3. Anthropometry and material properties for HBM development
A Human Body Model (HBM) is only as credible as the geometry and material data behind it. Two models can share the same solver, contact definitions, and restraint setup, yet produce injury predictions that differ by 30% or more if their anthropometry and tissue properties diverge. This chapter explains how HBM developers source, scale, and validate the two foundational inputs — body dimensions and material behavior — and how population variability is represented so that design decisions (restraint tuning, padding thickness, occupant packaging) are made against a defensible range rather than a single nominal body.
Anthropometry: from survey data to model geometry
HBM geometry is rarely digitized from one individual. The standard workflow starts with statistical anthropometry — typically stature, sitting height, biacromial breadth, chest depth, and limb segment lengths from large-scale surveys such as ANSUR II or CAESAR — and constructs a parametric body that matches a chosen percentile and body type. The GHBMC and THUMS families, for example, are built around a 50th-percentile male, a 5th-percentile female, and a 95th-percentile male, with obese variants added because BMI changes both mass distribution and belt interaction geometry.
The critical engineering point is that percentile selection is not a single scalar. A "50th-percentile" occupant is 50th in stature but may be 60th in chest depth and 40th in limb length. When a model is assembled, each segment should be assigned from the same multivariate distribution (using a correlation matrix from the survey) rather than independently picking 50th-percentile values for every dimension — otherwise the resulting body is a statistical fiction that no real occupant resembles.
Scaling laws: reusing a validated mesh
Building a full finite-element body from scratch is expensive, so developers scale an existing validated model to new dimensions. Two scaling families dominate:
- Geometric (isometric) scaling — all coordinates are multiplied by a length ratio derived from stature or a segment length. This preserves shape but ignores that tissue mass and stiffness do not scale linearly with length.
- Mass-consistent scaling — node positions are scaled geometrically, but nodal masses are rescaled so total body mass and segment inertias match the target anthropometry. This is essential for injury metrics that depend on momentum transfer.
A practical scaling factor for a segment is the ratio of target to source segment length, λ = Ltarget / Lsource. Nodal mass then scales as mtarget = msource · (Mtarget/Msource) · (Lsource/Ltarget) when mass is distributed proportionally to segment mass fraction. In practice, developers apply scaling per body region (head, thorax, abdomen, pelvis, limbs) because a single global factor distorts joint locations and belt anchor paths. After scaling, joint centers and belt routing must be re-checked against the target anthropometry — a step that is frequently skipped and is a common source of unrealistic belt fit in scaled models.
Tissue characterization: where the numbers come from
Material properties in HBMs are derived from three sources: cadaveric coupon testing, in vivo indentation or imaging, and inverse identification from whole-body impact tests. Each tissue type has a characteristic constitutive form and a characteristic test method:
| Tissue | Typical constitutive model | Primary characterization method | Key parameters |
|---|---|---|---|
| Cortical bone (rib, femur) | Elastic–plastic with damage | Three-point bending, coupon tension | E, yield stress, ultimate strain |
| Trabecular bone | Crushable foam | Uniaxial compression of cubes | Plateau stress, densification strain |
| Brain tissue | Viscoelastic / hyperelastic | In vitro shear, indentation | Shear modulus, relaxation time |
| Liver, spleen | Viscoelastic, nearly incompressible | Compression, indentation | Bulk modulus, relaxation |
| Muscle | Fiber-reinforced hyperelastic | Tensile testing along/across fibers | Fiber direction, stiffness ratio |
| Skin, fat | Hyperelastic | Indentation, suction | Initial stiffness, locking stretch |
For bone, the most decision-relevant property is not Young's modulus but the fracture strain and the rate dependence of strength. Rib cortical bone, for example, shows strength increasing with strain rate, which is why chest deflection tolerance curves are rate-dependent. A model calibrated only at quasi-static rates will under-predict rib fracture in a 40 km/h sled test. Developers therefore calibrate cortical bone to dynamic coupon data (typically 0.1–10 s⁻¹) and verify against whole-rib bending tests at matched rates.
For soft organs, inverse identification is common: a liver model's parameters are tuned until simulated compression force–deflection matches cadaveric organ compression data. This is legitimate but must be documented, because inverse-fitted parameters can compensate for geometric error and produce a model that matches one test while generalizing poorly.
Population variability: representing the range, not the average
Design decisions are made against a population, not a nominal body. HBM practice represents variability in three ways:
- Discrete variants — a small set of models (5th female, 50th male, 95th male, obese) that bracket the population. This is the most common approach and is sufficient for restraint geometry checks.
- Parametric families — a model whose dimensions and material properties are driven by a parameter set, allowing Monte Carlo sampling of anthropometry and tissue properties.
- Probabilistic material assignment — material parameters (e.g., bone strength) sampled from distributions derived from cadaveric scatter, propagated through the model to produce injury risk distributions.
The choice depends on the decision. For belt anchor placement, discrete variants are adequate. For setting an injury criterion threshold, probabilistic sampling is more defensible because it exposes the sensitivity of the predicted risk to tissue-property scatter — which is often larger than the sensitivity to geometry.
Putting it together: a worked example
Suppose a restraint team needs to check belt fit and chest deflection for a 5th-percentile female occupant in a 40 km/h frontal sled. The workflow is:
- Select the 5th-percentile female anthropometry from the survey correlation matrix (stature ≈ 151 cm, sitting height ≈ 79 cm, chest depth ≈ 19 cm — illustrative values).
- Scale the validated 50th-percentile male mesh geometrically to the target segment lengths, then rescale nodal masses to the target total mass (≈ 50 kg, illustrative).
- Re-assign material properties: do not inherit the male rib cortical thickness; use female-specific rib geometry and bone strength distributions from the literature.
- Run the sled simulation and extract chest deflection, belt force, and rib strain. Compare against the deflection-based injury criterion.
- Repeat with the 50th and 95th variants and with material parameters sampled from their distributions to obtain a risk range rather than a point estimate.
The expected observable is a deflection time history and a peak value; the judgment is whether the peak falls below the criterion with margin across the sampled population. A common failure symptom is a scaled model that produces unrealistically low deflection because the scaled rib cage is too stiff — trace this to inherited male cortical thickness and re-assign female bone properties before drawing conclusions.
Anthropometry and material properties are not independent inputs — they interact through scaling and through the structural behavior of the assembled body. Treating them as a coupled, documented, population-aware input set is what separates an HBM that supports design decisions from one that merely produces plausible-looking animations.