If refining the element size, are displacements change or not?

[suryamadduri]

If refining the element size, are displacements change or not? if not, what about the stress gradinet?

 

[cloa]

Try the FEA subforum and clarify what you want say- quite vague.

 

[Ron]

Agree with cloa, but consider the following:

When element are refined, the size of the element is usually made smaller. When that happens, the element displacement will change; however, the overall displacement will not change significantly; although it might be a bit more accurate depending on the algorithms used in the computation.

The same applies to the stress gradient; assuming you are in the elastic range of the material.

 

[Cockroach]

I have had this sme observation, similar toRon's comment. Refining the element refines the displacement. You can see this from Hooke's Law, F=kd, k=stiffness matrix. Since F=P\A, stress over area, then changing the stiffness matrix also influences displacement and stress.

Changing means refining, in the sense that stable models should exponentially move towards a limit. This is one of the checks for stability, a small change in element size should not yield wild swings in stress nor displacement. What you are doing is akin to adjusting an answer based upon significant digits.

But this should be a question in the FEA forum. You'd probably get better discussions.

Kenneth J Hueston, PEng
Principal
Sturni-Hueston Engineering Inc
Edmonton, Alberta Canada

 

[suryamadduri]

thanks RON and cockroach for Ur valuable answers.
But i have done experiment by changing the element size, i got much change in stress, negligible change in displacement.
but over thumb rule is stress is directly proportional to strain with E linear. can u clarify this two and concluded to one.

 

[Cockroach]

As all as you are doing is increasing the number of significant digits, if you will, by refining element size. Hooke's Law, yes, stress is linearly proportional to strain in the elastic zone. But I'm not sure of your question here.

Kenneth J Hueston, PEng
Principal
Sturni-Hueston Engineering Inc
Edmonton, Alberta Canada

 

[TERIO]

To get a useful answer to your question you need to provide more information, e.g.:
- geometry of what you are modeling
- what type of elements are you using
- where are the stresses and displacements you are looking at
- etc.

 

[GregLocock]

Typically if a automotive FEA engineer wants to know about stresses he uses about 10x the linear mesh density, compared with what is required for static displacements or modal analysis.

The reason is that the red bits are important for stress, but don't contribute much to the overall stiffness of the structure (for instance you could remove all fillets from a casting and the stiffness would hardly change, but the stresses would rise enormously).



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Gymmeh]

I would like to help but as TERIO said, we need more information.

A good start: What program are you using for the FEA?

 

[MintJulep]

Well it really depends on the macro part geometry and the initial and final mesh geometry doesn't it?

In the boring areas of a model all that you accomplish with a super fine mesh is slowing down the analysis.

Conversely, if your mesh is too course in the interesting pasts you risk missing something important.

 

[TheTick]

If displacement changes noticeably, it is because your mesh was too coarse the first time.

 

[izax1]

It is mainly the stresses that are changed (increased) with a finer mesh. To some extent the displacements are also increased, but if you have big geometric discontinuities (Stress concentrations) the finer mesh density will give you a more "correct" stress value, while teh displacements are not much influenced.