paper: Drivetrain Acceleration Model [Archive]

Andrew Schreiber

19-09-2013, 00:42

Very cool. Now if only we could convince you to use variable names longer than 2 characters :p

Am I correct in assuming that Tm = Ts*(1-V/Vfree); on line 55 could also be used to figure out approximate current draw? Or am I over simplifying this?

Ether

19-09-2013, 01:19

Am I correct in assuming that Tm = Ts*(1-V/Vfree); on line 55 could also be used to figure out approximate current draw?

Yes. Kt, the motor torque constant, is Newton_meters per amp. So you can calculate the motor amps A=Tm/Kt*. For a CIM, Kt≈0.018 Nm/A. You could modify the C code to output amps so it could be plotted.

However, the model does not account for voltage drops due to heavy currents through the wires, the wire connection points, the battery' s internal resistance, changes in motor resistance due to temperature, etc so the accuracy of Tm (and the current) will be affected. Also, Kt is not strictly constant - it varies somewhat with motor temperature.

Very cool. Now if only we could convince you to use variable names longer than 2 characters :p

I've been programming since 1969 and I've never liked long variable names. To me it makes the code much less readable. I like the code to look like the way I would write the equations. I know opinions differ on this matter.

* when not slipping. when slipping, you'd have to use the kinetic friction force and back-calculate the motor torque.

Andrew Schreiber

19-09-2013, 01:47

Yes. Kt, the motor torque constant, is Newton_meters per amp. So you can calculate the motor amps A=Tm/Kt. For a CIM, Kt≈0.018 Nm/A. You could easily modify the C code to output amps so it could be plotted.

However, the model does not account for voltage drops due to heavy currents through the wires, the wire connection points, the battery' s internal resistance, changes in motor resistance due to temperature, etc so the accuracy of Tm (and the current) will be affected. Also, Kt is not strictly constant - it varies somewhat with motor temperature.

So, close enough for determining if it's bad or not but not close enough to be an "accurate model".

I've been programming since 1969 and I've never liked long variable names. To me it makes the code much less readable. I like the code to look like the way I would write the equations. I know opinions differ on this matter.

Twas just a joke, I know the equations look like this but I've always found annotating them with english names helped me understand what was going on. Course, that was the second thing I did with this (first was read what heun integration was). Thanks for sharing this.

Nemo

19-09-2013, 10:28

This is cool. I wouldn't have looked up Heun's method if I hadn't seen this thing. I also wouldn't have looked up rolling resistance. This white paper has given me some homework. Hopefully I can include those two things into the calculator I've been playing with. Thanks very much for posting goodies like this.

Ether

19-09-2013, 11:36

Nathan Streeter

19-09-2013, 12:49

Can't look at this in-depth right now... but thank you very much for another great resource! This looks quite useful... particularly if k1, k2, and k3 are given real, semi-accurate values (not just assumptions)!

Andrew Schreiber

19-09-2013, 13:09

I posted an update/correction to the C code:

- fixed an error in the English-to-SI conversion constants

- modified the accel function to better reflect the relationship between friction and rolling resistance

- added total drive motor amps to output

Dangit man, I just finished converting to coffeescript and started putting a front end on it! Mind posting your diffs?

http://bl.ocks.org/schreiaj/raw/6604718/7bbf83cd0195ef2dc3b0244c446a78a8c98eda1b/ It's still missing the axis so it's kinda useless to get numbers out of but I'll post the updated version tonight once I get the axis un-inverted.

Current coffeescript port available at https://gist.github.com/schreiaj/6604718/7bbf83cd0195ef2dc3b0244c446a78a8c98eda1b/

Ether

19-09-2013, 13:14

Mind posting your diffs?

There's a CSdiff report in the posted ZIP file that contains the revised C code :-)

Andrew Schreiber

19-09-2013, 13:29

There's a CSdiff report in the posted ZIP file that contains the revised C code :-)

Thanks, I'll get those integrated sometime in the near future.

