On-demand tunable metamaterials design for noise attenuation with machine learning

Metamaterials with structure-dominated properties provide a new way to design structures to obtain desired performance. To achieve a wide range of applications, on-demand tunable metamaterials would fulfill various and changing needs. The design of on-demand tunable metamaterials requires a higher-level understanding of the relationship between the properties of the metamaterials and the geometrical parameters, which in many cases are complicated and implicit. With the advancement of machine learning and evolutionary methods, it becomes possible to design on-demand tunable metamaterials. This paper designs on-demand tunable acoustic metamaterials for noise attenuation at varying frequencies by employing a genetic algorithm based neural network method. The C-shaped acoustic metamaterials with slidable shells are combined with the specifically designed tri-stable origami-inspired metamaterials to realize the on-demand tunable structure. Experiments were conducted and showed that the designed tunable metamaterials exhibited desired characteristics in different targeting frequency ranges. The present general methodology is expected to provide a route for on-demand tunable design while exploring more possibilities for the application of metamaterials.