Reinforcement-learning empowered adaptive piezoelectric metamaterial for variable-frequency vibration attenuation

The complex environmental and variable-frequency excitation necessitate the adaptive capabilities of elastic metamaterials in vibration attenuation applications. This paper introduces an AI-empowered adaptive metamaterial featuring locally resonant units with each comprising a piezoelectric transducer, a voltage-controlled synthetic inductor and a digital control circuit. To overcome the frequency misalignment caused by uncertain variations in the equivalent capacitance of piezoelectric elements, a learning-based strategy is proposed for bandgap tuning to adapt to external changes. Specifically, a Twin Delayed Deep Deterministic policy gradient (TD3) agent is customized for the metamaterial tuning task, and a data-driven simulation environment is constructed based on it. Subsequently, the agent is trained offline to explore the optimal unified control strategy. To compensate for electrical differences among piezoelectric resonate units, a dual-stage strategy is tailored to deploy the learned policy. Experimental results confirm that the proposed method endows the metamaterial beam to achieve the optimal tuning of the vibration attenuation characteristics in the variable-frequency environment.