2. Human body model fundamentals for restraint engineering

Restraint system optimization depends on a validated computational surrogate of the occupant. Three families dominate automotive crash safety practice: THUMS (Total Human Model for Safety, Toyota Central R&D Labs), GHBMC (Global Human Body Models Consortium), and the SAFER HBM (Chalmers/SAFER, Sweden). They differ in mesh strategy, material formulation, joint representation, and — most importantly for engineering decisions — validation coverage. This section explains what each model actually contains, how to select one for a given restraint task, and how to judge whether its validation status supports your load case.

Mesh Architecture: What the Element Count Really Tells You

All three models are explicit-dynamics-oriented finite element (FE) assemblies, but their mesh philosophies diverge. THUMS (v5–v7) uses a predominantly hexahedral-dominant mesh with roughly 1.8–2.0 million elements for the full body, including solid organs, a detailed brain with sulci, and a full skeletal system. GHBMC M50-O (occupant, 50th percentile male) is a hexahedral-dominant model of about 2.2 million elements, with a separate simplified version (~1.0 million) for faster iteration. SAFER HBM is built on a different premise: a coarser, shell-and-solid hybrid mesh (~300,000–500,000 elements) designed for robustness and fast turnaround in parameter studies.

The practical consequence is a trade-off between biofidelity resolution and simulation cost. A THUMS or GHBMC full-body run at 1 ms time step on 32 cores typically requires 8–20 hours for a 150 ms frontal impact; SAFER HBM can complete the same event in 1–3 hours. For restraint optimization — where hundreds of design evaluations are needed — this cost difference is decisive. The recommended workflow is to use a coarse model (SAFER HBM or GHBMC simplified) for design-space exploration, then confirm the final candidate with a detailed model.

Element quality matters more than count. Check Jacobian ratio (>0.3), warpage (<15°), and aspect ratio (<5) in the deformable regions (brain, liver, lungs) before trusting any result. A 2-million-element model with poor organ elements is less useful than a 500,000-element model with clean mesh.

Material Models: Strain-Rate Sensitivity and Failure

Soft tissues are visco-hyperelastic. THUMS and GHBMC use Ogden or Mooney-Rivlin hyperelastic formulations for brain, liver, and lungs, with strain-rate-dependent scaling via tabulated stress–strain curves at multiple rates (0.01, 0.1, 1, 10, 100 s⁻¹). Bone uses a piecewise linear plasticity model with element deletion at a strain threshold (cortical bone ~1.5–2.0% for fracture initiation). SAFER HBM uses similar constitutive families but with fewer calibration points, which reduces fidelity in high-rate organ loading but improves numerical stability.

For restraint engineering, the critical material parameters are:

  • Rib cortical bone: elastic modulus ~11–15 GPa, yield ~80–120 MPa, failure strain ~1.5–2.5%. Rib fracture prediction is the primary injury metric for belt and airbag tuning.
  • Liver: hyperelastic with failure strain ~30–40% (Lagrangian). Liver injury correlates with belt loading in submarining.
  • Brain: shear modulus ~10–50 kPa (rate-dependent), used for diffuse axonal injury (DAI) prediction via strain or strain-rate criteria.

When you swap material cards between models, do not assume equivalence. A GHBMC rib card in a THUMS mesh will produce different fracture timing because element formulation and hourglass control differ. Always use the material set shipped with the model version.

Joints: The Most Under-Appreciated Modeling Choice

Joints determine how the occupant moves relative to the restraint. THUMS uses a combination of deformable ligaments (knee, shoulder) and kinematic constraints (spine, hip) with prescribed stiffness curves. GHBMC uses more detailed ligamentous structures with nonlinear force–displacement curves and contact-based joint capsules. SAFER HBM uses simplified revolute and spherical joints with rotational stiffness and damping, tuned to match overall kinematics rather than local ligament behavior.

For restraint optimization, the joint model affects:

  • Submarining risk: pelvis–femur joint stiffness and hip flexion range control whether the pelvis slides under the lap belt.
  • Chest deflection: thoracic spine stiffness and rib–spine joint compliance determine how much the chest compresses against the belt.
  • Head excursion: cervical spine stiffness and neck muscle representation (THUMS and GHBMC include active muscle lines; SAFER HBM uses passive stiffness) affect head–airbag interaction timing.
If your restraint design relies on active muscle response (e.g., pre-crash braking or occupant bracing), only THUMS and GHBMC offer the necessary muscle activation framework. SAFER HBM is not suitable for that load case.

Validation Status: What Each Model Has Been Proven Against

Validation is not a single number. It is a matrix of load cases, body regions, and injury metrics. The table below summarizes the published validation coverage for the three models as of their current public releases.

