A hybrid machine learning framework for predicting and optimizing the compressive strength and energy absorption of 3D-printed PLA lattices

This research employs hybrid machine learning techniques to optimize the mechanical properties of 3D-printed polylactic acid (PLA) lattice structures. By integrating random forest (RF) and XGBoost (XGB) algorithms with metaheuristic optimizers including genetic algorithm (GA), particle swarm optimization (PSO), and simulated annealing (SA), the study develops predictive models for compressive strength and energy absorption. Results show that the hybrid approach is much more accurate in any prediction. The XGBoost model using particle swarm optimization showed good performance in compressive strength (R2 = 0.905), whereas random forest using simulated annealing showed good predictions for both compressive strength (R2 = 0.769) and energy absorption (R2 = 0.799). Critically, the analysis showed that the effective printing parameters are highly geometry specific and for example Hex-Truss structures had optimal printing parameters of 60% infill density, while Octagon struts had optimal parameters with only 10% infill. These results give engineers a powerful data-driven framework for the design of application-specific lattice structures that allows for customized mechanical performance needs for aerospace, automotive and biomedical applications, without the requirement for expensive iterations of experimental designs. The research establishes hybrid machine learning as a robust data driven framework to support the development of additive manufacturing capabilities.