How to Read Shock Dyno Graphs Dirt
Probability Distribution Function
"There are 3 kinds of lies:
- Lies
- Damned lies
- And statistics" - Mark Twain
To get some item on fault analysis we have to rely on lies of the third type "statistics" to create what is know equally a probability distribution role. The probability distribution is created by examining each dyno measurement and asking the question "how far off is it". If that data betoken is off by 3% you put that betoken in the 3% error bin. If it is off past five% information technology goes in the 5% bin, and so on. At the end the number of data points in each bin are summed upward to effigy out how oft measurements are off at each mistake level.
For the 56 dyno runs on TT the probability distribution looks like this:
The first thing to check out in the error distribution is how peaked the curve is at 0% fault. If that curve were flat on top, say from -5% to +5% error, that would tell you the measurements lacked sensitivity. For case, if you tried to mensurate ane lbf with a 2,000 lbf load cell the instrument may not be sensitive plenty to measure those minor strength changes. In that example y'all would go random force measurements splattered over the +/- 5% error range and the PDF curve would be apartment on top. The data here shows no bear witness of that. The curve is nicely peaked effectually 0% fault indicating the dyno force and shaft velocity measurements are sensitive enough to resolve small damping force differences and generate a nicely peaked Gaussian shaped curve.
The second thing to await at is whether the data is skewed to one side or the other. If the fault distribution had more than data points on the left then the right that would tell y'all when the dyno went off it tended to always mensurate low. That might mean something like every bit the shock heated up in exam the temperature increase caused the oil viscosity to thin out and so the damping force measurements went low. If the information was skewed the other way with more data points on the right manus side that would hateful the dyno tended to read high. That might indicate the gas reservoir pressure went upwards as the stupor heated up in the test causing the measured damping force values to go high when the shock got hot.
Mistake distributions that are skewed to one side or the other give you lot some clues to what might be going wrong in test causing the errors. And that is a proficient matter. If you knew the dyno tended to ever read high or low you could try and effigy out what is causing the error and alter the test procedure or instrumentation to fix or at least minimize the measurement fault.
The information here shows no sign of being skewed to the left or the correct. That is a bad thing in terms of the errors are simply obviously old random noise with no rhyme or reason why the measurements went high or low.
Fault range
Figuring out the mistake range from statistical analysis becomes a big foggy mess. The usual method is pick some fraction of the data, the confidence level, and estimate the error range needed to bound that fraction of the data. The plot above shows the error range bounding 80% of the data.
Bounding 80% of the data sounds pretty practiced simply creates bug in dyno tuning. Suppose you lot wanted to increase damping force by 10%. To do that a series of dyno runs are fabricated adding confront shims and hacking around on the stack tapper to get to a 10% increase. If that takes v tests an 80% confidence interval means 4 out of v tests (fourscore%) will exist inside the error range and 1 exam in 5 (xx%) volition probably be outside the fault bounds. Hopefully the final test, the ane that demonstrated a ten% damping strength increment, is not the one test in five that was outside the normal mistake bounds.
Most of the data in the probability distribution are lumped around 0% error. So on any individual dyno run the most probable result is the data is practiced. Run 5 tests hacking around on the shock configuration and statistics will tell y'all 4 out of the 5 tests were probably good. But every once in awhile statistic as well tells yous the dyno is going to pull out some wacky value that is off in the ten to xv% error range.
That is the problem with statistical analysis, you lot just don't know for sure. The measurements are probably skilful. But one test in five is probably bad and every once in a while the dyno is going to pull out some wacky value out in the 10 to 15% error range. In the terminate, you just don't know for sure. That is what statistics shows yous.
Source: https://www.thumpertalk.com/forums/topic/1124437-dyno-data-%E2%80%93and-what-it-tells-us-about-how-to-tune-a-shim-stack-and-control-the-shape-of-the-damping-force-curve/page/19/
0 Response to "How to Read Shock Dyno Graphs Dirt"
Post a Comment