Simscape mechanics: angular acceleration not working

[zoltrix]

hello
I am a simscape mulibody user
take a four bars mechanism, no external load
you apply a constant torque to the crank
I would expect a periodic angular acceleration but acceleration keeps growing
anybody here using simscape mechanics ?
what's wrong ?

 

[erobz]

hello
I am a simscape mulibody user
take a four bars mechanism, no external load
you apply a constant torque to the crank
I would expect a periodic angular acceleration but acceleration keeps growing
anybody here using simscape mechanics ?
what's wrong ?

I don't know anything about the program you are using, but do the linkage components have mass?

 

[zoltrix]

yes, I even tried with the simscape examples
same story

 

[Filip Larsen]

Not familar with Simscape either, but if there is no friction or load on the mechanism the expectation would be to see an increase angular velocity of the crank, which again would give an increased linear acceleration of the bars. So when you say you observe an increase in acceleration do you mean linear or angular (or both) and for what part of the mechanism exactly?

Also, not sure if its relevant, but is the other pivot in rocker configuration?

 

[zoltrix]

I mean the angular acceleration of the crank
the moment of inertia of the mechanism referred to the motor shaft is not a costant but it should be periodic in a closed mechanism
the same for the angular acceleration if the torque is constant (see attached file)

 

[Filip Larsen]

But if the links has to be linear accelerated back and forth and no energy is lost and you put more in with that torque the plot is precisely what to be expected. Note that average acceleration seem to be constant consistent with your constant torque.

 

[jrmichler]

It would be very helpful if you posted a sketch of the linkage. Also label the units of both axes of the output plot.

Your results could be the result of a simulation error. A constant torque into a four bar linkage should show a linear acceleration with sinusoidal variation superimposed. The linear acceleration is from a constant torque into a system with mass, zero friction, and zero energy removed from the system. The sinusoidal variation is due to the accelerations of the various links, and should be periodic at each revolution of the input shaft.

Your plot does not show that. The total acceleration is increasing exponentially, not linearly. You should be able to estimate the linear acceleration from the inertia of the input shaft and the applied torque, and compare to the output plot. Your plot also does not show any sinusoidal variation in angular acceleration. Instead, it has spikes with amplitude that vary from revolution to revolution. The likely cause of both of these is that the time steps are far too small.

Whenever doing any sort of dynamic simulation, always check your time steps. For a four bar linkage, 100 time steps per revolution might be excessive, and 10 steps per revolution will almost certainly be insufficient. This plot looks like the time steps were specified as the number of calculation steps per revolution of the input shaft. In that case, change "time steps" to "calculation interval" or the appropriate term for your software.

 

[zoltrix]

But if the links has to be linear accelerated back and forth and no energy is lost and you put more in with that torque the plot is precisely what to be expected. Note that average acceleration seem to be constant consistent with your constant torque.

Thanks Filip and jrmichler

I suppose that the total acceleration should be periodic at each revolution since the moment of inertia at the shaft should depend only on the current geometry of the linkage
So I dont expect variable spikes whose amplitude not only vary but even increase from revolution to revolution

as I said I used the four bars Simscape multibody sample supplied with the software with standard settings (time steps etc )
the only modification I did was to apply a constant torque and to use the "transform sensing" block to sense the angular acceleration of the shaft

 

[berkeman]

as I said I used the four bars Simscape multibody sample supplied with the software with standard settings (time steps etc )

Does Simscape have an online forum? If so, this "example" should have triggered other discussions there if it has issues...

 

[Tom.G]

Constant torque, fixed mass, constant friction = Constant acceleration

Think of pushing a gallon jug of water across the floor. Normally you would push it harder to get it moving and then reduce the pushing force to what is required to overcome friction. If you maintain the initial pushing force, the jug will keep going faster and faster, it will keep accelerating.

Equivalently, think of opening a very heavy door (without a door-closer on it). The same thing happens. If you continue pushing it as hard as you pushed to get it moving, it his going to hit that door-stop very hard!

That is what is happening in your simulation.

Cheers,
Tom

 

[Baluncore]

... the same for the angular acceleration if the torque is constant (see attached file)

Without scales, it is meaningless. That plot may show an accumulated floating point round-off error, being auto-zeroed by an integrator, plotted against one to 500 turns of the crank.

 

[zoltrix]

Constant torque, fixed mass, constant friction = Constant acceleration

Think of pushing a gallon jug of water across the floor. Normally you would push it harder to get it moving and then reduce the pushing force to what is required to overcome friction. If you maintain the initial pushing force, the jug will keep going faster and faster, it will keep accelerating.

