Tunable energy absorption in 3D-printed data-driven diatom-inspired architected materials

Abstract

Boosted by additive manufacturing, architected materials have opened new opportunities to extend the performance of engineering materials. Yet, their development is held back by the intense efforts required to understand their complex property-structure-process-performance relationship. Therefore, data-driven biomimetic approaches are becoming increasingly popular to unveil such relationship. Here we mimic the functionally graded structures found in Coscinodiscus sp. diatom to understand the role of their shapes and define new guidelines for the design of novel architected honeycombs with tunable mechanical properties. Finite element simulations, validated on the outcome of a testing campaign performed on 3D-printed elastomeric samples, are used to build a dataset for machine learning algorithm training. Different machine learning techniques are used to link the geometric features of the designed biomimetic structures to their energy absorption properties and, in particular, to the specific absorbed energy divided by the peak force, here used as the performance index. The proposed approach leads to a novel design, which features a performance increase of 250% w.r.t. conventional honeycombs.