Hello,
I have to study the contact of a an elastic body over a rigid plane in Abaqus. The indenter is a cylinder so the shape of the contact area is circular. The plane is considered as infinite. The load applied is a gravity on the normal axis.
I am using this simple case to validate my way of defining contact properties, interaction definition and contact controls.
I am trying to understand the difference between several contraints enforcement methods to resolve normal contact. In the Abaqus manual, it is said that "The direct method can be used to strictly enforce a “hard” pressure-overclosure relationship. Lagrange multipliers are always used in this case".
I understand that in the case of hard contact, using the "direct method" or the "augmented lagrangien method" is equivalent (when all other parameters are constant). But in practise, I find significative differences in the results, particularly in the opening (COPEN) of the contact.
Do you have any idea why ?
Thanks a lot for your help
Rémi
Can you identify which plot is which? The second image (first COPEN plot) looks like noise (solver round-off) to me; all the values are around 10^-16 and I would call that "zero". How do CPRESS and CSHEAR look?
Have you tried analytical rigid surfaces rather than element based?
Have you tried a deformable but stiff substrate?
Glad to see you're running extensive tests on simple analogues. That's the only way to build a complicated analysis and have any confidence in the results!
Hi Graham,
thanks a lot for your answer.
The second image (first COPEN plot) is for the direct method, the other one for the augmented lagrangian method. I think you are right about the "noise" due to round-off.
How do you know about the order of magnitude of this round-off ? I looked in the Abaqus manual but I can't find any value.
I have posted some plots of CPRESS and CHSEAR. CPRESS seem to be the same but I don't have any CSHEAR because I only used a normal contact condition. I thought I would put some tangential load and contact condition after understanding this step.
About the substrate, I am using a deformable body but with a E=1e20 MPa et mu(Poisson) =0. I will try with an analytical rigid surface.
It is surprising how little tests like these can cause some difficulties...
Rémi
It's a function of machine precision. ABAQUS does its internal computations in double precision, which is carrying 15-17 significant digits in the mantissa. Depending on the magnitudes of numbers in your problem, if the numbers you're dealing with vary by 15-17 orders of magnitude you get round-off error. Without knowing the details of your inputs, CPRESS output on the order of 1E-4 and COPEN output on the order of 1E-16 might be round-off, especially looking at the noisy nature of the output in image 2 above.
CSHEAR is zero... I presume you have no friction on your interface? Is that going to be important to your analysis? You should get some slip and shear from the indenter expanding radially as it compresses axially.
I agree that small models can be tricky, but that's when you learn new things! Imagine trying to debug these contact nuances on your full model...
Ok I understand now. So I will try to apply a more important load or to soften the material to see how COPEN and CPRESS look like.
As you said, I have no friction at the moment so CSHEAR is zero. I will add friction later.
I am studying this case because I have to model a receding contact in order to get a simple model for the behaviour of a bolt when tightening. At the moment, the openings that occur during this phase are not taken into account in design. And I think it can have a significant effect on slip when the bolt carry an external load...
If cpress is non-zero, copen is zero and vice versa (That's the famous Kuhn-Tucker equation), at least when you use hard contact.
So the very small values for copen should be round-off errors. Keep in mind that 1e-11 is in your length units (m/mm?)
- Badges
- Science topic