Issues using Cyclic Symmetry

[jousefm]

I will have a look at your project later on Dale, hopefully I can find the culprit in the setup Maybe @ggiraldo has some additional tips and hints for you.

Best,

Jousef

[DaleKramer]

Thanks Jousef,

In the meantime I tried a 3rd sim run but still error…

Here was my logic for Run 3:

1. First, I am trying to find a sim setup that can be a validation of the theoretical final size that the plastic coin will get pressed into (simply to the diameter that my CAD program says is the final diameter of a cylinder of equal volume as the un-pressed coin but with a thickness of the final distance between the dies two contact surfaces after the full ram movement at iteration time 1.0s).
2. To achieve #1 in the sim setup, I decided to initially try to apply Ram Movement to the actual face that makes physical contact with the plastic coin, my brain said that this would likely be the best way to ensure that there would be no compression (or uneven compression) of the upper die to take into account. In a similar thought process I decided to support this whole press setup on the physical contact face of the lower die.

My run 3, simply changed the Ram movement face and the lower die support face to the upper face of the ram die and lower face of the lower die. (which would seem to add some die compression possibilities and make it harder to use as a validation setup)

But alas, no joy yet

[ggiraldo]

Hi @DaleKramer,

I took a look at your project, this is what I found:

1. Why are the three pieces not spanning the same arc? The middle body is wider. This could be a problem.
2. The elastic support parameter might be too low. Also, it is not constraining in the Z direction, which I think is the important one because X and Y are already constrained by the cyclic symmetry.
3. Again with the elastic support, if it doesn’t converge, maybe try the dynamic solver.

Please tell me if this helps or further help is needed.

Goof luck!

[DaleKramer]

@ggiraldo

Why are the three pieces not spanning the same arc? The middle body is wider. This could be a problem.

My CAD files have them all as 1 degree sectors. This might be a geometry display ‘bug’ I have started a topic on …

@ggiraldo

The elastic support parameter might be too low.

Just my attempt to lower it from Richard suggestion in the other topic …

@ggiraldo

Also, it is not constraining in the Z direction, which I think is the important one because X and Y are already constrained by the cyclic symmetry

Ah, I thought the Z was restrained by the Physical contact faces ???

But I never considered that the cyclic symmetry could be the X/Y constraint, yes off course though after I think about that. This could by it

@ggiraldo

Again with the elastic support, if it doesn’t converge, maybe try the dynamic solver.

Will do but will try removing elastic support all together first…

EDIT: Run 4 still not constrained, darn…

I love how few core hours I am using to zero in on this though …

[ggiraldo]

@DaleKramer

Ah, I thought the Z was restrained by the Physical contact faces ???

The problem is that at the very first solution step, contacts are not activated and then the body is not constrained, this is why we use the elastic constraint. It is numerical stuff and pointless to get into that details now.

Also wanted to point out, what are your plans for the material model? I saw you have a bi-linear plastic, but I think a hyperelastic model should be better suited.

Edit. Don’t worry about core hours, failed simulations do not count.

[DaleKramer]

@ggiraldo

The problem is that at the very first solution step, contacts are not activated and then the body is not constrained, this is why we use the elastic constraint. It is numerical stuff and pointless to get into that details now.

Aha, now I sorta understand Richards suggestion in first post, my brain was already thinking it was constrained by ‘Physical contact’, silly me

@ggiraldo

Also wanted to point out, what are your plans for the material model? I saw you have a bi-linear plastic, but I think a hyperelastic model should be better suited.

I was planning on correcting the plastic model and stress/strain curve once I got something working, I will definitely investigate hyperelastic model, much thanks on that

EDIT: Run 5 still no joy , unconstrained error by doing this:

[cjquijano]

Hi @DaleKramer,

I just took a look at run 5.

Your 1 Degree sector is causing really badly shaped elements in the Coin near the center. Maybe if you did 5 degrees the aspect ratio will not be so bad.

You Fixed the bottom of the coin in Z so this totally eliminates the bottom contact . I would remove this and go back to the elastic support.

I don’t think the top of the press is constrained on the initial calculation so I would try a Elastic Support on the top surface in the Z direction.

Let me know if that helps,
Christopher

[DaleKramer]

@cjquijano

Your 1 Degree sector is causing really badly shaped elements in the Coin near the center. Maybe if you did 5 degrees the aspect ratio will not be so bad.

I will try that but I am just assuming that long thin triangular faces would not be such a problem with FEA, is this face that tapers to the cyclic rotation center line the badly shaped element you see :

@cjquijano

You Fixed the bottom of the coin in Z so this totally eliminates the bottom contact . I would remove this and go back to the elastic support.

The about statement seems to me to conflict with this:

@ggiraldo

The elastic support parameter might be too low. Also, it is not constraining in the Z direction, which I think is the important one because X and Y are already constrained by the cyclic symmetry.

I am confused about the confliction of the above two statements, anyone care to straighten me out

@cjquijano

I don’t think the top of the press is constrained on the initial calculation so I would try a Elastic Support on the top surface in the Z direction.

