Bruce, I find your analysis helpful.

Indeed, it is metadata. However, if you have enough of it, a statistically significant picture emerges. But even so, metadata doesn't explain the detailed why and how of the picture it presents. That's the next step. At best it tells you where to look and what to look for.

I contend the 130,000 annual aero tows in the US and the related accidents are statistically significant. It's pretty simple to compare German winch launch to US aero tow for 2011. German winch launch and US aero tow suffered exactly the same number and type of accidents so the relative number of launches tells the story. 900,000 vs 130,000 is 7:1. The two countries can be said to be experts at their respective launch styles so it's a valid comparison. i.e. "Given equal levels of expertise, winch launch is 7 times safer than aero tow." I think both launch styles are equally sensitive to a lack of pilot skill.

Despite the truly excellent German winch safety record, Professor Dr. Richard Eppler at the University of Stuttgart (He of numerical airfoil analysis fame.) thinks it can be improved further and has written a paper on the subject for flight instructors. One of the insights from the paper is the concept of a minimum airspeed for winch launch that should be part of airworthiness documents and listed in flight manuals. DG was apparently listening so the DG1000 manual lists a minimum winch airspeed of 49 knots. So now in addition to Vw-max we have Vw-min. If pilots had simply ensured their airspeed was always above Vw-min, a significant number of accidents would not have happened.

Another not-so-rare accident type starts with a wing tip contacting the ground during the takeoff roll where it digs in and the rapidly accelerating glider "cartwheels". In the worst case, the glider ends up inverted and the result is fatal.

Counter-intuitively, the solution is even more acceleration to give the pilot aileron control before a wing can drop and then get him airborne where the wing can't touch the ground. More acceleration also gets the glider above Vw-min before the pilot can begin rotating into the climb. Vw-min assures enough elevator authority to prevent inertial pitch-up in high-CG, low hook gliders.

YouTube videos show a difference between UK practice and German practice. German ground rolls are pretty consistent at two seconds indicating 1G acceleration. By contrast, I've seen videos of very wobbly UK ground rolls taking as much as 15 seconds but that's an exception - most are 3 - 4 seconds.




On Wednesday, June 26, 2013 12:45:49 AM UTC-6, BruceGreeff wrote:
Meta data is data describing other data. In this case the metadata is

what is missing.



We don't have enough information for meaningful comparisons.



That is where I where I am going - and in fact gone.



You are using historical data, without enough context to draw

conclusions about current practice.



So the specific example I gave in fact demonstrates that the useful

information is often well hidden. If we use your approach you have a

statistic that supports the contention that winch launch is 100% more

dangerous than aerotow for this set of data.



Lots of problems -

1 the set is too small (the BGA vs LBA sets are statistically significant)

2 the metric and the generator are not linked. Launch method had nothing

to do with either incident. Poor airfield maintenance and pilot decision

on risk is the cause (both involved a groundloop because someone did not

cut the grass properly and the pilot decided to proceed anyway)

3 the outcome is biased by a "hidden" factor - the actual damage that

resulted in the report of damage to an airframe was traced to

progressive failure of a poor repair after an outlanding accident

decades back when the glider was a competitive two seater in Germany. It

was going to break some time soon. This flight just happened to involve

the trigger stress.



Did the launch method have much to do with the damage, not really. If

the poorly repaired wing failed during spin training, would it mean that

our spin training is dangerous?



So - to be very explicit - where I am going with my comment is that I am

concerned that you may be oversimplifying in an attempt to prove your

opinion. Where an hypothesis would be well enough constructed that is

could be proven.



If you said that the historical accident data for the period you

reviewed showed a difference in the relative safety of operations at

German versus British gliding clubs, that might even be an uncontestable

finding (102). If you compared the trends in accident rates and found

them to be stable over the period you might even be able to conclude

that this is representative of the current situation. (the numbers are

comparable)

If you had qualified the analysis by using the numbers where the

investigators concluded after careful consideration that the launch

method was the causal factor, then you could even conclude that "Winch

launch operations" are more dangerous in one than in the other.



Since you have too little "metadata" and are making conclusion without

all the painful logical and mathematical rigour that makes Business

Intelligence type metrics such a tedious business, I think it is

dangerous and misleading to make the kind of assertions you are making.



Unless you know the actual cause of every number in the set, you can't

make confident assertions.



(Humour alert) There are 10 kinds of people in the world. Those who

understand binary, and those who don't. (10 in binary = 2 in decimal) ;-)



As Bob says, is it possible? Definitely.

Does a simple integer comparison of two numbers prove the assertion? I

think you would have to question that.



Personally, I find winch launching risk to be more tractable. You can

control risk better.



But again - context is important. Winch launching an open class glass

ship with water ballast would probably be less safe than aerotowing the

same combination, all things being equal.



T59D #1771



Not sure where you're going with this, Bruce. Of course there are reasons for the differences and that's a a very important discussion to have but it doesn't affect the difference itself.



The raw accident count for the number of launches done is really metadata (To use a currently popular term.) Metadata is very valuable in seeing the big picture but not so much for analyzing the details for why it looks that way. Hopefully, the metadata will provoke that kind of detailed analysis.





--

Bruce Greeff

T59D #1771