Research on a Design Method of Generative Model for Cavity Section Shape of an Anechoic Coating

Abstract

One of the main challenges of the low frequency, broadband and strong sound absorption performance of an anechoic coating is the design of the cavity section shape. A generative-based design method was investigated in this study. Two generative-based models were trained to design the cavity section shape of anechoic coatings with a highly efficient absorption coefficient. The design results were compared in terms of mode collapse and average sound absorption coefficients. The results show that the WGAN-GP model can stably generate 100 new designs in only 3 seconds, dramatically speeding up the design process.