I have used a global tet mesh to derive cut boundary displacements to drive a hex submodel.
The model is a turbine blade, subject to uniform temperature, an angular velocity and restrained at it's root. The global model reactions agree well with the analytical solution, ie. mrw^2, where m=mass, r=radius of centre of mass, w=angular velocity in rad/sec.
I found that the reaction forces at the restraints of the submodel were not the same as the tet model by about 6%. Further investigation showed that the cut boundary reaction forces were in error by about this magnitude.
In the end I scaled the whole field by the value required to bring the force summation to what it should be. The reason I wish to do this is so that the cut boundary force field (shudders...) can then be applied to a blade-disc model without having to adjust the diplacements, or I can change material, or temperature etc.
My explanation for this inconsistency was that the inherent extra stiffness of tet elements over hex elements meant that the cut boundary displacements were lower than they should have been in reality, and therefore the reaction forces in the hex model were lower.
Brief test models of a bar in tension reveaed a 2% difference in reaction between models of hex and tet elements.
My questions to the community are:
How can I explain the extra discrepancy?
Could it be put down to mesh quality?
What further investigation should I be doing?
The mesh quality was called into question by the software vendors when I queried them. What guidelines - if any - exist for the mesh attributes of a global model, when used to drive a submodel?
Many thanks for your time,
Mark.