In a harmonic study of an unconstrained structure, the fist six eigenmodes will be rigid body modes having a frequency of zero (or very close to zero due to numerical errors). In attached zip-folder I have included a model of a random assembly (which looks like a ladder but it could be anything else) and in the model I have created two harmonic studies. In one study I use compatible meshing. In the other study I use incompatible meshing. Besides that, the two studies are identical. In the table below I have presented the first six eigenfrequencies of the two studies to compare. Obviously, the frequencies related to rigid body rotations (modes number 4, 5, and 6) when using incompatible meshing are not zero compared to the "zeros" of the frequencies related to the rigid body displacements (mode number 1, 2, and 3).
Rigid body mode shape
Frequency [Hz] - Compatible meshing
Frequency [Hz] - Incompatible meshing
Does anyone have an explanation for that? I do understand that a consequence of incompatible meshing is that derivatives of displacements (such as strains) will not be perfectly continuous across interfaces between parts since the continuity in displacements is obtained by means of interpolated element values. But this should just be local effects, or what? At least I did not expect that this could have such a significant global effect.
I hope that someone finds it interesting to share thoughts and comments. Any ideas are welcome.