Composite structures rarely fail in the clean, single-mode way assumed for isotropic metals. A laminate can survive a high global load while a single off-axis ply develops matrix cracking, or it can show an acceptable laminate stress average while a free edge initiates delamination. This guide to composite failure criteria focuses on the decisions that make failure predictions useful in a production FEA workflow: selecting an appropriate criterion, supplying defensible material data, interpreting solver output, and validating the result against the real structure.
Why composite failure criteria require engineering judgment
A failure criterion converts a ply-level stress or strain state into a measure of margin. In a linear static analysis, this is often reported as a failure index (FI), where a value of 1.0 indicates predicted initiation of the modeled failure condition. Some software instead reports a reserve factor (RF), where values greater than 1.0 indicate remaining capacity. These are not interchangeable unless the solver documentation defines their relationship for the selected criterion.
The central issue is that a unidirectional lamina has materially different behavior in the fiber direction, transverse direction, and through-thickness direction. Tensile fiber failure, fiber microbuckling in compression, matrix cracking, matrix crushing, and interlaminar failure are governed by different mechanisms. A criterion is therefore not merely a post-processing option. It represents an assumption about how those mechanisms interact.
The correct choice depends on the design question. A preliminary sizing study may need a fast screening method using in-plane shell stresses. A certification-supporting analysis of a compression-critical laminate may require mode-specific criteria, progressive damage modeling, geometric nonlinearity, manufacturing knockdowns, and correlation to representative test articles. The criterion must be proportional to the confidence required of the decision.
Start with the material system and allowable basis
No failure theory can compensate for incomplete material characterization. Before selecting a criterion, establish the lamina coordinate system, ply orientation convention, thickness definition, and material allowable basis. For an orthotropic ply, the minimum data set commonly includes longitudinal strengths X_t and X_c, transverse strengths Y_t and Y_c, in-plane shear strength S_12, and the associated elastic constants.
For three-dimensional solids or layered solid formulations, out-of-plane strengths and shears may also be needed: Z_t, Z_c, S_13, and S_23. Those values are often less available and more variable than in-plane properties, yet they control many local details near bolts, bonded joints, curved regions, and load introduction features.
The data source matters as much as the number entered into the material card. Coupon allowables may be mean values, statistically reduced B-basis or A-basis values, or internally derived design values with environmental and process knockdowns. A room-temperature, dry coupon strength is not a complete representation of a production laminate exposed to moisture, heat, fatigue, defects, and variable fiber volume fraction.
Use test data representative of the material form, resin system, cure cycle, thickness range, and intended service environment. Where the program relies on supplier data or a legacy database, document the assumptions and apply justified factors. This discipline prevents a precise-looking failure index from becoming an unsupported design claim.
Common composite failure criteria and where they fit
Maximum stress and maximum strain
Maximum stress and maximum strain criteria compare each stress or strain component independently with its corresponding allowable. They are easy to audit and identify the governing component directly. For example, a transverse tensile stress exceeding Y_t clearly points to matrix cracking risk.
Their limitation is that they do not capture interaction among normal and shear stresses. A ply under moderate transverse tension and shear may fail before either component independently reaches its limit. Still, these criteria remain useful as a baseline check, especially when material data are limited or when traceability and simple margins are the priority.
Tsai-Hill and Tsai-Wu
Tsai-Hill is an interactive polynomial criterion commonly used for plane-stress laminate analysis. It offers a compact way to account for combined loading, but it does not naturally distinguish tension from compression unless separate properties and implementation details address that distinction.
Tsai-Wu extends the polynomial form to include linear terms and different tensile and compressive strengths. It can represent a broader asymmetric failure envelope, making it attractive for general screening. Its weakness is practical rather than mathematical: the interaction coefficient must be chosen carefully. If it is assumed without test support, the predicted envelope may be either nonconservative or excessively restrictive in combined loading.
Tsai-Wu is often valuable for comparative design studies. It should not be treated as proof that the physical failure mode has been identified.
Hoffman
Hoffman is another polynomial interaction criterion that accommodates different tensile and compressive strengths. It is frequently used in established aerospace and industrial workflows because it is accessible and broadly implemented. Like Tsai-Wu, it returns a combined indicator rather than a direct statement of fiber, matrix, or delamination failure.
