Machine learning-enhanced finite element analysis for performance-optimized lattice structures in additive manufacturing

Common lattice unit cells, such as face-centered cubic (FCC) and body-centered cubic (BCC), find a broad range of applications in lightweight engineering but suffer from poor mechanical strength and stability when subjected to complex loading conditions. The present study seeks to address this gap by providing an analysis of the mechanical synergy of hybrid topologies through the design and analysis of advanced lattice configurations, including FCC with centre struts, hybrid BCC–FCC and hybrid BCC–FCC structures using both vertical and horizontal struts. To address the high costs associated with traditional computational testing, we have created a supervised machine learning (ML)-based method for predicting the stress-strain relationship using Gaussian Process Regression (GPR) through the use of a Matern 5/2 kernel. The data used for creating the predictive model were derived from Finite Element Analysis (FEA) simulation points. According to the results of FEA and GPR, the through hybrid BCC-FCC reinforced with central struts provides better functionality than all other combinations. Specifically, the highest maximum stress for the combined configuration of BCC and FCC hybrid (3520 MPa) was 23% greater than the highest load-bearing capacity for any of the unreinforced designs. In addition, the validation results showed that the predicted data correlated exceptionally well with the experimental data, with an average minimum error of 11.8% for standard lattices and 3.64% for BCC-FCC reinforced with central struts. Thus, the results indicate that GPR modeling can provide a cost-effective and energy-efficient means of optimizing the design of advanced metamaterials in additive manufacturing (AM), such as in high-performance aerospace and biomedical applications.