Learning Forum

[walter.araujo]

Hello,
I’m performing a 2D axisymmetric analysis. The simulation is related to a burst disc subjected to a pressure.

After analysis i found some disc penetration on sharp edge zone. The material disc is aluminium (NL - nonlinear) and contact uses normal lagrange formulation.

It seems this is related to characteristic of material. I found that on 2D axisymmetric analysis is not possible to adjust contact stiffness.

Anyone could help to improve my analysis and avoid this penetration? Any idea or suggestion are welcome

Thanks

[peteroznewman]

Put a small blend radius on the sharp corner. Put at least 6 elements on that blend edge. The contact algorithm will perform much better with the sharp corner removed.

[walter.araujo]

Many thanks for your prompt reply.
I already try blend edge, but i'm getting an instable behaviour. Check image bellow:

Note: frictionless contact type

[walter.araujo]

I had reduced the local mesh element size, and i get a converged solution, but i still have penetration.

 

I think that if I could increase the stiffness of the contact in some way, I could reduce the penetration....

[peteroznewman]

In the Details of the Contact is a row called Normal Stiffness. Change that to Factor and in the next row, put a Normal Stiffness Factor of 10 instead of 1.  That will reduce the penetration. In the row below that you can Update Stiffness on each iteration.

[walter.araujo]

Thanks again Peter.
It doesn't have that option, that's why I made this post.
I think unfortunately 2d axisymmetric does not have that option.

[peteroznewman]

Okay, what I said above is only for 3D models.

For your 2D Axisymmetric model, try setting Shell Thickness Effect to No.

[walter.araujo]

Hello Peter,
No difference!

Strange behaviour...

[Akshay Maniyar]

Hi Walter,

In addition to what Peter suggested, can you try changing the detection method to some nodal or to a combine method? 

Thanks,

Akshay Maniyar

[walter.araujo]

Hello Akshay,

Already tried:
 - Nodal-Normal From Contact --> no convergence

 - Nodal-Normal To Target --> penetration

 - Combined --> the disc moves away slightly from fixed part (picture bellow)

[peteroznewman]

I suggest you change the Behavior from Symmetric to Asymmetric.  Make sure the Target side of the contact is the flat plate and the Contact side of the contact is the corner blend. 

Then you can retry the two Nodal Normal settings. I can never remember which one is which.

[walter.araujo]

 

Hi Peter,
didn’t worked…

 – Nodal-Normal From Contact –> disc moves away slightly from fixed part and diverge

 – Nodal-Normal To Target –> penetration

 

[peteroznewman]

I see Small Sliding is On.  Turn that Off.

Under Analysis Settings is Large Deflection turned On?  That must be On.

Also the Frictionless Contact Behavior is set to Normal Lagrange.  You can try some other options.

[walter.araujo]

Hi Peter,
Thank you for your patience.

 - Large Deflection turned On? - yes, allways for nonlinear analysis.

 - Already tried different types of contact. Despite penetration, "Normal Lagrange" is perfoming better!

 - small sliding off - it seems to have a better behaviour but i am facing convergence issues... need more time.

 

[peteroznewman]

A good remediation for convergence issues is to force the solver to take many substeps. That means setting the Initial and Minimum Substeps to have a large number like 100 or 500 etc. Maximum Substeps should be 1000.

Another remediation is to allow the solver to continue iterations for more than 26 before taking a Bisection by adding the Command object with the code

NEQIT, 100

which will allow the iterations to go to 100 (or 500).

[walter.araujo]

Hi Pedro,
I tried to increase the number of substeps + 100 iterations before making a bisection. It was not successful.
No matter the number of sub-steps, the solution diverges at the same point.

load applied

 

less substeps - stop at 1.3

 

 

more substeps - stop at 1.3

[peteroznewman]

Another possibility is that the Ansys 2D system was broken at the beginning and is not working properly.

This can happen if you create 2D geometry and open that in Mechanical, but forget to set the Property of the Geometry cell in Workbench to 2D.  The geometry is imported into Mechanical as 3D. You build up the model, then realize it is not in the 2D mode so you can't set it to be Axisymmetric.  You might expect that you can simply go back to the Property setting of the Geometry cell in Workbench and flip it from 3D to 2D.

While this is possible to do, the outcome is not the same as setting the 2D propertry of the Geometry cell BEFORE ever opening the geometry in Mechanical. The only way to fix the problem is to start a new Static Structural and set the Geometry cell to 2D before you open the Model cell.  Then you can build the model and see if that improves the outcome.

[walter.araujo]

many thanks for you patience...

I had created a new study as your suggestion but i got not successfully results. Same behaviour!

small sliding switch off improve the study, but it is getting difficult to achieve convergence... 

[peteroznewman]

What is possibly happening is the structure becomes unstable at that point.
The static solver may not be able to find a stable equilibrium past that point.
You could try solving the model in either Transient Structural or Explicit Dynamics.

It might also be that a 3D disk wants to buckle out of an axisymmetric shape at some point so a 3D model will give different results in a Transient Structural or Explicit Dynamics solver than the 2D model.

[walter.araujo]

Hello Peter,

An explicit analysis is not practiable for me in 2D. It only uses one core and due to the elements size is a study that takes +/-40 hours.

 

Anyway, I think I have arrived at a balanced analysis.

It is true that from a certain point the disk jumps out of the rim, and it makes sense that the study diverges (for an implicit analysis).

But even diverging I can make an analysis at the moment that the disk jumps, which was actually what I wanted.

 

Conclusion:

-small sliding off issue helped me to check the pressure from which the disk release happens (no penetration, even for sharp edge).

-use of the transient study allows to "refine" the moment of the disk jump.

-In the small sliding there is another option "adaptive", which I think helped too!

 

Peter, thank you very much for your tips and help.