Algodoo simulation software

neufneufneuf

Hi,

I tested an example with Algodoo and I don't don't know if there is an error in the software but a curved object under gravity with balls inside move alone. Someone know this free software ?

http://www.algodoo.com/download/

If I take straight shape the object don't move alone (only vibration around a point). The force is not very big but enough to move it alone in one meter in 5 minutes. If I change mass of ball the speed is greater.

I attached the file with Algodoo scene in zip file.

The image give the scene I tested. You can reverse the object it move in the other direction.

dlouth15

That is interesting. It does appear to move. I'm not sure why that is but I think it might be numerical errors in the software. The balls don't seem to settle down properly in the area near the big disk.

In the options there's a setting called "simulation rate". The higher the rate the "finer" the simulation. In the real world the setting would be infinity but a computer simulation must break time down into discrete divisions. Experimenting with this I found that the balls settle down when the simulation rate is set very high.

If you set it low, however, the balls jiggle about and jump out of the container. This is not what you observe in the real world and therefore I think it is numerical errors causing the jiggling and large scale movement of the object.

There's forum on the Algodoo website where you could discuss this. Thanks for pointing out this software.

neufneufneuf

With 1200 Hz the object continue to move (more slowly, so it could say it is an error) but if you take straight line (not curved shape) or curved shapes in mirror axis the object don't move, with 1200 Hz it is 0 without vibration.

dlouth15

It is the jiggling about of the balls at low simulation rate that makes me suspicious. This doesn't happen in the real world. If you lower the simulation rate sufficiently the balls jump out of the container, but they don't do this if the simulation rate is high and they don't do this in the real world.

neufneufneuf

I attached another case where position is always 0 with 1200 Hz (start17). From 500 Hz the position is fixed.



Even I put friction to infinite and restitution at 0, the object continue to move (start19).

Sure, in the real world no vibration. But what's happens with gas with temperature ?

Morbert

This software looks interesting.

Does the software report the energy of the system? Are inelastic collisions permitted in your simulation?

dlouth15

You can specify a coefficient of restitution for any of the objects.

Morbert

dlouth15 wrote: »

You can specify a coefficient of restitution for any of the objects.

I'll give it a try when I get home. As the rate of the simulation is increased, does the centre of mass of the system approach/converge to having 0 velocity? If it does, that would imply numerical error.

dlouth15

Morbert wrote: »

I'll give it a try when I get home. As the rate of the simulation is increased, does the centre of mass of the system approach/converge to having 0 velocity? If it does, that would imply numerical error.

Yes, the velocity of the system as a whole converges to zero. I think the reason the simulation rate needs to be high is that you have very wide variations in forces. A small overshoot by one of the balls might bring it to overlap with another ball or one of the walls. It will then be subject to a very
large force to expel it.

With a coarse simulation the tendency to overshoot will increase and you will get greater repulsion. The system will be continually overreacting leading to instability.

With a fine simulation, the ball will never get far into another ball so these forces will not be so great so less instability.

What I'm not sure of is why there's an overall tendency for the system as a whole to move in a particular direction due to these effects. It might be that this overshoot is more likely to occur on the inside of the curved surface rather than the outside of it.

neufneufneuf

I had try another scene with small number of balls for watch if the result is changed but it's always in the same direction. Maybe a problem with sin/cos approx ? I gave attraction for balls for have "pressure".

dlouth15

What I would suggest you do is plot the velocity of the system as a whole against the simulation rate to see how the velocity is affected by the simulation rate. Pick about 4 or 5 simulation rates. For each simulation rate, let the simulation settle down and then note down the velocity at that rate. Then plot the results.

You should also, I suggest, discuss it in the forum for the software. Personally I think you're looking at a limitation that might exist in any simulation. If you set the simulation rate too low you can get non-physical results.

Why did you give the balls an attractive force with each other? What substance is this supposed to be modelling?

neufneufneuf

I added an attractive force for balls because the speed is constant, not a vibration like before, you can test for watch the difference.

I done your tests:

f1200 => 0.00023 m/s
f1000 => 0.00040 m/s
f800 => 0.0006 m/s
f600 => 0.00083 m/s
f400 => 0.00114 m/s
f200 => 0.0026 m/s

graphics in the image

neufneufneuf

I built another cases:

open_model: the velocity is around 0.025 m/s at f=1200Hz, the velocity change with frequency too.

closed_model: the velocity is 0

it's strange that in one case the software can find exactly 0 and other case not.

