ASME XIII Section 2 DBA: Mesh convergence for Plastic strains

[fem.fan]

Hi, i am performing an elastoplastic analysis for a vessel nozzle connection according to appendix 46 of ASME XIII Div. 1 using part 5 of ASME XIII div 2.
my problem is that when simulation for plastic collapse, i can´t get the plastic strains to converge when refining the mesh.
in other words, i keep refining the mesh and the plastic strains values keep changing. i can´t refine anymore as i run out of RAM (i have 128 GB) and it seems excessive.

i´m using SA 516 Gr 70 Stress strain curve provided by the code


can anyone provide any insight about some criteria used to decide if the results are sufficiently accurate.
this quetion is actually quite genreal as a FEA problem rather than a pressure vessels one.

i ran a very aimple problem of a "beam in cantilever" with a force at the tip using solid elements and i had to produce a highly refined mesh to achieve convergence.

thanks a lot,

 

[XL83NL]

Sec XIII (the new section for over-pressure protection), or sec VIII, Div 2?

Huub
- You never get what you expect, you only get what you inspect.

 

[fem.fan]

Div 2

 

[XL83NL]

Of section VIII I asumme? You haven't mentioned that

Huub
- You never get what you expect, you only get what you inspect.

 

[fem.fan]

i mentioned it, ASME XIII div 2. nevertheless, it is independent of the code, it´s a Finite elements analysis problem

 

[XL83NL]

There is no Sec XIII, Div 2. There's only Sec XIII. Part 5 of Sec XIII is about Requirements for pin devices.

Huub
- You never get what you expect, you only get what you inspect.

 

[fem.fan]

your right, sorry, i meant section VIII (8)

 

[TGS4]

Regardless of whether the plastic strains "converge", do you achieve convergence of the analysis with respect to static equilibrium?

Exactly how have you implemented the elastic-plastic curve from Appendix 3-D?

What element formulation are you using?

 

[fem.fan]

TGS4 - i achieve convergence with respect to static equilibrium.

i implemented the curve from appendix 3-D following the steps specified there. a multilinear curve.

i´m using 20 nodes bricks where i can and quadratic tetrahedrons where the geometry does not allow bricks for it´s complexity.

 

[TGS4]

So, I am not certain as to the problem here. Are you having an issue with Protection Against Local Failure?

 

[DriveMeNuts]

Photo's?

 

[fem.fan]

TGS4, this is not an issue with a failure mode, rather it is a numerical issue. in order to affirm that the results are not mesh dependent, i am performing successive refinements and compare the results.

this is the model:

this is the nodal fraction (ansys plot) that measures the percentage difference between the values of strain in a node coming from it´s adjacent elements (nearest gauss point in ansys):

as you can see, there is more than 20% difference in a lot of places.

 

[TGS4]

I do not think that you are showing what you think that you are showing.

When you modify the mesh, does your strain value change? Can you show plots of plastic strain with 2 or 3 mesh densities?

 

[fem.fan]

i have that too:

X axis is 1/volume of elements , and de y axis is de strain normalized being 1 the value of the coarser mesh.

however i showed the other plot since it is "more alarming" than this one.

 

[TGS4]

This last post seems to be typical h-element convergence. I see nothing concerning here.

 

[fem.fan]

i know, but what about the nodal fraction? shouldn´t one expect a smoother transition between elements?

 

[TGS4]

I am not familiar with "nodal fraction". Perhaps you should post this inquiry in the ANSYS forum for further discussion.