Gaussian process based surrogate modelling of acoustic systems

Abstract

The numerical simulation of acoustic problems is, for itself, a quite difficult task since the underlying systems are usually highly complex with a broad frequency range and high sensitivity. Due to this complexity and the corresponding computational burden, tasks like optimization and uncertainty quantification (UQ) are seldom performed in acoustics. Especially when dealing with polymorphic uncertainties where combined techniques of UQ might be required, a direct use of the model is not viable. To allow such engineering tasks, the construction of a cheap surrogate or reduced model is common practice in order to allow a large number of model evaluations at low costs.

For acoustic systems, the construction of a reasonably accurate surrogate model can become a challenging task since many systems operate in the frequency domain where phenomena like resonance and interference can cause highly nonlinear responses. In this paper, a surrogate model based on the combination of a parametric model for capturing local nonlinearities and a random process regression for modelling the global trend is presented. The basic procedure based on previous works of the authors is extended to a predict the system response both for unobserved parameters and frequencies. The procedure is demonstrated for a representative example namely the acoustic simulation of a car interior, and further improvements in accuracy and efficiency by the usage of multilevel information are discussed.