Unusual boundary conditions in FEA software

[feynman1]

For a 2D problem with unknown displacements u(x,y) and v(x,y), is it allowed to give such a set of BCs u(0,y)=1 and vy(0,y)=0, the former being a displacement BC, the latter being a force BC (vy is the y strain)?
How is this implemented in FEA software?

 

[Baluncore]

Can you control the “degree of freedom” of nodes in your model ?

 

[feynman1]

Can you control the “degree of freedom” of nodes in your model ?

the number of DOFs is fixed in a model, =2 in a 2D plane problem.

 

[Baluncore]

the number of DOFs is fixed in a model, =2 in a 2D plane problem.

Controlling the DoF of individual nodes is supported in some FEA software.

Search in the documentation for “anisotropic” or “directional properties”.
Maybe consider a "zero friction" layer between the material and the boundary.

Rather than considering generalities, it might help if you identified the particular FEA software package you are using.

 

[feynman1]

Controlling the DoF of individual nodes is supported in some FEA software.

Search in the documentation for “anisotropic” or “directional properties”.
Maybe consider a "zero friction" layer between the material and the boundary.

Rather than considering generalities, it might help if you identified the particular FEA software package you are using.

Ansys, abaqus etc. I don't know how “anisotropic” or “directional properties”, "zero friction" have sth to do with this topic. How do you control DOFs?

 

[FEAnalyst]

In FEA software it's pretty straightforward. You can apply boundary conditions to individual or grouped nodes (only CAD-embedded FEA modules are limited to geometry-based selection) and control each DOF (3 translations for 3D solid elements or 3 translations and 3 rotations for 3D shell or beam elements). Usually, you can specify local (user-defined) coordinate systems for boundary conditions, including cylindrical systems. There are also special constraints (called SPC and MPC that allow you to define more complex relations for degrees of freedom, for example, you can make selected DOFs for two nodes equal).

 

[feynman1]

In FEA software it's pretty straightforward. You can apply boundary conditions to individual or grouped nodes (only CAD-embedded FEA modules are limited to geometry-based selection) and control each DOF (3 translations for 3D solid elements or 3 translations and 3 rotations for 3D shell or beam elements). Usually, you can specify local (user-defined) coordinate systems for boundary conditions, including cylindrical systems. There are also special constraints (called SPC and MPC that allow you to define more complex relations for degrees of freedom, for example, you can make selected DOFs for two nodes equal).

many thanks but is there a specific example or demonstration of how this works as a response to my OP?

 

[FEAnalyst]

many thanks but is there a specific example or demonstration of how this works as a response to my OP?

You would have to focus on particular software. Personally, I recommend Abaqus where features like that are very well implemented. Abaqus documentation is very comprehensive and contains multiple examples for all options, including MPCs.

 

[feynman1]

If on one same boundary, a displacement BC is given along x but a force BC is given along y, is it allowed and what will the software do?

 

[FEAnalyst]

Sure, since those are separate DOFs, it's not a problem. In fact, such situations are quite common - for example, consider a block sliding on a surface. You can push the block towards the surface (force in the vertical direction) and slide it on this surface (prescribed displacement in the horizontal direction). This load and boundary condition can be applied to the same region of the block (i.e. its top face). In fact, it wouldn't be a problem to apply force and prescribed displacement to the same region and in the same DOF but it just wouldn't make much sense. The only thing that you can't do is apply conflicting boundary conditions - for example, fix the surface in one direction and make it move by 2 mm in that direction at the same time. In such a case the solver will give you an error message regarding redundant BCs.

 

[feynman1]

Sure, since those are separate DOFs, it's not a problem. In fact, such situations are quite common - for example, consider a block sliding on a surface. You can push the block towards the surface (force in the vertical direction) and slide it on this surface (prescribed displacement in the horizontal direction). This load and boundary condition can be applied to the same region of the block (i.e. its top face). In fact, it wouldn't be a problem to apply force and prescribed displacement to the same region and in the same DOF but it just wouldn't make much sense. The only thing that you can't do is apply conflicting boundary conditions - for example, fix the surface in one direction and make it move by 2 mm in that direction at the same time. In such a case the solver will give you an error message regarding redundant BCs.

When applying conflicting BCs, namely applying a displacement and force BC along the same direction on the same boundary, the software still can compute without an error message. Does that vary from software to software? In this case, which one will be overridden, displacement or force condition?

 

[FEAnalyst]

When applying conflicting BCs, namely applying a displacement and force BC along the same direction on the same boundary, the software still can compute without an error message. Does that vary from software to software? In this case, which one will be overridden, displacement or force condition?

What I meant by conflicting constraints, was limited to boundary conditions (prescribed displacements). Loads are somewhat different in FEA. Boundary conditions influence the left side of the equation while external forces form the right side of it: $$[K] \lbrace u \rbrace=\lbrace f \rbrace$$ where: $[K]$ - stiffness matrix, $\lbrace u \rbrace$ - displacement vector (from it we get the solutions), $\lbrace f \rbrace$ - external forces vector.

 

[feynman1]

What I meant by conflicting constraints, was limited to boundary conditions (prescribed displacements). Loads are somewhat different in FEA. Boundary conditions influence the left side of the equation while external forces form the right side of it: $$[K] \lbrace u \rbrace=\lbrace f \rbrace$$ where: $[K]$ - stiffness matrix, $\lbrace u \rbrace$ - displacement vector (from it we get the solutions), $\lbrace f \rbrace$ - external forces vector.

Right! Then perhaps the final FEA result depends on the sequential order of how the software loads the displacement and force BCs, that is the former (displacement/force) load will be overridden by the latter (displacement/force), and perhaps the software won't regard this as an error. Is it possible to investigate this sequential order, while I doubt software will release this trivial detail?

 

[FEAnalyst]

Right! Then perhaps the final FEA result depends on the sequential order of how the software loads the displacement and force BCs, that is the former (displacement/force) load will be overridden by the latter (displacement/force), and perhaps the software won't regard this as an error. Is it possible to investigate this sequential order, while I doubt software will release this trivial detail?

Yes, redundant features are sometimes treated this way in FEA software. However, this does not apply to such important stuff as boundary conditions - the program does not perform the analysis and shows an error so that the user realizes that he has defined conflicting boundary conditions. Redundancy is allowed only in cases when removal of unnecessary definition (the second one, solver leaves the one that’s first in the input file) won’t have significant impact on the solution. And even then there is a warning (but not error) message in output files.