Reaching a desirable metastructure for passive vibration attenuation by using a machine learning approach

The focus of this research is on designing a longitudinally excited lightweight metastructure that consists of external units distributed periodically, each enhanced with internal oscillators to serve as vibration absorbers. The metastructure initially exhibits uniformity, with all absorbers being identical to each other, being comprised of a cantilever that is integrated into the external parts of the metastructure, with each cantilever containing a concentrated mass block at its tip. Despite its simplicity and suitability for 3D printing, the design of the absorbers could not be kept in the original form when previous theoretical attempts were made with a view to achieving maximal vibration attenuation efficiency around the second resonance. To overcome this shortcoming and keep the absorbers in the original shape, this study undertakes a machine learning methodology to mitigate vibrations near the second resonant frequencies itself, as well as around the first and second resonant frequencies simultaneously. The newly designed metastructure is manufactured, and its advantageous vibration mitigation capabilities are experimentally verified qualitatively. Additionally, physical insight into the configuration and arrangement of the redesigned absorbers in the newly designed metastructure is provided.