The Mathematics of Linear FEA

2025-08-15
Understanding Linear FEA: The Fundamental Equation
Finite Element Analysis (FEA) for linear problems is built on a beautifully simple mathematical foundation:
This equation represents the equilibrium of a discretized structure, where:
- is the global stiffness matrix
- is the displacement vector
- is the force (load) vector
The Stiffness Matrix (K)
The stiffness matrix describes how the structure resists deformation. For structural problems, it relates applied forces to resulting displacements.
Building the Global Stiffness Matrix
The global stiffness matrix is assembled from individual element stiffness matrices:
Each element stiffness matrix is computed as:
Where:
- is the strain-displacement matrix (relates nodal displacements to strains)
- is the material constitutive matrix (relates stresses to strains via material properties)
- is the volume of element
Material Properties in the Stiffness Matrix
For linear elastic materials, the constitutive matrix depends on:
- Young's Modulus () - material stiffness
- Poisson's Ratio () - lateral strain response
For 3D isotropic elasticity:
How Mesh Elements Affect the Stiffness Matrix
Element Shape Functions
Each element type (triangles, quads, tetrahedra, hexahedra) uses shape functions to interpolate displacements within the element:
The strain-displacement matrix is derived from these shape functions:
Mesh Refinement Impact
Finer meshes (more elements, smaller size):
- Increase the size of (more degrees of freedom)
- Improve solution accuracy by better approximating curved geometries and stress gradients
- Capture localized effects like stress concentrations
- Increase computational cost (larger system to solve)
Coarser meshes (fewer elements, larger size):
- Reduce computational cost
- May miss important local behavior
- Can produce inaccurate results in high-gradient regions
Element Quality Matters
Poor element shapes (highly skewed or distorted) lead to:
- Ill-conditioned stiffness matrices
- Numerical errors in integration
- Inaccurate stress and strain calculations
The Force Vector (f)
The force vector contains all external loads applied to the structure:
Types of Loads
Point loads are directly assigned to nodes:
Distributed loads (pressure, surface tractions) are converted to nodal forces via integration:
Where is the pressure/traction on surface .
Body forces (gravity, acceleration) are volume loads:
Where is density and is gravitational acceleration.
The Displacement Vector (u)
The displacement vector contains the unknowns we're solving for:
For a 3D problem with nodes:
- are displacements in x, y, z directions at node
- Total degrees of freedom = (before applying boundary conditions)
Boundary Conditions
Essential (Dirichlet) boundary conditions fix certain displacements:
- Fixed support:
- Roller support: one or two components set to zero
These constraints are enforced by modifying the system before solving.
Solving the System
Once and are assembled and boundary conditions applied, we solve:
In practice, direct solvers (Cholesky, LU decomposition) or iterative solvers (Conjugate Gradient) are used because:
- is large and sparse (most entries are zero)
- is symmetric and positive definite (for stable structures)
Post-Processing
After solving for , we can compute:
- Strains:
- Stresses:
- Reaction forces:
Summary
Linear FEA reduces complex structural problems to a system of linear equations: . The stiffness matrix encodes geometry, material, and element properties. The force vector represents all loads. The displacement solution gives us the structure's response — and from there, all derived quantities.
The beauty of FEA is that this same mathematical framework extends to heat transfer, fluid flow, electromagnetics, and beyond — always returning to the same fundamental pattern of discretization, assembly, and solution.