Explainable machine learning determines effects on the sound absorption coefficient measured in the impedance tube

Measurements of acoustic properties sound absorbing materials in impedance tubes show poor reproducibility, which was demonstrated round robin tests. The tube measurements are standardized but lack precise definitions the actual measurement setup, specimen preparation, and other factors that introduce uncertainty practice. In this paper, machine learning models identify those mostly affect absorption coefficient from a large data set more than 3000 spectra measured one tube. specimens manufactured polyurethane foam, different cutting technologies, operators, diameters, thicknesses, two approaches to mount considered. Explainable techniques allow identification quantification most influential and, furthermore, frequency ranges affected by choice these setup factors. results indicate besides thickness, also operator affects directional non-random relationship. Hence, it needs be controlled carefully. method proves promising pathway for knowledge discovery using explainability models.