Heresy! Smooth Wheel Covers- Myth Busted!!!

[Vekke]

Originally Posted by JulianEdgar

From what he has posted here, he is doing a 3 minute run in each direction at 100 km/h, taking 5 instantaneous energy consumption measurements each way.

5 readings of the average consumtion display for each measurement point distance from start. Not a instant readings. I believe that method is described in the other topics clearly enough with videos and all.

Covers dont seem to work on the id3 oem alu wheels. I have tested now them 2 times and they were worse on each test. Taping with that big sheet takes only 15 minutes for 18" wheel. For 17" or smaller that would be even faster as it covers then whole rim in one shot. Its also faster to take out and at least in short test it does not leave any glue or pieces of tape to wheels. Just remember to clean the wheel before install properly otherwise they wont hold in place. That cleanup takes 30% of install time

[JulianEdgar]

Originally Posted by Vekke

5 readings of the average consumtion display for each measurement point distance from start. Not a instant readings. I believe that method is described in the other topics clearly enough with videos and all.

OK, I also covered that in a previous post:

So I have been thinking about Vekke's (wrong) way of making measurements.

If I have it correct, in short he drives for 5.5km in one direction, then 5.5km in the other direction, car warmed up and on cruise control.

He takes readings from the car's average fuel (energy) consumption display at points along each direction. It appears these readings are taken at about 800 metre intervals.

These readings get numerically closer together (ie have less variation) as he proceeds in each direction, so he thinks the reading is becoming more accurate.

So let's look at a parallel. We're measuring maximum daytime temperatures in a room, over a week. In degrees C, the temperatures are:

M - 23
T - 24
W - 21
T - 25
F - 18
S - 19
S - 23

Now let's show the average at each point:

T - Ave 23.5
W - Ave 22.7
T - Ave 23.3
F - Ave 21.5
S - Ave - 21.1
S - Ave - 21.9

As with any averaging, the the trend in the cumulative average is less variation as the number of items in the data set increases.

But this data tells us nothing about the accuracy of the thermometer, or of other aspects that might be impacting measured temperature. For example, maybe on Thursday a cat slept on the thermometer!

So it's not the number of interim, cumulative averaging measurements which is important, it's - as has been said here many times - the distance over which the average is taken. (eg if we measured max room temps over 2 months, Thursday's cat wouldn't even be noticeable in the average.)

And, for anyone who has ever tested fuel economy, two runs in opposite directions of only 5.5km isn't going to be very accurate - at least, not when tiny changes are trying to be measured.

It would be interesting to see the variation with (say) 5 runs done in this way, but with the car identically configured. Since in effect it's just the average of a 5km run (the interim points adding nothing to the accuracy) in each direction, I'd imagine the variation would be significant.

That's why no one else here does fuel economy runs over only 5km at 100 km/h.

[JulianEdgar]

Originally Posted by Vekke

Covers dont seem to work on the id3 oem alu wheels. I have tested now them 2 times and they were worse on each test. Taping with that big sheet takes only 15 minutes for 18" wheel. For 17" or smaller that would be even faster as it covers then whole rim in one shot. Its also faster to take out and at least in short test it does not leave any glue or pieces of tape to wheels. Just remember to clean the wheel before install properly otherwise they wont hold in place. That cleanup takes 30% of install time

I honestly can't see how you'd possibly know if they are better or worse after such a short test!

I've certainly never seen any technical reference that would support testing of the sort you are doing to find tiny changes in aero drag.

[California98Civic]

Originally Posted by Vekke

5 readings of the average consumtion display for each measurement point distance from start. Not a instant readings. I believe that method is described in the other topics clearly enough with videos and all.

Covers dont seem to work on the id3 oem alu wheels. I have tested now them 2 times and they were worse on each test. ...

Seems perfectly plausible to me. Thanks. We're at best thoughtful hobbyists churning familiar waters, and we don't get to have the materials for laboratory precision. Nonetheless, we can get a good sense of what works and what does not most of the time for our vehicle's specific configuration. Five runs and averages from a display, with other factors controlled as I know you do, is good enough for me. Did you consider combinations of mods, such as tire spats and the covers? Since we know they interact on some cars, and A-B-C test could have been tried. Also longer test loops.

[JulianEdgar]

Originally Posted by California98Civic

Five runs and averages from a display [...]

Um, that's not what he is doing. He is doing 5 readings of the averaging display during one run. The 5 readings add zero accuracy over doing 1 reading at the end.

So it's just one run of 5km in each direction at 100 km/h with an average at the end.

I just don't see how that technique is remotely going to give the accuracy needed to measure changes such as wheel covers.

[Vekke]

It's pretty easy to go on and try it for yourself with electric car. Or with diesel or petrol car with that pedal stopper but it won't be so accurate with them. Pedal stopper and seconds is accurate I should try how it works with this ev. I have used that route with my 15mg injection limited program with success by measuring the time changes.

