4. Human body models for children

Child occupant restraint simulation depends on a validated representation of the child's body — its mass distribution, joint stiffness, tissue compliance, and geometric interaction surfaces. Unlike the adult case, where a small number of well-established Human Body Models (HBMs) dominate, the child HBM landscape is fragmented: dedicated paediatric models exist but are few, and a large share of practical work relies on scaling adult models down to child anthropometry. This chapter examines the three dominant strategies — the THUMS child family, the PIPER framework, and morphing/scaling of adult HBMs — and explains how to select and prepare a model for a restraint interaction study.

Why a dedicated child HBM matters

A child is not a uniformly shrunk adult. The head is proportionally larger and heavier, the neck is shorter and more compliant, the rib cage is more cartilaginous and deflects further before fracture, and the pelvis ossification centres are incomplete. These differences change how a seat belt loads the body and how a child restraint system (CRS) shell interacts with the occupant. A scaled adult model inherits adult material properties and joint behaviour, so it can reproduce gross kinematics but will misrepresent injury-relevant local response. The practical consequence: use dedicated child HBMs where injury metrics matter, and treat scaled models as kinematic surrogates for restraint-fit and packaging studies.

The THUMS child family

THUMS (Total Human Model for Safety), developed by Toyota Central R&D Labs, is a finite-element HBM family distributed for LS-DYNA. The child variants — commonly the 3-year-old (3YO) and 6-year-old (6YO) models, with a 10YO variant in some releases — are full FE models with deformable head, neck, thorax, abdomen, pelvis, and limbs, plus a detailed skeletal structure. They are the reference choice when the study needs organ-level or skeletal injury prediction, for example belt-to-abdomen loading or head impact against a CRS shell.

Typical preparation steps for a THUMS child model in a frontal CRS simulation:

1. Check units and solver version. THUMS is normally distributed in mm-tonne-s (mm, t, s, N, MPa). Confirm the deck's unit system before merging with a CRS model built in kg-m-s; a silent unit mismatch produces absurd accelerations.
2. Position the model. Use the supplied positioning deck or the THUMS positioning tool to set the HBM into the seated posture matching the CRS. Do not hand-rotate limbs; joint definitions are calibrated and manual rotation corrupts them.
3. Initialise contact. Run a gravity-settle or a short pre-simulation so the HBM seats into the CRS cushion and belt before the crash pulse. A common failure is a large initial penetration that produces an artificial spike at t = 0.
4. Apply the restraint. Define belt segments with the correct slip-ring friction and retractor characteristics; the belt must route over the HBM's actual shoulder and lap geometry, not a simplified dummy surface.

The PIPER framework

PIPER (Positioning and Personalisation of Human Body Models) is an open-source framework, originally developed in the EU PIPER project, for positioning, morphing, and personalising HBMs. Its child-relevant value is that it provides a scalable, reproducible pipeline rather than a single fixed model: you can take a baseline child HBM and morph it toward a target anthropometry (stature, mass, BMI, or a target surface scan) using a statistical shape model or radial-basis-function (RBF) morphing.

A practical PIPER workflow for a child occupant study:

1. Define the target. Specify the child's anthropometry — for example, a 6YO target of 116 cm stature and 21 kg — or supply a surface mesh from a body scanner.
2. Select the morphing method. Use the statistical shape model for population-representative targets, or RBF/control-point morphing when you have a specific scanned surface.
3. Morph and check. PIPER warps the mesh and updates the FE model. Inspect element quality (Jacobian, aspect ratio) after morphing; severe distortion near joints is the usual defect.
4. Re-position and export. Use PIPER's positioning module to set the seated posture, then export the LS-DYNA or OpenRadioss deck.

PIPER's strength is transparency and reproducibility: the morphing parameters are recorded, so a study can be repeated or extended to a different child size. Its limitation is that morphing changes geometry, not material properties — a morphed model still carries the baseline's tissue behaviour, so it is best used for kinematic and fit studies rather than injury thresholds.

Scaling approaches

When no dedicated child HBM is available, the fallback is to scale an adult HBM (or a dummy) to child dimensions. Two families of methods exist:

• Geometric scaling (uniform or non-uniform). Apply a scale factor derived from anthropometric ratios — for example, scale the adult model by the ratio of child-to-adult stature in each axis. This is fast and preserves mesh topology, but it scales material properties and joint stiffness incorrectly unless you also adjust them.
• Morphing-based scaling. Use RBF or statistical shape models (as in PIPER) to warp the adult mesh to child surface targets. This gives better local geometry but still inherits adult tissue behaviour.

The critical caveat: scaling is a kinematic tool, not an injury tool. A scaled adult model will predict child head excursion and belt fit reasonably, but its rib deflection and head-injury metrics are not validated for children. Use scaled models for restraint packaging, belt routing, and CRS fit; switch to a dedicated child HBM (THUMS child or a validated PIPER-morphed child model) when the study reports injury criteria.

Choosing a model: a decision guide

The figure below summarises how the three strategies relate: a baseline child HBM can be morphed via PIPER, an adult HBM can be scaled geometrically or morphed, and the choice of path determines whether the output is suitable for injury assessment or only for kinematics.

Practical verification and common failure modes

Whichever model you use, verify the setup before trusting the crash response. A short checklist that catches most errors:

• Unit consistency. Confirm the HBM deck and the CRS/belt deck share the same unit system. Symptom of mismatch: accelerations off by orders of magnitude, or the model exploding at t = 0.
• Initial penetration. Check the contact output at t = 0. Large penetration between the HBM and the CRS cushion or belt indicates the settle step was skipped or too short.
• Element quality after morphing. Inspect Jacobian and aspect ratio near the neck, shoulders, and hips. Distorted elements cause local stiffness artefacts and premature termination.
• Belt routing. Confirm the belt passes over the actual shoulder and lap geometry, not a simplified surface. A belt that floats off the shoulder under-predicts restraint and over-predicts excursion.
• Posture match. Compare the HBM's seated posture to the CRS's intended occupant posture. A mismatch in hip angle or head position changes belt fit and head excursion.

For a reproducible example, consider a frontal 6YO CRS study. Input: THUMS 6YO positioned in a forward-facing CRS, belt routed with slip-ring friction 0.3, retractor force-limited at 4 kN, crash pulse a 50 km/h frontal deceleration. Execution: settle 50 ms under gravity, then apply the pulse for 120 ms. Expected observable: head forward excursion and belt force time history; check that the belt force rises smoothly without a t = 0 spike and that head excursion is within the CRS's design envelope. Typical failure: a spike at t = 0 from initial penetration — fix by extending the settle step or adjusting the initial HBM position. These values are illustrative and should be replaced with the study's actual pulse and restraint parameters.

The choice among THUMS child, PIPER, and scaling is ultimately a question of what the study must defend. If the output is an injury metric, use a dedicated child HBM. If the output is a geometric or kinematic comparison, a morphed or scaled model is defensible — provided the report states clearly which model was used and why.