Free vibration behaviour of pomelo peel–inspired sandwich composite plates with additively manufactured cores: an explainable machine learning approach

Sandwich composite plates with bioinspired cellular cores offer an effective means of achieving high stiffness-to-weight efficiency and improved vibration performance in lightweight structural applications. This work investigates the free-vibration behaviour of glass-fibre/epoxy sandwich plates incorporating a pomelo peel–inspired polymeric core (PIM02) through a combined experimental, numerical, and interpretable data-driven framework. The elastic properties of the facesheets were experimentally characterised using the impulse excitation technique (ASTM E1876), while the shear properties of the additively manufactured PLA core were determined using an alternate dynamic method. Experimental testing under multiple boundary conditions was carried out to validate a finite-element model developed in ANSYS APDL using a First-Order Shear Deformation Theory (FSDT) formulation, with deviations within approximately 6–7%. The validated model was then used to generate a parametric dataset by varying elastic properties, shear moduli, density, fibre weight fraction, and laminate configuration. Supervised machine-learning regression models were trained to predict the first-mode natural frequency, with the Extra Trees Regressor demonstrating the best predictive performance (R 2 ≈ 0.997). Explainable AI analysis revealed that the in-plane and out-of-plane shear moduli have the strongest influence on the vibration response, highlighting the governing role of shear stiffness and providing guidance for vibration-resistant design of pomelo-inspired sandwich plates.