Finite Element Analysis (FEA)

A structured learning path from FEA fundamentals through advanced nonlinear analysis. Master the finite element method for practical structural engineering applications.

Start Learning Full Handbook

Level 1

Beginner — Fundamentals of FEA

Start here if you are new to finite element analysis.

Fundamentals of FEA

The finite element method (FEM) is a numerical technique for solving partial differential equations by discretizing a continuous domain into smaller, simpler subdomains called finite elements. Each element connects to its neighbors at nodes, and the unknown field (displacement, temperature, etc.) is approximated using interpolation functions (shape functions). The element stiffness matrix [k] relates nodal forces to nodal displacements: {f} = [k]{u}. The global stiffness matrix [K] is assembled by summing element contributions at shared nodes, yielding the system equation [K]{U} = {F}.

The direct stiffness method is the foundation of all modern FEA software. The process involves: discretizing the structure into elements, deriving element stiffness matrices in local coordinates, transforming to global coordinates, assembling the global stiffness matrix, applying boundary conditions (eliminating rigid-body modes), solving for unknown displacements, and recovering element stresses and strains. Understanding this fundamental workflow is essential before using any commercial FEA package like ANSYS, ABAQUS, or SAP2000.

Element Types and Shape Functions

FEA offers a library of element types for different structural models. One-dimensional elements include truss elements (axial force only, 2 nodes, 2 DOF per node) and beam elements (axial + bending, 2 nodes, 3 DOF per node: ux, uy, rotation). Two-dimensional elements include CST (Constant Strain Triangle, 3 nodes, linear displacement), LST (Linear Strain Triangle, 6 nodes, quadratic), and quadrilateral elements (Q4, Q8, Q9). Three-dimensional elements include tetrahedral (4 nodes, linear; 10 nodes, quadratic) and hexahedral (8 nodes, linear; 20 nodes, quadratic) elements.

Shape functions define how the field variable varies within an element. Linear shape functions produce constant strain within an element, while quadratic functions allow linear strain variation. Isoparametric formulation uses the same shape functions for geometry and field interpolation, mapping a parent element in natural coordinates (ξ, η) to the actual element shape. The Jacobian matrix J transforms derivatives from natural to physical coordinates, and |J| relates differential areas or volumes. Poor element geometry leads to Jacobian singularity and inaccurate results.

Meshing Fundamentals

Meshing divides the geometry into finite elements. Mesh density (number of elements) directly affects accuracy — finer meshes capture stress gradients better but increase computational cost. Element quality metrics include aspect ratio (ideal: 1.0 for triangles/quads, <3 for most applications), Jacobian ratio (should be >0.5 for accurate integration), skewness (should be >0.7 on a 0-1 scale), orthogonal quality (should be >0.15), and warpage for quad elements. Poor quality elements cause numerical errors and convergence problems.

h-refinement increases accuracy by reducing element size (adding more elements), while p-refinement increases element polynomial order. A convergence study is essential: run the analysis with progressively finer meshes and observe how the quantity of interest (displacement, stress) converges to a stable value. The convergence rate depends on element type — quadratic elements converge faster than linear for smooth problems. Mesh transition regions (biasing, graded meshes) connect coarse to fine regions efficiently. Singularities at reentrant corners require special attention, often using submodeling or adaptive meshing.

Level 2

Intermediate — Modeling and Analysis Techniques

Build on fundamentals with practical modeling techniques.

Boundary Conditions and Loading

Boundary conditions (BCs) constrain structural degrees of freedom and must represent actual support conditions accurately. Fixed supports restrain all translational and rotational DOF. Pinned supports restrain translations only, allowing rotation. Roller supports allow sliding along one direction. Symmetry BCs reduce model size by exploiting structural and loading symmetry — requiring that displacements normal to the symmetry plane are zero and rotations about axes in the symmetry plane are zero. Elastic supports (spring constraints) model foundations, soil interaction, or flexible connections.

