Mesh is correct on one computer and completely different on another?

[wakka3d]

Hello!

I'm currently teaching my students how to use and understand Inventor 2013+2014 Stress Analysis. One of my students is getting a totally different mesh result (Different amount of nodes and elements), than myself and his other classmates. To describe the situation more specific, we're testing a steel tube with the dimentions 75x75x7,5mm, with an extrude of 1500 mm.

As you can see in the pictures below (Named Teacher Results and Student Result), the students results is nearly 2 times as much as me (the teacher).

When we look at the mesh settings, both settings are appearently set to the same values:

Average Element Size = 0,100

Average Element Size in Shells = 0,050

Minimum Element Size = 0,200

Grading Factor = 1,500

Maximum Turn Angle = 60°

(Also see this in the attached file "Meshing Defaults")

Am I completely missing something here? Is it just a simple thing I need to edit?

Hope I can get some help in this. The only solution I did with my student, was him changing computers, but he also have this problem on his computer at home.

Regards,

Peter

[wakka3d]

@ Cieslak.maciek - Please let him work the way he wants to. No need to have a somewhat rude undertone here. I appreciate both of your comments and feedback, as this issue kinda holds my student back from doing his assignments.

@ JDMather - Okay, so the student and me have ran through the files on his computer, done the meshing in the FEA, and correctly enough, it shows nearly the same results as on my own computer, and NOT the 2-3 times higher amounts with the before example. I'm attaching the pictures, as proof for that. 🙂

I feel that we're getting nearer and nearer some sort of solution, so I'll eagerly await what's coming next. 🙂

[JDMather]

---

@wakka3d wrote:

Another odd thing, is that all the students computers are a ghost of mine. So that -should- rule out, that Inventor has been installed differently on my students computers. 

---

OK, in the previous test we saw that the "same" geometry can return different results.  So we demonstrated how different creation methods of even a sketch can effect the results.  That is what we have experimentally established - (I will attempt an explanation of why in a response to follow). In response #6 I established that the sketches for your part examples were not identical.  Now although we have established different results from "same" geometry - we should expect the results to be within the "same ballpark", an insignificant difference.  You are reporting a significant difference that as yet is unexplained.

Lets go through the other facts that have been established.

In my lab our computers are also ghosted images.  But the iProperties of the files you attached indicate that your installs are not the same.

One of the files was created (and last saved) in r2013 and one was created r2013 SP2.

So this establishes another unreported difference.

Differences

1. In response 4 I showed that you were not testing the same file on different machines.

2. In response 6 I showed that the geometry was not created identically.

3. And now a third difference - not the same SP of the software.

But I don't think 1,2 or 3 is explains the significant difference you are seeing (on your files - note that you did get equivalent (in fact identical) results on my files)).  There is still at least one more difference that we have not identified. I have a suspect in mind, but I want to do some testing myself to see if I can reproduce the behavior that my suspect would cause this significant difference.

I can tell you, that in every case I have tracked down with my students getting a significant difference in results - there was a reason.

The issue of getting "insignificant differences" on "same" geometry deserves more discussion (to follow) as this leads users to not trust FEA analysis. 

Back in a while.

[JDMather]

Can you post screen shot of Solver tab in Stress Analysis Settings from both machines?

[cieslak.maciek]

To be honest, I am a little surprised, that Inventor FEA when meshing takes sketch differences into account. I thought it would be a little bit better designed.

At the same time I disagree with your statement "to not trust FEA analysis". Inventor has very poor FEA, good for teaching basics, worthless in real life analyses. FEA is not idiotproof. You put in crappy boundry conditions, it will spit out crappy results, with nice colors on top of it

I don't know how it's done in the US, but on my University you can't take FEA classes unless you completed prerequisites such as: 2 semesters of Mechanics, 2 semesters of Strength of Materials, Differential Equations, 2 semesters of Material Science and 2 semesters of Physics.

PS @JDMather - Sorry for being impatient, sometimes I'm a bit hot-headed.

[JDMather]

---

@cieslak.maciek wrote:
 Inventor has very poor FEA, good for teaching basics, worthless in real life analyses.  

