Manufacturing-aware design of multi-material asymmetric acoustic absorbers

Abstract

Advances in additive manufacturing technology are promising for the creation of materials to control acoustic waves. This work presents the design of an acoustic metasurface (AMS) created from a periodic two-dimensional array of multi-material scatterers embedded in an elastomeric matrix. Finite element analysis (FEA) of the AMS shows that the absorption of acoustic energy is dependent on direction of incident wave propagation if the material properties are asymmetric within the scatterers while the acoustic transmission remains symmetric. Asymmetry of acoustic absorption is highly sensitive to small changes in scatterer geometry and property distribution. The AMS design must therefore consider inherent variability in manufacturing processes and the resultant stochastic asymmetric absorption performance of the as-built AMS, which is computationally expensive to determine using FEA. A machine learning classifier is trained to replace FEA and enable efficient robust design of the AMS via Monte Carlo simulations. The classifier is trained to be valid for a variety of characteristic manufacturing variations and the trivial expense to call it leads to significantly faster design with awareness of manufacturing variability. This work demonstrates how robust design approaches can be used to select fabrication methods suitable for acoustic materials. [Work supported by NSF and ONR.]