How to exclude or workaround the points of concentration of stress?

[trezvy]

I'm trying to calculate the maximum stress occuring in a tube exposed to the bending load.
After making the calculations, I've obtained von Mises Stress, Max Shear stress and Max Principle stress give me a very different pictures of stress (and also different max values).
Max Principle stress picture shows the maximum stress at the botton side of the tube, as it is expected to be under the load.
While von Mises Stress and Max Shear stress give me the little points of concentration at the top side of the tube.
In the real construction the plate shown at the top side of the tube will be a thick (~4 mm) solid steel plate, which I suppose will not produce the little points of concentration of stress.
But in the model I have to replace the solid plate with the shell, simulated with "Facebinder" tool.
I suppose, the little points of concentration of stress (shown on the upper side of the tube) hides (or "overrides") the real stress which occurs at the bottom side of the tube.
I'm interested in calculating max stress at the bottom side of the tube, not at the top side.
So, how to exclude them?
May be I can exclude the thick plate from the model, but after that I don't know how to apply the load at the exact area of the tube.

FreeCAD v0.19.2, Gmesh v4.9.2.

[NewJoker]

It seems that your model consists of only shell elements (you can export and check the input file to see this).

To hide these stress concentrations so they don’t hinder viewing values in other locations, you can adjust the legend’s range. Add post-processing pipeline to the results, select the stress field and double click the legend to modify its range. In this case, you should specify a lower maximum value to cut off the high stresses.

[trezvy]

I was able to visualize von Mises stress at the bottom of the pipe by adjusting the legend.
But I suspect there is something wrong in my model. I was expected that the top rectangular plate (modeled as a shell, as I don't know how to do FEM on a model containing both solids and shells) will distribute the load evenly along the tube. But there is stress concentrations at the corners of the plate. How can I modify the model to avoid this kind of defects in the future?

[thschrader]

My 2 cent:
file with cleared mesh

plain_tube_clean_ts.FCStd

(15.98 KiB) Downloaded 65 times

shell_thickness_partial.JPG (88.97 KiB) Viewed 1844 times

[trezvy]

After correcting the displacement constraints it gives the more realistic stress picture, as the stress at the top have become more evenly distributed.