Von Mises Precision

[Anonymous]

What level of precision should you shoot for?  Thanks.

[Anonymous]

Hi,

In my experience, 10% is a rule of thumb that I have used for convergence on other simulation softwares, which would be equivalent to 0.05 for von Mises precision.

Although, It usually depends on how much accuracy you are looking for in an analysis, and your model setup, for example, if you have areas in your model where there is an edge on surface contact - that is mathematically a singularity, you may get locally higher von Mises precision values there than the rest of the model, so you will have to use your judgement to determine how much you can trust the results.

Hope this helps.

[Anonymous]

Thanks ssenbore!  I got this response from an Autodesk support guy:

  We don’t place any criteria on what should be considered acceptable or not, however, perhaps I can clarify that it ranges from 0 to .5 where .5 is the worst.  So, it is a gauge for a node… each element calculates a stress for that node.  Theoretically, each element that shares that node (with some exceptions for part boundaries, boundary conditions, material changes and thickness changes) should calculate the same stress at that node.  The larger the precision value indicates the higher the difference the elements are reporting for that node.  How much?  Presume that at a particular node you have a precision value of .5 (typically bad) and the max von Mises in the model is 10,000.  The .5 is read as 50%, 50% of that 10,000psi or – 5,000psi… one of the elements is reporting 5,000psi different than what one of the other elements is reporting.  So, you need to consider what the maximum von Mises is to know how much the difference is and then is there any reason (explanation) for the poor precision in that area.  If there is no way to explain it (near bc, material, thickness change etc.), then it is likely due to element quality.  If the element precision is poor and there is no explanation, then it likely warrants trying a better/finer mesh.  If on the other hand, the precision shows midrange or low values (0 to, say, .25) and the von Mises stress in the model are overall very low (well within the acceptable limits), well, then does it matter whether the stress at the node is 1000psi or 1250 if your modeling criteria is simply that you remain below a yield of 36000.  Probably not.  If on the other hand, the goal of the analysis is to try and precisely find the stress result, regardless of where it lays with respect to the yield, well, then yes, I would consider a 25% increase  (or variation) in the stress a big deal.  Does that help make sense?  So, I don’t think I can answer the question directly, but I perhaps help to explain whether it should be a concern or not?