The interaction between dislocations and obstacles, such as voids and precipitates, is crucial for the strength of body-centered cubic (BCC) iron. Due to the nanoscale size of dislocation cores, atomic modeling is necessary to study these interactions, typically involving around 100,000 atoms. Empirical atomic potentials like the embedded atom method (EAM) are commonly used, but they often fail to accurately predict the structure and energetics of dislocation cores compared to density functional theory (DFT) calculations. To address this, an artificial neural network (ANN) framework was employed to construct atomic potentials with DFT-level accuracy [1,2]. Using this ANN potential, a molecular dynamics simulation involving one million atoms was conducted to investigate the interaction between edge dislocations and obstacles in BCC iron. The study revealed the formation of an Orowan loop and its subsequent disintegration into smaller loops during interactions with rigid spheres, highlighting the detailed atomic-level mechanisms involved.