When you test in hilly course you will see more variation which is good. If you just test in flat ground you will lose your ass feeling of the car you can feel and see from readings it's doing better. Like it will hold same reading longer and before than with other configuration. When you do a lot of testing you can guess where an when it will change the reading numbers and how they should change normally. You can see also that with more changes in the car point of lowest numbers have changed to further in the course than when started the testing with the car.

The readings at flat will show the same during all test if changes are minimal (or that 0.1 better which is much more than the ev resolution of 0.1 from 15.9kWh average mpg display is more accurate use that if possible. On hilly route they change on measurement points. On my data the last 400m is uphill slope on first direction. Previous point is lowest in the course. In most cases the readings in final uphill change up 0.1 units for every 50 meters or so. When you reach the goal if last 2 points are both lower you can be pretty sure the configuration is better. If over half is lower more certain. If you change the parts on and off every second run it averages the wind and temp changes. On my route through the woods I have seen that even little 3-5m/s wind is not yet a big reliability issue.

Id3 speed varies +-1kmh on the route and that creates biggest variation to results. On fossil cars you see more speed variation if you use your cruise control which is not good as it will over accelerate to hills lot more.

[JulianEdgar]

Originally Posted by Vekke

It's pretty easy to go on and try it for yourself with electric car. Or with diesel or petrol car with that pedal stopper but it won't be so accurate with them. Pedal stopper and seconds is accurate I should try how it works with this ev. I have used that route with my 15mg injection limited program with success by measuring the time changes.

When you test in hilly course you will see more variation which is good. If you just test in flat ground you will lose your ass feeling of the car you can feel and see from readings it's doing better. Like it will hold same reading longer and before than with other configuration. When you do a lot of testing you can guess where an when it will change the reading numbers and how they should change normally. You can see also that with more changes in the car point of lowest numbers have changed to further in the course than when started the testing with the car.

The readings at flat will show the same during all test if changes are minimal (or that 0.1 better which is much more than the ev resolution of 0.1 from 15.9kWh average mpg display is more accurate use that if possible. On hilly route they change on measurement points. On my data the last 400m is uphill slope on first direction. Previous point is lowest in the course. In most cases the readings in final uphill change up 0.1 units for every 50 meters or so. When you reach the goal if last 2 points are both lower you can be pretty sure the configuration is better. If over half is lower more certain. If you change the parts on and off every second run it averages the wind and temp changes. On my route through the woods I have seen that even little 3-5m/s wind is not yet a big reliability issue.

Id3 speed varies +-1kmh on the route and that creates biggest variation to results. On fossil cars you see more speed variation if you use your cruise control which is not good as it will over accelerate to hills lot more.

I don't know if it's a language thing but I am sorry, I can't understand what you are saying. As far as I can see, nearly every point you make will result in less accuracy.

Certainly, your windows up/down results didn't match what I'd expect from the technical literature (ie the increase in drag that wind tunnel results show on a variety of cars).

As I have said, why not do multiple tests (say 5) with the car absolutely unchanged and see the spread of the results?

[AeroMcAeroFace]

I agree with Julian here, without knowing your margin of error it is impossible to claim that something does or doesn't work.

You need to test your test regime before you test your variables.

If the scatter or variation in the results with no change to what you are testing is large, you have a large margin of error. If the scatter or variation is small you have a low margin of error.

Currently we don't know what the margin of error is.

Also, if the display is averaging doesn't that mean that results are double counted?

Reading one, will be the fuel economy on the journey so far,
Reading two, will be the fuel economy on the journey so far,
same with three, four and five. if you do the reading every km then you are double counting.

The first km and reading will include only the first km, the second will include the first km and the second km, the third will be the first,second, and third km.

[aerohead]

'If the eye offend thee, pluck it out. Pluck it out!'

[JulianEdgar]

Originally Posted by AeroMcAeroFace

I agree with Julian here, without knowing your margin of error it is impossible to claim that something does or doesn't work.

You need to test your test regime before you test your variables.

If the scatter or variation in the results with no change to what you are testing is large, you have a large margin of error. If the scatter or variation is small you have a low margin of error.

Currently we don't know what the margin of error is.

Also, if the display is averaging doesn't that mean that results are double counted?

Reading one, will be the fuel economy on the journey so far,
Reading two, will be the fuel economy on the journey so far,
same with three, four and five. if you do the reading every km then you are double counting.

The first km and reading will include only the first km, the second will include the first km and the second km, the third will be the first,second, and third km.

Yeah, I am a bit puzzled that the first basic of testing - doing multiple runs in the same configuration and ensuring the results vary less than the variation we're trying to measure - wasn't done.

The second basic is to do a test with a known change and then check that the measured change matches expectations based on the tech literature. (That was done but the result doesn't match the tech lit.)

The third basic is well known here - A/B/A testing with and without the modification.

If we could road test with accuracy down to 0.6 per cent drag variation then it would be fantastic - but based on the tech lit and my own testing, I don't think that's possible.