Abstract:Computational physics simulation can be a powerful tool to accelerate industry deployment of new scientific technologies. However, it must address the challenge of computationally tractable, moderately accurate prediction at large industry scales, and training a model without data at such large scales. A recently proposed component reduced order modeling (CROM) tackles this challenge by combining reduced order modeling (ROM) with discontinuous Galerkin domain decomposition (DG-DD). While it can build a component ROM at small scales that can be assembled into a large scale system, its application is limited to linear physics equations. In this work, we extend CROM to nonlinear steady Navier-Stokes flow equation. Nonlinear advection term is evaluated via tensorial approach or empirical quadrature procedure. Application to flow past an array of objects at moderate Reynolds number demonstrates $\sim23.7$ times faster solutions with a relative error of $\sim 2.3\%$, even at scales $256$ times larger than the original problem.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: in large - scale industrial applications, how to accelerate the deployment of new technologies through computational physics simulation, especially in the absence of large - scale data, to achieve scalable and accurate predictions. Specifically, the paper aims to extend the existing Component - Reduced - Order - Model (CROM) method to handle the nonlinear steady - state Navier - Stokes equations, so as to be able to perform efficient fluid - dynamics simulations in large - scale systems. ### Problem Background 1. **Challenges at the Industrial Scale** - The expansion of new technologies from the laboratory scale to the industrial scale usually requires physical pilot plants, which are both expensive and time - consuming. - Traditional numerical methods are computationally infeasible when dealing with industrial - scale problems. - Sub - grid - scale approximation methods (such as volume averaging or large - eddy - simulation closure models) may significantly reduce accuracy. - Large - scale applications also pose challenges to data - driven methods because of the lack of large - scale data, and predictions must be made through scale extrapolation. 2. **Limitations of Existing Methods** - The Component - Reduced - Order - Model (CROM), which combines Reduced - Order - Modeling (ROM) and Discontinuous Galerkin Domain Decomposition (DG - DD), can construct component ROMs on a small scale and assemble them into large - scale systems. - Previous applications of CROM were limited to linear physical equations. ### Goals of the Paper - **Extend CROM to Nonlinear Equations**: Especially the steady - state incompressible Navier - Stokes equations. - **Efficiently Handle Nonlinear Terms**: Evaluate the nonlinear convection terms through the tensor method or the Empirical Quadrature Procedure (EQP). - **Avoid Spurious Pressure Modes**: Expand the velocity basis functions so that they satisfy the relevant inf - sup conditions. ### Main Contributions - A new method is proposed that can perform reliable scale extrapolation from small - scale sample data in the absence of large - scale data. - For the nonlinear Navier - Stokes equations, two effective methods for handling nonlinear terms are proposed: the tensor method and EQP. - By expanding the velocity basis functions, the problem of spurious pressure modes caused by the divergence - free property of the velocity field is solved. - Verification was carried out in the porous - media - flow problem, demonstrating the effectiveness of this method in a large - scale 16×16 system, achieving an acceleration of approximately 23.7 times with a relative error of approximately 2.3%. ### Conclusions The method proposed in the paper is applicable not only to the steady - state Navier - Stokes equations but also can be extended to other types of nonlinear physical equations. Future work can further expand the treatment methods of interface penalty terms to improve the applicability to problems such as unsteady hyperbolic conservation laws. ### Formula Summary 1. **Steady - State Incompressible Navier - Stokes Equations** \[ -\nu \nabla^2 \tilde{u}_m + \nabla \tilde{p}_m + \tilde{u}_m \cdot \nabla \tilde{u}_m = f_m \quad (1a) \] \[ \nabla \cdot \tilde{u}_m = 0 \quad (1b) \] 2. **Evaluating Nonlinear Convection Terms by the Tensor Method** \[ \hat{C}_m[\hat{u}_m]_i = \sum_{j,k} \hat{C}_{ijk} \hat{u}_m,j \hat{u}_m,k \equiv \sum_{j,k} \langle i \phi_{u,m,j} \phi_{u,m} \cdot \nabla_k \phi_{u,m} \rangle_{\Omega_m} \hat{u}_m,j \hat{u}_m,k \quad (7) \] 3. **Evaluating Nonlinear Convection Terms by the Empirical Quadrature Procedure (EQP)** \[ \hat{C}_m[\hat{u}_m]_i = \sum_{q = 1}^{N_q} \phi^\top_{u,m}(x_q) w_q [u_m(x_q) \cdot \nab

Scalable physics-guided data-driven component model reduction for steady Navier-Stokes flow

Scaled-up prediction of steady Navier-Stokes equation with component reduced order modeling

Train small, model big: Scalable physics simulators via reduced order modeling and domain decomposition

Data-driven Nonlinear Constitutive Relations for Rarefied Flow Computations

An Undecomposed Hybrid Algorithm for Nonlinear Coupled Constitutive Relations of Rarefied Gas Dynamics

Accelerating Flow Simulations using Online Dynamic Mode Decomposition

A non-overlapping optimization-based domain decomposition approach to component-based model reduction of incompressible flows

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

An optimisation-based domain-decomposition reduced order model for the incompressible Navier-Stokes equations

Data-driven kinematics-consistent model order reduction of fluid-structure interaction problems: application to deformable microcapsules in a Stokes flow

CROM: Continuous Reduced-Order Modeling of PDEs Using Implicit Neural Representations

Reduced-Order Modeling Toward Solving Inverse Problems in Solid Mechanics and Fluid Dynamics

Enhancing high-fidelity nonlinear solver with reduced order model

Adjacency-based, Non-intrusive Reduced-order Modeling for Fluid-Structure Interactions

Reduced order modeling for flow and transport problems with Barlow Twins self-supervised learning

Approximating a branch of solutions to the Navier--Stokes equations by reduced-order modeling

A LSTM-enhanced surrogate model to simulate the dynamics of particle-laden fluid systems

Deep-learning-based reduced-order modeling for subsurface flow simulation

Projection-Based Reduced Order Model for Simulations of Nonlinear Flows with Multiple Moving Objects