---

We run the exact same FEA labs in 4 different software programs and get equivalent results in all 4?
Many of the labs we do are based on classical formulas (results match our hand calculations).

I guess we don't know what we are doing.  Just enough to be dangerous. (or make pretty pictures)

[wakka3d]

@ JDMather - Very important thing you've found out here. I didn't know where I could find it out, if the computers had SP2 or not, but I can see from all the computers I've looked through at the school, only 3 of them had SP2 (And my student used one of them). This claims that not all of the computers are ghosted (or it has been possible on some of the computers to install the SP2). I'll notify the IT-support on the school of this, and tell them that their ghost has been busted, so to speak. 😛

So, thanks for finding this vital problem! 😄

Now that you've found out this problem, I assume that you then know how to solve the main problem, so we can somehow configure the "SP2" versions to get same mesh.

I'll get to the Stress Analysis Solver Settings tommorow, as I'm not at work right now. 🙂

[ReACT]

This has been an interesting discussion, so far. I'm hoping you can also explain the difference in FEA results between a simple part, and that same single part inside an assembly file.

I was playing with a simple, rectangular cross-section beam, of isotropic material (steel), fixed at one end, and pressure applied at the other. The results I see when looking at this object as a part are slightly different than when the same part is placed in an assembly file, and the same contraints and forces are applied. I have played with the "h Refinements" and the "Stop Criteria" of the FEA solver, and can see the results converge in a logical way, but they never seem to match, part versus assembly.

[JDMather]

---

@wakka3d wrote:
...., so we can somehow configure the "SP2" versions to get same mesh.

I'll get to the Stress Analysis Solver Settings tommorow, as I'm not at work right now. 🙂

---

I would be surprised if the Service Pack makes any difference.
I think it is something else (running Convergence?).

[JDMather]

---

@Anonymous wrote:

... FEA results between a simple part, and that same single part inside an assembly file.

... and pressure applied at the other.

and can see the results converge

---

You should have started a new thread and attached your dataset.

Pressure?

I didn't even go into the topic of Convergence on this yet (especially on my example files)).

[ReACT]

Sorry, JD

Please continue your valuable explanations.

I just mentioned what I believe to be a similar, but much simpler observation of the meshing behaviour problem: the same part on the same computer meshes and gives slightly different results (more so for stress than for displacement), depending on whether the part is analyzed from the part file or from an assembly file. This can be easily seen with any simple problem - basic constraint and basic pressure or force. I'm sure your thorough explanations will make sense of it all, without having to bother with a new thread.

[TheHangman]

Guys to be honest, I can't see wood for the tree in these comments.

I read although every single comment here and with the variation of machine to machine and software version to version it obviouly could be expected. 

Although, I run a simple part several times in the same computer and the same software and same preset simulation, and most of the time the results vary. So why is this phenomenon happening, one of the true thought that may even even sound stupid, but I though about the tolerance change upon each evaluation, But I am not sure too.   

Is inventor super inteligent that behind the scenes calculate the air temperature based on the time of the day and recalculate the material properties 🙂 or what? Come on Autodesk enlight us here... 

[TheHangman]

Amending my previous post, after I done my homework,

Understanding this phenomenon 🙂 it is something that the software is programmed to automatically process, the variation occur upon update of the mesh position differ from element/mesh size, resulting in slight or bad update of nodes and consequently its elements.

For instance if a simulation is save, the software generates some assets related to the simulation into one specific folder (Same Part Name > AIP > FEA> Simulation related files) Now to share the file and its FEA assets, the part file will have the simulation pre-set within the system already, right. If the Simulation runs again, then results are very likely to be the same as previously system until the mesh is updated.

By all means, if the mesh is updated independent of computer, software version and so forth, the results are likely to change even if the parametre set is not altered. The reason is, that the software tries to optimise the position of the elements in relation to the area of the part, and may not be visible in the eyes. Thus:

3 nodes = 1 Element > Few Element “Triangles” = Mesh.   

Hope this helps and gives a better understanding 

[kmeldfreyssinet]

@ReACT

Hi,

Please read here stress analysis warning.

This is my old thread about stress analysis and exactly the same thing - different results in assembly and part.

there is an explanation from Autodesk.

Cris.