Accelerating hybrid lattice structures design with machine learning
Chenxi Peng, Phuong T. Tran, Erich Rutz
Abstract
Lattice structures inspired by triply periodic minimal surfaces (TPMS) have attracted increasing attention due to their lightweight properties and high mechanical performance. Recent research showed that hybrid structures based on the topology of two or more types of TPMS can present interesting multifunctional properties. However, the complexity of TPMS-based lattice designs presents challenges in both design and evaluation. To address these challenges, this study was designed to explore the integration of the machine learning method to predict the mechanical properties of hybrid lattice structures inspired by TPMS based on their patterns. A back propagation neural network (BPNN) was designed and trained on a dataset generated through finite element (FE) simulations and homogenization methods. The BPNN demonstrated robustness in predicting elastic modulus and Poisson’s ratio of TPMS hybrid lattice structures, offering rapid and efficient predictions. Validation against FE simulations confirmed the accuracy and reliability of the BPNN predictions, proving its potential as a valuable tool for accelerating the design and evaluation of complex hybrid lattice structures.