Mesh size

[JoseAndres24]

Just one quick question.

How do you determine mesh size? Far as I know, general rule is that element cannot be smaller than t x t (t being thickness of plate that is meshed).

Thank you!

 

[rb1957]

the elements can be as small as you want, though their accuracy may suffer. run some simple tests ... axially loaded, bending (in-plane), ...

another day in paradise, or is paradise one day closer ?

 

[mikeneal]

That is a open ended question and not easily answered. Basically the mesh has to be fine enough to capture the solution with the accuracy you require. How small is that? It depends on many things and cant be easily answered here. Obviously mesh must be better in regions where stress concentrations/constraints/loads exist. Typically I personally start large to make sure model behaves properly and refine in regions you care about until the solution stops changing. After awhile you get some knowledge based on experience.

 

[BlasMolero]

Dear Jose Andres,
The answer is in your hand: you need to prepare different studies and compare results, with double DOF, reduce the element size on each study to have more or less the double DOF. Run the stress analysis and compare stress results in different points of you FE model (for instance, use vonMises nodal stress for comparison), if the solution do not vary more than 5% (others say 8%) then the solution is "mesh undependant", ie, your solution has converged to correct stress results.

Please note convergence in displacements is achieved inmediately, but in stresses your need to perform a few iterations. Also, select points to read stresses not coincident with hot spots, for instance a corner point. The reason: the local stress concentration at this point is a stress singularity, so if you continue refining mesh the stress will never converge, and then this point is not valid for assessment of stress analysis. This is because the theoretical elastic stress at the corner is infinite.

Best regards,
Blas.

~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director

IBERISA
48004 BILBAO (SPAIN)
WEB:

http://www.iberisa.com

Blog de FEMAP & NX Nastran:

http://iberisa.wordpress.com/