linear stress simulation problems

hello, I'm studying a support element (midplane section below). On the left, a vertical load is applied and the two red elements are radial supports.
<?xml version="1.0" encoding="UTF-8"?>
Herebelow a view of half-model (symmetric with respect to the XY plane below).
<?xml version="1.0" encoding="UTF-8"?>
Calculation performed on standard fem software show smooth stress concentration in the throat section.
<?xml version="1.0" encoding="UTF-8"?>
With simsolid:
1 - first problem: the hemi-cylindrical surface is imported as a single surfaces (even if it is separated in order to apply the support restraints).
<?xml version="1.0" encoding="UTF-8"?>
2 - stress distribution is unrealistic, even worse if I set forther refinements (activate feature adaptation or increase the number of adaptive solutions). Here is the von Mises stress with default options
<?xml version="1.0" encoding="UTF-8"?>
Any idea? here annexed the model (test.zip).
Tyia
Answers
-
Couple of things:
- Wrong boundary condition. You were using elastic foundation. Better just to apply sliding on 3 surfaces
- Under converged. When looking for stress, better to use 4-passes and check on Adapt to features. See results below.
<?xml version="1.0" encoding="UTF-8"?>
Now stress are located in the slot as expected. Note: due to the fact SimSolid uses a fixed results plotting grid for contour display, stress distribution can appear more fragmented in narrow slots as compared to FEA. This is an artifact of the viz and not of the solution.
Here is one more plot with the legend adjusted close to that used in FEA.
0 -
Thank you Ken for your kind reply.
In my opinion, soft spring elastic foundation is a proper way in order not to influence excessively the stress distribution near the throat. Anyway this is due to the fact that I'm unable to import correctly the support faces, that are properly defined (in red, herebelow). Probably a simsolid limitation. I attach the geometry file (test.stp), if you want to try debugging the problem (or add a feature in simsolid to avoid collapsing surfaces when importing).
With a slider on three faces (that in my opinion is less accurate near the throat section, at least on a standard FEA calculation), 4 passes and feature refinement, I obtain only a qualitative approximation of the result, that I would not approve as a design calculation (I subsequently perform a strain-life fatigue assessment, so I need the peak stress value, or better, the peak value about 0.5 mm, where the crack enucleates). In simsolid the peak stress 516 MPa is well above the real stress (about 350 MPa).
<?xml version="1.0" encoding="UTF-8"?>
If I try to refine the solution (during postprocess) on the surface I do not obtain any better/smoother stress distribution.
If I perform another simulation, increasing the number of adaptations to 5 (herebelow), I do not obtain a smoother solution. The peak stresses in the throat are better in line with FEA but the distribution is unacceptable: the results in the throat are too much scattered, and not uniformly varying.
The artefact (within the revision cloud) is probably due to the slider condition that, in my opinion, is not a good condition in this case.
<?xml version="1.0" encoding="UTF-8"?>
My conclusion is that, at the moment, for this kind of calculations, I will better rely on a standard FEM model.
Any consideration/suggestion?
Best
Tyia
0 -
-
Hello Mario
Altair Forum User said:you can try to use a 'Spot' for a region to apply boundary conditions and loads. See attached picture.
Ok I see.
But why a soft spring is a wrong boundary condition if I'm just interested in the maximum stress in the throat? I'm a PhD mechanical engineer with more than 20 years of experience on FEA, and I wrote FEA codes myself: I can state that in standard FEA it is a perfectly acceptable condition.
Altair Forum User said:use more passes and use 'Adapt to feature'
Yes, the results shown in my post of yesterday are with 'Adapt to feature' turned on and with 4 and 5 passes, as I explicitly wrote. So you suggest more than 5 passes?
Ok, a simulation with 6 passes (apart using 3.5 GB of RAM, that is too much for that kind of solution). Stress are unacceptably scattered in the throat section. I don't see any benefit if compared to the 5 passes solution previously shown. In classical FEA there exists a concept called grid convergence. I'm not sure if it is demonstrated that increasing the passes, the linear stress simulation performed with simsolid converges to the theoretical exact solution.
<?xml version="1.0" encoding="UTF-8"?>
Best,
Tyia
0