About Integration Points Stresses

[sunnyqzhang]

Sorry to bother you guys again...

Drej mentioned "...I would say that the most accurate stresses should be obtained at the INTEGRATION points..."

I have problem to understand:
1)IP are for sure used in Matrix calculation (volum/surface integration), how are they also used in stress calculation? why stresses are more accurate at IP than nodal points? Aren't they both calculated in the same way that from displacement function?

2)How IP stresses connect to nodal stresses?
If nodal stresses can be obtained by interpolation from IP stresses, there will still be stress discontinuity at the conection nodes. Which one to pick then? or average them?

Hope my questions are understandable! Thanks a lot!

-Sunny

 

[gwolf]

The most accurate stresses are at the integration points, the most easy to use stresses are at the nodes. Nodal stresses have been derived from integration point stresses and are not as accurate. However, if you want to know the stress on the surface of a structure you must use nodal stresses.

Average or not?: If your mesh is really good w.r.t the stress gradient, adjacent elements will both predict the same stress at their joining nodes. Then averaged or unaveraged values can be used with equanimity. In practice expect within 5%, if the difference is any larger you probably need a finer mesh, unless you work in civil engineering.

regards,

gwolf.

 

[sunnyqzhang]

gwolf,

I got it. Thank you very much.

 

[workharder]

There some special methods to smooth the nodal stresses(e.g. SPR method). You can look into the corresponding papers.