Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
To accelerate parametric PDE solves without sacrificing fidelity, this work trains a neural network error-gating model to predict when computationally expensive boundary laws can be safely swapped for simplified asymptotic forms. Stiff Robin and nonlinear boundary conditions severely bottleneck multiscale simulations in singular limit regimes, but naive substitution introduces uncontrolled error. Evaluated on galvanic corrosion and nonlinear evolution benchmarks, the learned gating mechanism reliably routes to cheaper Dirichlet limits whenever both domain and boundary errors fall strictly within user-defined tolerances.
Stiff boundary conditions don't always need full-fidelity simulation: learned error gates determine exactly when a solver can safely drop to cheaper Dirichlet limits without violating numerical tolerances.
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.