Numerical Analysis Seminar: Deep Learning of Nonlinear Flame Fronts Development Due to Darrieus-Landau Instability
Rixin Yu, Associate professor in Fluid Mechanics, Lund University. Title: Deep Learning of Nonlinear Flame Fronts Development Due to Darrieus-Landau Instability
Abstract: The problem of nonlinear flame development in a channel subjected to Darrieus Landau instability is studied using a data-driven, deep neural network approach. The task is setup to learn a time-advancement operator mapping any given flame front to a future time. A recurrent application of such operator rolls out a long sequence of predicted flame fronts, it is required a learned operator not only makes accurate short term predictions but also reproduces characteristic nonlinear behavior such as fractal front structures and detached flame pockets. Using two datasets of flame fronts solutions obtained from a heavy-duty direct numerical simulation and a light-duty modeling equation, we compare the performance of three state-of-art operator-regression network methods: Convolutional neural networks(CNN), Fourier Neural Operator (FNO) and Deep Operator Network (DeepONet). We show FNO gives best recurrent predictions in both short and long term. A consistent extension allowing the operator-regression networks to handle complicate flame fronts shape is achieved through representing the latter as an implicit curve.
Tid: 2022-11-18 10:15 till 11:00
alexandros [dot] sopasakis [at] math [dot] lth [dot] se