Abstract:Advection-dominated problems are commonly noticed in nature, engineering systems, and a wide range of industrial processes. For these problems, linear approximation methods (proper orthogonal decomposition and reduced basis method) are not suitable, as the Kolmogorov $n$-width decay is slow, leading to inefficient and inaccurate reduced order models. There are few non-linear approaches to accelerate the Kolmogorov $n$-width decay. In this work, we use a neural-network shift augmented transformation technique, that employs automatic-shit detection and detects the optimal non-linear transformation of the full-order model solution manifold $\mathcal{M}$. We exploit a deep-learning framework to derive parameter-dependent bijective mapping between the manifold $\mathcal{M}$ and the transformed manifold $\tilde{\mathcal{M}}$. It consists of two neural networks, 1) ShiftNet, to employ automatic-shift detection by learning the shift-operator, which finds the optimal shifts for numerous snapshots of the full-order solution manifold, to accelerate the Kolmogorov $n$-width decay, and 2) InterpNet, which learns the reference configuration and can reconstruct the field values of the same, for each shifted grid distribution. We construct non-intrusive reduced order models on the resulting transformed linear subspaces and employ automatic-shift detection for predictions. We test our methodology on advection-dominated problems, such as 1D travelling waves, 2D isentropic convective vortex and 2D two-phase flow test cases. This work leads to the development of the complete NNsPOD-ROM algorithm for model reduction of advection-dominated problems, comprising both offline-online stages.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is that for advection - dominated problems, traditional linear approximation methods (such as POD and RB methods) show inefficiency and insufficient accuracy when constructing Reduced Order Models (ROMs). The characteristics of such problems are that high - dimensional features lead to slow decay of Kolmogorov n - width, which means that there is no low - dimensional linear approximation subspace that can effectively approximate the solution of the Full - Order Model (FOM). Therefore, traditional linear model reduction methods are not applicable in this case, resulting in the constructed linearly - projected - based ROMs being both inefficient and inaccurate. To solve this problem, the author proposes a nonlinear transformation technique based on neural networks - Neural Network Shift Augmented Manifold Transformations. This technique finds the optimal nonlinear transformation of the full - order model solution manifold $M$ through automatic shift detection, thereby accelerating the decay of Kolmogorov n - width. Specifically, this method uses a deep - learning framework to derive a parameter - dependent bijective mapping, mapping the manifold $M$ to the transformed manifold $\tilde{M}$. The whole process consists of two neural networks: 1. **ShiftNet**: Used for automatic shift detection, it finds the optimal shift of multiple snapshots by learning the shift operator to accelerate the decay of Kolmogorov n - width. 2. **InterpNet**: Learns the reference configuration and can reconstruct the field values of each shifted grid distribution. Through these techniques, the author constructs Non - Intrusive Reduced Order Models (NI - ROMs) and makes predictions on the transformed linear subspaces. This method is not only applicable to advection - dominated problems such as 1D traveling waves, 2D isentropic convective vortices, and 2D two - phase flows, but also develops a complete NNsPOD - ROM algorithm, covering the offline and online stages, ensuring the efficiency and accuracy of the model.

Non-intrusive model reduction of advection-dominated hyperbolic problems using neural network shift augmented manifold transformations

The Neural Network shifted-proper orthogonal decomposition: A machine learning approach for non-linear reduction of hyperbolic equations

Explicable hyper-reduced order models on nonlinearly approximated solution manifolds of compressible and incompressible Navier-Stokes equations

An artificial neural network framework for reduced order modeling of transient flows

Model Reduction for Advection Dominated Hyperbolic Problems in an ALE Framework: Offline and Online Phases

Non-Intrusive Reduced Order Modeling of Convection Dominated Flows Using Artificial Neural Networks with Application to Rayleigh-Taylor Instability

Non-intrusive model reduction of large-scale, nonlinear dynamical systems using deep learning

Neural-Network-Augmented Projection-Based Model Order Reduction for Mitigating the Kolmogorov Barrier to Reducibility of CFD Models

A fast and accurate physics-informed neural network reduced order model with shallow masked autoencoder

Towards a machine learning pipeline in reduced order modelling for inverse problems: neural networks for boundary parametrization, dimensionality reduction and solution manifold approximation

Reduced-order modeling of advection-dominated systems with recurrent neural networks and convolutional autoencoders

Hyper-Reduced Autoencoders for Efficient and Accurate Nonlinear Model Reductions

An optimisation-based domain-decomposition reduced order model for parameter-dependent non-stationary fluid dynamics problems

SNF-ROM: Projection-based nonlinear reduced order modeling with smooth neural fields

Proper Orthogonal Decomposition and Physical Field Reconstruction with Artificial Neural Networks (ANN) for Supercritical Flow Problems

Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders

Non-intrusive inference reduced order model for fluids using linear multistep neural network

Non-linear Manifold Reduced-Order Models with Convolutional Autoencoders and Reduced Over-Collocation Method

Nonlinear Manifold Learning and Model Reduction for Transonic Flows

Nonlinear model order reduction for problems with microstructure using mesh informed neural networks

Nonlinear model reduction with Neural Galerkin schemes on quadratic manifolds