I added a file "2balls" for isolate the problem. With 2 balls and 1 object.

I found something, when frequency is at 1200 Hz, the altitude (y position) is constant but when the frequency is lower the altitude vibrates maybe the error is relative at this vibration.

dlouth15

neufneufneuf wrote: »

I done your tests:

f1200 => 0.00023 m/s
f1000 => 0.00040 m/s
f800 => 0.0006 m/s
f600 => 0.00083 m/s
f400 => 0.00114 m/s
f200 => 0.0026 m/s

graphics in the image

Would you agree that it looks like the velocity is converging to zero as the simulation rate increases? What conclusions might you draw from this?

neufneufneuf

Yes, it seems to converge even it's not a straight line and it's not possible to increase frequency more than 1200 Hz. The problem coming when forces are in vibration and ONLY when friction is not 0, when friction of object is 0, all forces are stable and the system don't move (or don't accelerate because friction is at 0). But I can find some cases where frequency don't change the velocity, it's difficult to understand where is the problem in software.

I put a scene where frequency don't change the speed. If you ask for look forces, the object go in the other direction !

neufneufneuf

I had try to give all forces with 2 balls and one object. I have a question about my image, (-Fa, Bn) don't give a torque on ball2 ? Maybe this give a vibration in software and the error come from this. Balls are turning in the simulation. Maybe for that the friction is important. If sofware compute attraction with a matrix for the object, ball with slope contact is attract with asymmetric forces and maybe this give a torque on this ball. The ball turn and with friction this move object.

I gave forces from software and algodoo scene. I don't understand T forces at bottom, these forces come from friction ? And the sum seems to be different of 0.

Scene "2balls v3 nb" don't change speed with frequency. It's possible to see a rotation for each ball, but they don't turn at the same rotational speed, maybe the error come from this.

dlouth15

I have it moving with 1 ball:

neufneufneuf

True, because the ball turn and friction is not to 0. It can be possible because there is solid object/ball. Fixed object (not ball) on your scene and see like ball turn around itself even there friction (put friction different to 0). The angular velocity is fixed from 500 to 1200 Hz.

I have a question about the image attached. Gravity attract from each point to each point. Ball with slope seems to have a torque for me.

dlouth15

neufneufneuf wrote: »

I have a question about the image attached. Gravity attract from each point to each point. Ball with slope seems to have a torque for me.

In the Algodoo software?

neufneufneuf

No, in general physics. Look at black forces, they are more canceled by wall of object compared than red forces.

Have you look at your Algodoo scene with fixed object ? the balls rotate all the time with constant speed.

srsly78

OP the cause is tiny errors in floating point maths being compounded as time goes along.

neufneufneuf

@srsly78: with 2 objects it's possible to have rotationnal speed at 1e-5 rd/s so why in this model ball turn at 0.005 rd/s ?

srsly78

Because it's bad software and not using enough precision apparently.

neufneufneuf

could you explain forces in image (forces2.png) ?

dlouth15

neufneufneuf wrote: »

No, in general physics. Look at black forces, they are more canceled by wall of object compared than red forces.

No, I would not expect any torque on either of the balls in the real world.

Have you look at your Algodoo scene with fixed object ? the balls rotate all the time with constant speed.

The balls should not rotate with either the fixed or non-fixed models in the real world. My own opinion is that these are errors in the simulation, not because there's anything wrong with the software itself but that the situations are stretching it beyond the sorts of things it is designed for. But I would have a word on the forums for the software. Maybe some of the developers might comment.

neufneufneuf

I wrote a message in the Algodoo's forum, thanks for the idea, maybe it's a known error.

In general physics, look at the image please, the sum of black forces is G, but these forces don't give the same torque with the force N. The force N can't move, the contact is fixed. So, how all torques can give 0 ?

neufneufneuf

I understood, it's because "mg" and "N" are not exactly parallel, look at image. I built another scene (different shapes) and the error is repeat again. And like the attraction is in law 1/d² the attraction in half disk is different than other.

