Static stress simulation : average stress off

[Anonymous]

Hello everyone,

Is it possible to switch off stress averaging ? I would like to see on my linear static results real calculcated stresses.

Thanks!

Regards,

Sebastian

[TheCADWhisperer]

No pictures?

No files?

A picture might be worth a thousand words.

I file might be worth a thousand pictures.

I'll try one picture first.

Is this what you are after?

 

[Anonymous]

Thanks for your answer.

No, that's not what I am looking for. I want raw solver stresses. The stresses shown on the standard Fusion 360 view are averaged. Even if it's not specified, I am sure of this because the transition between one element and the other is perfectly gradual. Raw results from solver are not gradual (even with very fine meshes), steps should be seen between and element and the one besides.

See example image enclosed. Normally FEA solvers let you switch between averaged (nicer looking) and raw stress results, I suppose (I hope at least) this should be possible on Fusion.

[Anonymous]

Just out of curiosity what is the application for using unaveraged results? I'm not sure I've seen that used much if at all...but that's going to be limited to the industries and applications I've worked with.

When you do a surface probe. Its probing the node so I'm not sure if there is averaging at the peak... meaning the results are useful. I'm curious as to the feedback on this one regardless.

[Anonymous]

Let me rephrase the question: are the linear static stresses shown on Fusion 360 Simulation averaged? If yes, is it possible to switch off averaging?

@Anonymous: peak stresses can be lowered because of the averaging, so it's a bad idea to do so when making fatigue estimations for example. Also singularity points can be smoothen out. The only good point about averaged results is that they are prettier to present, but depending of your audience this could be a problem. At least, this is the way I was taught to do simulations : always unaveraged.

Regarding probes: I suppose that if you're doing a surface probe over an averaged result, the probe should show the averaged stress. The averaging happens at the node, in order for it to be equal between neighbour elements. That's all the problem.

[TheCADWhisperer]

---

@Anonymous wrote:

... The only good point about averaged results is that they are prettier to present, but depending of your audience this could be a problem. At least, this is the way I was taught …..

---

What is the purpose of doing a simulation?

[TheCADWhisperer]

@Anonymous

Where did you go?  I have a lot of questions.

How do you validate your digital analysis set-up against the real world?

Do you do one real world test or do you gather empirical evidence from a number of real world tests?

Why Different?

[Anonymous · Accepted solution]

To close this topic, taking into account the different answers, I conclude that averaging is always on and it's impossible to switch it off (please correct me if I'm wrong). As I believe in this software package, I would like to propose this improvement for future updates. Would you know where and/or how can I do this?

@TheCADWhisperer, I'm here. Send me a private message with a concrete example and I'll try to help you out. Or just open another topic and send me the link.

Regards,

Sebastian

[TheCADWhisperer]

Here are two files.

Same geometry, identical volume.  Same analysis set-up.  Slightly different results?

(You didn't answer my question about how you validate your model (In any analysis software)?)

[Anonymous]

@TheCADWhisperer: I didn't answer because your question is out of subject. But as you insist, I'll answer you on this topic. The difference on the result is explained by the meshes being slightly different. Even if the mesh settings are identical, the software calculated two different meshes for the same part. I invite you to put the two meshes side by side to see it by yourself. The two models aren't using the same sketches (even if resultant geometry is the same), maybe this can make a difference for Fusion's meshing algorithm.

By the way, I don't know why you're analysing this problem but given the unrealistic 90° angle you have a nice example of a singularity, so the max stress you'll find has no sense as convergence is impossible.

[TheCADWhisperer]

In the real world it is not possible to manufacture perfect parts or identical parts.

There will be some variation from part-to-part.

As you have explained - even a variation of the difference in technique of modeling identical geometry results in slightly different mesh.

I explain this to my FEA class as imagining they are a cave spider with no vision.

The only thing they can detect is 1 and 0.

The spider must examine the code of ones and zeros in the order in which they are encountered and build a web (mesh).  If the code encountered is not exactly the same - the web will not be exactly the same. As we have seen, if modelers take a different approach to reaching the same end goal (finished solid model) then there might be a slightly different analysis result.  The question is, is this difference statistically significant.

Now we go back to the real world and much more variability of material/manufacture enter the equation.

I think in the computer world we have a tendency to lean towards treating problems as pure 2+2=4 math problems.

Averaging results more closely represents the real world than 2+2=4 statements.

If you validate your digital model against real world empirical evidence - my guess is that you will plot variable results and average.

[Anonymous]

I see, you asked me the question in order to explain me something about FEA and statistics... nice, thanks for your contribution to the answering of the topic's question. I don't understand why you don't just answer me directly can we do it? yes or no and how. Instead you're trying to make me change my mind about averaging. Moreover, the fact that I don't want averaging on doesn't means that I don't understand that the computer simulation is an approximation...

It's nothing to do with who you are or who I am. I am trying to find a solution and you're just making me loose time. Thanks for your effort, you won!