Abstract:This study investigates the application of machine learning, specifically Fourier Neural Operator (FNO) and Convolutional Neural Network (CNN), to learn time-advancement operators for parametric partial differential equations (PDEs). Our focus is on extending existing operator learning methods to handle additional inputs representing PDE parameters. The goal is to create a unified learning approach that accurately predicts short-term solutions and provides robust long-term statistics under diverse parameter conditions, facilitating computational cost savings and accelerating development in engineering simulations. We develop and compare parametric learning methods based on FNO and CNN, evaluating their effectiveness in learning parametric-dependent solution time-advancement operators for one-dimensional PDEs and realistic flame front evolution data obtained from direct numerical simulations of the Navier-Stokes equations.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to use machine - learning methods (especially Fourier Neural Operator (FNO) and Convolutional Neural Network (CNN)) to learn time - advancing operators with parameters, in order to accurately predict the short - time solutions of the evolution of unstable flame fronts and provide robust long - time statistics under different parameter conditions. Specifically, the authors focus on extending the existing operator - learning methods to handle additional input parameters, which represent variables in partial differential equations (PDEs). The goal is to create a unified learning method that can accurately predict short - term solutions under multiple parameter conditions and provide robust long - term statistical results. This helps to save computational costs and accelerate the development of engineering simulations. ### Main problems: 1. **Evolution of unstable flame fronts**: The evolution of unstable flame fronts involved in the study is a complex non - linear process, usually described by partial differential equations. 2. **Parameter dependence**: Different parameter values will affect the evolution pattern of the flame front, so a method that can handle parameter changes is required. 3. **Challenges of long - time prediction**: Due to the chaotic characteristics of the system, long - time prediction is very difficult, especially for non - linear systems that are sensitive to initial conditions. ### Solutions: - **Fourier Neural Operator (FNO)** and **Convolutional Neural Network (CNN)** are used to learn time - advancing operators with parameters. - **Parameterization extension**: By introducing additional input parameters, the model can capture the evolution of the flame front under different parameter conditions. - **Short - term and long - term prediction**: It is necessary not only to accurately predict short - term behavior, but also to ensure that the long - term statistical characteristics are similar to the real solutions. ### Methods: 1. **Problem setting**: Define the time - advancing operator $\hat{G}$ with parameters, which maps the input function $v(x)$ and the parameter $\gamma$ to another function $v'(x')$. \[ \hat{G}: (v(x), \gamma) \rightarrow v'(x') \] 2. **Training method**: Use the recursive training method to optimize the model so that it can make multiple consecutive predictions from a single input function. 3. **Model design**: - **Parameterized Convolutional Neural Network (pCNN)**: Consider the influence of parameters by expanding the encoder block. - **Parameterized Fourier Neural Operator (pFNO)**: Include the influence of parameters in different wavenumber ranges by introducing additional learnable complex weights. ### Results: - **Short - term prediction**: All models (pFNO, pFNO*, pCNN) perform well in short - term prediction with relatively small relative errors. - **Long - term statistical characteristics**: pFNO performs excellently in capturing long - term statistical characteristics, especially in the chaotic DT front evolution. - **DL front evolution**: For DL front evolution, pFNO* and pCNN are superior to pFNO in long - term prediction, especially when the noise influence is large. Through these methods, the authors have successfully developed a unified learning framework that can handle parameter changes, providing a new approach for efficient computation in complex engineering simulations.

Parametric Learning of Time-Advancement Operators for Unstable Flame Evolution

Parametric learning of time-advancement operators for unstable flame evolution

Koopman Theory-Inspired Method for Learning Time Advancement Operators in Unstable Flame Front Evolution

Learning Flame Evolution Operator under Hybrid Darrieus Landau and Diffusive Thermal Instability

FlamePINN-1D: Physics-informed neural networks to solve forward and inverse problems of 1D laminar flames

Learning stiff chemical kinetics using extended deep neural operators

Finite Operator Learning: Bridging Neural Operators and Numerical Methods for Efficient Parametric Solution and Optimization of PDEs

Fourier neural operators for spatiotemporal dynamics in two-dimensional turbulence

Integration of physics-informed operator learning and finite element method for parametric learning of partial differential equations

Wavelet Neural Operator for solving parametric partial differential equations in computational mechanics problems

Enhancing Solutions for Complex PDEs: Introducing Complementary Convolution and Equivariant Attention in Fourier Neural Operators

Model-Parallel Fourier Neural Operators as Learned Surrogates for Large-Scale Parametric PDEs

Linear and nonlinear flame response prediction of turbulent flames using neural network models

Predictive Models for Flame Evolution Using Machine Learning: A Priori Assessment in Turbulent Flames Without and with Mean Shear

DeepFlame: A deep learning empowered open-source platform for reacting flow simulations

Wavelet neural operator: a neural operator for parametric partial differential equations

Neural Control of Parametric Solutions for High-dimensional Evolution PDEs

Augmenting filtered flame front displacement models for LES using machine learning with a posteriori simulations

Multiscale Neural Operator: Learning Fast and Grid-independent PDE Solvers

Adaptive physics-informed neural operator for coarse-grained non-equilibrium flows

Learning nonlinear operators in latent spaces for real-time predictions of complex dynamics in physical systems