cantilever beam v7 versus v9

[Chris74]

hi all,
I noticed a difference between 2 versions of Elmer (V7/V9) about Von Mises results at the end of a simple cantilver beam : VM = 110 in V7, 123 in V9
The displacement are identical. I wonder if you've ever noticed that.

Note : In v7, the nodes coordinates are in mm, and in meters in V9. So I applied a conversion (in particular for E, rho)

Best regards
Chris

[kevinarden]

It is not surprising that v7 and v9 are different as the code is updating daily with enhancements and bug fixes. The question is what is the correct answer? I would not trust any code for stress analyses using a mesh of linear tetrahedrons, those elements formulations just do not support accurate stress analyses. I looked at the geometry in the v9 definition and a 1 meter enforce deflection at the end of a 0.1 meter beam with a cross section of 0.003x0.001 is a highly non-linear solution, and the ElasticSolver takes into account geometric non-linear terms. It is possible to calculate what the stress should be using linear theory, but to match that in Elmer you would need to use higher order tetrahedrons or switch to hexahedrons, I would still reccomend higher order elements for stress. You would need to switch to StressSolver to match the linear case.

[kevinarden]

This is a validation of a cantilever point load at the end using stresssolver. I don't have a reference solution on hand for the non-linear case using elasticsolver.

the mesh uses higher order hexahedron elements

results.png (150.65 KiB) Viewed 1129 times

compare.zip

(508.61 KiB) Downloaded 131 times

[Chris74]

Thank you Kevin