mesh refinements... 1

[caviac]

HI all,
I need some help on a problem that I'm not sure to explain.

I have a plate with a rib.
I put a normal force to the rib.

The finer my mesh will be, the higher the stress I have... I Thought it would stop but no....
I think it's a singularity (perfect sharp angle 90°) but not sure...

This is not converging...after several refinement (exagerated)...

Thx all

Caviac

 

[feadude]

Normal force could be causing it. Try distributed load.

 

[ahad29]

Try adding a small radius at the foot and see if the solution converges.

 

[johnhors]

Your model description is somewhat lacking.

assuming:-

1). your high stress is at the join of the rib to the plate

2). element type used is shells

3). rib is normal to the plate

3). normal force, is normal to the rib i.e. bending the rib

Then effectively the plate is seeing loading along a knife edge, this is a singularity and stress will not converge with mesh refinement.

Instead, model the structure in solids and include fillets and welds... then apply the loading as a distributed pressure.

 

[caviac]

in shell it's converging....not in solid with 90° and no fillet.
Try with a simple model you will see....
I will try with a fillet, I suppose it would be OK.
Thx
caviac

 

[pja]

If you by normal force you mean point load then of course you have a singularity......

 

[caviac]

No....

I have this kind of geometry :

_
| |
| | <- force distrib. on the surface.
| |
-------------- -------------
clamped | |clamped surface
------------------------------

A base with a rib on it...

And now you mesh it with 3D elements solid.
You solve it once for one element in the thickness for example.
You note the results.
You refine the mesh and solve it again.
...etc....

it seems the model gets high stress at the base that increase each time without converging ...

 

[johnhors]

It should (will !) converge with a fillet at the rib to plate join. I assume that your clamped (yuck !) surfaces are sufficiently far away not to interfere and that the pressure on the rib stops before the join/fillet , ie. invoke St. Venants rule.

What elements are you using? linear or parabolic ?, hex or tet ?

 

[crisb]

Have a look at this link, I think it describes your problem.

http://www.machinedesign.com/BDE/cadcam/bdecad3/bdecad3_4.html

 

[caviac]

Thanks for your responses all !
The last one is interesting and we have made the same conclusions and verify it in theory/practices on calculations.
It's OK for us.
Thx again !