(coastdown-related) So I remembered why I paused dev work on MPGuino for 3 years...

[JRMichler]

Originally Posted by t vago

Turns out that coastdown testing calculation is not easy!

Right now I am looking at using this Instructable as a start, but with using a starting speed of 115 kph and an ending speed of 15 kph instead.

I remember the discussion about there actually existing a V term that corresponded to drivetrain-related viscous drag, so there's a complication there. Not really sure how I'd calculate a C(d) either, since people have pointed out that cars have a negative lift component at speed which is going to throw off the V^2 term calculation. Hm.. at least C(rr) could still be reliably calculated... perhaps.

I think the best to be hoped for is to just do a sort of stopwatch function that starts when the vehicle slows down and passes 115 KPH, and stops when the vehicle reaches 15 kph. One could do an average of several passes in that fashion, and store the resultant elapsed time for A-B-A comparisons.

Your approach is good, and would be better with one change. If you want to separate aerodynamic drag (proportional to V^2) from viscous drag (proportional to V) from constant drag (fixed amount, not proportional to speed), you need a wider range of speeds.

In my coastdown testing (https://ecomodder.com/forum/showthre...yon-20405.html), I started at 60 MPH and coasted down to zero speed. I did not try to pull a viscous drag number from that data because the speed range was not large enough. I would have needed to coast down from about 80 MPH to get good enough data for that.

The effect of a barely noticeable breeze surprised me. I would not have thought that a 5-7 MPH breeze would be so large. That breeze was barely detectable. Another thing to watch is the road slope. A slope of 0.1% (5 feet per mile) will have a noticeable effect on the drag, especially the rolling resistance.

[freebeard]

...you want to separate aerodynamic drag (proportional to V^2) from viscous drag (proportional to V) from constant drag (fixed amount, not proportional to speed)...

I can imagine aerodynamic and skin friction drag, but what is constant drag?

Another thing to watch is the road slope.

I'm now on board with level, no up or downgrade is going to be constant.

The only locations I'm aware of that might get to 1 in 1000 slope would be the salt flats in Utah and Bolivia.

Wind is another matter -- maybe a fixed course between two weather stations?

[t vago]

Originally Posted by JRMichler

Your approach is good, and would be better with one change. If you want to separate aerodynamic drag (proportional to V^2) from viscous drag (proportional to V) from constant drag (fixed amount, not proportional to speed), you need a wider range of speeds.

Yah, that is most likely the case. Also, how would one separate out the aero drag due to lift? That would cause vehicle weight to fluctuate at speed, dependent on speed of course, and would make determining C(rr) that much harder...

Originally Posted by JRMichler

In my coastdown testing (https://ecomodder.com/forum/showthre...yon-20405.html), I started at 60 MPH and coasted down to zero speed. I did not try to pull a viscous drag number from that data because the speed range was not large enough. I would have needed to coast down from about 80 MPH to get good enough data for that.

The effect of a barely noticeable breeze surprised me. I would not have thought that a 5-7 MPH breeze would be so large. That breeze was barely detectable. Another thing to watch is the road slope. A slope of 0.1% (5 feet per mile) will have a noticeable effect on the drag, especially the rolling resistance.

That just goes to show that proper coastdown testing is really hard to do. Folks like yourself and aerohead really deserve recognition for actually doing coastdown testing.

Originally Posted by freebeard

I can imagine aerodynamic and skin friction drag, but what is constant drag?

That would be C(rr), which is ideally dependent on the rolling resistance of your wheels.

Hm... maybe I could rig up a few different test schedules - one at low speed to solely determine C(rr), and then another to determine C(d) and C(l)? I will have to look into that

Originally Posted by freebeard

I'm now on board with level, no up or downgrade is going to be constant.

You can kind of mitigate around this by doing a coastdown run between point A and point B, and doing a reciprocal course between point B and point A. However, the averaging is only accurate during those portions of the runs where vehicle speed is approximately equal. However, that would have to assume the slope was in fact constant for the length of the run.

One could, I suppose, add a slope angle parameter into the equation to be solved for, along with another parameter to specify whether the vehicle is going up the slope or down it. Again, that would have to assume the slope was in fact constant for the length of the run.

It would be much more ideal to find some path that is known to be reasonably flat. Less computations that way.

[aerohead]

I've watched this thread while out of town, and been thinking how you might 'linearize' the testing, which dates to something CAR and DRIVER talked about back in 1978.
If you had access to a constant-grade portion of some mountain highway, say 3%, and there was some slack time where traffic would permit it, a fully-warmed vehicle, in neutral, would reach a 'terminal velocity' signifying a force equal to 3% of the mass of the vehicle.
This velocity would just compensate for aero and R-R.
Knowing the original Cd and frontal area would allow one to calculate the rolling force coefficient for the tires.
--------------------------------------------------------------------------------------
If the terminal velocity increased after drag were reduced, while holding vehicle weight constant, the new R-R portion could be calculated for the higher speed, and factored in to isolate the only Cd which could explain the speed increase.
-------------------------------------------------------------------------------------
It would get rid of all the differential calculus required to solve for two unknowns simultaneously.
Gravitational constants don't change.
Air density wouldn't change appreciably, if the aero mods could be added/subtracted in a short time period.
The car would still be 'warmed up', so non- thermally-equilibrated testing wouldn't rear it's ugly head.
Wind might not change appreciably in a short time domain.
--------------------------------------------------------------------------------------
I drove from Pest-Buda, north, into Slovakia and the Taurus Mountains once. I wasn't thinking about it then, but there might have been places which might have lent themselves to testing like this.
Austria, Slovenia, Romania, Serbia, etc., all might offer opportunities.
Just a thought.
Coastdowns remain a daunting task. One year, there were only eleven days where the weather was good enough to do any outdoor testing in all of England.
It's why we have wind tunnels and CFD.