Machine learning helps swap complex boundary conditions efficiently
Selective boundary condition reduction via learned error gating
Machine Learning
Summary
When solving physics problems with equations, the edges or boundaries often need special rules that can be hard and slow to use. The researchers created a method where a computer learns when it’s okay to use simpler edge rules instead without losing accuracy. It uses pairs of solutions to train a model that predicts errors and only switches to simpler ones if the predicted mistakes are small. They tested this on corrosion and other physical problems and showed it can save effort while keeping good results.
What this means in practice
- •For computational engineers: Replace complex boundary conditions with simpler ones adaptively to reduce simulation time while maintaining accuracy in engineering models.
- •For industrial corrosion analysts: Improve simulation efficiency in galvanic corrosion models by selectively using simpler boundary rules when accurate enough.
Authors
Daniel Fernández, Dominik Penk, Dominik Riedelbauch
Abstract
Parametric PDEs can admit different boundary conditions with different accuracy and computational cost. We introduce a framework for learning when one reduced boundary condition can replace another: paired solutions train a neural network to estimate the resulting domain and boundary errors, and the simpler condition is used only when both predicted errors meet prescribed tolerances. We focus on singular limits in applications, in which a stiff Robin or nonlinear boundary law is replaced by its limiting Dirichlet form. We evaluate the method on a galvanic corrosion problem and other nonlinear stationary and evolution problems.