Investigation of Nonlinear Model Order Reduction of the Quasigeostrophic Equations through a Physics-Informed Convolutional Autoencoder
arXiv:2108.12344
Abstract
Reduced order modeling (ROM) is a field of techniques that approximates complex physics-based models of real-world processes by inexpensive surrogates that capture important dynamical characteristics with a smaller number of degrees of freedom. Traditional ROM techniques such as proper orthogonal decomposition (POD) focus on linear projections of the dynamics onto a set of spectral features. In this paper we explore the construction of ROM using autoencoders (AE) that perform nonlinear projections of the system dynamics onto a low dimensional manifold learned from data. The approach uses convolutional neural networks (CNN) to learn spatial features as opposed to spectral, and utilize a physics informed (PI) cost function in order to capture temporal features as well. Our investigation using the quasi-geostrophic equations reveals that while the PI cost function helps with spatial reconstruction, spatial features are less powerful than spectral features, and that construction of ROMs through machine learning-based methods requires significant investigation into novel non-standard methodologies.
References in corpus (13)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Theory-guided Data Science: A New Paradigm for Scientific Discovery from Data
- Proper Orthogonal Decomposition Closure Models For Turbulent Flows: A Numerical Comparison
- Extreme learning machine for reduced order modeling of turbulent geophysical flows
- POD/DEIM Reduced-Order Strategies for Efficient Four Dimensional Variational Data Assimilation
- Physics-informed Autoencoders for Lyapunov-stable Fluid Flow Prediction
- A stabilized proper orthogonal decomposition reduced-order model for large scale quasigeostrophic ocean circulation
- The Reduced-Order Hybrid Monte Carlo Sampling Smoother
- Forecasting Sequential Data using Consistent Koopman Autoencoders
- ODE Test Problems: a MATLAB suite of initial value problems
- Enforcing Deterministic Constraints on Generative Adversarial Networks for Emulating Physical Systems
- Multifidelity Ensemble Kalman Filtering Using Surrogate Models Defined by Physics-Informed Autoencoders
- Neural network closures for nonlinear model order reduction