2D stress plot directly from computed stresses at gauss points

[femfunm]

Is it possible to create a 2D stress plot directly from computed stresses at gauss points in each finite element? Without extrapolating and avaraging techniques to corresponding nodes!

 

[johnbridge231]

How would you calculate then the stresses at locations different from Gauss points? Extra/Inter polation techniques are needed in my opinion.

Analysis and Design of arbitrary cross sections
Reinforcement design to all major codes
Moment Curvature analysis

http://www.engissol.com/cross-section-analysis-design.html

 

[jhardy1]

It depends on the capabilities of your post-processing software.

For example, Strand7 gives the following contouring options:

. "Average - Extrapolate to Nodes" - yields the smoothest contours, by calculating and plotting average node values, using the adjacent gauss point values of all connected elements

. "Never Average - Extrapolate to Nodes" - extrapolates from the gauss points to the corner nodes of each element, but doesn't calculate average node values, so you get small "jumps" in the contours at element edges

. "Never Average - Don't Extrapolate to Nodes" - plots a single value for each element, being the average value at the centroid of each element - yields a "blocky" contour plot in areas with significant gradients

http://julianh72.blogspot.com

 

[flash3780]

As was mentioned, different programs handle post processing differently. ANSYS can display an elementwise solution, which is not averaged to the nodes. It's good practice to compare your elementwise solution to your nodal averaged solution to ensure that you have enough mesh refinement.

It's probably possible to output stress results at gauss points in ANSYS, but I haven't had a need to.

 

[femfunm]

Thanks guys for all your answers...

Lets assume now that we follow "Average - Extrapolate to Nodes". As jhardy1 said, this rule yields to the smoothest contours, by calculating and plotting average node values, using the adjacent gauss point values of all connected elements.

Right! But I wonder if the weight for this averaging rule must be equal to 1.0 for all elements connected to this specific node that we want to find the average stress. Is it scientifically acceptable or is a rough approximation for our purposes?

 

[jhardy1]

Right! But I wonder if the weight for this averaging rule must be equal to 1.0 for all elements connected to this specific node that we want to find the average stress. Is it scientifically acceptable or is a rough approximation for our purposes?

Again, it depends on the plotting algorithm used by the software. I believe Strand7 just uses the arithmetic mean of the adjacent gauss point stresses for all connected elements to determine the average node stress, but you could argue that the average node stresses should actually be calculated using weighted averages in proportion to the tributary areas of the respective gauss points.

What is more important, though, is to look at things like mesh quality and convergence. If you are seeing big differences in the adjacent tributary gauss point stresses, your mesh quality / convergence is probably not good enough to be worrying too much about the node-averaging algorithm which is used to plot the stresses. If you have a nice converged mesh, all adjacent gauss point stresses should be very similar in magnitude anyway, so it shouldn't matter whether you are using simple arithmetic averages or weighed averages to plot the stress results.

http://julianh72.blogspot.com