Machine Learning Models for Jet Noise Analysis
Loading...
Date
2019-05
Authors
Shah, Madhav
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Jet noise has been an active area of research since the inception of jets and although understanding of the mechanisms behind jet noise are more understood, there is still a lack of knowledge to predict jet noise from first principles. The extraction of jet noise from flow data is a difficult process, but recently, a method has been developed that works well on computational fluid dynamics (CFD) data. This method, Doak's Decomposition, extracts the relevant pressure perturbations to get the acoustic response of the jet. The method involves solving a differential equation however, which makes it unusable on a point by point basis, which limits its application to CFD results only. The goal of this project is to use machine learning to learn the correlations between various jet flow parameters, including velocity, density and pressure perturbations, and the acoustic response of the system. Machine learning models were trained on the acoustic component, which was extracted from well validated CFD data using Doak's Decomposition, to see if it could learn to output the acoustics on a point by point basis. The machine learning was implemented using open source libraries for Python and the results of this project will further the understanding between flow parameters and the acoustics of jets.
Description
Keywords
Aeroacoustics, Machine Learning, Jet noise, CFD