2. Vehicle body-in-white digital twin architecture

A body-in-white (BIW) digital twin for crashworthiness is not a single model. It is a governed chain of representations: the CAD master that defines nominal geometry, the CAE model that discretizes it, the material and connection data that give it physical meaning, and the traceability layer that keeps every downstream result tied to a specific design revision. The architecture question is therefore not "which solver do we use" but "how do we keep geometry, mesh, materials, and boundary conditions consistent across hundreds of load cases and dozens of design iterations."

CAD/CAE Integration: From Nominal Geometry to Simulation-Ready Model

The CAD master for a BIW typically arrives as a multi-body assembly: floor pan, front longitudinal rails, shock towers, A/B/C pillars, roof rails, rockers, cross members, and closure reinforcements. Crash CAE does not consume this geometry directly. The integration layer must perform a controlled transformation:

• Geometry simplification: remove non-structural fillets below a threshold (commonly 3–5 mm), suppress bolt holes that do not affect crush mode, and defeat small embossments that would force excessive mesh refinement.
• Midsurface extraction: for stamped steel and aluminum panels with thickness-to-span ratios below roughly 1:10, extract midsurfaces and assign shell elements. Thick castings and extrusions remain solid.
• Feature preservation: keep beads, ribs, and spot-weld flanges intact — these control folding patterns and must not be idealized away.
• Assembly positioning: apply the vehicle coordinate system (X forward, Y lateral, Z vertical, origin at front axle centerline or a defined datum) consistently across CAD, CAE, and test data.

A practical integration pattern is a revision-locked pipeline. The CAD assembly is exported to a neutral format (STEP AP242 or JT), tagged with a revision ID, and imported into the pre-processor. The CAE model references that revision ID in its metadata. When the CAD changes, the pipeline flags affected parts rather than silently rebuilding the entire model. This prevents the classic failure mode where a mesh was generated from an obsolete geometry revision and the resulting crash result is physically meaningless.

Material Models for Crashworthiness

Crash simulation is dominated by plasticity, strain-rate sensitivity, and failure. The material model must capture these, not just elastic stiffness. For BIW steels, the standard choice is an elastoplastic model with isotropic hardening and strain-rate scaling, typically *MAT_PIECEWISE_LINEAR_PLASTICITY (MAT_24) in LS-DYNA or an equivalent in Abaqus/Pam-Crash. For advanced high-strength steels (AHSS) such as DP780 or boron steel in hot-stamped pillars, the hardening curve must be defined to large plastic strains, often beyond 0.5 true strain, using data from tensile and bulge tests.

Key parameters and their typical sources:

For aluminum extrusions and castings, the same framework applies but with lower ductility and different strain-rate sensitivity. For spot welds, the material model is not a continuum — it is a connector. A common approach is a beam or solid element cluster with a failure criterion based on normal and shear force, calibrated from lap-shear and coach-peel coupon tests. The weld model must reproduce both pull-out and interfacial fracture modes.

Strain-rate effects matter most in the first 20–50 ms of a frontal impact, where strain rates in the front rails can reach 100–1000 s⁻¹. Ignoring rate sensitivity typically underpredicts energy absorption by 10–20% in mild steel and less in AHSS, but the error is design-relevant because it shifts the crash pulse.

Mesh Strategy for Crashworthiness

Mesh strategy is where model fidelity and computational cost are traded. For BIW crash models, the dominant element type is the Belytschko–Tsay or fully integrated shell with 5 integration points through thickness. Solid elements are reserved for castings, extrusions, and weld nuggets.

Recommended baseline targets for a full-vehicle frontal impact model:

• Shell element size: 5–8 mm in crush zones (front rails, bumper, crash boxes), 10–15 mm in non-crush zones (floor, roof, rear structure).
• Aspect ratio: below 3:1 in crush zones, below 5:1 elsewhere.
• Warpage: below 15° for shells.
• Jacobian: above 0.6 for all elements; below 0.3 is a hard reject.
• Hourglass energy: below 5% of internal energy for the whole model; investigate any part exceeding 10%.
• Time step: controlled by the smallest element; a 5 mm steel shell gives roughly 1 µs, so mass scaling is often applied — but keep added mass below 1–2% of total vehicle mass.

Mesh refinement must follow the expected deformation mode. In a front rail, the folding wavelength is typically 30–80 mm, so at least 6–10 elements across one fold are needed to capture the hinge lines. A 5 mm mesh gives 6–16 elements per fold — adequate. A 15 mm mesh gives 2–5 — too coarse, and the rail will fold artificially or not at all.

Connection modeling is part of mesh strategy. Spot welds are represented by one beam or solid element per weld, with the weld diameter matched to the physical nugget (typically 5–7 mm for 1–2 mm sheet). Adhesive bonds are modeled with cohesive elements or tied contact with a failure criterion. The mesh at flanges must be aligned so that weld elements connect coincident nodes or a tied interface — misaligned flanges create artificial compliance and change load paths.

Putting the Architecture Together

The three pillars — CAD/CAE integration, material models, and mesh strategy — are not independent. A mesh decision changes the effective material response: coarse elements delay localization and artificially stiffen the structure. A material card calibrated on 5 mm coupons may not be valid for a 15 mm mesh without regularization. A CAD simplification that removes a bead changes the folding mode, which changes the strain-rate history, which changes the material response.

The practical consequence is that the digital twin must be validated as a system. A recommended workflow for a new BIW program:

1. Build the CAD master and lock a revision.
2. Generate the mesh with the zone-based sizing rules above and run a mesh quality audit.
3. Assign material cards from the approved material database, with strain-rate and failure data traceable to test reports.
4. Run a component-level sled test simulation (e.g., front rail + bumper) and correlate against physical sled data. Adjust mesh density or material regularization if force–displacement correlation error exceeds 10%.
5. Assemble the full-vehicle model, run the frontal ODB and side MDB load cases, and check hourglass energy, added mass, and energy balance before interpreting occupant or structural results.
6. Store the validated model as a revision-locked digital twin artifact with its manifest.

When this architecture is in place, the BIW digital twin becomes reusable: a design change to the front rail triggers a targeted mesh and material update, not a full rebuild, and every result remains traceable to a known CAD revision, material card, and mesh quality state. That traceability is what separates a digital twin from a one-off simulation.