When a Hoffman index governs, examine the underlying ply stresses and strains. A single scalar result is insufficient for deciding whether the model needs local refinement, a material change, or a different load path.
Hashin, Puck, and mode-specific methods
Hashin-type criteria separate fiber tension, fiber compression, matrix tension, and matrix compression modes. This distinction is especially useful when a design team needs to understand what is expected to fail first and whether a proposed change addresses that mechanism. Many progressive damage workflows use Hashin-style initiation criteria followed by stiffness degradation rules.
Puck-based methods place greater emphasis on physically motivated matrix and inter-fiber failure planes. They can provide stronger insight for off-axis plies and compression-dominated states, but they also demand more material inputs and a careful understanding of the software implementation.
Neither approach eliminates the need for test correlation. Mode-specific output can improve engineering judgment, yet it also creates more opportunities to apply an unsupported parameter or damage-evolution law.
A guide to composite failure criteria in FEA
The quality of the stress recovery is often more consequential than the name of the criterion. In a shell laminate model, confirm that ply stresses are recovered at the correct layer and integration point, in the local material axes. A global Sxx value is not automatically the fiber-direction stress for a 45-degree ply.
Check the following before relying on a contour plot:
- Ply angles and element material axes are consistent across the model, including mirrored geometry and coordinate-system transitions.
- The laminate stack sequence, ply thicknesses, dropped plies, and reference plane match the released design definition.
- The solver is reporting stresses at the appropriate ply surface or integration point, not a thickness-averaged value that conceals a critical surface stress.
- Singular stresses at point constraints, rigid connections, sharp re-entrant corners, or idealized contact are separated from meaningful structural hot spots.
- Units for elastic constants, strengths, thickness, and applied load are internally consistent.
A shell model is efficient for broad laminate behavior, but it cannot fully resolve every three-dimensional failure mechanism. Through-thickness stresses near a fastener, a bonded overlap termination, a thick laminate transition, or a curved flange may require a local solid model, cohesive elements, or a submodel driven by the global response. The objective is not to make every region three-dimensional. It is to apply local fidelity where the governing physics requires it.
Interpret failure index with margin, not alarm
An FI above 1.0 is a signal to investigate, not an automatic declaration that the assembly will fail at that exact load. First, verify whether the result occurs over a meaningful area and whether mesh refinement stabilizes it. Then inspect the governing ply, load case, stress components, and failure mode. A narrow peak in an adhesive-adjacent ply may call for local modeling. A broad region of matrix-compression failure in several plies may point to a genuine laminate sizing issue.
Likewise, an FI below 1.0 does not establish acceptable performance by itself. A laminate can accumulate damage below ultimate initiation values under repeated loading. Buckling can alter stress distribution before material failure occurs. Manufacturing defects and open-hole effects may govern rather than pristine coupon strengths. If the structure has damage tolerance, fatigue, impact, or post-buckling requirements, those conditions must be represented separately.
For nonlinear progressive damage analysis, the damage law deserves the same scrutiny as the initiation criterion. Element stiffness reduction can cause mesh-sensitive energy dissipation and unrealistic load redistribution if fracture energy, characteristic length, and degradation parameters are not calibrated. A converged nonlinear run is not necessarily a validated failure prediction.
Build validation into the workflow
A defensible composite analysis follows a hierarchy. Begin by validating material definitions against coupon behavior. Validate laminate stiffness and strain response against simple panels. Then correlate subcomponent models for representative holes, joints, curvature, load introduction, or impact conditions before relying on full-assembly predictions.
At each level, compare quantities that can reveal the source of disagreement: displacement, strain-gage response, buckling load, failure location, failure mode, and load at first detectable damage. Matching ultimate load while missing the failure mechanism may still lead to a poor design decision.
Experienced teams also preserve traceability. Record the criterion equation used by the solver, strength values, interaction coefficients, coordinate convention, recovery location, allowable source, and applied knockdowns. This record makes peer review faster and prevents a later program phase from inheriting an unverified assumption. eNastran Engineering applies this type of model and workflow validation when supporting Nastran-based composite programs.
The most useful failure criterion is the one whose assumptions match the material data, model fidelity, and decision at hand. Treat the index as evidence within a validated engineering process, and it becomes a practical tool for reducing test risk and building confidence in the laminate before hardware reaches the floor.