Notifications
Clear all

Non-local model in ABAQUS VUMAT

4 Posts
3 Users
1 Likes
1,779 Views
Posts: 2
Topic starter
(@zab1323)
New Member
Joined: 4 years ago

Local damage models are mesh sensitive so i am going to add non locality in my local model (in Abaqus VUMAT ). In order to do so I need weighted average which requires data for all material point or integration point at an increment. However, ABAQUS VUMAT only gives local information. Is there any way to get access to global information in VUMAT?

3 Replies
Posts: 3993
(@jorgen)
Member
Joined: 4 years ago

I have not tried to do that, but you may be able to do it by using global variables. That is, each time the VUMAT is called it can save whatever state variables you like into the global variables. Then you can probe those global variables to get the non-local info that you need for your integration. If you are careful, you should be able to do this in a thread-safe way.

I realize, however, that this is not very "elegant".

Another way is to write your own user-elements. Which is also annoying...

What material and behaviors are you studying?  (I have not seen a need to use a non-local damage model yet)

-Jorgen

Reply
Posts: 2
Topic starter
(@zab1323)
New Member
Joined: 4 years ago

Hi Jorgen, 

 

Thank you for he reply. I am studying dynamic crack propagation and crack branching in brittle material. I have a Continuum damage model which leads to localization of damage. Hence, I am trying to introduce an internal length scale to prevent localization.

My initial thought was to use VGETVRM utility in VUSDFLD and pass the data to VUMAT, which seems little too complicated now. Any idea would be greatly appreciated. Thanks

 

-Zaber 

Reply
1 Reply
(@familyguy)
Joined: 4 years ago

New Member
Posts: 2

@zab1323,

 

non-local models using all "data for all material point or integration point" are certainly impractical and expensive. Having said this, even calling only points in a neighbourhood with many elements could be cumbersome, for me at least (that is, I do not know of any acceptably efficient way to do it for anything larger than an academic 100 elements computation).

If the non-locality is sufficiently "small", you could sometimes get away with the strain and its first derivative, staying within a single element.

If not, you could let a lot the mathematical formulation do a lot of the hard work, and consider phase-fields model, which perfectly regularise local damage formulations, e.g.  https://www.sciencedirect.com/science/article/abs/pii/S0927025614004133

Reply
Share: