Uncertainty Quantification in hybrid approaches for the prediction of broadband fan noise Seminar

In order to reduce the production costs associated to aeroacoustic model testing, industries are increasingly using numerical simulations. However, the estimation of the confidence level of such simulations is still a difficult task, especially if the equipment is to be certified over a wide operating range. Uncertainty quantification (UQ) permits propagating the effect of uncertainties associated to a physical or computational model, towards variations of the predicted performance. Due to growing computational resources, UQ methods can nowadays be applied in the field of Computational Fluid Dynamics, and it has recently tackled the field of computational aeroacoustics.

In this talk, UQ will be applied to the prediction of the noise emitted by static and rotating blades, relevant to low-speed cooling fans used in automotive industry. The sound generation mechanism considered here involves the scattering of vortical perturbations convected past the trailing edge of a controlled diffusion airfoil. Scale-resolved unsteady flow modelling as as well as statistical approaches will be discussed, in the framework of aeroacoustic analogies such as Curle's analogy and Amiet's linearized airfoil theory.

We will consider in particular the influence of uncertain inputs on the flow solution for different flow modelling approaches (RANS or LES), on different wall-pressure spectrum reconstruction methods from steady RANS computations, and on the noise propagation.