Static and Dynamic Behavior of Novel Y-Shaped Sandwich Beams Subjected to Compressive Loadings: Integration of Supervised Learning and Experimentation

In this article, an in-depth investigation into the mechanical response of novel Y-shaped core sandwich beams under static and dynamic compressive loading conditions is presented. Utilizing deep feed-forward neural networks (DFNNs) as the primary supervised learning scheme, the compressive behavior of these advanced structures is predicted. The trained DFNN model demonstrates high fidelity in capturing the stress-strain relationships, as evidenced by the close alignment of predicted and experimental results. Key design parameters of the cores of the sandwich beams are varied to understand their influence on the beams' linear, plateau, and densification regions, where higher values of design parameters contribute to increased stiffness, prolonged plateau regions, and higher densification points. Additionally, the impact of loading rates (1, 7, and 14 mm min-1) on the mechanical performance is analyzed, revealing significant rate-dependent behaviors. The decision tree algorithm exhibits superior classification performance with a 99.79% accuracy, further validating the robustness of the predictive model. In contrast, the support vector machine algorithm with radial basis function shows moderate accuracy at 75.12%. Through these findings, the potential of DFNNs in predictive modeling and the importance of design parameters and loading rates in optimizing the mechanical performance of novel Y-shaped core sandwich beams is proposed.