Doesn’t the Ram movement constrain x/y and force movement in the z direction only

@cjquijano Thanks for input here

[DaleKramer]

@ggiraldo

Again with the elastic support, if it doesn’t converge, maybe try the dynamic solver.

Where do I find a Dynamic solver or is it one of these :

[cjquijano]

See my comments below.

@DaleKramer

@cjquijano

Your 1 Degree sector is causing really badly shaped elements in the Coin near the center. Maybe if you did 5 degrees the aspect ratio will not be so bad.

I will try that but I am just assuming that long thin triangular faces would not be such a problem with FEA, is this face that tapers to the cyclic rotation center line the badly shaped element you see :

• The elements are 4 noded Tetrahedral elements. Looking at the triangle on the surface it looks good but the third dimension is very short. Ideally, you want an aspect ratio one to one. Meaning the ration between the longest edge and the shortest edge is 1. In areas of interest I like to have at least a 5:1 ratio and in other areas it could be up to 20:1. Pretty much all of your disk is critical and the first element has an aspect ratio of roughly 50:1.

@cjquijano

You Fixed the bottom of the coin in Z so this totally eliminates the bottom contact . I would remove this and go back to the elastic support.

• I think @ggiraldo and I are still agreeing. Think of it this way. Before any loads or physical contacts are applied you want your model to be stable. The bottom portion of your press is Fixed so it’s not going to move on its own. The coin has two sides that are constrained so it is only free to move in Z. Adding the Elastic Support in Z will keep the coin stable until the physical contacts start to engage. The top part of your press is the same as the coin. Because of the two constraints it can only move in Z. Adding the elastic support will hold it into position for the initial calculation. The top elastic support is probably not needed is all scenarios but it can help rule out some issues.

• In your model you have the bottom press part fixed, contact defined between the coin and the bottom press, and then you have the bottom of the coin fixed in Z. In this setup you are never engaging the physical contact or the entire bottom part.

• The spring support is also spring to ground connection. You want the springs to be strong enough to hold the part but not strong enough to affect the stiffness. If the spring is not strong enough, when you apply a force, the coin could move too far into the bottom press that the software can not figure out the contact. If the spring is too strong it will affect your results. Since you not applying a force but a displacement this is not as big as an issue.

The about statement seems to me to conflict with this:

@ggiraldo

The elastic support parameter might be too low. Also, it is not constraining in the Z direction, which I think is the important one because X and Y are already constrained by the cyclic symmetry.

I am confused about the confliction of the above two statements, anyone care to straighten me out

@cjquijano

I don’t think the top of the press is constrained on the initial calculation so I would try a Elastic Support on the top surface in the Z direction.

Doesn’t the Ram movement constrain x/y and force movement in the z direction only

@cjquijano Thanks for input here

[cjquijano]

@DaleKramer

Dynamic is the solution Type.

[DaleKramer]

Ah, a dynamic analysis, not a dynamic solver … UGH another Analysis type to learn

[DaleKramer]

Excellent reply to some of my questions!

@cjquijano

The elements are 4 noded Tetrahedral elements. Looking at the triangle on the surface it looks good but the third dimension is very short. Ideally, you want an aspect ratio one to one. Meaning the ration between the longest edge and the shortest edge is 1. In areas of interest I like to have at least a 5:1 ratio and in other areas it could be up to 20:1. Pretty much all of your disk is critical and the first element has an aspect ratio of roughly 50:1.

So if 1 degree slice gives 50/1, and with this center slice in our ‘area of interest’, where we want 5/1, wouldn’t I try ~10 degree slices

@cjquijano

• I think @ggiraldo and I are still agreeing. Think of it this way. Before any loads or physical contacts are applied you want your model to be stable. The bottom portion of your press is Fixed so it’s not going to move on its own. The coin has two sides that are constrained so it is only free to move in Z. Adding the Elastic Support in Z will keep the coin stable until the physical contacts start to engage. The top part of your press is the same as the coin. Because of the two constraints it can only move in Z. Adding the elastic support will hold it into position for the initial calculation. The top elastic support is probably not needed is all scenarios but it can help rule out some issues.
• In your model you have the bottom press part fixed, contact defined between the coin and the bottom press, and then you have the bottom of the coin fixed in Z. In this setup you are never engaging the physical contact or the entire bottom part.
• The spring support is also spring to ground connection. You want the springs to be strong enough to hold the part but not strong enough to affect the stiffness. If the spring is not strong enough, when you apply a force, the coin could move too far into the bottom press that the software can not figure out the contact. If the spring is too strong it will affect your results. Since you not applying a force but a displacement this is not as big as an issue.