Equivalently, think of opening a very heavy door (without a door-closer on it). The same thing happens. If you continue pushing it as hard as you pushed to get it moving, it his going to hit that door-stop very hard!

That is what is happening in your simulation.

Cheers,
Tom

it seems to me that you are speaking of velocity not of acceleration

 

[zoltrix]

Without scales, it is meaningless. That plot may show an accumulated floating point round-off error, being auto-zeroed by an integrator, plotted against one to 500 turns of the crank.

if I click on "angular acceleration" of the "transform sensing" block I expect the current angular acceleration to be plotted i.e : instantaneous accel vs time
it should be a periodic curve, IMHO

 

[Baluncore]

if I click on "angular acceleration" of the "transform sensing" block I expect the current angular acceleration to be plotted i.e : instantaneous accel vs time
it should be a periodic curve, IMHO

The plot is meaningless without a scale and units.

 

[zoltrix]

The plot is meaningless without a scale and units.

the "scope" block of simscape never quote the scale and units , you must set them elsewhere
in any case a continuosly going up angular acceleration with a constant torque is meaningless,in my opinion,, regardless of units and scale

 

[Baluncore]

in any case a continuosly going up angular acceleration with a constant torque is meaningless,in my opinion,, regardless of units and scale

How can you possibly know that it is "continuously going up angular acceleration"? That may be real acceleration data, or it may be floating point round-off noise showing up when a constant angular acceleration is being plotted with auto-scaling.
If you gave it a vertical scale, we would have some idea of what was being plotted.

 

[zoltrix]

I attach the scaled plot c/w with units
the trend is the same as the previous one...of course

 

[erobz]

You still never shared the linkage with us that @jrmichler asked of you. If we had it, we could do some actual physics/mathematics to examine it.

If $t=0$, and the torque $T$ is constant, $\alpha$ should be some non-zero initial value - your plot shows $0 [1/s^2]$@ $t= 0$ ? Also, this whole plot occurs over 1 second...

 

[zoltrix]

Does Simscape have an online forum? If so, this "example" should have triggered other discussions there if it has issues...

they have ... but...uhm...Simscape is a great software , in my opinion, but support is not the best, to the least...
however I agree with you, it is a so common and so important feature that I dont think it is a bug...
probably I am doing or expecting something wrong, but what ?
I supposed that on mechanical engineering forum many people were using Simscape on regular basis
I wonder if someone came across the same issue and how did it fix it.

 

[zoltrix]

You still never shared the linkage with us that @jrmichler asked of you. If we had it, we could do some actual physics/mathematics to examine it.

If $t=0$, and the torque $T$ is constant, $\alpha$ should be some non-zero initial value - your plot shows $0 [1/s^2]$@ $t= 0$ ? Also, this whole plot occurs over 1 second...

yes I was expecting some non zero initial value
it is what I have been repeating and repeating...
i did not share the linkage ?
come on , it is just a four bar linkages ....never seen it ?
I am not speaking of some esoteric device

 

[Filip Larsen]

Looks like numerical instability as jrmichler mentioned.

I assume the model does not include elastic dynamics (either distributed on the links or as explicit linear/torsional springs), but if it does include such elements then the fast oscillations could also be a result of (possibly unstable integrated) mechanical elasticity.

 

[erobz]

yes I was expecting some non zero initial value
it was I have been repeating and repeating...

Your plot starts at zero, and its over a 1 second.

I would advise against being nasty, you've only just shared the units with us right now! How the hechk do we know you didn't bungle that?

come on , it is just a four bar linkages ....never seen it ?
I am not speaking of some esoteric device

Hey, buddy. I meant give us the diagram with masses, lengths, etc... so we can crunch some numbers. But now since your acting like a spoiled brat you can ... read between the lines. Bye!

 

[jrmichler]

Thread locked while everybody cools down. To the OP: You were requested to provide certain information, much of which you have not yet provided. Please read this entire thread very carefully, taking your time doing so. Then PM me when you are ready to fully answer every question that has been posted, after which I will reopen the thread. Take your time because I will not be back until this evening.

Hint: When running a simulation with a torque or force input, and nothing that dissipates the input power, shorten the simulation to show only the first five or so cycles or revolutions.

 

[jrmichler]

After a PM discussion with the OP, the thread will stay closed.