How to obtain resonable nodal stress values? 1

[sunnyqzhang]

Hi folks,

Since nodal stresses are not continuous across the element boundary, in what way can I obtain the resonable values? For example, average the stress values at the nodes within different elements?

Thanks a lot!

Sunny

 

[mrgoldthorpe]

sunnyqzhang,

The most accurate stresses are at the element centroids
(*EL PRINT, POSITION=CENTROIDAL). If you want nodal stresses ABAQUS will provide them
(*EL PRINT, POSITION=AVERAGED AT NODES)
Of course, they are averages from adjoining elements if the element type and material properties are the same.

MRG

 

[Drej]

MRG - Am sure you mean the "Gauss points" rather than the "centroid" of the element. Hence the command should be *EL PRINT,POSITION=INTEGRATION POINTS (default case) for the most accurate stresses to be printed.

sunny - The gauss points stresses should always be printed/output if the most accurate stresses/etc. are needed since the stresses here are obtained directly from the displacements at the nodes (third order quantity after strain) without extrapolation. If the mesh is adequate you should find little difference (say 5%) between element-averaged and nodal-averaged values anyway. In fact, this methodology is used to estimate mesh errors/norms due to high displacement gradients.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[corus]

In Viewer go to result/options and you can uncheck the averaging at nodes option. I think the same result can be obtained by moving the averaging threshold slide bar down to zero.

corus

 

[mrgoldthorpe]

Drej,

Centroidal stresses are generally regarded as most accurate. Well, at least when I first learnt FEA I was told/read this!

Gauss points are used to integrate stiffness matrix etc.

MRG

 

[Drej]

Have to be careful with terminology here: in a shell I would refer to integration points (IP) through-thickness, but in a planar element I would refer to either gauss or integration points. Again, I would say that the most accurate stresses should be obtained at the INTEGRATION points (which I also rather lazily call GAUSS points, and you're quite right in picking up on that). However, if you have a single rectangular/square element with a single integration point - located at the centroid of the element - then in this instance you would expect the stresses to be as accurate as the IP stresses. It all depends on the element IP structure. For a multi-IP element, the centroidal stresses I'm sure will be an extrapolation of the IP values, hence not as accurate.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[gwolf]

Most accurate stresses are from gauss points, sometimes called integration points. This is because as Drej says they have not been extrapolated to nodes.

All that Drej says is completely correct.