[Gmsh] Stress on Gauss Points

[Jeremy Theler]

Hey Max

On Fri, 2019-02-15 at 11:01 -0500, Max Orok wrote:


> My impression going in was that the Gauss point steps were part of
> the simulation and the goal was just to plot the data. 

Well, there is a fine line here but it can be considered that the
modeling [1] is finished when the stresses (or the heat fluxes, etc)
are computed at the gauss points.

Actually, I think the original question was: "given that I have data at
the Gauss points (and not at the nodes nor at the cell centers), how
can I plot them on Gmsh?"

Indeed, I think Gaetano wanted to have Gmsh to do something like to re-
mesh the base grid and have new cells around the gauss points and plot
the data as cell-centered values. I do not know if I am clear, but for
a 8-node hexa with 2 Gauss points per dimensions that would be to
replace the original element with 8 little hexas, one in each octant
and then assign a cell value from the dataset to each of them. For
other elements it would be more complex.

This is definetely in the post-processing step of the worflow.


> I think for this data set, the shared node data does repeat, but a
> single element will vary as a function of its  nodes. 

As far as I can see, the data corresponding to the shared nodes does
not repeat. Again, this is what is expected from secondary fields
computed as derivatives of a primary field at the Gauss points.

> For any averaging etc., perhaps this could be done during the
> simulation? 

Yes, it could. This is why I am interested in this issue, and raised
the question on how to average if the solver has to output nodal values
of stresses. Weighted average? With the element's volume? With the
element's quality?



[1] https://www.seamplex.com/blog/say-modeling-not-simulation.html


--
Jeremy Theler
www.seamplex.com id="-x-evo-selection-start-marker">
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/gmsh/attachments/20190216/2c8bfe8d/attachment.html>