Machine learning-assisted early ignition prediction in a complex flow

Pavel P. Popov, David A. Buchta, Michael J. Anderson, Luca Massa, Jesse Capecelatro, Daniel J Bodony, Jonathan Freund

Research output: Contribution to journalArticle

Abstract

Machine learning methods are used to improve the efficiency by which turbulence-resolved simulations predict whether a hydrogen jet in air crossflow will successfully ignite. The flush-mounted jet issues perpendicularly from the wall of a low-speed wind tunnel into a turbulent boundary layer wherein a laser-induced optical breakdown (LIB)hotspot is deposited. A detailed hydrogen chemical mechanism is used to model the radicals and any subsequent chemical reactions. A dielectric-barrier discharge actuator generates body forces and hydrogen radicals near the jet orifice. We focus on the success or not of the ignition based on LIB location. A challenge is that definitive determination of this requires long simulations, up to 440 µs after the LIB deposition. This is particularly expensive since multiple simulations are required to find the threshold. To reduce the computational effort, three short-time (91 µs)criteria are proposed, evaluated, and compared: a constructed criterion based on detailed observations of radicals near the stoichiometric surface and two machine learning approaches, each trained on 38 realizations. The constructed criterion provides a low-cost estimate of the ignition boundary that is unambiguous in 45 of the 50 training and test trials, so only 10% of them would need to be simulated longer. The trained neural networks correctly predict outcomes in all cases evaluated, with the more automated procedure—a convolutional neural network (CNN)trained on two-dimensional images—providing the most definitive outcome prediction. From the CNN, a sensitivity analysis is used to determine which kernel features from the two-dimensional input data, as defined by intermediate-layer network weights, are important for identifying ignition.

Original languageEnglish (US)
Pages (from-to)451-466
Number of pages16
JournalCombustion and Flame
Volume206
DOIs
StatePublished - Aug 2019

Fingerprint

machine learning
ignition
Ignition
Learning systems
Hydrogen
Neural networks
Lasers
breakdown
predictions
hydrogen
Network layers
cost estimates
low speed wind tunnels
Orifices
lasers
Sensitivity analysis
Wind tunnels
simulation
Chemical reactions
turbulent boundary layer

Keywords

  • Ignition
  • Ignition prediction
  • Machine learning
  • Plasma-coupled combustion

ASJC Scopus subject areas

  • Chemistry(all)
  • Chemical Engineering(all)
  • Fuel Technology
  • Energy Engineering and Power Technology
  • Physics and Astronomy(all)

Cite this

Machine learning-assisted early ignition prediction in a complex flow. / Popov, Pavel P.; Buchta, David A.; Anderson, Michael J.; Massa, Luca; Capecelatro, Jesse; Bodony, Daniel J; Freund, Jonathan.

In: Combustion and Flame, Vol. 206, 08.2019, p. 451-466.

Research output: Contribution to journalArticle

Popov, Pavel P. ; Buchta, David A. ; Anderson, Michael J. ; Massa, Luca ; Capecelatro, Jesse ; Bodony, Daniel J ; Freund, Jonathan. / Machine learning-assisted early ignition prediction in a complex flow. In: Combustion and Flame. 2019 ; Vol. 206. pp. 451-466.
@article{b366972a3ddf418abfb8d654c9bfa84b,
title = "Machine learning-assisted early ignition prediction in a complex flow",
abstract = "Machine learning methods are used to improve the efficiency by which turbulence-resolved simulations predict whether a hydrogen jet in air crossflow will successfully ignite. The flush-mounted jet issues perpendicularly from the wall of a low-speed wind tunnel into a turbulent boundary layer wherein a laser-induced optical breakdown (LIB)hotspot is deposited. A detailed hydrogen chemical mechanism is used to model the radicals and any subsequent chemical reactions. A dielectric-barrier discharge actuator generates body forces and hydrogen radicals near the jet orifice. We focus on the success or not of the ignition based on LIB location. A challenge is that definitive determination of this requires long simulations, up to 440 µs after the LIB deposition. This is particularly expensive since multiple simulations are required to find the threshold. To reduce the computational effort, three short-time (91 µs)criteria are proposed, evaluated, and compared: a constructed criterion based on detailed observations of radicals near the stoichiometric surface and two machine learning approaches, each trained on 38 realizations. The constructed criterion provides a low-cost estimate of the ignition boundary that is unambiguous in 45 of the 50 training and test trials, so only 10{\%} of them would need to be simulated longer. The trained neural networks correctly predict outcomes in all cases evaluated, with the more automated procedure—a convolutional neural network (CNN)trained on two-dimensional images—providing the most definitive outcome prediction. From the CNN, a sensitivity analysis is used to determine which kernel features from the two-dimensional input data, as defined by intermediate-layer network weights, are important for identifying ignition.",
keywords = "Ignition, Ignition prediction, Machine learning, Plasma-coupled combustion",
author = "Popov, {Pavel P.} and Buchta, {David A.} and Anderson, {Michael J.} and Luca Massa and Jesse Capecelatro and Bodony, {Daniel J} and Jonathan Freund",
year = "2019",
month = "8",
doi = "10.1016/j.combustflame.2019.05.014",
language = "English (US)",
volume = "206",
pages = "451--466",
journal = "Combustion and Flame",
issn = "0010-2180",
publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Machine learning-assisted early ignition prediction in a complex flow

