more filtering options including raw data.
- If you click any of the links to the final report and don't get a result, it is because there isn't one issued yet. I'm just building that URL based on the accident number.
- I'm certainly open to feedback/suggestions but this is just something I'm playing with so I can't promise I'll do anything else with it. I am hoping to publish some other variations at a later date.
- I don't care if you share the link.
- If a moderator wishes to move or delete this post, I'm fine with that, too.

I hope this is educational. Blue skies.

John

[carusoam]

Hey John...

I clicked on the M20R on the pie chart... but got a few Bravo responses...

Any ideas?

I really like the idea, the ease of use, and the knowledge gained...

I really like not being on the list.

Best regards,

-a-

[mike_elliott]

Neat concept John. I noticed the data doesnt include one accident I am intimately familiar with, however. How accurate is your harvest of the DB?

[neilpilot]

8 minutes ago, mike_elliott said:

Neat concept John. I noticed the data doesnt include one accident I am intimately familiar with, however. How accurate is your harvest of the DB?

An aircraft other than N72FG, which is in the data?

[mike_elliott]

5 minutes ago, neilpilot said:

An aircraft other than N72FG, which is in the data?

I must have missed it not using the M20M filter. I do have to agree with them, the damage was "substantial" to say the least.

[kris_adams]

15 minutes ago, mike_elliott said:

Neat concept John. I noticed the data doesnt include one accident I am intimately familiar with, however. How accurate is your harvest of the DB?

Same here.  I was looking for a departure accident, but that was from quite a long time ago.

I enjoy looking through it btw!

[jmbaute]

So, the model data is human entered so there is not any consistency to the pattern they use. Sometimes it is M20R, M20-R, etc. I have a script that attempts to normalize this. I'll have to check why some of them are getting mixed in.

As for the missing aircraft, I'm not sure on that one. The data is current at least of as Friday. Might be something with the make or model pattern. I'll look into it later and reply back. Thanks for the feedback.

[warren.huisman]

I find it interesting that the 2nd highest accident rates occurred at cruise.  Generally, this is considered a pretty safe phase of flight.  Any ideas what are the common causes during this phase?  Fuel shortage perhaps?

[jmbaute]

3 hours ago, mike_elliott said:

Neat concept John. I noticed the data doesnt include one accident I am intimately familiar with, however. How accurate is your harvest of the DB?

Can you tell me a tail number of the missing plane? I will investigate against source.

[jmbaute]

3 hours ago, carusoam said:

Hey John...

I clicked on the M20R on the pie chart... but got a few Bravo responses...

Any ideas?

I really like the idea, the ease of use, and the knowledge gained...

I really like not being on the list.

Best regards,

-a-

So, I think we have a poor visual here- the donut chart's scale makes the exterior labels a bit misleading and they run together. You can verify your selection by the tooltip that shows when you hover over a slice, or expand the donut chart with the icon that shows up in the top right when you hover on the chart (it says "focus mode"). The R and M are neighboring slices and it is easy to pick the wrong one. I checked the filter and it seems to be working as designed...if that isn't the issue then let me know.

[neilpilot]

1 hour ago, jmbaute said:

Can you tell me a tail number of the missing plane? I will investigate against source.

As I indicated above I think it wasn't missing, just overlooked.  N72FG.

[carusoam]

John,

I think the issue I encountered is the label for the M20M is missing...

The line for the M20R is so close to the label for the M22 it appeared that the line belonged to the M22 and not the R...

I am using an iPad Air....

Keep up the good work... I have some accident reviews to go through... a necessary, homework assignment to avoid the errors of the past.

Best regards,

-a-

[carusoam]

https://app.ntsb.gov/pdfgenerator/ReportGeneratorFile.ashx?EventID=20130815X14705&AKey=1&RType=HTML&IType=CA

I found an interesting one....  M20R gear up caused by a Battery failure... G1000 system went dark, pilot successfully navigated by iPad...

20/20 hind-sight... click that battery switch and everything will be OK...

The pilot was interested in being on the ground, not saving somebody else's plane...  I get it...

 

But, I don't get what I am looking for out of the reports, photos, interviews...

All I want to know is what Battery failed, Brand and model would be nice... did the charging system kill it?

Did I miss something?

Or is the NTSB only interested in engine, prop, airframe and pilot details...? 

Wanting to prevent a similar problem... all I want to know is Battery make and model, and what killed it...

 

Back on topic...

John's system has done a great job of organizing the accident reports.  The NTSB seems to have put a lot of data in public view including photos and supporting docs.

in a quick review of the M20R list.  There were VFR into IMC, dark night similar to IMC, IFR In snow that was like dark night, and a whole bunch of going too fast on landing, bounces, and failed go arounds on too short of a runway....

Nice work, John!

Best regards,

-a-

[jmbaute]

8 hours ago, carusoam said:

John,

I think the issue I encountered is the label for the M20M is missing...

The line for the M20R is so close to the label for the M22 it appeared that the line belonged to the M22 and not the R...

I am using an iPad Air....

Keep up the good work... I have some accident reviews to go through... a necessary, homework assignment to avoid the errors of the past.

Best regards,

-a-

Glad you're finding it useful; the rendering on mobile devices is not ideal. I'll see if there is a better way to scale this.

[Marauder]

So, I think we have a poor visual here- the donut chart's scale makes the exterior labels a bit misleading and they run together. You can verify your selection by the tooltip that shows when you hover over a slice, or expand the donut chart with the icon that shows up in the top right when you hover on the chart (it says "focus mode"). The R and M are neighboring slices and it is easy to pick the wrong one. I checked the filter and it seems to be working as designed...if that isn't the issue then let me know.

You may be able to use the serial number prefix to denote model number instead of the model number. My F model starts with 22-XXXX. The Js I think are 24-XXXX.


Sent from my iPad using Tapatalk Pro

[bcwiseguy]

An interesting look. Thanks for sharing this. Some of the data surprises me....like the number of accidents during cruise. It would be nice to be able to see a graph that compares IMC to VMC. I was expecting a larger occurrence of IMC so that was a bit of a surprise.

[rbridges]

I didn't get to spend too much time with it, but it looks pretty cool.  Thanks for taking time to do it and sharing!

[Marauder]

Another factor that isn’t shown is the number of flights that made it back without an incident or accident report. It would be nice to have those available to look at.

Back in the 1990s when the NTSB made these electronically available, I did an analysis of the M20F accidents. My primary concern was engine failure. It showed that less than 15% were related to mechanical failure (I suspect it is a lot higher because a number were able to successfully get the plane back on the ground). The rest were the usual stupid pilot tricks. Running out of fuel, IMC encounters and landing accidents.


Sent from my iPad using Tapatalk Pro

[Marauder]

36 minutes ago, Marauder said:

Another factor that isn’t shown is the number of flights that made it back without an incident or accident report. It would be nice to have those available to look at.

Like this: https://www.msn.com/en-us/news/us/pilot-lands-plane-on-costa-mesa-freeway-i-saw-an-opening-on-the-highway-and-i-went-for-it-right-away/ar-BBIn7Nn?li=BBnb7Kz

[jmbaute]

38 minutes ago, Marauder said:

Another factor that isn’t shown is the number of flights that made it back without an incident or accident report. It would be nice to have those available to look at.
 

The NTSB splits these into two categories, accidents and incidents. This report is filtered to accidents. I could probably add another filter to allow you to choose.