Machine Learning-Guided Design of Shell-Based Multistable Lattices for Superior Energy Absorption and Reusability

Hollow-tube micro/nanolattices are widely used as energy-absorbing materials due to their excellent recoverability and energy absorption capabilities. However, they often suffer from permanent node damage, resulting in limited reusability. Multistable metamaterials provide a promising alternative for reusable energy absorption. Yet, their load-bearing and energy-absorbing capacities are frequently constrained by instability mechanisms inherent to tilted/curved beams. Here, a new type of shell-based unit cell with straw-like geometry and continuously varying thickness is designed by integrating finite element simulations and machine learning. These unit cells exhibit tunable multistable behaviors governed by their geometric parameters and can be readily fabricated via 3D printing. Multistable microlattices comprising these unit cells, fabricated using projection microstereolithography, demonstrate large deformability, excellent recoverability, reusability, and a remarkable combination of high strength (≈91.7 kPa) and superior energy absorption (≈6.2 × 10⁴ J m^-3), surpassing nearly all previously reported hollow-tube lattices. The versatility and scalability of the design are further highlighted by fabricating shell-based lattices from various constituent materials via different 3D printing techniques, as well as creating 3D hierarchical shell-based lattices exhibiting multistability in all three spatial directions. This study establishes a mechanomaterial route for designing multistable metamaterials through shell-based architectures, showcasing their potential for exceptional energy absorption and multistability.