This paper presents an active learning framework to accelerate predictions of galvanic corrosion. The framework uses a low-cost surrogate model (Gaussian Process and Neural Network) calibrated on Finite Element (FE) simulations to identify optimal input configurations (environmental and geometric parameters) for improving predictions of cathodic current. A staggered workflow refines predictions and identifies information-rich data points, enabling expansion to a larger parameter space. The framework significantly reduces computational costs and improves prediction accuracy.
Publisher
npj Materials Degradation
Published On
May 21, 2024
Authors
Aditya Venkatraman, Ryan Michael Katona, Demitri Maestas, Matthew Roop, Philip Noell, David Montes de Oca Zapiain
Tags
active learning
galvanic corrosion
surrogate model
Gaussian Process
Neural Network
Finite Element simulations
cathodic current
Related Publications
Explore these studies to deepen your understanding of the subject.
