Graded honeycombs with high impact resistance through machine learning-based optimization

Abstract

Gradient structures with enhanced performance are ubiquitously observed in nature and in engineering materials. In this paper, we studied the impact resistance of two types of broadly used honeycomb structures (HCSs), a hexagonal HCS and an auxetic HCS. We developed a neural network (NN) which could effectively help to find an optimal gradient design for energy absorption of HCSs in contrast with their uniform counterpart. The optimal density gradient for both hexagonal HCS and auxetic HCS was identified, which are 66% and 40% higher in energy absorption than their respective uniform control. Followed finite-element analysis revealed that density gradient of HCSs enables loading transfer among a greater deformation zone, consequentially more cells involving in energy absorption. The initially graded sample promotes a de-gradient process and leads to more homogeneous density; conversely, a uniform sample develops localized deformation when subject to impact loading. Such an equal-load-partition (ELP) strategy in graded HCSs is responsible for their supreme energy absorption. The developed machine learning (ML) method for impact resistance optimization and the revealed deformation mechanisms in graded HCSs would be meaningful for the design of new advanced graded materials.