Learning Forum

[VVCZ]

Hello all,

 

I have a following problem. My task is to simulate compression test of metal foam and afterwards, for specific thickness of the specimen (block shape, rectangular plate with thickness) simulate tensile test just to reveal initial resistance (i.e. elastic modulus) of the compressed specimen.

Because of severe deformation during the compression test I was forced to use explicit solver via LS-Dyna, where I artificially increase time step time via mass scaling + I compress the foam specimen in a way smaller time (plasticity model does not involve rate dependency + kinetic energy is much smaller than internal energy throughout the simulation, so I conclude I can do that).

What bothers me is if I want to compress the foam and afterwards I want to pull the nodes (in a direction which is transversal to the compression direction) in order to get force reaction in boundary conditions, there is a problem with elastic deformation. After compression, there is an elastic deformation and I am of course supposed to wait until the elastic deformation vanishes and then try to pull the nodes. But that leads to big simulation times + even after waiting quite a long time and then pulling, I cannot distinguish the linear part from which I need to acquire elastic modulus.

Another solution might be only save deformed state in LS-Dyna with appropriate thickness and then, in a new simulation, pull the nodes. I tried that, but LS-Dyna is unable to finish the solution due to too big mass increase.

So I want to ask whether there is a possibility to save the deformed mesh with stress-strain history and import this into Workbench? I know I can export the mesh and also get nodal displacements, but somehow I don't know how to carry these results into Workbench.

[Ushnish Basu]

You can try using INTERFACE_SPRINGBACK_LSDYNA with NHSV specified to create a dynain file at the end of the run - this dynain file will contain stresses and all history variables (see beginning of the *MAT section in the material manual for a count of history variables for each material). Since you can run LS-DYNA from within workbench, presumably you can import this file also.

To address your root issue, make sure you are using DAMPING to damp out the transient oscillations from your run. Once you plot your force-deformation history, you can use Oper -> differentiate in LS-Prepost to compute the derivative, which should give you the initial modulus.

Hope that helps.

[VVCZ]

Dear ubasu,

thank you kindly for your answer. Meanwhile I came with a different idea. Just to make it more clearer for marking entities, the specimen is in block shape (2x4x4) mm, so the node sets on the surfaces are named (X0,X2;Y0,Y4;Z0,Z4). I am performing one simulation (1 keyword input deck) and am dividing it into 3 steps.

• In step one explicit scheme is used, and I am compressing the foam specimen. (displacement applied on tool 1, via contact and the tool 2 which is fixed, the specimen is compressed)


• In step 2 I am switching to implicit scheme (tools are gone, foam has X2 node set fixed to allow relaxation in the opposite direction of compression).


• In step 3 I am again switching back to explicit scheme while releasing the X2 node set and at the same time I fix other Y0 node set and the opposite one (Y4) is pulled.

What does not appear that good is the X-force reaction from the X2 nodes, it seems like during the implicit step the specimen is not relaxing (e.g. the X-force reaction curve should jump when step 2 and also X2 BC are applied and then it should get closer to zero with time, but it doesn't happen...).

You correctly mentioned damping and I am using damping coefficient (damping constant = CDAMP * (2 * omega_min) = CDAMP * (4 * pi * FREQ) - got from this link) on the specimen throughout the whole simulation to damp the specimen which is stimulated by relatively high speed displacement (160 mm/ms).

Now I must say you really made my day with that Oper -> differentiate in LS-Prepost, I was really importing the data in python to do regression. Which may seem funny but as I think about that, I really can't see into algorithm of LS-Prepost to see how exactly it differentiates, so let's say the python way wasn't that bad

Anyway thank you very much for the kind reply.

V