Abstract:SummaryIn this article, a model reduction technique is presented to solve nonlinear multiscale parabolic problems using dynamic mode decomposition. The multiple scales and nonlinearity bring great challenges for simulating the problems. To overcome this difficulty, we develop a model reduction method for the nonlinear multiscale dynamic problems by integrating constraint energy minimizing generalized multiscale finite element method (CEM‐GMsFEM) with dynamic mode decomposition (DMD). CEM‐GMsFEM has shown great efficiency to solve linear multiscale problems in a coarse space. However, using CEM‐GMsFEM to directly solve multiscale nonlinear parabolic models involves dynamically computing the residual and the Jacobian on a fine grid. This may be very computationally expensive because the evaluation of the nonlinear term is implemented in a high‐dimensional fine scale space. As a data‐driven method, DMD can use observation data and give an explicit expression to accurately describe the underlying nonlinear dynamic system. To efficiently compute the multiscale nonlinear parabolic problems, we propose a CEM‐DMD model reduction by combing CEM‐GMsFEM and DMD. The CEM‐DMD reduced model is a coarsen linear model, which avoids the nonlinear solver in the fine space. It is crucial to judiciously choose observation in DMD. Only proper observation can render an accurate DMD model. In the context of CEM‐DMD, we introduce two different observations: fine scale observation and coarse scale observation. In the construction of DMD model, the coarse scale observation requires much less computation than the fine scale observation. The CEM‐DMD model using the coarse scale observation gives a complete coarse model for the nonlinear multiscale dynamic systems and significantly improves the computation efficiency. To show the performance of the CEM‐DMD using the different observations, we present a few numerical results for the nonlinear multiscale parabolic problems in heterogeneous porous media.

An Adaptive Method Based on Local Dynamic Mode Decomposition for Parametric Dynamical Systems

A model reduction method for parametric dynamical systems defined on complex geometries

Direct Simulation for CAD Models Undergoing Parametric Modifications

Parametric Dynamic Mode Decomposition for nonlinear parametric dynamical systems

Model Reduction via Dynamic Mode Decomposition

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

A Survey of Projection-Based Model Reduction Methods for Parametric Dynamical Systems

Data-driven models of nonautonomous systems

A Dynamic Mode Decomposition Based Reduced-Order Model for Parameterized Time-Dependent Partial Differential Equations

Active-Learning-Driven Surrogate Modeling for Efficient Simulation of Parametric Nonlinear Systems

PF-DMD: Physics-fusion dynamic mode decomposition for accurate and robust forecasting of dynamical systems with imperfect data and physics

Model Reduction for Nonlinear Multiscale Parabolic Problems Using Dynamic Mode Decomposition

Learning Nonautonomous Systems via Dynamic Mode Decomposition

A Condensed Algorithm For Adaptive Component Mode Synthesis Of Viscoelastic Flexible Multibody Dynamics

Non-intrusive reduced-order model for time-dependent stochastic partial differential equations utilizing dynamic mode decomposition and polynomial chaos expansion

A reduced order method for nonlinear parameterized partial differential equations using dynamic mode decomposition coupled with <i>k</i>-nearest-neighbors regression

Reduced-order Modelling of Parameter-Dependent, Linear and Nonlinear Dynamic Partial Differential Equation Models.

A parametric framework for kernel-based dynamic mode decomposition using deep learning

Dictionary-based online-adaptive structure-preserving model order reduction for parametric Hamiltonian systems

A Dynamic Process Optimization Method Based on Nonlinear Programming and Hybrid Automatic Differentiation

Transient surrogate modeling of modally reduced structures with discontinuous loads and damping