Preconditioned FEM-based neural networks for solving incompressible fluid flows and related inverse problems

The numerical simulation and optimization of technical systems described by partial differential equations is expensive, especially in multi-query scenarios in which the underlying equations have to be solved for different parameters. A comparatively new approach in this context is to combine the go...

Full description

Saved in:
Bibliographic Details
Published in:Journal of computational and applied mathematics Vol. 469; p. 116663
Main Authors: Griese, Franziska, Hoppe, Fabian, Rüttgers, Alexander, Knechtges, Philipp
Format: Journal Article
Language:English
Published: Elsevier B.V 01.12.2025
Subjects:
FEM-based neural network
Finite elements
Machine learning
Navier–Stokes
Parametric PDEs
Preconditioning
Stokes
Finite elements
Stokes
FEM-based neural network
Machine learning
Parametric PDEs
Navier–Stokes
Preconditioning
ISSN:0377-0427
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The numerical simulation and optimization of technical systems described by partial differential equations is expensive, especially in multi-query scenarios in which the underlying equations have to be solved for different parameters. A comparatively new approach in this context is to combine the good approximation properties of neural networks (for parameter dependence) with the classical finite element method (for discretization). However, instead of considering the solution mapping of the PDE from the parameter space into the FEM-discretized solution space as a purely data-driven regression problem, so-called physically informed regression problems have proven to be useful. In these, the equation residual is minimized during the training of the neural network, i.e., the neural network “learns” the physics underlying the problem. In this paper, we extend this approach to saddle-point and non-linear fluid dynamics problems, respectively, namely stationary Stokes and stationary Navier–Stokes equations. In particular, we propose a modification of the existing approach: Instead of minimizing the plain vanilla equation residual during training, we minimize the equation residual modified by a preconditioner. By analogy with the linear case, this also improves the condition in the present non-linear case. Our numerical examples demonstrate that this approach significantly reduces the training effort and greatly increases accuracy and generalizability. Finally, we show the application of the resulting parameterized model to a related inverse problem. •We advance FEM-based neural networks by adding a preconditioner to the physics-informed loss function.•Preconditioning significantly reduces the training effort and greatly improves accuracy and generalizability.•Numerical examples for stationary 2D Stokes/Navier–Stokes equations and a related inverse problem are provided.
ISSN:0377-0427
DOI:10.1016/j.cam.2025.116663