I had message from forum but it's not logical for me...I put a new scene where a disk accelerate, it's a big error in the software !

http://www.algodoo.com/forum/viewtopic.php?f=18&t=10523

Edit: Maybe the error come from the method for calculate the torque. Maybe Algodoo compute torque for each part and AFTER do the sum. But this don't explain the case of a disk full with small balls.

dlouth15

neufneufneuf wrote: »

I understood, it's because "mg" and "N" are not exactly parallel, look at image. I built another scene (different shapes) and the error is repeat again. And like the attraction is in law 1/d² the attraction in half disk is different than other.

I had message from forum but it's not logical for me...I put a new scene where a disk accelerate, it's a big error in the software !

Yes, the responses so far don't make a huge amount of sense. It looks to me that the people responding in that forum don't understand the physics. There should be no perpetual rotation in any of these models in the real world.

neufneufneuf

I drawn forces in a real case, look at image. N1 and N2 are normal forces, f1 and f2 are friction forces, this give 2 black forces. I don't know how to built the torque because it's not only center of mass but friction too.

first image, the torque is not good
second image is good ?

dlouth15

neufneufneuf wrote: »

I drawn forces in a real case, look at image. N1 and N2 are normal forces, f1 and f2 are friction forces, this give 2 black forces. I don't know how to built the torque because it's not only center of mass but friction too.

first image, the torque is not good
second image is good ?

On the second diagram, what is the pink vector pointing upwards from the lower plane?

neufneufneuf

It's an error. But I find something interesting in Algodoo, there are 2 forces T in the scene, I think the software give one T (motor I think) in one direction and other T in other direction. Maybe the error come from this.

I give bnj2 scene, the only parameter that change T is the mass of curved object and the attraction. It's possible to see it. But this don't change the speed, it's very strange an error must increase with force.

And if you change the altitude of the object, it's possible to see modification of T, pass from a maximum. And if friction is cutting, the speed increase more and more.

I added bnj3, it uses only straight line, it's the same, one moment I thought the error come from the circle. The scene is built again easily, and the same result all the time, so it's not a manipulation, it's really an error. The force is very small.

neufneufneuf

I studied with one point for attraction for isolate the error. I drawn all forces, I must be wrong too because I see a torque. I gave 2 possible construction but it's not the same torque...I added point3.png where no torque. How physics can do for choose the good one ? Is friction change something ?

dlouth15

neufneufneuf wrote: »

I studied with one point for attraction for isolate the error. I drawn all forces, I must be wrong too because I see a torque. I gave 2 possible construction but it's not the same torque...I added point3.png where no torque. How physics can do for choose the good one ? Is friction change something ?

It is getting a bit confusing at this point. What I would suggest therefore is to start again with the simplest system you can think of to demonstrate the idea. I think you are going in this direction already. It is much easier to analyse 2 or 3 objects than a big pile of them.

Label everything.

So for example if the objects have mass, specify what that mass is e.g. 2 Kg, 3Kg. If there is gravity, indicate it and its direction. If there are attractive forces note them and any surfaces that have friction.

I think it is best to forget the Algodoo software for the moment as it appears to have limitations.

Don't draw in force vectors just yet but say what you think should happen to the objects themselves. Do you think they should move? Do you think they should rotate etc.

Stick with the one model until it is fully understood before moving to the next one.

Do this on a new thread. Then we can try and figure out what the physics says.

neufneufneuf

OK, I will start a new thread and explain more my images. I drawn a figure where if disk turn on left wall before meet the up wall, if there is friction the up wall give less force and the sum of forces is a torque on disk. Maybe it's the error Algodoo done.

dlouth15

neufneufneuf wrote: »

OK, I will start a new thread and explain more my images. I drawn a figure where if disk turn on left wall before meet the up wall, if there is friction the up wall give less force and the sum of forces is a torque on disk. Maybe it's the error Algodoo done.

I would ignore Algodoo for the moment. The problem I have with that diagram is that I assume N1 is a the normal force from one of the walls but it is drawn incorrectly. I will explain on the new thread. Leave out the forces initially on the new thread as we will draw them later.

neufneufneuf

I give new scene where energy reach 10000 J in 10 minutes. It's strange that the software give a so big error ! It's very linear.

neufneufneuf

It's very interesting because if rectangle attract disk: no rotation. But if disk attrack rectangle there is rotation ! I think the "add center axle" is with holes, and this is matter except where there is holes, this is the asymmetric function. Maybe it's for that the rotationnal speed is like a see-saw.