Data-Driven Shape Optimization of Lattice Structures for Low-Frequency Broadband Bandgaps

Low-frequency vibration control is an important challenge in aerospace, energy, and precision engineering.However, conventional damping devices are limited by frequency-dependent loss characteristics and size constraints.Phononic metamaterials can provide effective vibration attenuation by forming bandgaps that suppress elastic wave propagation; however, designing lattice structures with low-frequency broadband bandgaps requires efficient exploration of a high-dimensional geometric design space.In this paper, a data-driven shape optimization framework combining Gaussian process regression and Bayesian optimization is proposed.The strut geometry is parameterized using symmetric Bézier curves to allow continuous radius variation, and the dispersion relation is evaluated through finite element analysis.Consequently, the optimized structure forms a first complete bandgap from the normalized frequency f n = 0.1767 to 1.1463, corresponding to a relative bandgap of 143.17%, whereas the cylindrical lattice does not exhibit a complete bandgap.Transmission loss analysis further confirms that the predicted bandgap region agrees well with the attenuation range of the finite lattice structure.In terms of structural performance, the optimized structure reduces the normalized maximum von Mises stress by 33.1% compared with the discrete resonant structure.These results demonstrate that Bézier curve-based shape optimization improves the balance between broadband vibration attenuation and structural stability.