Composite Failure Model Calibration

What Is It?

Calibration is the process of determining the material parameters that a composite failure model needs to reproduce the observed material behaviour. A progressive damage model requires strengths (for initiation criteria), fracture energies (for evolution laws), stiffness properties (for the elastic response) and potentially additional parameters for the damage evolution formulation. Calibration uses material test data — from coupon-level tests designed to isolate specific failure modes — to identify these parameters. The calibration is the foundation of the model: if the parameters are wrong, no amount of analysis sophistication will produce credible predictions.

Why It Matters

The credibility of a composite damage analysis depends fundamentally on the quality of the calibration. A model calibrated with inappropriate or insufficient data will produce non-physical predictions — wrong failure modes, wrong loads, wrong damage patterns. Calibration is not a single step but a systematic process that follows the material characterisation hierarchy, isolating each failure mode and measuring the relevant properties. Without this systematic approach, the model parameters are guessed, not calibrated — and the predictions cannot be trusted.

Calibration is the foundation of a credible damage model. If the parameters are wrong, the predictions are wrong. Calibration must follow a systematic material characterisation hierarchy — not be reverse-fitted from a single component test.

Material Test Hierarchy

The material test hierarchy is the structured approach to characterising composite material behaviour. It progresses from basic material characterisation through coupons, elements, subcomponents, components and full-scale structure. Each level builds on the previous: material characterisation provides the basic properties, coupons isolate failure modes, elements test structural details, subcomponents test features, and components/full-scale validate the overall approach. Calibration of the damage model parameters occurs primarily at the coupon level, with verification at higher levels.

Coupon Data for Strengths

The initiation criteria require strengths for each failure mode. These are obtained from coupon tests designed to isolate each mode: unidirectional tension along the fibre direction for fibre tensile strength (X_t), unidirectional compression along the fibre for fibre compressive strength (X_c), transverse tension for matrix tensile strength (Y_t), transverse compression for matrix compressive strength (Y_c), and in-plane or interlaminar shear for shear strength (S_12, S_23). Each test must be designed to produce the intended failure mode — if a "transverse tension" test fails by delamination rather than transverse matrix cracking, the data are not valid for Y_t.

• Fibre tension (X_t): unidirectional 0° tension test
• Fibre compression (X_c): unidirectional 0° compression test — anti-buckling fixture
• Matrix tension (Y_t): transverse 90° tension test
• Matrix compression (Y_c): transverse 90° compression test
• In-plane shear (S_12): ±45° tension or V-notched rail shear
• Interlaminar shear (S_23): short-beam shear or similar
• Each test must produce the intended failure mode — verify by post-test inspection

Fracture Energy Characterisation

The damage evolution laws require fracture energies for each failure mode. The interlaminar fracture energies are obtained from standard fracture mechanics tests: DCB for Mode I (G_Ic), ENF for Mode II (G_IIc), and MMB for mixed mode. The intralaminar fracture energies (fibre tension, fibre compression, matrix tension, matrix compression) are more difficult to obtain and may require compact tension, compact compression or specialised tests. The fracture energy is a critical parameter for mesh-objective damage evolution — it must be measured, not assumed. Do not invent fracture toughness or damage parameters.

Fracture energies must be measured from fracture tests — not assumed or invented. G_Ic from DCB, G_IIc from ENF, mixed-mode from MMB. Intralaminar fracture energies require specialised tests. Without measured fracture energies, the damage evolution is not calibrated.

Damage Evolution Parameter Identification

Beyond the strengths and fracture energies, the damage evolution law may have additional parameters — the softening shape (linear, exponential, bilinear), the characteristic length definition and any mode-interaction parameters. These are identified by fitting the model response to the observed test behaviour. For example, the softening shape can be identified from the post-peak load-displacement curve of a notched specimen. The mode-interaction parameters (e.g. the B-K exponent η for mixed-mode delamination) are identified from MMB tests at multiple mode mixes.

