Different results with Explicit method and Implicit method

[Angel Ochoa]

Hello everybody,

I´m simulating a problem of indentation of a tungsten needle tip on an aluminum layer. Before I was using just the module "Static Structural" from ANSYS, which is based on an implicit solver. Now I wanted to do the same simulation with the module "Explicit Dynamics" of ANSYS, performing a quasi-static simulation. So my aim is to obtain the same results with both modules.

The process is rather slow and I have checked that the kinetic energy is negligible (<<5%). Most deformation is plastic. I have tried different stress-strain curves for the aluminum, changing the strain hardening during the plastic deformation.

Results:
1- When using large strain hardenings I get the same results from both modules, which I expected.

2- HOWEVER, when using small strain hardenings, the results differ a lot between both modules. This is to say: the more plastic I do the aluminum, the higher the differences in the results between both modules.

Question:

Is this something due to the physics or could I "fix" this difference by tunning the parameters in the explicit module?

Some ideas:

I know that the plastic waves travel slower than the elastic ones. The lower the hardening, the slower the plastic wave is. Maybe this is the reason for the observed results.

I have already changed many parameters (time step size, mass scaling, damping factor and loading rate) and they do not have influence in the observed behavior for lower strain hardening.
------------------
I would appreciate a lot any idea or reference in relation to this.

Best regards!

 

[afreiden]

Are your simulations dynamic (e.x. applying the indentation over a time of 5 seconds)? I assume the answer is yes, in which case perhaps the timestep you're using for implicit is too large, which is causing error in your results. I'm not that familiar with ANSYS, but usually software with explicit dynamics capability will choose a stable timestep for you, and accuracy is guaranteed. So if your solutions are different, then the implicit solution is probably wrong. What is the total time of your simulation (e.x. 5 seconds) and what is the computational cost you're seeing when doing implicit vs. explicit (e.x. implicit 20min vs. explicit 3hrs)?

I've done plenty of quasi-static simulations using both implicit and explicit (LS-DYNA) and often assume elastic-perfectly-plastic material behavior. I've never seen any significant difference in the solution, but I use a very small timestep when doing implicit. For simulations slower than a second or so (i.e. "quasi-static") implicit is still much more computationally cheap for me.

 

[Angel Ochoa]

Dear afreiden,

thanks a lot for your response.

The simulations are actually quasi-static. The indentation is applied over 1 second, but the elements are so small that their inertial effects can be neglected ( I have already checked this and it is widely proven also in literature). In addition, if I apply high strain hardenings the results are identical between both problems. The differences in the results arise when I reduce the strain hardening of the models.

Regarding the implicit model, I am quite confident I am using steps that are small enough, because I have also tried to variate their size.

The question is, why do I only observe differences in the results when "increasing" the plasticity of the material (and the differences increase proportionally with the decrease in the strain hardening)?

Best regards. I appreciate all your answers.