Von Mises Precision - Problem even on tutorial files!

[scudelari]

Hello everyone,

I am trying to use Autodesk Simulation to simulate the stuff we require and I simply can't seem to do it.

I went throught the simulation in this tutorial, the Bolted Flange Static Stress Analysis. The resutls are the same as shown at the end of the tutorial.

It is the next picture that worries me:

The nodes, close to the bolt, are presenting a very high imprecision (around 30%) rendering the results useless. And on top of that there is this weird peak on one single node.  To quote the help of the application:

--------

There is no guarantee that actual stresses will fall within the indicated range. The range in which they fall depends on how well the model is meshed, how well the actual part is modeled, and other factors. You should not consider differences smaller than the indicated precision significant. For example, if the precision is .1 and the maximum stress is 200, then differences of less than 20 stress units are not significant.

-----------

I am having the same kind of results in my models and I cannot seem to make this work properly. I also have in my models, since they are also very thin layers. I also have spikes of imprecision appearing randomly....

The questions are:

1- Are the results presented trustworthy even if the precision is so bad?

2- How can I improve the precision?

3- Why is the precision so bad in a tutorial file?! Are the results obtained in the tutorial totally useless? If so, I suggest changind this tutorial so that people wouldn't trust wrong results in their engineering!

Thank you so much for your assistance, guys. This forum is indeed fantastic. 

[S.LI · Accepted solution]

The essential part of your question is about the error of FEA analysis.

There are lots of places introducing errors into FEA analysis/results, such as modeling, meshing, numerical calculation, etc.

Focusing on this model, several truss/beam elements are used to simulate bolts. This way is simple and cheap. If you only care about results far from bolts, I think this model is good enough. If the local region is really your points, the modeling setting here for bolts might not be accurate enough. The stress precision shows us this in some way. You need other accurate/reliable way to simulate bolts. You have to zoom into the local region to provide more accurate description for material, loads and constrains over there. Then the modeling error would drop down.

 Also, you can have finer mesh around the bolts to reduce errors by meshing.