Discretization Error from ANSYS, Equalvalent with Simulation?

[Anonymous]

Hello,

I'm working on a collaborative project in which one group is running FEA with ANSYS and I'm running FEA with Simulation Mechanical 2014. The results will be put into a DOE with the expectation of developing a transfer function that will be used in programming designs. The stress and strain results of particular locations in the model are of interest. The group running ANSYS is providing results and using the discretization error (PRERR) approach to identify nodes that represent the stress results in the areas of interest. More information on the discretization error approach is found at http://www.ansys.com/staticassets/ANSYS/staticassets/resourcelibrary/confpaper/2004-Int-ANSYS-Conf-5... Is there an equivalent method that can be used with Simulation Mechanical 2014?

My first thought was to use the "Smooth Results" view and grab a point at a similar distance from a contact; however, I do not have any number describing if that point has a similar discretization error. I do not see anything that stands out as an error analysis in the three Results tabs. The alternative to ensure similar results would be to force node to node connection in the ANSYS model at the lines of contact between parts. This method would increase their computation time which is not preferred.

Thank you,

Jeramie

[KubliJ]

Hi Jeramie,

If I understand correctly the discretization error of that article, the error is the difference between the actual results for the node and the average results of that node.  While we do not have a direct means of reporting that, I think you can still get the values and calculate yourself.  If you select a node and inquire results, with smoothing turned on it will display the average results for that node, disable smoothing and you will get a breakdown of the results per element for that node.

For error checking we use von Mises Precision, it is similar to discretization, it looks at the change in stress from one element to another and ouptus a ratio.  The higher the ratio the more dramatic the change in stress.

Thanks,

James

[Anonymous]

Hello James,

Thank you for the feedback.  It looks like the von Mises Precision is very similar to discretization process used in ANSYS.  If the group using ANSYS has a limit of 10% using their discretization approach, and because the ANSYS process uses a percent of 100 for E, the it would seem logical to use a limit of .05 for the von Mises Precision because it uses a percent of .5.  This logical approach would hold true only if the von Mises Precision process is equal to or very similar to the ANSYS  PRERR process.  Do you have a resource that has compared the similitude through a statistical process?

I noticed the von Mises Precision description is under the Linear Results portion of help.  Will this process provide reliable results for non-linear analysis?  I am running these models in non-linear to account for the thin body geometry that is present.

Thank you,

Jeramie

[KubliJ]

Hi Jeramie,

Unfortunately we do not have any comparison documentation regarding precision vs discretization.

The von Mises precision should work for both non-linear and linear.

Thanks,

James