Mesh refinement limit 2

[davidrdguez]

Hi there!

I've been making a simple simulation of an aluminium plate.

In a previous post I already asked how to get the reactions on the supports.

My problem now is the more I refine the mesh, the higher the stresses get. This happens in a very specific area. I mean. Everything is "blue" and suddenly a big gradient in the maximum stress area.

In a first approximation I got an error of 10% with something like 30 MPa, now i have 116 MPa with a 6% global error.

I'm sure if I make the mesh finer the maximum stress will get larger and the error smaller. But, where is the limit?

thank you
regards,

http://img135.imageshack.us/g/20492407.jpg/

some pictures

 

[rstupplebeen]

In FEA stress risers such as corners or point loadings will increase with mesh refinement. It is up to the analyst to determine if these localized stresses can be ignored or if more modeling rigor is needed such as a plane strain detailed model. I hope this helps.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/

 

[davidrdguez]

umm...

it helps, but i'm not experienced enough to decide if i can neglect this tension.

what do you mean with "plane strain detailed model"?

thank you!

 

[franc11HD]

THis usually occurs when using some constraints at nodes specific nodes when the model is not fully in equilibrium. For example, if you model a straigth rectangular beam with 3D elements. You load it in tension at one end and clamp it at the other end. You stress should be accurate along the beam, but more you get closer to the clamp, the stres will raise because of this "feature" clamp.

One way to avoid this for simple structures is to apply force in equilibrium. So you apply tension at one end and equalt tension at the other end. Then, you are in equilibirum. But you still need to constrain the model to avoid rigid body motion. You restrain one node with (Tx, Ty, Tz), then in the same plane, you restain a second node (Ty, Tz), and another node in Tz. (I don'T know if the order is right, but you need to constain all 6 DOF. Once you run your model, those three restraint should not react any significant force, because you model is loaded in equilibirum.

This usually applies easily for simple model. For more complicated model, I should suggest to put restraints as far away as possible of the region of interest and make sure that those big stresses are not detrimental (see with more experiences users).

 

[rstupplebeen]

A plane strain detailed model is a 2D model that you can apply boundary conditions developed from your 3D model. The benefit is that the 2d model can be very detailed with many more elements.

I assume from your picture that you have 2 materials or a step in the model. For the first is there a glue, is it press fit... For the second applying a fillet of appropriate size and then significantly refining the mesh may also be an option. However conventional knowledge would say not to include the fillets.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/

 

[davidrdguez]

Thank you for the answers.

I'm busy with something else now, but I will try your ideas!

cheers,