Ah, I am beginning to ‘see’ all this in my brain, watch out for fallout

So, perhaps I would have seen this earlier had I simply placed the 3 solids 1 mm apart from each other in the CAD file…

In all cases, ‘Physical contacts’ do not engage until some movement of objects has occurred or pressure applied, even if the part faces are coincidental to begin with …

Before that contact happens (that is infinitesimally small movement with respect to coincidental faces to begin with), all parts must be fully constrained except in the direction we want them to move …

So, the simulation starts to move the ram (or apply pressure), at some point contact is made with the coin, then those 2 parts move together until the coin hits the lower die where ‘contact’ is made and the coin should start expanding until the limit of ram movement.

Does my brain now ‘see’ this correctly

If so I think my next attempt may work better

Footnote: Is this elastic spring rate for an ideal spring of infinite length and the geometry parts are assumed to have no mass

Do the elastic springs ever impart loads to the face they are attached to that would affect say, how big a diameter the coin gets pressed to

P.S. Stll unanswered:

@DaleKramer

Doesn’t the Ram movement constrain x/y and force movement in the z direction only

The Ram movement is dx =0 dy = 0 and dz is a t dependent formula…

[cjquijano]

So if 1 degree slice gives 50/1, and with this center slice in our ‘area of interest’, where we want 5/1, wouldn’t I try ~10 degree slices

It also depends on the size of your elements as well. Personally, I would do about 10 degrees and it should still be much quicker than the full model.

Does my brain now ‘see’ this correctly

At a high level I think you have it.

Footnote: Is this elastic spring rate for an ideal spring of infinite length and the geometry parts are assumed to have no mass

Correct Massless.

Do the elastic springs ever impart loads to the face they are attached to that would affect say, how big a diameter the coin gets pressed to

Yes, the springs can affect the results. that is why you want the smallest value that still stabilizes the parts.

Doesn’t the Ram movement constrain x/y and force movement in the z direction only

The applied displacement does constrain the Ram in the Z direction.

[DaleKramer]

@cjquijano

So, in Run 6, I have tried adding the elastic support on the top and bottom round faces of the coin at spring rates of x0 y0 and z1000 N/m. This still results in unconstrained error

Then, I imported a new CAD geometry of a 10 degree sector and meshed it under this simulation, there are potential issues with the mesh as described in this topic…

I ran the Run 1 in that sim with the mesh of concern, but had to cancel the run after ~25 minutes at 4%

Have I constrained the plastic coin as you suggested

Any ideas on why the simulations still fail

[cjquijano]

Hi @DaleKramer and @rszoeke,

So, I gave up with the cyclic symmetry. I think I know what is going on and maybe @rszoeke can correct me if I am wrong. As far as I can tell the Cyclic Symmetry boundary condition does not stop the model from rotating about the axis. Normally what I would do is create a local (r, theta, z) coordinate system at the center of rotation and then constrain the side faces to theta=0. Another option to constrain the model would be to apply a boundary condition to a node or point but in WB 2.0 we don’t have this option anymore.

I went back to a quarter model and aligned the side to the global coordinate system and I came up wit the following results.

You can see the Quarter Model Here.

It ran in about 6 minutes.

[DaleKramer]

That is GREAT, lets see if Richard agrees with you on cyclic symmetry.

BUT, perhaps that cyclic error on the last 10 degree sector was related the the RHINO bug.

Did you try a run on your Parasolid 10 degree sector with cyclic, just to make sure our cyclic problem is not RHINO bug related

I the meantime I used your Quarter model and made the plastic ~10 times ‘softer’ so the coin diameter would increase due to less ram force which causes less flexing in the die plates.

Ideally I want a uniform 0.002" think plastic coin.

Here are the dimensions of your run in inches at various points:

And with my run with 10x ‘softer’ plastic (this got an extra 1/2" radius on the coin and very close to 0.002" thick almost everwhere ).

Yeah, I think that horse was won the race

[cjquijano]

Finally!

[DaleKramer]

But there is still no answer to this topic, since there is no solution for a cyclic symmetry setup that will finish a simulation run, still need help there please …

[rszoeke]

Hi @DaleKramer,
seems like I lost track of this project…

@cjquijano is actually right. What the cyclic symmetry condition does is the following: it works as a bonded contact, but actually bonds t the slave in a rotated coordinate system to the master.

This means that it can still allow the system to rotate around the cyclic symmetry axis if this is not prevented at the master surface.

So what we usually do to prevent this is to add a fixed value condition at the master and fix there the direction which is normal to the master. In some cases you would need to swap master and slave (and change the sign for the rotation axis accordingly) so we use the face which is perpendicular to a coordinate axis as master.
Alternatively we can use a symmetry condition on the master.

I copied Christopher’s project and applied the above changes (additionally I replace the fixed support at the bottom with a fixed value leaving only the axial direction constrained): https://www.simscale.com/workbench/?pid=8247173219304259707&rru=b8c12682-ac90-4c1f-bb6a-3de738fd81dc&ci=29c8ecf9-587e-43c1-a871-841ead00b9d3&ct=SOLUTION_FIELD&mt=SIMULATION_RESULT

The case finished easily in a couple of minutes and the deformation looks fine:

Best,
Richard