Simulate Early, Simulate Often... In Rhino
A simple test with my cylinder--2
The same model,with the same conditions.
When you clip the resolution button,get more elements.
The solutions are different,sometimes the differences may be 'large',
It depends on the model.We have to try more ‘resolution’ to see the displacement converged.
Here are my conditions;
Material:Steel, Stainless (ferritic)
Bcs:the inside circular surface 123fixed,
Loads:10Mpa pressure on the left outside surface.
elements=10,000
elements=100,000
Then i tried&got;
a simple sample compared to see the convergence of the max displacement;
The ‘resolution’ button is really mean it,but i think that we may use the element size to control the resolutions,
Can we?
In this sample the 'element size' transferred
from
| element size | {2.22434,2.22434,2.22434} |
| Computation Box | |
|---|---|
| min. corner | {-45.6035,-45.2646,-16.6861} |
| max. corner | {45.5945,65.9525,-1.11575} |
to
| lement size | {0.941332,0.941332,0.941332} |
| Computation Box | |
|---|---|
| min. corner | {-45.1819,-45.1981,-16.4237} |
| max. corner | {45.186,65.8791,-1.36239} |
But,the 'Computation Box' was changed(very small), too. Why?
More test coming ...
Views: 217
Replies to This Discussion
-
Permalink Reply by Vadim Shapiro on
-
Richard, thank you for posting this example and your study of the solution. It is a perfect example of how to validate Scan&Solve (or any other FEA) solutions. Your table and graph suggest that the max displacement solution converged to a value around 3.78E-03 and the max stress fluctuates in the range 62-65 MP.
Normally, it is much easier to get convergence in displacement values, because stress varies so rapidly
in the neighborhood of stress concentrations point. But looks like in your case, both displacement and stresses converged. Well done!
To answer your question about the "computation box" -- it is always an integer multiple of the element size, because our "mesh" always covers the geometric shape with identical elements. So if you change the element size (or resolution), the computation box is adjusted too. Hope this answers your question.
-
Permalink Reply by Cristian Luca Atzori on
-
I found Richard's example to be very clear showing the meaning of "convergence".
It would also be a natural evolution, in SnS, to allow an automatic reiteration of the solution at different resolutions in order to calculate the convergence, or at least to show the results in a table/graph.
Maybe a simple button to choose from the normal and the "accurate" solution?
Thank you
-
Permalink Reply by Richard.L on
-
Thank you ,Cristian！
Your idea is nice, i think the designer or engineer may be very familiar with its own models&conditions.He can find a suitable resolutions, by solving several models.
But it's not easy to the S&S . The size&shap of the models may vary greatly！Optimization of numerical algorithms must be used to choose a suitable resolutions quickly.
Of course it‘s very convenient for newer.
-
-
If we only could launch a multi-job and see the results of a complete set of solutions (at 10'000, 20'000....100'000) maybe we could also discard the results with the max deviation.
Actually I already do that, solving at different detail levels....as you know nobody can take for good a single result, even if it's at the max resolution.
-
-
Right! I think so.
I’d like discuss with you!
1st,as one case ,S&S do not check the convergence automaticly, we should do it manually.
2nd i think the max resolution is not necessary every time,for saving computing resources！
3rd ,also the max resolution(now is 100,000 elements) is not enough.It's just ower beta release.Far more is needed in many cases.
I'd like to have the functionality of batch tasks input,not only the set of solutions as you mentioned, but also the set of load-locations(10kN,15kN,20kN...)，the different restrictions，a series of the product...etc.
If multi-job is allowed to launched , multi-result is also allowed to take put into one sheet,
we can write a .txt or a .bat file，submit it when leaving out,then we can get the results to discuss the next morning.
May be we could design a plus-in for S&S ,:)
Btw,
How do you choose the set of solutions by now ,at 10'000, 20'000....100'000?
Would you like sharing with us?
© 2024 Created by Michael Freytag.
Powered by
