Accelerated Structural Design of Cellular Materials for Compressive Deformation Using a Machine-Learning

It is well known that cellular materials (including porous materials) are widely observed in engineered and nature systems, because their mechanical performance is excellent, such as compressive deformation and energy absorption against impact loading. The mechanical response is significantly dependent on their inherent cellular structure, i.e., geometric arrangement pattern. A nonuniform arrangement could provide a significant variation of mechanical performance, and then material selection and geometrical designs are challenge. This study established machine-learning (ML) based framework to design geometrical arrangement (architecture) in cellular material to achieve better mechanical performance against uniaxial compression. Especially, we investigated peak force at plateau region and work of energy absorption until structural densification. Cellular material having various pattern of internal geometry was modeled using finite element method (FEM), and we simulated uniaxial deformation behavior, which was used as training data (teaching data) for machine learning method. This study employed neural network (NN) for machine learning method, which connects cellular geometric pattern with mechanical performance (force - displacement curve and peak force - work of energy absorption relationship). Our results showed that the proposed framework is capable of predicting the mechanical response of any given geometric pattern within the domain of our setting. Thus, it is useful to discover cellular structure in order to achieve desired mechanical response.