Boundary conditions: constraints and loads in FEA

When building a Finite Element Analysis (FEA) model, defining the boundary conditions is the step where the most conceptual mistakes are made. You can have a perfect CAD geometry, a well-characterised material and a refined mesh with flawless quality metrics, but if the constraints or the loads do not faithfully represent the real physics, the model will return completely wrong results.

Boundary conditions are the interface that connects the numerical simulation with the physical world: they tell the software where the structure is supported, how it interacts with neighbouring components and how the external forces are applied.

1. The mathematical problem: singular matrix and rigid body motion

To understand why constraints are mandatory, we have to go back to the underlying linear algebra. The global static equation of the FEM is written as:

Before any constraint is applied, the global stiffness matrix [K] of an unattached structure is singular: its determinant is exactly zero (det([K]) = 0) and therefore it has no inverse.

Physically, this means the system has no fixed reference point in space. If a force were applied to the unconstrained structure, it would undergo rigid body motion: it would translate or rotate through space indefinitely without any elastic deformation or internal stress.

Degrees of freedom and rigid body modes

A three-dimensional solid body in space has 6 rigid body modes:

• 3 pure translations along the x, y, z axes (T_x, T_y, T_z).
• 3 pure rotations about the x, y, z axes (R_x, R_y, R_z).

For the matrix [K] to become invertible and the solver to compute the displacements, the analyst must apply enough kinematic constraints to remove at least these 6 rigid body modes.

2. Classification of boundary conditions

Boundary conditions are formally divided into two primary mathematical categories:

1. Essential or Dirichlet conditions (primary): impose known values on the displacements or rotations at the boundary of the domain (for example, u = 0 at a fixed support).
2. Natural or Neumann conditions (secondary): impose known values on the forces, pressures or derivatives of the field variable (for example, a point load F or a distributed pressure p).

3. The art of not over-constraining: isostatic supports and the 3-2-1 rule

A common modelling mistake is over-constraining (over-stiffening). It happens when more constraints than necessary are applied, or when nodes are rigidly fixed to ground in regions that in reality should be free to deform.

For example, if all translations are locked at the base of a block subjected to bending or thermal expansion, the transverse contraction due to the Poisson effect (ε_y = −ν · ε_x) or the natural expansion of the material is prevented, introducing artificial parasitic stresses that do not exist in the real part.

The isostatic solution: the 3-2-1 rule

To avoid over-stiffening and keep the structure in static equilibrium without adding extra stiffness, isostatic restraints are used through the 3-2-1 rule. This rule removes the 6 rigid body degrees of freedom (3 translations and 3 rotations) using exactly three support points on the reference plane:

1. At node 1 (point A): the 3 translations [T_x, T_y, T_z] are constrained. This cancels the global translations of the structure but still allows free rotations about point A.
2. At node 2 (point B, not coincident with A): 2 translations [T_y, T_z] are constrained. Combined with node 1, this removes the rotations about the axes perpendicular to the line A-B.
3. At node 3 (point C, off the line A-B): 1 translation [T_z] is constrained. This blocks the last remaining rotation mode, the rotation about the line A-B.

In this way the 6 rigid body degrees of freedom are strictly eliminated without adding any redundancy to the model, letting the part contract, expand or bend naturally under load.

4. Compatibility between elements and constraints

A critical point is to check the compatibility between the element type and the constrained degrees of freedom:

• 3D solid elements (TET4, TET10, HEX8, HEX20): they only have 3 translational degrees of freedom per node (u, v, w). They have no rotational stiffness at their nodes. If you try to apply a fixed support by constraining rotations (θ_x, θ_y, θ_z) on a single row of solid nodes, the solver will ignore the rotational constraint and the model will behave like a cylindrical hinge, producing an unstable mechanism. More detail in 3D solid elements.
• Shell and beam elements: they have 5 to 6 degrees of freedom per node (translations and rotations). With these elements it is physically possible and valid to constrain moments or rotations directly at the nodes. See Shell elements and bars and beams.

5. Applying realistic loads: avoiding singularities

Just as constraints must be realistic, the way external loads are applied conditions the validity of the local stresses.

A. The danger of nodal point loads

In elasticity theory, applying a concentrated force F at a single node (area A = 0) theoretically produces an infinite stress (σ = F/0 = ∞). In the mesh, this creates a stress singularity that grows without bound as the mesh is refined. Learn to recognise it in the convergence study.

B. Using R-Type coupling elements (RBE2 and RBE3)

To apply remote forces or moments without creating singularities at a single node and without artificially altering the stiffness of the loaded region, R-Type connectors are used:

• RBE2 (rigid N-node element): connects a master node to a set of dependent nodes through infinite stiffness. It is used to simulate rigid welded joints or extremely massive load-spreading plates. Note: it locks the dependent nodes relative to each other, preventing local deformation.
• RBE3 (interpolation element): distributes the force and moment applied at a reference node over the dependent nodes through a weighted average, without adding stiffness to the model. It lets the structure deform, bend or warp freely under the load. It is the ideal tool for applying resultant forces or torques on holes and faces.

6. Verification and quality control of boundary conditions

Before accepting the stress results as valid, the analyst should apply two fundamental quality-control checks:

1. Reaction equilibrium check: the sum of the applied external forces and moments must be equal and opposite to the sum of the reactions computed at the constrained nodes:
   ΣF_applied = −ΣF_reactions
   If there is a discrepancy, the model contains inconsistencies in the loads or the matrix has not converged properly.
2. Free-free modal check: run a modal analysis of the model with no boundary conditions and no loads. If the model is properly connected internally, the first 6 vibration modes must have numerically zero frequencies (0 Hz). If the analysis shows more than 6 zero modes, there are disconnected elements or “loose” internal mechanisms in the structure that will fail when the static solution is run.

Conclusion

Applying the 3-2-1 rule to avoid over-constraining, respecting the degree-of-freedom compatibility of the elements and using RBE3 interpolation elements to distribute the loads are the practices that separate a stable, reliable numerical model from a useless one.