The 35th annual conference on Astronomical Data Analysis Software & Systems 2025 in Görlitz, Germany

Identifying pulsar signals from radio telescope data archives poses a major big data challenge. Although several efficient algorithms have been developed to tackle this problem, our software package introduces an innovative approach: a machine learning–based framework that employs training data generated through Digital Twins derived from theoretical physics models, combined with a U-Net–based neural network for pulsar signal segmentation.This framework aims to fascilitate the verification and discovery of unknown pulsars and other astronomical signals grounded by physical theory, thereby advancing the capabilities of astronomers and physicists in their exploration of the universe. This poster shows the status of the framework.