Skip to main content

A Machine Learning Method for Acoustic Decomposition

Time: Thu 2021-01-14 10.30 - 11.30

Location: FLOW eSeminar (Zoom)

Participating: Stefan Jacob (KTH)

Export to calendar

Abstract. Acoustic mode decomposition is commonly used to separate pressure wave components in flow ducts, and established methods relay on mathematical descriptions of the wave motion. I will present our new approach to mode decomposition which uses artificial neural networks to separate the acoustic mode content. A neural network is trained with data gained from a small set of numeric solutions of the Linearized Navier-Stokes equation in the frequency domain in a straight duct. The network is tested on relevant experimental data. Good agreement with the analytical method is demonstrated in two applications with different flow conditions. We conclude that using machine learning for mode decomposition is a promising alternative to the standard method as it is applicable to a much broader set of aero-acoustic problems.

Page responsible:Ardeshir Hanifi
Belongs to: FLOW
Last changed: Jan 11, 2021