INTRODUCTION:
Helmets and other protective equipment must injury risk over a wide range of impact severities, but current liner design still relies heavily on trial-and-error and single-condition optimization. This work introduces a general data-driven inverse material design framework for helmet liners that begins with biomechanically defined target stress–strain curves representing desired macroscopic behavior under different impact energies. These targets, derived from head impact simulations and injury-oriented objectives, are then used to guide the search for materials and structures that can reproduce the required mechanical response.
METHODS:
The framework first uses a validated head–neck finite element model and AI-based optimization to generate target stress–strain curves for helmet liners at multiple impact energies, deriving desired trade-offs between strength, energy absorption, and allowable accelerations. A dynamic compression database of conventional foams and architected metamaterials is then built from pendulum impact tests, and each candidate liner material is scored with a multi-objective metric that combines peak stress error, specific energy absorption error, and curve similarity across elastic and plastic regimes. Top-ranked materials and layered liner architectures are subsequently implemented in the same helmet impact simulation environment, where their performance is quantified using peak linear and rotational head accelerations and summarized with a generalization score that captures how consistently a given design performs across all targeted impact conditions.
RESULTS:
Across three impact energies (50, 100, and 300 J), the framework selected architected materials that closely matched the target stress–strain curves and substantially improved head kinematics in simulation compared with EPS foams. For single-layer configurations, the best candidates reduced peak linear acceleration by up to 63% at 50 J, 44% at 100 J, and 18% at 300 J, and reduced peak rotational acceleration by 36%, 24%, and 10% at the same energies, relative to conventional foam baselines. When extended to layered architectures, two-layer lattices achieved high specific energy absorption (up to 6.30 MJ·m⁻³) with reduced peak stress (down to 1.78 MPa) and showed the lowest generalization scores for both linear (212%) and rotational (93%) accelerations, indicating robust performance across all impact conditions.
CONCLUSION:
This study introduces a transferable inverse design methodology that links AI-defined macroscopic targets based on biomechanics, experimental material databases, numerical scoring, and head impact simulations into a single selection and evaluation workflow. By focusing on performance metrics and target curves rather than on any specific material technology, the framework can be extended to other sports, impact scenarios, and material systems, and can guide future development of protective equipment that is tailored to sport-specific and multi-impact requirements.