this is from FreeCAD forums were i’m interested also, frequently i used in designing of concrete slab on ground (dowel lines).
i try reproduced their problem with simplification of models. a simple supported continuous beam with uniform loads, hinged at mid spans of the right beams. without MPC this model is unstable and lead to rigid body motions.
offset node coordinates of short beams i made for visually clarification and it be overwrites by keyword during calculation in my models. as can be seen in result below, stress in longitudinal direction of beams are zero at the point since rotation/moment is released.
actually this is a simple case to be solved in CalculiX CCX, however the user required to be defined manually of MPC for each of nodes.
may this task can be simplified trough GUI by definition of each group nodes to be connected by MPC (translation, rotation, or both) using coincident or nearest by distance (tolerances).
In order to enable relative movement between two mesh parts, the geometry must be split into two parts or created as two separate parts. Then the mesh might not be coincident after meshing and could be very different in size. Then a less experienced user would have many problems obtaining a correct solution. Maybe an edge on a combined, single geometry part could be selected and defined as a hinge. Then automatic MPC creation would probably work better since the meshes would coincide. I only wonder, how robust are MPC connections in combination with shell elements in CalculiX.
thanks for noticing me about the problem of MPC implementation due to different node of element since unstructured mesh may exist.
regarding to the robustness in application, it seems to be general approach to connect two parts. each 3 translations and 3 rotations can be set separately, however this method is required the nodes to be coincident or duplicated (zero length).
may this is not a must in CCX, as many FE with the same capabilities. i try reproduced this problem using manually defined MPC (6dofs is constrained) based on nearest distance. it’s run, solved and display result as expected.
try to understand these workflow, it seems the best approach since the master and slaved nodes will guarantee to be coincident or duplicated nodes(zero length).
on my second examples, the model proposed is only an approach due to some offset distance. actually the left side group of nodes should share the force transfer also. this can be eliminates for a model with part merged/combined.
