4. Finite element modeling of airbag seams
An airbag seam is not a geometric feature that can be meshed like a panel edge. It is a narrow, multi-layer, stitched or welded joint whose mechanical response — load transfer, peel resistance, progressive rupture — governs whether the cushion holds pressure during deployment. Finite element (FE) modeling of seams therefore has to answer three coupled questions: which element family carries the seam, how the seam line is represented in the mesh, and how fine the discretization must be before the predicted burst pressure or seam-opening load stops changing. This section addresses all three, with emphasis on the practical decisions an analyst makes in LS-DYNA, Abaqus/Explicit, or PAM-CRASH.
4.1 Shell versus solid element strategies
Airbag fabric is a thin, nearly incompressible woven membrane. The default and usually correct choice is a shell element formulation — Belytschko–Tsay, Belytschko–Leviathan, or a fully integrated shell with hourglass control — because the fabric bending stiffness is negligible and through-thickness stress is small except inside the seam. Shells keep the model tractable: a driver-side cushion with a 3–5 mm target element size typically contains 40,000–120,000 shell elements, which is acceptable for explicit time integration.
The seam itself, however, is a three-dimensional problem. Two fabric plies are folded, overlapped, and joined by a row of stitches or a welded bead; the load path runs through the thickness of both plies and through the stitch columns. A single shell layer cannot represent this. Three strategies are used in practice:
- Shell-only with a seam constitutive law. The two plies are modeled as separate shell surfaces, and the seam is represented by a tied contact, a shared node row, or a cohesive zone inserted between coincident shell nodes. This is the cheapest approach and is adequate when the seam is not the failure initiator.
- Shell plies plus solid seam strip. A narrow band of solid elements (one or two elements through the thickness) replaces the shell in the seam region, and the shell plies are tied to the solid strip. This captures through-thickness peel and stitch pull-out, at the cost of a non-conforming mesh transition.
- Full solid seam model. Both plies and the stitch columns are meshed with solids. This is reserved for coupon-level seam specimens or for validating a simplified model, because the element count and stable time step become prohibitive for a full cushion.
The choice is driven by the failure mode under investigation. If the observed failure is fabric rupture adjacent to the seam, a shell model with a calibrated fabric failure strain is sufficient. If the failure is stitch tear-out or seam opening, the through-thickness load path must be present, and a solid seam strip is the minimum viable representation.
4.2 Representing the seam line in the mesh
A seam line is a curve on the fabric surface, not a straight edge. In a flat-pattern cushion it follows the stitch path; in a folded or rolled cushion it becomes a three-dimensional curve after the folding operation. The mesh must conform to this curve, otherwise the seam load is smeared over adjacent elements and the predicted opening load is non-physical.
Two meshing conventions are common. In the node-aligned approach, the seam line is seeded first, and the surrounding surface mesh is generated outward from that seed so that element edges lie exactly on the stitch path. This gives a clean, single-row seam and is preferred for coupon models. In the projected approach, the surface is meshed independently and the seam is defined by a set of node pairs identified through a geometric search tolerance (typically 0.5–1.0 mm). The projected approach is faster to set up but introduces a tolerance-dependent load path; the tolerance must be reported and its effect checked.
For stitched seams, each stitch column should be represented by at least one element edge along the stitch direction, and the stitch spacing (commonly 2.5–4.0 mm in automotive airbag fabric) should be resolved by at least two element edges if stitch-by-stitch rupture is to be captured. If the stitch spacing is smaller than the target element size, the seam is homogenized: the discrete stitches are replaced by a continuous cohesive or tied interface with equivalent stiffness and strength per unit length.
4.3 Mesh convergence for seam models
Seam models converge more slowly than the surrounding membrane because the stress gradient across the seam is steep. A convergence study should vary the element size in the seam region while holding the far-field mesh and the material parameters fixed, and should track two quantities: the peak seam-opening load (or burst pressure) and the crack-initiation location. The load typically converges monotonically; the initiation location may jump between element rows, which is a sign that the mesh is still too coarse to resolve the stitch column.
A practical convergence criterion for a coupon-level seam model is that doubling the seam-region element density changes the peak load by less than 5% and does not move the initiation site by more than one element row. For a full-cushion model, where a coupon-level study is not affordable, the seam-region element size is usually fixed at 1.0–2.0 mm and the convergence check is performed on a representative seam sub-model that is later embedded in the full cushion.
The stable time step is the binding constraint. For a shell with characteristic length L, density ρ, and modulus E, the explicit stable step scales as L√(ρ/E). Halving the seam element size halves the global time step and doubles the run time. This is why seam refinement is applied locally, with mass scaling or subcycling used only when the added mass is verified to be below 1–2% of the local model mass.
4.4 A worked convergence check
Consider a seam coupon: two 50 mm × 100 mm fabric plies, 0.35 mm thick, overlapped by 15 mm, joined by three stitch columns at 3.0 mm spacing. The coupon is clamped at both ends and pulled in tension at 100 mm/s. The fabric is modeled with a linear-elastic orthotropic shell (E1 = 9.0 GPa, E2 = 8.5 GPa, ν = 0.3, ρ = 1.15 × 10−9 t/mm3), and the seam is represented by a cohesive interface with a peel strength of 4.5 N/mm and a fracture energy of 0.8 N/mm.
Run the model at seam-region element sizes of 4.0, 2.0, 1.0, 0.5, and 0.25 mm, keeping the far-field mesh at 4.0 mm. Record the peak reaction force and the element row where the cohesive damage first reaches 1.0. The expected trend is a rising peak force that flattens between 1.0 mm and 0.5 mm, with the initiation row stabilizing at the same stitch column. If the peak force is still rising at 0.5 mm, the cohesive strength is being under-resolved and the interface parameters should be re-examined before the mesh is refined further.
4.5 Practical checklist
| Decision | Recommended practice | Verification |
|---|---|---|
| Element family | Shell for plies; solid strip or cohesive interface for seam | Compare coupon peak load against test |
| Seam line meshing | Node-aligned seeding for coupons; projected with reported tolerance for full cushion | Check seam load path continuity |
| Stitch resolution | ≥2 element edges per stitch spacing, or homogenize with calibrated interface | Stitch-by-stitch rupture vs. homogenized response |
| Convergence | Halve seam element size until peak load changes <5% and initiation row is stable | Convergence plot of peak load vs. element size |
| Time step | Local refinement only; verify mass scaling <1–2% of local mass | Energy balance and added-mass report |
The seam model is only as trustworthy as its calibration. A shell-only model with a tied seam can predict the global cushion pressure history well while completely missing a stitch tear-out that occurs at 80% of burst pressure. The analyst should therefore decide early which failure mode the model must capture, choose the element strategy accordingly, and document the mesh convergence evidence that supports the chosen seam-region element size.