Adding Viscoelastac Effects (Prony Series) 1

[aphilic]

Hi guys,

I have met a problem and could not think of anything. If anyone can help to show where i have done a mistake i will appreciate.

I have a very simple model in Ansys. It is a cube shape element. I give a displacement on a node and constrained the opposite site no movement (0 DOF). I just want to see the difference between the reaction forces after 1 sec and 10 secs.

I applied to displacement defining "time at the end of load step=1" and run it and I had no problem. But then when I chance the time at the end of load step to 10 and tried to run the program terminated itself leaving two different messages at each try of mine:

1. The solution did not converge
2. The element has turned inside out

In both cases when I plot the element at the last substep before terminating, its shape was something so much different. I cannot even explain. It is not even similar to cube or have any shape. Maybe I can show a pic of that if anyone asks.

Where did I do the mistake? I think it is something about the viscoelastic constants that I added to the model. I just added one a1(this is alpha i guess) and t1 values to the share response choice.

Thanks in advance.

Regards,

Can

 

[cbrn]

Hi,
being your material property a function of time, I find it normal that you may run into non-convergence for some combination of loads and time.
The fact your shape is "absurd" is due to the fact that no compatible displacement field has been found in order to respect equilibrated force field.
How to remedy? I don't precisely know, but if you're only doing tryouts in order to practice viscoelasticity, I'd take TIME=1 and NSUB=10, then ask for "save all results at every substep": thus you'd have 10 substeps at 0.1, 0.2, ..., 1.0 time units; If you think 1 "time-unit" = 10 "seconds", then you're done! WARNING: by doing that, i.e. by using a "time" frame which doesn't correspond to reality, it's the same thing as scaling some viscoelastic parameter!

Regards

 

[aphilic]

Hi cbrn,

Well you are right, I am just doing some tryouts to find out the effect of the viscoelastic parameters. However, one of the parameters i have to enter is relaxation time constant which is in sec unit actually. So, I don't think I can assume 1 "time-unit" = 10 "seconds" for if I assume that, I should enter relaxation constant accordingly and at the end it would make no difference from me just entering normal value of ti (relaxation constant) and 1 sec as time unit. Don't you think so?

One more question, lets assume that I want to apply loading as displacement in 2 load-steps. In the first load step I want 1 unit of elongation to be applied at one node in ramp shape within 1 sec which is quite easy to apply. Then, for nine more secs, I want that node is kept one unit elongated (I mean no more elongation but keep the previous elongation valid). At this point i should run the model again but should I change anything else but changing "time at the end of load" = 10?

Thanks in advance

Regards,

Can

 

[cbrn]

Hi,
1- You're right. It seems the relaxation of the material is making your model collapse after a certain amount of time. This may be a scenario; the last converged substep will give you the situation just before the collapse. If you want convergence over all the time frame, I think you have to diminish the imposed displacement.

2- Your procedure seems correct.

Regards

 

[aphilic]

Hi,

I have a question about viscoelasticity. In Ansys, when determining a viscoelasticity effects using prony option, should i use volumetric and share response dataset together or is it enough to fill one response dataset? I mean, if i just enter the "alpha" and "t" terms of shear response of prony; it should be enough, shouldn't it?

 

[mdiany]

Hi,
Can any one tel me how to calculat prony constants for viscoelasticity models.