A steal roller and a sheet of rubber that is going to get rolled. I’m using PPM v1.4.0, analysis is dynamic implicit (I couldn’t get explicit to run). The roller is tied to a reference point that is constrained in the x and y direction (i tied the outer surface of the roller, to make it rigid). There is a symmetry BC on the roller and rubber sheet in the z direction. For now the roller and the rubber sheet are modeled as linear elastic, until i solve the other issues (in reality the rubber is uncured (unvulcanized) which I know will be another problem to model). There is a hard surface to surface contact between the roller and the sheet (mortar wouldn’t run). The underside of the sheet is constrained in the x direction, to simulate a table. I applied a initial velocity to the rubber in the -y direction, so it would move through the roller.
Now the problems:
I would also like to add initial rotational velocity to the roller and i don’t know how. i tried adding a BC with 1 rad rotation around the z axsis, but I get strange results. I think what is happening is the roller accelerates form 0 with really high acceleration and therefore gets distorted. Any suggestions on how to get the roller spinning?
Now also when I try to remove the rotation the simulation fails after about 2 mm of movement of the rubber. How could I try making the simulation more stable or how do I get the explicit dynamic to run (I think it would be more suitable for such a analysis).
The goal of the analysis is mainly to get the necessary torque and power of the roller to deform the rubber and to get the force with which the roller is getting pushed upwards. Secondary is also to visualize the movement of rubber.
Apply rigid body constraint to the whole roller to save time. Then symmetry constraint won’t be needed for the roller - just fix its reference point in the Z direction.
Try adding this in the keyword editor to apply initial rotational velocity to the roller:
Of course, you will have to check the number of the ROT NODE of a rigid body constraint in the keyword editor first. Fix the other rotational DOFs of the rigid body reference point.
If you want to apply a load or a boundary condition in such a way that it increases gradually during some time, you have to use the amplitude feature available in PrePoMax.
Ultimately, you will have to use a hyperelastic material model for the rubber.
Dynamic explicit is quite limited and difficult to use in CalculiX but normally it could be the best choice for this type of analysis (if you manage to properly simulate quasi-static conditions). You could also try with a static analysis first. Anyway, don’t forget to enable Nlgeom.
Proper friction definition for contact will be crucial here.
So, I tried implementing your points, adding that keyword worked, so maybe that could be implemented in the UI of PrePoMax in the velocity initial condition definition. I also added displacements of the roller and rubber, so that the simulation is more stable. The simulation now runs as dynamic explicit, and it seems to be more stable than implicit. I also greatly reduced the element count, by removing unnecessary parts of the roller.
The problem now is that the simulation takes waay to long, it took about 2 and a half hours to calculate 3,3e-6 s of the total 1 s analysis at which point the simulation aborted because the maximum number of iterations (1000) was reached. Do you have any advice on how to speed it up? I’m already using the Pardiso. I would like to try PaStiX, but I don’t know where to get it.
Another problem is the material. Now I know that rubber is difficult to model, uncured rubber at 120°C even more so. I looked through some research papers and it seems that it is most commonly modeled using elastic visco-plastic material models. For now, I tried simplifying it a lot to just linear elasticity and linear plasticity, but when I add plasticity to the material definition the simulation won’t run anymore. It gives the error that no result exists. Do you have any idea what could be wrong? I could try modeling visco-plasticity using the *CREEP and *CYCLIC HARDENING keycards, but I don’t really understand enough of how they work, and I don’t have any values to use. And even if I manage that I doubt it would help the simulation if regular plasticity doesn’t work.
I’m also attaching the .pmx file so you can have a look. Thanks. Roller_die-analysis2.pmx (1.5 MB)
You can speed up an explicit dynamics analysis in CalculiX by setting the minimum time increment to the desired value higher than the stable increment - this will enable mass scaling. Just make sure that the results are correct and kinetic energy is not too large.
Regarding material properties, hyperelasticity is the most common model for rubbers and you could find some approximate data for that. Some applications require more advanced material models such as viscoelasticity or even Parallel Rheological Framework but this is not supported by CalculiX and it’s more difficult to find good data in literature. Here you can find an interesting discussion about advanced material models in CalculiX: ViscoPlastic Model - CalculiX
There is an issue that makes many (or even most) explicit dynamics analyses in CalculiX fail when plasticity is added. Recently, I had a discussion about this with the developer of CalculiX and I sent him my example. He said that he would take a look at it. Moreover, the next release of CalculiX will feature the Johnson-Cook plasticity model meant for explicit dynamics (but rather for high-speed dynamics).
Thanks, I will try increasing the minimum time increment today.
I know hyperelastic models, since I used them in already when I was designing a pneumatically actuated gripper finger. But I don’t think it would be applicable here, since uncured rubber at high temperatures behaves very differently. I read the post you shared, it seems interesting, but I doubt it is worth pursuing, if there is a problem that makes most analyses fail when plasticity is introduced, since plastic deformation is what I’m after.
I found two papers, where the researchers developed very complex material models, for similar materials:
But I guess this is not possible in Caluclix. For now, I just wanted to approximate bi-linearly the first loading curve for one strain rate of the stress-strain curves from the second paper, since I don’t have cyclic loading. Kinda like this, my lines are green:
Or do you think that a more complex model might have more luck in running than this simple model? Do you know approximately when the next release of Calculix is coming? Is it more likely a year od a month from now?
It might be better to start with a simple material model, especially when it comes to explicit dynamics in CalculiX. Then you could even consider using custom model based on the literature and implemented via umat.f subroutine.
Based on what Guido said some time ago, it should be released this summer. He even considered releasing it at the end of July but I guess that there will be some delay.
I noticed your post today that a new of CalculiX was released which is really incredible timing. I’m wondering where can I find it, since the official CalucliX page still shows v2.20 for me and how can I replace it in PrePoMax if it’s possible. I would like to test if plasticity will work now, since i got the model working fine without it. i just need to change the material model now.
Hi Cimek,
My name is Nadda. I’m a new PrePoMax user and trying to set up a rolling model as below image. I am trying to download your Roller_die-analysis2.pmx model for reference but always got error (download not complete). Can you please share it again?
