A deep learning approach to jet noise prediction

Andrew S. Tenney, Mark N Glauser, Jacques Lewalle, Christopher J. Ruscher

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

This study is an exploration of a deep learning technique as it applies to jet noise prediction and modeling. A database was created describing the near-field and far-field conditions of a complex non-axisymmetric jet flow, with Mach numbers ranging from 1.0 to 1.6. The problem was posed as a form of multivariate nonlinear regression, and a Deep Neural Network (DNN) was used to create a model. Google’s TensorFlow was utilized in network construction and training, and a feature space consisting of plausible predictors of far-field directional SPL was defined, based on previous fundamental studies and jet noise scaling laws. The effects of feature space, learning rate, and network architecture modification were also studied and quantified. On average, the DNN was able to predict directional far-field SPL within ±0.75dB, surpassing our original goals. In addition, we suggest ways to extend this technique to encompass a greater range of nozzle geometries and flow conditions.

Original languageEnglish (US)
Title of host publicationAIAA Aerospace Sciences Meeting
PublisherAmerican Institute of Aeronautics and Astronautics Inc, AIAA
Edition210059
ISBN (Print)9781624105241
DOIs
StatePublished - Jan 1 2018
EventAIAA Aerospace Sciences Meeting, 2018 - Kissimmee, United States
Duration: Jan 8 2018Jan 12 2018

Other

OtherAIAA Aerospace Sciences Meeting, 2018
CountryUnited States
CityKissimmee
Period1/8/181/12/18

Fingerprint

Scaling laws
Network architecture
Mach number
Nozzles
Geometry
Deep learning
Deep neural networks

ASJC Scopus subject areas

  • Aerospace Engineering

Cite this

Tenney, A. S., Glauser, M. N., Lewalle, J., & Ruscher, C. J. (2018). A deep learning approach to jet noise prediction. In AIAA Aerospace Sciences Meeting (210059 ed.). American Institute of Aeronautics and Astronautics Inc, AIAA. https://doi.org/10.2514/6.2018-1736

A deep learning approach to jet noise prediction. / Tenney, Andrew S.; Glauser, Mark N; Lewalle, Jacques; Ruscher, Christopher J.

AIAA Aerospace Sciences Meeting. 210059. ed. American Institute of Aeronautics and Astronautics Inc, AIAA, 2018.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Tenney, AS, Glauser, MN, Lewalle, J & Ruscher, CJ 2018, A deep learning approach to jet noise prediction. in AIAA Aerospace Sciences Meeting. 210059 edn, American Institute of Aeronautics and Astronautics Inc, AIAA, AIAA Aerospace Sciences Meeting, 2018, Kissimmee, United States, 1/8/18. https://doi.org/10.2514/6.2018-1736
Tenney AS, Glauser MN, Lewalle J, Ruscher CJ. A deep learning approach to jet noise prediction. In AIAA Aerospace Sciences Meeting. 210059 ed. American Institute of Aeronautics and Astronautics Inc, AIAA. 2018 https://doi.org/10.2514/6.2018-1736
Tenney, Andrew S. ; Glauser, Mark N ; Lewalle, Jacques ; Ruscher, Christopher J. / A deep learning approach to jet noise prediction. AIAA Aerospace Sciences Meeting. 210059. ed. American Institute of Aeronautics and Astronautics Inc, AIAA, 2018.
@inproceedings{2dc48250518849bb87e8ce6533e46538,
title = "A deep learning approach to jet noise prediction",
abstract = "This study is an exploration of a deep learning technique as it applies to jet noise prediction and modeling. A database was created describing the near-field and far-field conditions of a complex non-axisymmetric jet flow, with Mach numbers ranging from 1.0 to 1.6. The problem was posed as a form of multivariate nonlinear regression, and a Deep Neural Network (DNN) was used to create a model. Google’s TensorFlow was utilized in network construction and training, and a feature space consisting of plausible predictors of far-field directional SPL was defined, based on previous fundamental studies and jet noise scaling laws. The effects of feature space, learning rate, and network architecture modification were also studied and quantified. On average, the DNN was able to predict directional far-field SPL within ±0.75dB, surpassing our original goals. In addition, we suggest ways to extend this technique to encompass a greater range of nozzle geometries and flow conditions.",
author = "Tenney, {Andrew S.} and Glauser, {Mark N} and Jacques Lewalle and Ruscher, {Christopher J.}",
year = "2018",
month = "1",
day = "1",
doi = "10.2514/6.2018-1736",
language = "English (US)",
isbn = "9781624105241",
booktitle = "AIAA Aerospace Sciences Meeting",
publisher = "American Institute of Aeronautics and Astronautics Inc, AIAA",
edition = "210059",

}

TY - GEN

T1 - A deep learning approach to jet noise prediction

AU - Tenney, Andrew S.

AU - Glauser, Mark N

AU - Lewalle, Jacques

AU - Ruscher, Christopher J.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - This study is an exploration of a deep learning technique as it applies to jet noise prediction and modeling. A database was created describing the near-field and far-field conditions of a complex non-axisymmetric jet flow, with Mach numbers ranging from 1.0 to 1.6. The problem was posed as a form of multivariate nonlinear regression, and a Deep Neural Network (DNN) was used to create a model. Google’s TensorFlow was utilized in network construction and training, and a feature space consisting of plausible predictors of far-field directional SPL was defined, based on previous fundamental studies and jet noise scaling laws. The effects of feature space, learning rate, and network architecture modification were also studied and quantified. On average, the DNN was able to predict directional far-field SPL within ±0.75dB, surpassing our original goals. In addition, we suggest ways to extend this technique to encompass a greater range of nozzle geometries and flow conditions.

AB - This study is an exploration of a deep learning technique as it applies to jet noise prediction and modeling. A database was created describing the near-field and far-field conditions of a complex non-axisymmetric jet flow, with Mach numbers ranging from 1.0 to 1.6. The problem was posed as a form of multivariate nonlinear regression, and a Deep Neural Network (DNN) was used to create a model. Google’s TensorFlow was utilized in network construction and training, and a feature space consisting of plausible predictors of far-field directional SPL was defined, based on previous fundamental studies and jet noise scaling laws. The effects of feature space, learning rate, and network architecture modification were also studied and quantified. On average, the DNN was able to predict directional far-field SPL within ±0.75dB, surpassing our original goals. In addition, we suggest ways to extend this technique to encompass a greater range of nozzle geometries and flow conditions.

UR - http://www.scopus.com/inward/record.url?scp=85044409663&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044409663&partnerID=8YFLogxK

U2 - 10.2514/6.2018-1736

DO - 10.2514/6.2018-1736

M3 - Conference contribution

AN - SCOPUS:85044409663

SN - 9781624105241

BT - AIAA Aerospace Sciences Meeting

PB - American Institute of Aeronautics and Astronautics Inc, AIAA

ER -