Reconstruction of Molten Pool Flow Fields from High-Speed Video Using Physics-Informed Neural Networks

This research paper targets molten pool flow reconstruction in wire-arc directed energy deposition using high-speed video and physics-informed neural networks. It addresses the difficulty of direct flow sensing under extreme temperatures, intense arc illumination, and complex flow patterns. The work also motivates a data-driven alternative to computationally intensive fluid dynamics simulations that depend on assumed boundary conditions.