Asymmetric solution in thermal stress analysis

paulz168

Hello,

I did a thermal stress analysis on a square with center at (0,0), and expected the square will expand symmetrically in all directions. However the solution gave a square that gets rotated and displaced from (0,0).
The Mecway file is attached.
Any help is greately appreciated.

Thanks
Paul

Victor

This is rigid body motion because it's not constrained. The solver tries to find any static solution but here there are many possible solutions for different positions and orientations and it just picks one randomly, which is not unreasonable because it doesn't know that you want the center to be fixed.

I've added constraints for all 6 rigid body degrees of freedom so now it doesn't move or rotate.

paulz168

Thank you for the explanation!

disla

Good afternoon,

I would like to use this example to make a question that has been around my head for some time. Excuse me if it is not strictly related with Mecway operation.

I have read that shell/membrane elements are good elements to discretize objects in which the thickness is “much more” small than the other two dimensions.
By other hand, remeshing is normally recommended to check mesh convergence and provides more accuracy but…. ¿Aren’t we losing this good relationship between proportions on each individual shell element?.
In paulz168 plate, individual shell elements are .125 mm side x .1mm thickness. ¿Is this a valid shell element to mesh the model ?
Is it inappropriate to excessively remesh models build with shell/membrane elements

Thanks in advance.

Victor

Interesting question that puzzled me a while ago too. The thickness to size ratio refers to the size of the overall part, not the element. So refining a shell mesh doesn't change that ratio.

Be aware though, that the usefulness of shells is usually not because they're more accurate but that their worse accuracy (vs solids) is less objectionable the thinner they are. So if it's easier to model with shells, you can get away with using them to save modelling time. There might be some saving of solver time too but it's not clearly significant.