Karlsruhe 2024 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 15: Neutrino astronomy 1
T 15.2: Vortrag
Montag, 4. März 2024, 16:15–16:30, Geb. 30.23: 6/1
Improving DSNB Event Detection at JUNO: Advancements Through 3D Convolutional Neural Networks — •David Maksimović1, Daniel Tobias Schmid1, Dhaval J. Ajana2, Michael Wurm1, Marcel Büchner1, Arshak Jafar1, George Parker1, Oliver Pilarczyk1, and Tim Charisse1 — 1Johannes Gutenberg-University Mainz, Institute of Physics — 2Department of Physics, Florida State University, Tallahassee, FL 32306, USA
The detection and analysis of the Diffuse Supernova Neutrino Background (DSNB) pose a significant challenge in neutrino astronomy, primarily due to backgrounds mimicking the Inverse Beta Decay (IBD) signature events. The Jiangmen Underground Neutrino Observatory (JUNO) uses a liquid scintillator to detect these neutrinos, especially challenged by Neutral-Current (NC) interactions of atmospheric neutrinos in the 12 to 30 MeV range.
In this talk, we introduce a novel method employing 3D Convolutional Neural Networks (3D CNNs) for better discrimination of DSNB events from these backgrounds. This technique analyses time-sequenced data from photomultiplier tube (PMT) hit patterns, arranged in frames like a movie, capturing the spatial-temporal dynamics of particle interactions. Simulation studies within the JUNO detector environment show our 3D CNN method significantly improves background reduction. Compared to previous applied machine learning methods, our approach shows a 30% reduction in background levels and a 17% improvement in detection accuracy.
Keywords: Machine Learning; Diffuse Supernova Neutrino Background; JUNO; 3D Convolutional Neural Networks; Liquid Scintillators