Search arXiv⌕ Search

arXiv · 2211.13418

AI-augmented stabilized finite element method

Abstract

An artificial intelligence-augmented Streamline Upwind/Petrov-Galerkin finite element scheme (AiStab-FEM) is proposed for solving singularly perturbed partial differential equations. In particular, an artificial neural network framework is proposed to predict optimal values for the stabilization parameter. The neural network is trained by minimizing a physics-informed cost function, where the equation's mesh and physical parameters are used as input features. Further, the predicted stabilization parameter is normalized with the gradient of the Galerkin solution to treat the boundary/interior layer region adequately. The proposed approach suppresses the undershoots and overshoots in the stabilized finite element solution and outperforms the existing neural network-based partial differential equation solvers such as Physics-Informed Neural Networks and Variational Neural Networks.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Sangeeta Yadav, Sashikumaar Ganesan. 2022-11-24. AI-augmented stabilized finite element method. https://arxiv.org/abs/2211.13418

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Large friction limit of compressible Navier--Stokes equations with Navier boundary conditions in a half-space

We study the large-friction limit for the three-dimensional barotropic compressible Navier-Stokes equations in a half-space. The velocity satisfies a Navier boundary condition with friction coefficient $α>0$, while the limiting problem satisfies the no-slip boundary condition. We establish estimates for local-in-time smooth solutions that are uniform in $α$ and prove strong convergence of the density and velocity as $α\to\infty$. For weak solutions, we use the Lagrangian flow maps associated with the two velocities to compare the densities and construct suitable transported test functions. This yields weak convergence to the no-slip solution. Our results provide a compressible counterpart of the large-friction limit for incompressible flows.

math.AP↗

Bochner-Riesz means for critical magnetic Schrödinger operators in ${\mathbb R^2}$

We study $L^p$-boundedness of the Bochner-Riesz means for critical magnetic Schrödinger operators $\LL_{\A}$ in ${\mathbb R^2}$, which involve the {physical} Aharonov-Bohm potential. We show that for $1\leq p\leq +\infty$ and $p\not= 2$, the Bochner-Riesz operator $S_λ^δ(\LL_{\A})$ of order $δ$ is bounded on $L^p(\R^2)$ if and only if $δ>\max\big\{0, 2\big|1/2-1/p\big|-1/2\big\}$. The new ingredient {in} the proof is to obtain the localized $L^4(\R^2)$ estimate of $S_λ^δ(\LL_{\A})$, whose kernel is heavily affected by the physical magnetic diffraction, and more singular than the classical Bochner-Riesz means $S_λ^δ(Δ)$ for the Laplacian $Δ$ in ${\mathbb R}^2$.

math.AP↗

Solitons, scattering and blow-up for the nonlinear Schrödinger equation with combined power-type nonlinearities on $\mathbb{R}^d\times\mathbb{T}$

We investigate the long time dynamics of the nonlinear Schrödinger equation (NLS) with combined powers on the waveguide manifold $\mathbb{R}^d\times\mathbb{T}$. Different from the previously studied NLS-models with single power on the waveguide manifolds, where the non-scale-invariance is mainly due to the mixed nature of the underlying domain, the non-scale-invariance of the present model is both geometrical and structural. By considering different combinations of the nonlinearities, we establish both qualitative and quantitative properties of the soliton, scattering and blow-up solutions. As one of the main novelties of the paper compared to the previous results for the NLS with single power, we particularly construct two different rescaled families of variational problems, which leads to an NLS with single power in different limiting profiles respectively, to establish the periodic dependence results.

math.AP↗