Static Analysis of Beam

[mikefenomeno]

Hi I just started using CosmosWorks. I have a question for whoever is good in meshing and cosmos tricks.
I am studying the static analysis of a cantilever beam (rectangular section) fixed on one edge to the wall (Immovible REstraint) and on the other edge a load of 10KN is applied. When i get the von misses stresses, normal stressses and shear stresses the results are different from the theoretical values but not too much difference (Slight error) and it also gives me very high stresses (singularitites on the fixed edges).
But if I split the beam using three plans going the neutral axes (view attached pic) the results are very similar to the theoretical results.
Can somebody please explain to me the effect of splitting on the stress results. what does it do with the meshing? it obviously improve it but how?
Regards

 

[jte]

First, I will assume that you have recognized that the behavior of this cantilever beam probably is predominantly governed by bending, but due to its height:length ratio it will also have some influence from shear.

Second: Ignore stresses at first. First look at displacements. If your displacements are good, then your stresses have a chance. Remember, FEA determines stresses only as a function of strain which is derived from displacement. So how do the displacement values compare with theoretical?

With the split plane on the neutral axis, it will force Cosmos to place nodes directly on the plane. Thus, the stresses there are not interpolated based on the element shape function. This alone may explain your issues.

But... How far off are your results? Remember, engineering is a game where "close enough" needs to be recognized as adequate, and "perfection" is often not economically viable.

Also... I presume you are using solid high order tet's for the meshing. What aspect ratio are you keeping the elements at, and have you played with the element size to see how the results vary? For what its worth, my experience with Cosmos is that their elements are very robust - not hugely sensitive to having high aspect ratios, but nonetheless, you need a reasonable mesh.

jt

 

[corus]

I imagine that by splitting the model you've increased the number of elements through the thickness. This will improve your results when you have bending stresses present.

corus

 

[ENzo36]

check your defined restraints and make sure they used as you want,