Ether

19-09-2013, 13:46

Thanks, I'll get those integrated sometime in the near future.

I've got visitors coming over in 15 minutes.

Later this evening I'll strip out all the non-functional changes and give you a report of the minimum changes needed.

It's not a lot, once I get rid of the non-functional changes.

Ether

19-09-2013, 16:37

http://bl.ocks.org/schreiaj/raw/6604718/7bbf83cd0195ef2dc3b0244c446a78a8c98eda1b/... I'll post the updated version tonight...

Thanks for doing this. I'm hoping your user-friendly interface will generate more interest and technical discussion, and maybe even inspire teams to conduct tests and make their data available.

Andrew Schreiber

19-09-2013, 17:04

Thanks for doing this. I'm hoping your user-friendly interface will generate more interest and technical discussion, and maybe even inspire teams to conduct tests and make their data available.

Here's the updated version (haven't put your code changes in yet) but I did make it a little easier to read.

http://bl.ocks.org/schreiaj/raw/6604718/81de5905aa072e9666bee0c0e8f06fc207f7a970/

IKE

21-09-2013, 08:51

Thanks for putting this together.

I put together this "Drag Estimator" for my HS supermileage team years ago. It is based off of the Empirical rolling resistance and wind resistance test given in the Bosch automotive handbook.

http://www.epulaski.k12.in.us/tech/supermileage/design/design.htm

For rolling resistance, this gives a constant coefficient and a v^2 coefficient.

Ether

21-09-2013, 13:53

For rolling resistance, this gives a constant coefficient and a v^2 coefficient.

I'm glad you brought this up.

I structured the code to make it very simple to change the rolling resistance model if you want to. It takes only 2 keystrokes to change the rolling resistance model from
L = Kro+Krv*V; (line #68 in the 9/19/2013_1112a code)
to
L = Kro+Krv*V*V;

In fact, you could make the rolling resistance any function of V, or even use a look-up table if you have experimental data.

However, I think the V*V term only comes into play due to air resistance at speeds much higher than FRC robot speeds. It would be enlightening to see some actual data of rolling resistance vs speed for a typical FRC robot (with drivetrain mechanically disengaged from wheels) on FRC carpet surface.

SoftwareBug2.0

21-09-2013, 15:57

I structured the code to make it very simple to change the rolling resistance model if you want to. It takes only 2 keystrokes (^2) to change the rolling resistance model from
L = Kro+Krv*V; (line #68 in the 9/19/2013_1112a code)
to
L = Kro+Krv*V^2;

Um... "^" is xor in C.

Tom Line

21-09-2013, 17:41

Um... "^" is xor in C.

Someone slipped into excel mode in the middle of a line of C. :D

Ether

21-09-2013, 21:04

Someone slipped into excel mode in the middle of a line of C. :D

Actually...
Octave (http://www.angelfire.com/nd/edu/FRC/Octave.png), Maxima (http://www.angelfire.com/nd/edu/FRC/Maxima.png), SciLab (http://www.angelfire.com/nd/edu/FRC/SciLab.png), AWK (http://www.angelfire.com/nd/edu/FRC/AWK.png), and RLaB (http://www.angelfire.com/nd/edu/FRC/RLaB.png) :-)

Andrew Schreiber

21-09-2013, 23:38

Actually...
Octave (http://www.angelfire.com/nd/edu/FRC/Octave.png), Maxima (http://www.angelfire.com/nd/edu/FRC/Maxima.png), SciLab (http://www.angelfire.com/nd/edu/FRC/SciLab.png), AWK (http://www.angelfire.com/nd/edu/FRC/AWK.png), and RLaB (http://www.angelfire.com/nd/edu/FRC/RLaB.png) :-)

Which is still less weird than the weird case of coffeescript that I found while porting this by accident...

Turns out "variable *" compiles to "variable * variable" as opposed to throwing a compile error.

