Implemented FreeWheel

[AHaumer]

As mentioned in #3977:
I see that we have a rather sophisticated Mechanics.Rotational.Components.OneWayClutch, but no simple ideal freewheel (like the Electrical.Analog.Ideal.Diode). We could introduce such a component for usage e.g. in bicycle models.

[christiankral]

@AHaumer Please check the following changes and re-invite me for a review afterwards.

1. I added SI units the parameters in 8c6ba79
2. I updated the documentation of the example in 73078a6

[beutlich]

• The new model is missing in ModelicaTest.Rotational.AllComponents.
• The new example misses the required file Modelica/Resources/Reference/Modelica/Mechanics/Rotational/Examples/DemoFreeWheel/comparisonSignals.txt

[beutlich]

• The new model is missing in ModelicaTest.Rotational.AllComponents.
• The new example misses the required file Modelica/Resources/Reference/Modelica/Mechanics/Rotational/Examples/DemoFreeWheel/comparisonSignals.txt

[tobolar]

Is there any reason to not extend from Modelica.Mechanics.Rotational.Interfaces.PartialCompliantWithRelativeStates as e.g. Rotational.Components.Clutch does?

[tobolar]

Is there any reason to not extend from Modelica.Mechanics.Rotational.Interfaces.PartialCompliantWithRelativeStates as e.g. Rotational.Components.Clutch does?

With 7e6f667, the freewheel extends now from Mechanics.Rotational.Interfaces.PartialCompliant (without the state selection since it could be problematic). Thus, the sign convention of phi_rel (and w_rel) established for all rotational components is assured now. Consequently, the sign of w_rel and tau changes - compared to the original implementation. The example models work fine.

[christiankral]

@tobolar Would the following wording make it better on your opinion:

parameter Real(final unit="rad/N/m/s") residualAngularVelocity =1e-5 "Residual angular velocity coefficient (free = false)";
...
  w_rel = s*unitTorque         *(if free then 1 else residualAngularVelocity);

[tobolar]

Would the following wording make it better on your opinion:..

Sounds good. The following could be a reasonable implementation now:

parameter Real residualFriction=1e-5 "Residual friction coefficient (if free = true)";
parameter Real residualAngularVelocity=1e-5 "Residual angular velocity coefficient (if free = false)";
...
w_rel = s*unitAngularVelocity * (if free then 1 else residualAngularVelocity);
tau   = s*unitTorque          * (if free then residualFriction else 1);

The documentation shall be modified then:

free = false: flange_a is driving ...
...
tau = w_rel / residualAngularVelocity;

Regarding the two equations for tauin the documentation (where neither unitAngularVelocity nor unitTorque is used), the correct units of the two parameters shall be:

• residualFriction (unit="N.s")
• residualAngularVelocity (unit="1/N/s")

but it seems to be strange since one could, in general, expect both friction and a coefficient being unitless.

[christiankral]

w_rel = s*unitAngularVelocity * (if free then 1 else residualAngularVelocity);
tau   = s*unitTorque          * (if free then residualFriction else 1);

On my understanding the units of the proposed equations are not correct. They shall rather be:

parameter residualFriction (unit="N.m.s")=1e-5 "Residual friction coefficient (if free = true)";
parameter residualAngularVelocity (unit="1/N/m/s")=1e-5 "Residual angular velocity coefficient (if free = false)";
w_rel = s*unitTorque          * (if free then 1 else residualAngularVelocity);
tau   = s*unitAngularVelocity * (if free then residualFriction else 1);

Note, that I swaped unitTorque and unitAngularVelocity in the two equations.

[christiankral]

• residualFriction (unit="N.s")
• residualAngularVelocity (unit="1/N/s")

@tobolar As we deal with torques (not forces) the units shall rather be N.m.s and 1/N/m/s, isn't is?

[christiankral]

Alternatively we could consider the unit rad, but I am not sure if this makes sense.

parameter residualFriction (unit="N.m.s/rad")=1e-5 "Residual friction coefficient (if free = true)";
parameter residualAngularVelocity (unit="rad/N/m/s")=1e-5 "Residual angular velocity coefficient (if free = false)";

[tobolar]

@christiankral
IMO, one would expect regarding a mechanical component:

1. "friction coefficient" (mu overall in MSL) is generally a unitless quantity,
2. and so is probably also with "angular velocity coefficient".

This is why I tend to have no units. Optionally, we schould think again of changing "residualFriction" name and description. Not sure if this also holds for residualAngularVelocity.

[HansOlsson]

Note that unitless can mean two different things:

• A variable without a specified unit.
• A variable with unit="1", also known as dimensionless.

To me the friction coefficient should be of the second kind for translational friction, but I notice that often in MSL it is of the first kind. (And it is generally not possible to deduce its unit because it appears as mu*cgeo where the geometric coefficient also is without unit.)

For rotational friction it is not clear to me if mu or cgeo should handle the conversion between force and torque.

[tobolar]

@HansOlsson Need to open a new issue?
(See Sec. 2.3.3 Dimensions of quantities of the BIPM: The International System of Units – 9th edition (2019) - probably you know it already. )

[HansOlsson]

@HansOlsson Need to open a new issue?

Possibly. You decide.

[christiankral]

@HansOlsson Is using unitAngularVelocity and unitTorque still "state of the art" in how we handle the diode and freewheel implementation?

[christiankral]

The freewheel model and example look OK on my opinion.

I think the parameter names and units are fine. @HansOlsson do you agree?

[HansOlsson]

@HansOlsson Is using unitAngularVelocity and unitTorque still "state of the art" in how we handle the diode and freewheel implementation?

Yes. We are working on better solutions, but not yet.