4. Constitutive modeling of airbag fabrics

Airbag fabric is not an isotropic sheet. It is a woven structure in which warp and weft yarns interlace, coated or uncoated, and respond to load in a strongly direction-dependent way. A constitutive model is the mathematical bridge between the measured behavior of that fabric and the stress state computed inside an explicit finite element solver. Choosing the wrong modeling philosophy — or the wrong parameter set for the right philosophy — produces a simulation that may converge yet predict the wrong deployment pressure, the wrong bag shape, or the wrong seam load. This section compares the three dominant modeling families and shows how to select, calibrate, and validate them.

Why Anisotropy Cannot Be Ignored

In a plain-weave nylon 6.6 fabric, the yarns are nearly inextensible along their own axis but the weave crimp allows large shear deformation. Consequently, the in-plane response is governed by three distinct mechanisms: yarn extension along warp and weft, trellising (angular shear between yarn families), and coating or film contribution. A uniaxial strip cut at 0° (warp) may reach 20–25% strain at 200 N/cm, while a strip cut at 45° can exceed 60% strain at the same load because it deforms primarily by trellising. Any model that assigns a single modulus to the fabric will misrepresent both the bag's inflation trajectory and its contact pressure distribution.

Before selecting a model, define the reference frame explicitly. Throughout this course, the material coordinate system is: 1-axis = warp direction, 2-axis = weft direction, 3-axis = through-thickness normal. Positive shear strain γ₁₂ denotes a decrease in the angle between warp and weft (trellising toward 90° → acute). All models below use this convention consistently.

The Three Modeling Families

Continuum (phenomenological anisotropic hyperelasticity). The fabric is treated as a homogeneous continuum with a strain-energy density function W that depends on invariants of the right Cauchy–Green tensor and on pseudo-invariants that encode the warp and weft directions. A representative form is:

W = W_iso(I1) + W_aniso(I4, I6) + W_shear(I8)
I4 = a0 · C · a0        (warp fiber stretch)
I6 = b0 · C · b0        (weft fiber stretch)
I8 = a0 · C · b0        (warp–weft coupling / trellising)

where a₀ and b₀ are unit vectors along warp and weft in the reference configuration. This approach is computationally cheap, integrates directly into shell elements, and is the standard choice for full-vehicle occupant simulations. Its weakness is that the fitted parameters are not physically measurable — they are regression coefficients, and they extrapolate poorly outside the calibration strain range.

Mesoscale (yarn-level or unit-cell). Here the actual weave architecture is represented: yarn cross-sections, crimp amplitude, weave pattern (plain, twill, basket), and coating thickness. The unit cell is solved either analytically (e.g., pin-jointed or energy-minimization models) or numerically with beam/solid elements. Mesoscale models predict the trellising lock-up angle, the nonlinear shear stiffness, and the coupling between yarn tension and fabric shear. They are essential when the fabric architecture itself is a design variable, but they are too expensive to embed in a full bag simulation. Their practical role is to generate virtual test data that calibrates a continuum model.

Phenomenological (empirical / test-driven). The model is defined directly by the test curves it must reproduce — typically uniaxial tension in warp, weft, and bias directions, plus picture-frame shear. A common implementation is an orthotropic elastic–plastic or piecewise-linear formulation in which each direction has its own modulus and yield point, and shear is defined by a tabulated stress–strain curve. This is the most transparent approach for validation because every parameter maps to a measurable quantity, but it does not naturally capture coupling or rate effects.

Selecting the Right Approach

The choice is driven by the question being asked, not by model sophistication. Use the following decision logic:

• Full-vehicle occupant simulation, bag shape and pressure history: continuum anisotropic hyperelastic shell model. Calibrate against coupon tests, accept the loss of yarn-level detail.
• Fabric architecture trade study (weave, denier, coating): mesoscale unit-cell model. Run a parametric sweep, extract effective shear and tensile curves.
• Coupon-level validation, seam and joint analysis, regulatory documentation: phenomenological orthotropic model with tabulated curves. Every parameter traceable to a test report.
• Rate-dependent deployment (high-speed inflation): continuum model extended with viscoelastic or viscoplastic terms, or phenomenological model with rate-dependent tabulated curves.

Calibration Workflow and Practical Parameters

Regardless of family, calibration follows the same logic: measure, fit, verify against an independent test. A representative workflow for a coated nylon 6.6 plain-weave fabric (470 dtex, 20 × 20 yarns/cm, silicone-coated) is given below. The values are illustrative teaching examples, not measured results.

For the continuum model, fit W_aniso to the warp and weft curves simultaneously, then fit W_shear to the bias and picture-frame data. A common pitfall is fitting all directions at once with a single objective function; this allows the optimizer to trade warp accuracy for shear accuracy. Fit sequentially and check each direction independently.

For the phenomenological model, extract the tangent modulus at 1%, 5%, 10%, and 20% strain from each curve and build a piecewise-linear table. The shear curve should be entered as engineering shear stress versus γ₁₂, with the lock-up angle (typically 60–70° for plain weave, i.e., γ₁₂ ≈ 0.5–0.7) marked as the point where stiffness rises sharply.

Common Failure Modes and Troubleshooting

Symptom: bag inflates too slowly in simulation. Likely cause: warp/weft moduli overestimated, or shear stiffness too high. Check the bias test fit first — shear dominates early inflation.

Symptom: excessive element distortion or negative volume in explicit solver. Likely cause: missing lock-up in the shear curve, allowing unphysical trellising. Add a stiffening term beyond γ₁₂ = 0.6.

Symptom: seam loads much higher than test. Likely cause: continuum model too stiff in the through-thickness or coupling direction. Verify I₈ contribution and consider reducing W_shear coupling coefficient.

Symptom: simulation converges but bag shape is wrong. Likely cause: incorrect material orientation assignment. Confirm that the 1-axis in the model matches the warp direction in the physical cut pattern. A 90° orientation error swaps warp and weft and produces a visibly asymmetric bag.

Practical Deliverable: Calibration Parameter Sheet

The following sheet is a reusable template for documenting a fabric model calibration. Fill one row per test and one column per parameter. It forces traceability from test report to solver input.

Once the sheet is complete, the model is ready for coupon-level verification. Only after that verification passes should the parameters be promoted to full-vehicle simulation. Skipping the coupon check is the most common reason a deployment simulation produces plausible-looking but quantitatively wrong results.