Data-driven optimization design of a novel combined metamaterial for robotic grinding vibration control

In order to suppress low-frequency vibration induced by low structural stiffness and rapid point-to-point motion in robotic grinding, a novel combined mechanical metamaterial is proposed. Two cosine beams are integrated into a previously developed lattice and are aligned along the load direction, enabling near-zero load conditions within geometric constraints. A data-driven bi-objective optimization algorithm is applied, utilizing adaptive constraints to maximize effective Young's modulus and to tailor the band gap. A two-output machine learning model is trained to separately predict the upper and lower boundaries of the band gap. To address the discontinuity caused by band gap jumps, the model incorporates a jump feature and a target-weighted loss, thereby achieving stable and accurate predictions. Compared with the original lattice, the optimized metamaterial is shown to exhibit a 2.5-fold increase in effective Young's modulus (from 150.34 Pa to 374.38 Pa), a reduction in the lower bound of the first band gap (from 18.33 Hz to 9.47 Hz), and an expansion of both band gap range and count. Experimental validation is conducted, demonstrating an approximately double reduction in minimum vibration transmission values, thereby confirming the effectiveness of the design. The provided efficient and accurate data-driven framework is potentially applicable to the simultaneous optimization of two properties in metamaterials.