AU - Popov, Pavel P.

AU - Buchta, David A.

AU - Anderson, Michael J.

AU - Massa, Luca

AU - Capecelatro, Jesse

AU - Bodony, Daniel J

AU - Freund, Jonathan

PY - 2019/8

Y1 - 2019/8

N2 - Machine learning methods are used to improve the efficiency by which turbulence-resolved simulations predict whether a hydrogen jet in air crossflow will successfully ignite. The flush-mounted jet issues perpendicularly from the wall of a low-speed wind tunnel into a turbulent boundary layer wherein a laser-induced optical breakdown (LIB)hotspot is deposited. A detailed hydrogen chemical mechanism is used to model the radicals and any subsequent chemical reactions. A dielectric-barrier discharge actuator generates body forces and hydrogen radicals near the jet orifice. We focus on the success or not of the ignition based on LIB location. A challenge is that definitive determination of this requires long simulations, up to 440 µs after the LIB deposition. This is particularly expensive since multiple simulations are required to find the threshold. To reduce the computational effort, three short-time (91 µs)criteria are proposed, evaluated, and compared: a constructed criterion based on detailed observations of radicals near the stoichiometric surface and two machine learning approaches, each trained on 38 realizations. The constructed criterion provides a low-cost estimate of the ignition boundary that is unambiguous in 45 of the 50 training and test trials, so only 10% of them would need to be simulated longer. The trained neural networks correctly predict outcomes in all cases evaluated, with the more automated procedure—a convolutional neural network (CNN)trained on two-dimensional images—providing the most definitive outcome prediction. From the CNN, a sensitivity analysis is used to determine which kernel features from the two-dimensional input data, as defined by intermediate-layer network weights, are important for identifying ignition.

AB - Machine learning methods are used to improve the efficiency by which turbulence-resolved simulations predict whether a hydrogen jet in air crossflow will successfully ignite. The flush-mounted jet issues perpendicularly from the wall of a low-speed wind tunnel into a turbulent boundary layer wherein a laser-induced optical breakdown (LIB)hotspot is deposited. A detailed hydrogen chemical mechanism is used to model the radicals and any subsequent chemical reactions. A dielectric-barrier discharge actuator generates body forces and hydrogen radicals near the jet orifice. We focus on the success or not of the ignition based on LIB location. A challenge is that definitive determination of this requires long simulations, up to 440 µs after the LIB deposition. This is particularly expensive since multiple simulations are required to find the threshold. To reduce the computational effort, three short-time (91 µs)criteria are proposed, evaluated, and compared: a constructed criterion based on detailed observations of radicals near the stoichiometric surface and two machine learning approaches, each trained on 38 realizations. The constructed criterion provides a low-cost estimate of the ignition boundary that is unambiguous in 45 of the 50 training and test trials, so only 10% of them would need to be simulated longer. The trained neural networks correctly predict outcomes in all cases evaluated, with the more automated procedure—a convolutional neural network (CNN)trained on two-dimensional images—providing the most definitive outcome prediction. From the CNN, a sensitivity analysis is used to determine which kernel features from the two-dimensional input data, as defined by intermediate-layer network weights, are important for identifying ignition.

KW - Ignition

KW - Ignition prediction

KW - Machine learning

KW - Plasma-coupled combustion

UR - http://www.scopus.com/inward/record.url?scp=85066262922&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85066262922&partnerID=8YFLogxK

U2 - 10.1016/j.combustflame.2019.05.014

DO - 10.1016/j.combustflame.2019.05.014

M3 - Article

AN - SCOPUS:85066262922

VL - 206

SP - 451

EP - 466

JO - Combustion and Flame

JF - Combustion and Flame

SN - 0010-2180

ER -