volumetric elements or plates ?

[Innovec]

I am looking for published references on mesh selection and error related to improper mesh.

I currently work on an excavator chassis and parts made of 8 to 20 mm steel plates welded one to the other.

My client has previously made some analysis using thetrahedral elements with only one element through the plate thickness. Tipically those analysis where made with a limited FEA design tools added to a CAD package.

To my opinion this method can lead to incorrect results. I prefer to use plate elements and to analyse the welding with the plate stresses output and the linear forces output.

It takes more time to create the mesh with plate because we have to intersect mid planes (I use Femap modeler tools to do this).

I read a note on this forum from Bradh suggesting a minimum of 4 solid elements thru a plate thickness to have some good results. I tend to aggree with Bradh. However in the case of a chassis such a fine solid model would be quite oversized.

Please let me know if you have some experiences on the consequences of plate vs. solid element selection on similar cases.

Regards

Robertcossette@wanadoo.fr










Robert Cossette

 

[tld23]

Dear Innovec
NAFEMS (

http://www.NAFEMS.org)

have published several articles on meshing problems, the last one was in the July 200 BENCHmark. I also have a book that gives quite a few details on meshing pitfalls - but it is at home and I cannot remember the title - I will find this out and then repost.

Although it depends on where they are and how they are loaded, you are right to be suspicious of the models that used Tets. There again be suspicious of all models - they are wrong until proven correct! Depending on the size of the structure and what results you want, plate elements (Shells) are probable the sensible choice. However, if you are looking for detailed stresses at the intersections (eg at welds) shells may not be the best option and bricks may have to be used.

The general rule about 4 bricks through the thickness depends on the element type and internal loading (stresses).
· For uniform stress in the section a single 8-node brick is adequate.
· For linear bending stress in the section a single 20-node brick is ok
· To capture the shear stress distribution across the section >10 20-node bricks will be needed!
You just have to use a bit of engineering judgement and a healthy dose of validation.


TERRY

 

[brad]

TERRY--You stole my thunder!!
I'm beginning to wonder whether we were separated at birth . . .

Innovec--I agree with Terry on these points. I would also recommend NAFEMS as a source for these types of things. Some of the FEA manufacturers have their own benchmarks which discuss these problems; some of them also include NAFEMS benchmarks in their discussions. Check in your user's manuals.

Brad