TY - GEN
T1 - Large eddy simulation with flamelet progress variable approach combined with artificial neural network acceleration
AU - Angelilli, Lorenzo
AU - Ciottoli, Pietro Paolo
AU - Malpica Galassi, Riccardo
AU - Hernandez Perez, Francisco
AU - Soldan, Mattia
AU - Lu, Zhen
AU - Valorani, Mauro
AU - Im, Hong G.
N1 - 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).
PY - 2021/1/4
Y1 - 2021/1/4
N2 - 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.
AB - 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.
UR - http://hdl.handle.net/10754/667594
UR - https://arc.aiaa.org/doi/10.2514/6.2021-0412
U2 - 10.2514/6.2021-0412
DO - 10.2514/6.2021-0412
M3 - Conference contribution
SN - 9781624106095
BT - AIAA Scitech 2021 Forum
PB - American Institute of Aeronautics and Astronautics
ER -