[brunomramoa]

Hello to all,


I am trying to create a bouncing ball simulation with solids4foam. Something similar to this https://www.youtube.com/watch?v=WwZj8JXrLIo.




I started to assemble this case from the impact bar tutorial but now I am having some trouble with the initialization of the case.


I used the setFields to set the ball velocity to a very small value (1e-6) but it will not gain velocity due to the acceleration of gravity.


Does anyone have experience with this?


link for the case: https://drive.google.com/file/d/1Agw...ew?usp=sharing


Additionally, is it possible to get the kinetic and internal energy from solids4Foam?

[bigphil]

Originally Posted by brunomramoa

Hello to all,


I am trying to create a bouncing ball simulation with solids4foam. Something similar to this https://www.youtube.com/watch?v=WwZj8JXrLIo.




I started to assemble this case from the impact bar tutorial but now I am having some trouble with the initialization of the case.


I used the setFields to set the ball velocity to a very small value (1e-6) but it will not gain velocity due to the acceleration of gravity.


Does anyone have experience with this?


link for the case: https://drive.google.com/file/d/1Agw...ew?usp=sharing


Additionally, is it possible to get the kinetic and internal energy from solids4Foam?

When there is rigid body motion, I have found that using equation relaxation is a bad idea as it essentially eradicates rigid motion. Try the following relaxation settings in system/fvSolution:

relaxationFactors
{
    // Under-relaxatr ing the D equation by 0.99-0.9999 can improve convergence in                                                                                                                                              
    // some cases, in particular when there are solidContact boundaries                                                                                                                                                         
    equations
    {
        //DD           0.9999;                                                                                                                                                                                                  
    }

    // Under-relaxating the D field by 0.1-0.9 can improve convergence in some                                                                                                                                                  
    // cases                                                                                                                                                                                                                    
    fields
    {
        DD              0.7;
    }
}

Also for testing you might find it easier to start with a linear geometry solid model (e.g. linearGeometry or linearGeometryTotalDisplacement): they will have better convergence and allow you to learn the "good" settings for your case before changing to non-linear geometry.

Also for rigid motion, the solver struggles to define a reasonable normalisation factor for calculating the residuals so sometimes a time-step will go to the max correctors but if you look at the solution for different correctors it is not changing (or changing by a very small amount).

Finally, I am not sure how important the solution tolerance is in these cases but it is something to be aware of.

EDIT: Finally finally, you might need to use a finer plate mesh near the contact region as the contact quick-check algorithm can miss contact when the master and slave faces are very different in size; also a smaller time-step may help.

Philip

[brunomramoa]

Originally Posted by bigphil

When there is rigid body motion, I have found that using equation relaxation is a bad idea as it essentially eradicates rigid motion. Try the following relaxation settings in system/fvSolution:

relaxationFactors
{
    // Under-relaxatr ing the D equation by 0.99-0.9999 can improve convergence in                                                                                                                                              
    // some cases, in particular when there are solidContact boundaries                                                                                                                                                         
    equations
    {
        //DD           0.9999;                                                                                                                                                                                                  
    }

    // Under-relaxating the D field by 0.1-0.9 can improve convergence in some                                                                                                                                                  
    // cases                                                                                                                                                                                                                    
    fields
    {
        DD              0.7;
    }
}

Also for testing you might find it easier to start with a linear geometry solid model (e.g. linearGeometry or linearGeometryTotalDisplacement): they will have better convergence and allow you to learn the "good" settings for your case before changing to non-linear geometry.

Also for rigid motion, the solver struggles to define a reasonable normalisation factor for calculating the residuals so sometimes a time-step will go to the max correctors but if you look at the solution for different correctors it is not changing (or changing by a very small amount).

Finally, I am not sure how important the solution tolerance is in these cases but it is something to be aware of.

EDIT: Finally finally, you might need to use a finer plate mesh near the contact region as the contact quick-check algorithm can miss contact when the master and slave faces are very different in size; also a smaller time-step may help.

Philip

What about the latter part? Is there any function to keep track of the kinetic and potential energy in the system?


Best Regards!

[bigphil]

I have just pushed a commit (55bd6bb2) to the development branch which adds a new functionObject called solidKineticEnergy and the tutorial tutorials/solids/linearElasticity/plateHole/bouncingPlateHole.

You can see from the code that it is relatively straight-forward to do. I aim to add some other energy function objects when I have time (but you might want to try it yourself).

Philip

[bigphil]

I pushed the solidPotentialEnergy function object to the development branch (20e0f235), to be merged with the master in due course.

This potential energy is defined a rho*g*h as opposed to internal strain energy (stress && strain), though I will aim to add that too when I have time.

Philip

[Hylandjo]

I'm moreover looking for a reply, because i am new here.

[Hylandjo]

I also like to gain experience in this



BallSportsGames