Machine Learning Guided Optimization of Biomimetic Polymeric Lattice Structures

Abstract

In this article, we report that, machine learning, an artificial intelligent technique, is used to optimize biomimetic rods and lattice structures. Various structures available in nature such as plant stems and roots that exhibit better buckling resistance are mimicked and modeled using finite element analysis, to obtain a training dataset. For validating the finite element analysis, uniaxial compression to buckling of additive manufactured biomimetic rods using a polymeric ink is performed. These model results are then formed into a dataset. Forward design and data filtering are conducted by machine learning to optimize the biomimetic rods from the dataset. The results show that the machine learning assisted rod designs have 150% better buckling resistance than all the rods in the training dataset, i.e., better than the nature’s counterparts. These optimal rods can be used in designing structures with superior buckling resistance such as in bridges, buildings, lattice structures, etc. Using these biomimetic rods, lattice structures with better structural performance are manufactured. While lattice unit cells such as octahedron, tetrahedron, octet, etc., have been previously proposed for lightweight structures, it is plausible that more optimal unit cells exist which might perform better than the existing counterparts. Machine learning technique is used to discover new optimal cells. Uniaxial compression tests using ANSYS are performed to form a dataset, which is used to train machine learning algorithms and form predictive model. The predictive model is then used to identify a total of 20 optimal symmetric unit cells. These new unit cells show 51%–57% higher capacity than octet cell. Particularly, if the porous biomimetic rods are used to construct the unit cells, an additional 130%–160% increase in buckling resistance is achieved. New lattice unit cells exhibit a buckling load of 261%–308% higher than the classical octet unit cell. Sandwich structures manufactured by 3D printing these optimal symmetric unit cells show 13%–35% higher flexural strength. This study opens up new opportunities to design high-performance metamaterials combining biomimetics and machine learning.