mesh change resulted in considerable stress change in ansys

[amirreza]

HI
I have a problem in ansys that made me confused...I have modeled a conceret dam and meshed it with solid 45 elements...after that I have assigned the available surface pressure on the dam to the model...in another modeling, I changed the mesh size only and analyzed the the model with the same loading conditions...but the results have been changed...this change is considerable...for example, when the number of the elements increases 2 time, the stress s values increase 4-5 times...!!!! why the results are depended on the mesh size in this situation? is there any relation with surface pressure loading and mesh number...?! what should i do...? with kind regards...

[tgman2]

I will at least give a semi serious answer:

Generally more elements leads to a more accurate model. FEA software (ansys) calculates the forces and displacements at each element in the model. It just assumes everything is constant between the elements. The downside to increasing the number of elements and as a result calculations is that the time required to process your results increases very quickly. The way I normally think about it is imagine a space frame construction and you have the calcs done at each joint, anywhere with complex geometry will require more joints to fit the shape and anywhere where a force is applied will require members to react that force. From this look at you model and work out where you need lots of elements and as a result areas where only a few elements are required.

While the number of elements you should use varies greatly 45 is far far to low, throw away your results. Tens of thousands are generally required to make a decent model. Up to a million if you are doing something very complex. Increase the number of elements until the difference between sets of results is only a small percentage.

EDIT: Put some stuff in that I thought I already had

[Rex Chan]

I haven't done much FEA for FSAE, but I'll just add that only make the mesh more detailed in areas that have more complex design features (i.e. changes in geometry). In Solidworks, its called mesh control.

In response to your situation, add more elements where you get the most differences. On the other hand, I've seen cases where the actual mesh design influences the results greatly. It's a very complex area, so I'd find more info from the maker of Ansys.

[billywight]

I believe he's using SOLID45 element formulation, not 45 elements. Are you seeing peak stresses increase between the two models, or stress overall increase? Do an iso-clipping of the stress contour of both and compare. If you see small areas of the larger stress, then you are probaly seeing the effect of singularities and decreasing mesh size near them. If you stress differences throughout the model (at every element), it is more than likely the pressure distribution that is incorrectly applied.

[Pico]

Adding to the above, before running your model, you need to find out if your mesh size gives you a converging result. You run a benchmark model (simplified version) while decreasing the mesh size, under the same load and see if your stress/strain results converge. If they don't converge you either have to rethink your choice of element or solution type for modelling the structure or your boundary conditions. I assume you're doing a linear model?
On the other hand you should make a hand calculation to check what range of results you're looking for; you can never go wrong with good pencil and paper checks.

[Vaibhav Kumar]

@ amir
Your question is "How small do i make the elements before i can trust the results?"

The answer is convergence test. As Pico said, you should make a simplified version of your model.

Mesh it with element that you believe is most suitable. refer element library and read the description.

solve and note the results like max stress, max strain etc. You can also take a screenshot of contour plot with legends.

reduce the element size and run the simulation again. see the same things again. results must have changed.

reduce the size again and check results. repeat this till you reach the element size where results do not change significantly even after reducing element size. This is the size that you must use for your analysis.

Vaibhav
Formula Manipal Alumni