Damage evolution parameter identification:

Required parameters:
  Strengths:    X_t, X_c, Y_t, Y_c, S_12, S_23
  Fracture energies: G_Ic, G_IIc, G_c(fibre tension),
                    G_c(fibre compression),
                    G_c(matrix tension),
                    G_c(matrix compression)
  Softening shape: linear, exponential, bilinear
  Mixed-mode exponent: η (from MMB tests)
  Characteristic length: solver-defined (verify)

Identification method:
  Strengths → coupon tests (each mode isolated)
  G_Ic, G_IIc → DCB, ENF tests
  G_c(intralaminar) → compact tension/compression
  η → MMB tests at multiple mode mixes
  Softening shape → post-peak curve fitting

Uncertainty in Calibration Parameters

Calibration parameters carry uncertainty from multiple sources: material variability (specimen-to-specimen scatter), test method uncertainty (gauge accuracy, alignment, boundary conditions), parameter identification uncertainty (fitting scatter), and the inherent variability of composite materials. The strengths and fracture energies should be reported with their scatter (mean, standard deviation, B-basis or A-basis allowables). The uncertainty should be propagated through the analysis — for critical structure, the effect of parameter uncertainty on the predicted failure load and mode should be assessed through sensitivity analysis.

• Material variability: specimen-to-specimen scatter in strengths and fracture energies
• Test method uncertainty: gauge accuracy, alignment, boundary condition effects
• Parameter identification: fitting scatter when extracting parameters from test data
• Report parameters with scatter — mean, standard deviation, B-basis or A-basis allowables
• Propagate uncertainty through analysis — sensitivity study on key parameters
• For critical structure, use lower-bound (B-basis or A-basis) values, not mean values

Why Not Calibrate Against Final Component Failure Load?

A common but fundamentally flawed approach is to calibrate the damage model parameters by reverse-fitting to match the final failure load of a component test. This is problematic for several reasons. First, the component failure involves multiple failure modes interacting — adjusting parameters to match the final load does not ensure the individual modes are correctly characterised. Second, the model may predict the right load for the wrong failure mode — the parameters are not uniquely identified. Third, the calibrated parameters are specific to that component and cannot be transferred to other geometries or load cases. Finally, the approach provides no physical understanding — it is curve-fitting, not calibration.

Do NOT calibrate a damage model only against a final component failure load. This does not correctly identify the individual failure mode parameters, may produce the right load for the wrong failure mode, cannot be transferred to other geometries, and provides no physical understanding. Calibrate against coupon-level data that isolates each failure mode.

Correct Calibration Workflow

The correct calibration workflow follows the building block from the bottom up. Material characterisation provides elastic properties. Coupon tests isolate each failure mode and provide strengths and fracture energies. The damage model parameters are identified from these coupon data. Element and subcomponent tests then verify that the calibrated model predicts the correct behaviour at the structural detail level — including the failure mode, load-displacement response and damage pattern. Only after this verification is the model used for component-level predictions.

1. Material characterisation: elastic properties (E1, E2, G12, ν12)
2. Coupon tests: isolate each failure mode — strengths (Xt, Xc, Yt, Yc, S) and fracture energies (G_Ic, G_IIc, G_c)
3. Parameter identification: fit damage evolution parameters from coupon post-peak behaviour
4. Element tests: verify model at structural detail (open-hole, joint, etc.)
5. Subcomponent tests: verify model at feature level (stiffened panel, etc.)
6. Component tests: validate overall prediction — not calibrate parameters

Key Takeaways

• Calibration is the foundation of a credible damage model — wrong parameters produce wrong predictions
• Follow the material test hierarchy: material characterisation → coupon → element → subcomponent → component → full-scale
• Strengths are from coupon tests that isolate each failure mode — verify the intended mode occurs
• Fracture energies must be measured from fracture tests (DCB, ENF, MMB) — not assumed or invented
• Damage evolution parameters (softening shape, mode interaction) are identified from post-peak behaviour
• Uncertainty must be quantified and propagated — use B-basis or A-basis for critical structure
• Do NOT calibrate only against a final component failure load — it does not correctly identify parameters
• Calibrate at coupon level, verify at element and subcomponent level, validate at component level