Neural networks speed up solving Helmholtz equation systems
Neural Network Preconditioning: A Case Study for the Solution of the Parametric Helmholtz Equation
Machine Learning
Summary
Solving certain equations that model waves can be very slow with traditional computers. The authors show that a special kind of artificial intelligence called a neural network can help speed this up by assisting the usual solving method. The network learns by itself how to improve the solving process without needing example answers. While it alone can't solve the problem, using it together with an existing solver works much faster. The approach works well even when conditions change, like different wave speeds or sizes.
What this means in practice
- •For computational engineers: Speed up iterative solvers for wave equation simulations in engineering designs using a neural network preconditioner.
- •For acoustic simulation teams: Reduce computation time for acoustic wave modeling by integrating a neural preconditioner in solver toolchains.
Authors
Luc Giraud, Carola Kruse, Paul Mycek, Maksym Shpakovych, Yanfei Xiang
Abstract
This work presents a hybrid numerical approach for solving linear systems arising from the discretization of the two-dimensional parametric Helmholtz equation. A convolutional neural network based on the U-Net architecture is trained in an self-supervised manner (i.e., without explicitly generating reference solutions) to approximate the inverse of the discretized Helmholtz operator, using a loss function involving the residual norm of the linear system. The trained network is used as a nonlinear preconditioner within the flexible GMRES (FGMRES) algorithm. Numerical experiments show that while the neural network is not accurate enough to act as a standalone solver, it significantly improves the convergence of FGMRES when employed as a preconditioner. The neural preconditioner demonstrates robust performance and generalization capabilities with respect to variations in the velocity field and the domain size. Comparisons with algebraic preconditioners, based on LU factorizations of approximations of the original system matrices, indicate the superior efficiency of the neural approach under equivalent conditions. We believe that the proposed method is not tied to a specific neural architecture and can be extended to other parametric PDEs.