Hello Everyone,
I am currently trying to simulate a fiber filament (diameter 15 mikrometer and 16 cm long) that being fixed in one side (thus free at the other side) under gravity in Abaqus. The filament is inside a container and is put perpendicular to the gravity. So the fiber should, at the end of simulation, leaning down to the down side of container, while it is still hanged.
The problem is: Abaqus need 18 minutes to simulate one filament. That's unexceptable for such a simple simulation. I think it took abaqus so long to simulate it, because the elongation is really small compare to bending (and torsion in my future simulation) and also the ratio of diamter/length is really small. Therefore the abaqus need a really small "stable time" for each step, to be able to calculate the elongation. But elongation is not my concern, my concern is the bending and also torsion.
I think, it would be faster if the material definition is modified somehow, so that the elongation will not be considered in the simulation. At my last simulation I used isotropic elastic material definition. I had also tried using vumat and make the first three rows and columns of tensor matrix 0 and the rest diagonal elements of tensor matrix filled with G (Shear Modulus). But after do that and run the simulation, it stoped, because of distorted mesh element.
Could somebody advice me, if there is a way to make this simulation efficient?
Thank you for your interest in my problem and input.
I attached as well my input file for the simulation for 5 filaments with isotropic material definition as well the vumat file.
Best regards
Gunardi
The stable time increment in Explicit is determined by speed of sound and element size of the smallest element.
If dynamic forces are not relevant for you, you can use mass scaling.
Or you could think about switching to implicit which is unconditionally stable (if it converges).
Hello Martin Baeker, thank you for your interest in my problems.
"The stable time increment in Explicit is determined by speed of sound and element size of the smallest element."
I think the simulation was really slow because the element size of the filament is really slow. I can't use any larger element size, because the filament has 15 um diameter. I can't apply beam elements for the fiber filament, because eulerian analysis (that I am planning to do latter on) does not support beam elements. That's why I have to use other than beam and truss elements.
"If dynamic forces are not relevant for you, you can use mass scaling."
The mass scalling is not appropriate to be applied to the filaments, because the filaments are the main purpose of this simulation.
"Or you could think about switching to implicit which is unconditionally stable (if it converges)."
Abaqus implicit does not support eulerian analysis, which is why it is not applicable here.
I was thinking that may be I could overcome the limitation of eulerian analysis by using co-simulation (The first simulation for the eulerian analysis in dynamic explicit step and the second one for the filaments using beam elements in dynamic implicit step (or maybe in other steps). But now there is a new problem, which is: what should I define to be the connecting "elements" (could be either surface or nodes) between these two simulations in step "interaction" >> "standard-explicit co-simulation".
I tried yesterday to do the simulation with 5 filaments. The simulated stopped because the elements of the filaments are distorted after some steps. I can always use smaller elements for the filaments but the simulations would take too much time.
Bottom line, it all comes down to this following question: how could I model an object properly that has a really small diameter/length ratio (in my case fiber filaments with diameter 15 micrometer and length 17 cm)?
Thank you again for any feedbacks :)
Gunardi
"The mass scalling is not appropriate to be applied to the filaments, because the filaments are the main purpose of this simulation."
I don't understand that statement - as long as dynamic effects are not relevant, mass scaling should cause no problems, it basically means artificially speeding up the process.
If you cannot use mass scaling and are forced to model the filaments, I think you will have to forego the eulerian analysis and switch to implicit. (Depending on what you are actually trying to do.)
- Badges
- Science topic