Quote:
Originally Posted by 3-Wheeler
Jim,
One way that I compensated for daily shifts, is to continue to collect data on the same mods for many numerous days.
I did this back in 2009 and 2010, and collected data all winter into summer, with basically the same setup on the car. Who is going to dramatically change their Eco-Mods in the middle of the winter? Right?
Anyway, my data acquisition methods turned out to be too coarse and not repetitive enough. The sampling rate was not high enough.
Looking at the data did start to show just how much the Crr changes in cold weather versus summer, but I was not convinced enough of the accuracy to post anything here at EM.
For those that like to speculate, the 50 coast down runs showed at least a 17% increase in rolling drag for winter temps versus summer, but again, I was not convinced that from a statistical standpoint the data was that valid.
If one were to test a MOD over many days, and see the same decrease in Cd or the same increase in mileage, then one could assume that environmental conditions were properly addressed, not just one run over the coarse of a single day of testing.
Sorry to say, but without a wind tunnel, I think that is what we have to resort to. Multiple tests over multiple days.
Actually it's rather interesting to perform multiple tests in this way, as it shows just how consistent (or inconsistent) tests of this nature really are in the real world, so we learn something even though we may not actually know how big of an improvement our latest aero mod is worth.
Jim.
|
But, in the end you basically failed to get enough stable data
I think though that you might be right on trying to get lots of data over several days, at least for "small" improvements. One could then treat the improvement as a statistical variable with a mean and standard deviation.
In the future, I want to try this approach to coastdown testing(at least for small improvements):
1. Pick only summer days, 1 to 4 p.m. when the temperature is high and relatively stable.
2. Pick only summer days, 1 to 4 p.m. when the wind is most likely to be nearly dead.
3. Do two way testing.
4. Average the two "A" segments to null any linear gradient in conditions.
5. Compare the "B" test to 4 above.
6. Treat the calculated improvement as a statistical variable.
7. Test over severl/many afternoons.
8. And, let me add, disable any assist/regen.
I still think that doing actual 2 way fuel economy test drives is the most direct way to measure experiments, but here again several procedural constraints must be obeyed:
1. Two way testing.
2. Moderate length (10mi minimum) so that driver instability is minimized.
3. Carefully controlled target average speed. This can be monitored with any number of instruments including the scangauge.
4. Re Insight, constant or near constant lean burn.
5. Summer day, little wind, stable temperature.
6. Standardized driving technique.
7. Level or near level course.
8. Disableany assist/regen.
Before my standardized testing course was destroyed last fall by construction, I was able to drive very tightly grouped runs using the above techniques.
So, I think we are left with three techniques which can be made to work under very good conditions and with careful driving:
1. Coastdown(Yielding indirect results)
2. Long course driving for MPG(Yielding direct results)
3. Tuft testing.(Yielding VERY indirect results)
JMHO as always