Reinforcement Learning approach of switching bi-stable oscillators to adapt bandgaps of 1D-meta-structures

Meta-structures with dynamic vibrational resonators (DVRs) are programmed to control the propagation of waves and attenuate vibrations over a broadband frequency spectrum. Attributes of DVRs, such as their resonant frequency and mass, determine the location and width of the bandgap, respectively. As a result, to adaptively program bandgaps, one has to modify or tune the eigenvalues of individual DVRs, and a popular approach is to vary the stiffness of each resonator. However, the tunable range of bandgaps is often restricted to maximum change in DVRs' stiffness. This work presents a novel approach to adaptively program bandgaps of a 1D flexural meta-structure over a broad frequency bandwidth. DVRs with two stable configurations are attached to a beam in developing the meta-structure. A numerical model is developed to illustrate the scope of the novel approach. An experimental investigation then validates the simulated results and shows the extent of the vibration absorption capabilities of the meta -structure. A reinforced learning approach is used to adaptively tune the bandgap over 220 Hz to 840 Hz.

Automated summary

Meta-structures with dynamic vibrational resonators are used to control wave propagation and attenuate vibrations over a broadband frequency spectrum. A novel approach is presented to adaptively program bandgaps of a 1D flexural meta-structure, using DVRs with two stable configurations. The proposed method is validated through numerical modeling and experimental investigation.