Experimental validation and uncertainty quantification in wave-based computational room acoustics

It is well known that input data uncertainty has a major influence on the correctness of room acoustic simulations. This paper proposes a comprehensive framework for experimental validation and uncertainty quantification in room acoustic simulations. The sources of uncertainty in room acoustic simulations are many, but especially the boundary conditions present significant uncertainty. The input data uncertainty is propagated through the simulation model, which in this work is a wave-based time-domain discontinuous Galerkin finite element method (DGFEM) simulation scheme, and is then presented as a `probability box' (p-box), expressing both aleatory and epistemic input data uncertainties and model form uncertainties. The p-box can then be compared against the perceptual limits of the human auditory system to gauge the severity of the uncertainty. The framework is applied on a small room with an absorbing wall for the experimental validation of the simulation results and the uncertainty quantification of three boundary parameters: the flow resistivity of the porous absorber, the thickness of the porous absorber, and the absorption characteristics of the remaining hard walls. It is found that the input data uncertainty and the model form uncertainty are too high for the model to be able to predict several room acoustic parameters within just noticeable thresholds. (C) 2021 Elsevier Ltd. All rights reserved.