Microstructural design of rigid porous materials using a Bayesian optimization method

This study presents a methodology for designing the periodic unit cell (PUC) to optimize the sound absorption properties of rigid porous materials using Bayesian optimization (BO). BO is a machine learning algorithm proficient in identifying the global optimum of a ‘black-box’ objective function with a limited set of observations. This BO methodology was applied in the design of two distinct types of PUCs: a body-centered cubic structure and a Kelvin cell structure. To verify the efficiency and robustness of the proposed approach, the optimization process was executed multiple times, each with varying initial random samples. In the case of the body-centered cubic design, the optimal PUC was ascertained by examining a mere 2.08 % of the total candidate designs; for the Kelvin cell design, this value was 4.33 %. The effectiveness of the BO-driven approach was validated by comparing it with random sampling method.