Load application methods depend on the load type: nodal loads are applied directly at nodes; distributed loads (pressure, line loads) on element faces are converted to equivalent nodal loads by the FEA software using consistent load vectors; thermal loads require temperature-dependent material properties and reference temperature; gravity loads require mass density and gravitational acceleration; prestress loads are applied as initial stress conditions. Multipoint constraints (MPCs) couple degrees of freedom between nodes, useful for rigid connections, mesh transitions, and applying loads at eccentric locations.

Structural Modeling with FEA

Modeling decisions significantly affect FEA accuracy and efficiency. Beams and frames are modeled with 1D beam elements (Timoshenko or Euler-Bernoulli theory), capturing axial, bending, shear, and torsional behavior. Thin-walled plates and shells use shell elements (based on Kirchhoff or Mindlin-Reissner theory), with 5 or 6 DOF per node. Thick components require solid elements (tetrahedral or hexahedral). Mitigating shear locking in bending-dominated problems requires reduced integration or higher-order elements. Hourglassing (zero-energy modes) occurs with reduced integration, necessitating hourglass control or stabilization.

Connecting different element types requires careful treatment: shell-to-solid connections use MPCs or transitional elements; beam-to-shell connections account for offset between beam neutral axis and shell midsurface. Submodeling analyzes a local region with a finer mesh using boundary conditions from a global coarse model. Rigid and flexible connections in joints are modeled using constraint equations or contact elements. Spot welds, bolts, and rivets can be modeled with beam elements, rigid links, or specialized connector elements. Proper idealization balances accuracy with computational efficiency.

Convergence and Error Estimation

FEA solutions are approximate, and error estimation is critical for reliable results. Three convergence paths exist: h-convergence reduces element size (h) while keeping element order constant; p-convergence increases polynomial order while keeping mesh fixed; r-convergence moves node positions to optimize element distribution. The Zienkiewicz-Zhu (Z2) error estimator recovers improved stresses by superconvergent patch recovery, then computes the energy norm error. The relative error in energy norm should typically be below 5-10% for acceptable engineering accuracy.

Sources of FEA error include: discretization error (insufficient mesh density), modeling error (incorrect BCs, material properties, or element type), numerical error (round-off, integration tolerance, solver tolerance), and idealization error (ignoring nonlinearity, simplifying geometry). Verification confirms that the mathematical model is solved correctly (code verification + calculation verification). Validation confirms that the mathematical model accurately represents the physical reality. A mesh convergence study plotting a key result (stress, displacement) vs. element count or DOF is standard practice for every FEA analysis.

Level 3

Advanced — Nonlinear FEA and Dynamics

For senior students and practicing engineers.

Nonlinear FEA

Nonlinear FEA addresses problems where the stiffness matrix changes during the analysis. Geometric nonlinearity arises from large deformations (strain >5%), buckling, and follower forces — requiring the element stiffness to include initial stress (geometric) stiffness. Material nonlinearity includes plasticity (yield criteria: von Mises, Tresca, Mohr-Coulomb), hardening rules (isotropic, kinematic), and hyperelasticity (rubber, foam). Contact nonlinearity involves changing boundary conditions as surfaces interact, including friction (Coulomb, stick-slip), separation, and sliding.

The Newton-Raphson method solves nonlinear equations iteratively: compute the residual (out-of-balance force vector), calculate the tangent stiffness matrix, solve for displacement increment, update configuration, and check convergence. Arc-length (Riks) methods trace unstable post-buckling paths. Convergence criteria include force-based (norm of residual < tolerance), displacement-based (norm of displacement increment < tolerance), and energy-based (combined force and displacement check). Under-relaxation and line search improve convergence for highly nonlinear problems.

Dynamics and Vibration with FEA

Dynamic FEA accounts for inertial and damping forces. The dynamic equation is [M]{ü} + [C]{u} + [K]{u} = {F(t)}. The mass matrix can be consistent (same shape functions as stiffness, producing off-diagonal terms) or lumped (diagonal mass matrix, computationally efficient). Eigenvalue analysis solves [K - ω²M]{φ} = {0} for natural frequencies ω and mode shapes { φ }.

