Hi all
I have carried out two different studies of statics on the same set.
One where I simplify the part using symmetry planes and the other where I don't simplify the model.
With the same conditions and the same mesh, I get a difference of about 10% on the maximum VON MISES stress.
Shouldn't simplification only affect the calculation time and not the results?
Thanks in advance
That's the complete and simplified package. In fact, as the whole thing is symmetrical, I kept only a quarter to reduce the calculation time.
No no problem with the symmetry planes when meshing and executing the solver, that's why I found it strange. I even double-checked that all the symmetry faces were selected and I hadn't forgotten any.
I'll do the test with a single plane of symmetry tomorrow I think
The calculation must also be with a load respecting the symmetry, (so here either axial or radial).
A common omission is indeed to have one of the symmetry surfaces that is not selected (but since you have checked it, it is not that).
It happens quite frequently that local values are "outliers" on the symmetry planes (also related to the local detail of the mesh on the plane). In your case, the simulation is to be analyzed only in the central part of the 1/4 of the room: there is a chance that the 2 models will give very similar results.
Also try the curvature mesh option, this often results in cleaner meshes for revolved parts (the base mesher often does anything when it arrives on the verticaus and horizontal axes of the part).
Indeed, you have to look at where the gap is to see if it is annoying.
On the other hand, 10% difference may seem like a lot, but how much does it represent in absolute terms? Between 10 and 11 MPa there is a 10% difference but we can't say that it's significant.
The force is also applied symmetrically on its face so it shouldn't come from there.
It's true that by targeting the central area there is less difference and as you say there are some aberrant points in the planes of symmetry, the error must probably come from here. Thank you!
I had already used a curvature-based mesh for this study. It is true that for the pieces of revolution, the results immediately seem more coherent and the meshing is above all easier.
And last question, what is the difference between the different solvers that can be used? Direct, FFEPlus, Direct of voluminous problem...