Parametric uncertainty quantification of sound transmission measurement procedures with a maximum entropy approach

An important issue in transmission loss assessment is the significant variability of the performance across a range of possible facilities, similar to what is called reproducibility in the literature. Instead of performing expensive experimental round robin testing, virtual round robin testing can be carried out to assess the inherent uncertainty of a given measurement procedure. In this paper, the combined effect of multiple uncertain parameters is assessed with a probabilistic framework in which their joint probability distribution is constructed by means of a maximum entropy approach. Using a Monte Carlo simulation, the variability of the predicted sound transmission loss across a range of possible facilities is quantified. For frequencies for which the emitting and receiving room can be modeled as diffuse fields, a nonparametric probabilistic approach based on the Gaussian orthogonal ensemble can be used. The methodology is applied to a calcium silicate block wall. The results agree well with measurements and with indicative values of the reproducibility in standards and literature.