Transient dynamic analysis methods include: modal superposition (efficient for linear problems, uses mode shapes to decouple equations), implicit direct integration (Newmark-beta, HHT-alpha — unconditionally stable for structural dynamics), and explicit direct integration (central difference — conditionally stable, suitable for wave propagation and impact). Rayleigh damping models damping as [C] = α[M] + β[K]. Frequency response analysis computes steady-state response to harmonic loading. Response spectrum analysis combines modal maxima using SRSS or CQC methods for seismic design.

Practical FEA Workflow and Validation

A systematic FEA workflow ensures reliable results: (1) define analysis objectives and acceptance criteria; (2) create or import geometry, simplify details (remove small features, fillets, holes not in stress zones); (3) assign material properties (elastic, plastic, density, thermal); (4) define element types and mesh with appropriate density and quality; (5) apply boundary conditions and loads; (6) select solver settings (linear/nonlinear, time step, convergence tolerances); (7) run analysis and monitor convergence; (8) post-process results and check for errors — stress singularities, mesh dependency, unrealistic deformations; (9) validate against hand calculations, test data, or code provisions; (10) document assumptions and results.

Common FEA pitfalls include: stress singularities at reentrant corners (not physical, ignore or use submodeling), mesh dependency (results change significantly with mesh size — run convergence study), hourglassing (check element energies, use hourglass control), spurious zero-energy modes (check for mechanisms, apply adequate constraints), incorrect material units, and misinterpretation of element stress output (element centroidal vs. nodal averaged vs. extrapolated stresses). Always verify FEA results with simple hand calculations, free-body checks, or the structural analysis tools on this site.

Practice Exercises

Exercise 1: Truss FEA

A 2D 3-bar truss with E=200 GPa, cross-sectional areas 1000mm², loaded 50kN at joint. Write element stiffness matrices for each bar in local coordinates, transform to global coordinates, assemble the global stiffness matrix, apply boundary conditions (pinned at supports), solve for unknown displacements, and recover member forces. Verify your results with the Truss Analysis Calculator.

Exercise 2: Mesh Convergence Study

A cantilever beam 2m long, 0.2m x 0.3m cross-section, carries a 10kN end load. Run FEA with 4, 8, 16, 32, and 64 elements along the length. Plot tip deflection vs. element count. Determine the mesh density required to achieve less than 1% error compared to the Euler-Bernoulli analytical solution. Use the Cantilever Beam Calculator for verification.

Exercise 3: Plate with Hole Stress Concentration

A 200mm x 100mm plate with a 20mm diameter central hole is subjected to 100MPa tensile stress. Model one-quarter symmetry to reduce computational cost. Compare the maximum stress concentration factor with the theoretical value Kt=2.5 from Peterson's formulas. Study the effect of element size refinement around the hole perimeter on the peak stress value.

Exercise 4: Modal Analysis of a Frame

A 2-bay 2-story steel frame with W250x89 columns and W360x110 beams, bay width 8m, story height 4m. Using FEA, extract the first 3 natural frequencies and corresponding mode shapes. Compare the fundamental frequency with the simplified Rayleigh method approximation. Discuss how the mode shapes influence the seismic design of this frame.

References

Zienkiewicz, O.C., Taylor, R.L., and Zhu, J.Z. The Finite Element Method: Its Basis and Fundamentals. 7th ed., Butterworth-Heinemann, 2013.
Bathe, K.J. Finite Element Procedures. 2nd ed., Klaus-Jurgen Bathe, 2014.
Cook, R.D. et al. Concepts and Applications of Finite Element Analysis. 4th ed., Wiley, 2001.
Logan, D.L. A First Course in the Finite Element Method. 6th ed., Cengage Learning, 2016.
Reddy, J.N. Introduction to the Finite Element Method. 4th ed., McGraw-Hill, 2019.
Civil Engineering Handbook — FEA fundamentals chapter.
Engineering Formula Library — Stiffness matrix and element formulation formulas.
Engineering Standards Reference — AISC, ACI, and Eurocode provisions.
Engineering Glossary — Definitions of FEA terms.