Pediatric Crash-Test Neck

A neck is only "right" if it bends like a child's, so each team defined response corridors: acceptable ranges of force or moment versus displacement or angle. I built our team's bending corridor in two parts:

• Passive response (low loads): segment-by-segment moment–angle curves from 6-year-old cadaver data (Luck, 2019), combined into a full cervical spine curve.
• Active response (higher loads): muscle-driven behavior fit through maximum range-of-motion data from children (Arbogast et al., 2007) and maximum voluntary neck moments (Vincent et al., 2006).
• Blending the two: a sigmoid activation function (after Vasavada et al., 1998) shifts weight from the passive curve to the active curve as moment increases, mimicking how muscles are recruited under load.

The approach held up at low loads but over-extrapolated to non-physical angles at high moments. For validation, the class combined every team's bounds with the published Hybrid III 6-year-old corridor (Irwin & Mertz, 1997) into the final corridor shown here, which drew most heavily on my team's corridor.

The final design keeps the Hybrid III's mounting interfaces and exact 109 mm length so it drops into the existing dummy, but rebuilds the column from 12 parts: butyl rubber segments with cut slits, chloroprene rods, and aluminum plates. The key idea was to make the response change over the course of a crash, the way a real neck does:

• Catch mechanism: the top plate hangs free of a lip until tension builds, so early bending lets the head lag naturally and the catch only engages later to stop the head from whipping forward.
• Stopper: a rigid lip contacts the upper disk only after enough rotation, adding a late moment that drives further rotation.
• Compression rods: stiffer chloroprene rods sit just short of the top plate, so they add stiffness only in compression and stay out of the way in bending and tension.
• Slit geometry: cutouts in the rubber segments set how easily each level bends. The lowest, most loaded segment has smaller cuts and a wider rear section so it doesn't overextend.

Every design was run through six standardized LS-DYNA simulations (tension, compression, flexion, extension, a frontal sled pulse (NBDL), and a CHOP sled test) using the class's automated MATLAB/LS-PrePost pipeline, then scored against the corridors with a weighted objective rating. We iterated through 38 designs.

Making iteration fast enough. Early models used 2 mm elements around the fine, curved slit geometry, and a single NBDL run took 14 hours. I redesigned the curved slit and rod features as rectangular geometry so the neck could mesh cleanly at larger element sizes, then ran a mesh-size study to find the limit: 8 mm elements preserved the expected response, while 10 mm was too coarse for the simulations to run. At 8 mm, the NBDL run dropped from 14 hours to about 20 minutes, which is what made 38 iterations practical.