| Microperforated panel (MPP) sound absorbers are capable of providing high sound absorption coefficients without the use of fibrous materials. However, they typically function in limited frequency ranges. By combining multiple MPPs into a multilayer absorber system, the absorption bandwidth can be increased while maintaining high absorption coefficients. Each additional MPP layer added to such a multilayer absorber increases the overall complexity. Therefore, Bayesian model selection, which quantitatively embodies the principle of parsimony, is well-suited to the design of multilayer absorber systems to achieve a minimum number of layers. To demonstrate the Bayesian design framework, a practical multilayer absorber is designed. Based on this predicted design scheme, MPP samples are fabricated and used to conduct normal-incidence sound absorption measurements in an impedance tube. Fabrication uncertainties may lead to imprecise final absorption results, since each layer must accurately reflect four specific MPP parameters in order to match the desired design scheme. Therefore, quantifying fabrication uncertainty is of practical significance to satisfy application-specific requirements. In this work, Bayesian parameter estimation is used to infer the actual effective MPP parameters of the manufactured samples, from the experimentally measured absorption data. These results demonstrate that the design scheme can be well satisfied with Bayesian probabilistic inference. As a result, the MPP fabrication process is controlled by adjusting original design values to attain the ideal design scheme. |
