Re: [Jsbsim-devel] C-172 Performance

[Curtis L. O. <cur...@fl...>]

Bill Galbraith wrote:

>This is of particular interest to me. Detailed testing of the C-172 would
>require detailed flight test results. In general, you would set up a script
>to EXACTLY mimic the flight test procedure, such as weight and blance (crew,
>equipment, ballast for CG control), power and control application (need time
>histories of control inputs), and atmospheric test conditions (temp, wind,
>pressure).
>
>Let me digress for a moment. I have a lot of experience in flight test data
>reduction and application to simulation validation. For standard tests,
>there are standard test procedures that the pilot is SUPPOSED to follow,
>although they don't always follow. For an in-flight test, a particular
>dynamic test might require the pilot to trim the aircraft to a particular
>condition before performing a dynamic maneuver. The pilot might be sloppy,
>winds may affect ease of trim, or the test point might be particularly
>difficult to trim to. If you have time history information available to see
>if he was trimmed, and you trim your simulator to the exact same conditions
>(a slight rate of decent if that's what the pilot flew, etc), then apply the
>same input that the pilot did. If your model is accurate and you have
>faithfully replicated the pilot actions, you should get the same response.
>
>Think about this: If you applied full elevator deflection from the very
>start of a takeoff run, versus applying it at 45 knots, you will get
>different results, right?
>
>So, back to the C172 issue. Unless your shirt says "Cessna" on it, the
>chances that you have this flight test data available are pretty slim.
>
>So, what is the solution here?  Well, I've know guys that will do "bar stool
>and yard stick" flight tests (yeah, really using a bar stool and a yard
>stick for their instrumentation). You spend 10-20 hours in the air, with
>some simple measuring tools (stop watch, hand-held force guage) and a
>clipboard and record data this way. Probably not going to happen here. 
>
>Therefore, you are stuck digging through POHs, Jane's All The World's
>Aircraft, and anything else that mentions the words "Cessna" and "C-172".
>You have to decide if the information provided is valid for your aircraft,
>and if there is enough information to be useable (set-up, conditions,
>procedure, etc). You compare your results with generalize curves in the POH,
>keeping in mind that you might perform better than the POH predicts.
>Phugoid, short period, steady heading sideslip, flap response... Yeah, good
>luck. You won't find any useful data.
>
>
>
>Just as an aside (since the fingers are typing fast this morning), buying a
>flight test package doesn't mean that the data is good. I worked on a
>simulator for an H-60 helicopter. We were fortunate enough to get the raw
>flight test data digitally, so we could look at ALL the channels of data
>collected (128 channels or more, if I remember correctly). When the aircraft
>was supposed to be trimmed, some channels indicated that it wasn't. We
>traced it down to accelerometers that didn't read zero when sitting on the
>ground, so we had to subtract out these biases from the flight test data. If
>this wasn't done, you'd get some very different results. The pilot might be
>asked to input a 1 inch step input and hold it, but flight test data showed
>that he might overshoot to 1.2 inches, then release back to 1 inch, but the
>stick would continue to drift back to 0.5 inches. If you input a perfect 1
>inch step rather than what the pilot input, you'd get different results. By
>mimicing the exact setup and inputs as best as we could, the resulting
>simulator turned out to be an excellent simulator, and the flight model has
>been used on subsequent H-60 simulators without change.
>  
>

Bill,

I have nothing useful to add here other than it is clear you are 
speaking from experience. :-)

I went through the same thing when validated the ATC single engine 
trainer last spring.  The baseline data was gathered with video tape and 
force gauges and rulers and levels and such.  But even in the final 
validatation step, we were still finding goofy things in the flight test 
data where the results made it clear the pilot didn't follow the 
specified test procedure.  Like you say, trimmed for 100 fpm decent can 
be a lot different than trimmed for level flight, but it might look 
"good enough" when the test is being run.  There were a few cases where 
we had to deduce what the pilot was really doing from the real flight 
test data so we could build the automated simulator test to input the 
same control sequence.

I'm not an "aero" guy so I found the whole process incredibly facinating 
and enjoyable, although it was very tedious and time consuming as well.  
Fortunately I'm wired for tedious and time consuming so it worked out 
fine. :-)

Curt.

-- 
Curtis Olson        http://www.flightgear.org/~curt
HumanFIRST Program  http://www.humanfirst.umn.edu/
FlightGear Project  http://www.flightgear.org
Unique text:        2f585eeea02e2c79d7b1d8c4963bae2d