Ether

23-09-2013, 16:51

- added total drive motor amps to output

There's a small error in the build 9/19/2013_1112a code.

At line 75, "A*4/10" should be "A*n/10".

Andrew Schreiber

23-09-2013, 17:03

My current JS port has some minor typos that I haven't had the chance to hunt down yet. I'm working with Emscripten to compile the C code to Javascript which should mean I can integrate your changes faster too.

Ether

23-09-2013, 17:22

...should mean I can integrate your changes faster too.

That would be great.

I want to add battery voltage drop so mentors can use the model as a visual teaching tool to show students why low resistance in the motor wires and connections really does matter.

Ether

24-09-2013, 23:13

the model does not account for voltage drops due to heavy currents through the wires, the wire connection points, the battery' s internal resistance...

9/24/2013 added updated C code (http://www.chiefdelphi.com/media/papers/download/3771) (build 2013-09-24_2231) which includes motor voltage drops due to circuit resistance.

9/24/2013 added an explanation how the voltage drop model was derived (http://www.chiefdelphi.com/media/papers/download/3776).

Richard100

13-10-2013, 15:03

Ether

13-10-2013, 15:11

Ether - Very nice, thanks for developing this. Do you think it could be extended to include wheelbase parameters and predict motor current draw when turning (say differential steering)? We've had trouble in this area during drive system design.

Possibly. You can find several links here (http://www.chiefdelphi.com/forums/showthread.php?p=1295286#post1295286) for static analysis of turning force (which relates to motor current) for skid-steer vehicles.

cadandcookies

18-12-2013, 22:51

The executable works very well functionally-- I was able to run it with no problems and create some nice graphs based on it.

The only problem I've noticed is that I really have no idea what any of the variables are until I look back into the original program for comments.

I don't think I've quite expressed my appreciation for this program enough-- it is really nice to have such a good model for generating a good estimate of how fast a given drive train setup will move.

brennonbrimhall

02-01-2014, 15:44

Based on the discussion on this thread (http://www.chiefdelphi.com/forums/showthread.php?t=123111&page=2) (see posts #24 and #28), I've ported Ether's/Andrew Schreiber's C code to Java, and made a GUI front-end with charting. It's not the most pretty thing ever coded, but it may help to make the calculator more accessible by non-programmers. You can find the source code here (https://github.com/brennonbrimhall/DrivetrainCalculator), with .jar file releases here (https://github.com/brennonbrimhall/DrivetrainCalculator/releases).

I'd appreciate any feedback you might have.

Ether

17-09-2014, 09:20

You can find several links here (http://www.chiefdelphi.com/forums/showthread.php?p=1295286#post1295286) for static analysis of turning force (which relates to motor current) for skid-steer vehicles.

Due to changes made by the web host, many of those links are now broken.

So I collected everything together in one place and posted it here (http://www.chiefdelphi.com/media/papers/3022).

lemiant

04-10-2014, 15:21

Thank you so much for making this. You're my go-to resource for FRC physics and the community wouldn't be the same without you.
On line 84 voltage drop is calculated, but you don't ever seem to account for it when calculating the torque of the motors, is that correct?

Ether

04-10-2014, 16:19

On line 84 voltage drop is calculated, but you don't ever seem to account for it when calculating the torque of the motors, is that correct?

No. It is accounted for in the motor torque calculation. See the last paragraph in the voltage drop model rev E.pdf (http://www.chiefdelphi.com/media/papers/download/3780) document:

In the C code, there is an accel(V) function which computes the vehicle acceleration given its speed V. The vehicle speed V is directly related to the motor speed Wm through the wheels and the transmission, assuming no wheel slip. This motor speed Wm is then used in the formula Tm=Toffset-Tslope*Wm derived above to obtain the motor torque (see line 77 in the code). By using this formula for motor torque, the voltage drops due to current are automatically included in the calculation.

If you change the values of Rone and/or Rcom, you will definitely see the effect of changed voltage drop in the model's output.

Thanks for the words of encouragement.