Large eddy simulation with flamelet progress variable approach combined with artificial neural network acceleration

Lorenzo Angelilli, Pietro Paolo Ciottoli, Riccardo Malpica Galassi, Francisco Hernandez Perez, Mattia Soldan, Zhen Lu, Mauro Valorani, Hong G. Im

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

Abstract

In the context of large eddy simulation of turbulent reacting flows, flamelet-based models are key to affordable simulations of large and complex systems. However, as the complexity of the problem increases, higher-dimensional look-up tables are required, rendering the conventional look-up procedure too demanding. This work focuses on accelerating the estimation of flamelet- based data for the flamelet/progress variable model via an artificial neural network. The neural network hyper-parameters are defined by a Bayesian optimization and two different architectures are selected for comparison against the classical look-up procedure on the well known Sandia flame D. The performance in terms of execution time and accuracy are analyzed, showing that the neural network model reduces the computational time by 30%, as compared to the traditional table look-up, while retaining comparable accuracy.
Original languageEnglish (US)
Title of host publicationAIAA Scitech 2021 Forum
PublisherAmerican Institute of Aeronautics and Astronautics
ISBN (Print)9781624106095
DOIs
StatePublished - Jan 4 2021

Bibliographical note

KAUST Repository Item: Exported on 2021-02-23
Acknowledgements: The authors acknowledge the support of King Abdullah University of Science and Technology (KAUST). Computational resources were provided by the KAUST Supercomputing Laboratory (KSL). This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 682383).

Fingerprint

Dive into the research topics of 'Large eddy simulation with flamelet progress variable approach combined with artificial neural network acceleration'. Together they form a unique fingerprint.

Cite this