To enhance the vibration reduction performance of ships and underwater equipment in the medium-low frequency range, this work explores a method of collaborative optimization using deep neural networks (DNNs) and genetic algorithms (GAs) for locally resonant metamaterial plates (LRMPs). Addressing the shortcomings of conventional design methods, which rely on manual experience, show low computational efficiency, and feature narrow bandgaps of resonance units, an optimization framework that integrates surrogate models with intelligent algorithms is proposed. By establishing a finite element model of the LRMP, a DNN is trained to learn the implicit relationship between geometric parameters and vibration responses, which is utilized as a surrogate model to accelerate the fitness evaluation of the GA. Additionally, AutoML is introduced to construct an automated design framework, enabling the inverse design of the geometric parameters of the resonance units. Compared with conventional methods, this reduces computational and time costs by 95%. The optimized LRMPs exhibit more outstanding vibration attenuation performance in the target frequency band. This research provides new insights into the rapid optimization of complex acoustic metamaterials.

Keywords: Local resonant metamaterial plates; DNN; GA; optimization design; AutoML