ModelFrontal sledLateral impactRear impactPedestrianBelt loadingAirbag interaction
THUMS v5–v7Extensive (PMHS, sled)ModerateLimitedExtensiveExtensiveModerate
GHBMC M50-OExtensiveExtensiveModerateModerateExtensiveExtensive
SAFER HBMModerateModerateLimitedLimitedModerateLimited

“Extensive” means multiple published PMHS (post-mortem human subject) comparisons with corridor-based rating (e.g., CORA or ISO/TR 9790). “Moderate” means at least one validated load case with reasonable correlation. “Limited” means the model can run the case but published validation is sparse or absent.

For a frontal restraint optimization study, GHBMC M50-O and THUMS are both defensible. For lateral impact (side airbag, door trim), GHBMC has the strongest validation base. For pedestrian protection, THUMS is the de facto standard. For rapid design sweeps where absolute injury prediction is secondary to relative ranking, SAFER HBM is appropriate — but you must state that limitation in your report.

Selection Workflow for Restraint Engineers

The following decision logic is recommended before any optimization run:

  1. Define the load case and injury metric. Frontal sled with belt + airbag → rib fracture and chest deflection. Lateral → rib fracture and pelvic acceleration. Rear → neck injury (Nij).
  2. Check validation coverage. If your load case is not in the validated matrix, either switch models or plan a validation sub-study.
  3. Match model fidelity to iteration budget. >200 design evaluations → start with SAFER HBM or GHBMC simplified. <50 evaluations → use detailed THUMS or GHBMC.
  4. Verify joint and material consistency. Do not mix material cards across model families. Confirm that active muscle settings match your pre-crash scenario.
  5. Document the model version and validation basis. THUMS v5 vs. v7, GHBMC M50-O v2.3 vs. v2.4, and SAFER HBM v9 vs. v10 have different mesh and material updates. Your optimization result is only reproducible if the version is recorded.
HBM Selection Logic for Restraint Optimization 1. Define Load Case Frontal / Lateral / Rear / Pedestrian Injury metric: rib fracture, Nij, etc. 2. Check Validation Matrix Is your load case covered? If not → switch model or validate 3. Match Fidelity to Budget >200 runs → coarse model <50 runs → detailed model THUMS v5–v7 ~1.8–2.0M elements, hex-dominant Active muscle, detailed brain Best: frontal, pedestrian Cost: 8–20 h / 150 ms run Use for: final confirmation GHBMC M50-O ~2.2M elements (full) / ~1.0M (simple) Detailed ligaments, active muscle Best: frontal, lateral, airbag Cost: 6–18 h / 150 ms run Use for: lateral + frontal design SAFER HBM ~300–500K elements, hybrid mesh Simplified joints, passive muscle Best: rapid design sweeps Cost: 1–3 h / 150 ms run Use for: ranking, not absolute injury 4. Verify Joint & Material Consistency → 5. Document Version & Validation Basis Do not mix material cards across model families. Record THUMS v5 vs v7, GHBMC v2.3 vs v2.4, SAFER v9 vs v10. Detailed / high fidelity Balanced Coarse / fast All run times are indicative for 32-core explicit solver, 150 ms event.

Practical Notes on Model Setup

Before running any restraint optimization, verify the following in your deck:

  • Gravity settling: Run a 50–100 ms gravity ramp before impact. Without it, the occupant starts in an unnatural position and belt fit is wrong.
  • Belt routing: The belt must be modeled as a 1D seatbelt element with slipring and retractor. THUMS and GHBMC both provide belt routing templates; SAFER HBM requires manual routing.
  • Contact definitions: Use penalty-based contact with a scale factor of 0.1–0.3 for belt-to-occupant. Too stiff → numerical instability; too soft → belt penetration.
  • Time step: The smallest element controls the stable time step. For THUMS, typical Δt is 0.5–1.0 μs. Use mass scaling only in non-critical regions (e.g., lower legs) and document the added mass.

These checks are not optional. A model that runs but has a 5 mm belt penetration or a 10% mass-scaled torso will produce injury metrics that look plausible but are not valid for restraint design decisions.

Summary

THUMS, GHBMC, and SAFER HBM are not interchangeable. THUMS offers the highest biofidelity for frontal and pedestrian loading with active muscle. GHBMC provides the broadest validation matrix, especially for lateral impact and airbag interaction. SAFER HBM trades fidelity for speed, making it suitable for design-space exploration but not for final injury prediction. The engineer’s task is to match the model’s validated load cases and fidelity level to the restraint problem at hand, then document that choice so the optimization result is reproducible and defensible.

← Previous 📖 Contents Next →