Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 1: AKPIK I

AKPIK 1.3: Vortrag

Mittwoch, 1. April 2020, 17:00–17:15, H-HS XII

Real-time Data Analysis and Automation using Deep Learning at Accelerator-based light sources — •Mohammed Bawatna, Jan Deinert, and Sergey Kovalev — Institute of Radiation Physics, HZDR, Dresden, Germany

Accelerator-based light sources are constantly advancing and offer insights into the world of molecules, atoms, and particles on the ever shorter length and timescales. This goes along with a rapid and highly accurate transformation of analog quantities into discrete values for electronic storage and processing with exponentially increasing amounts of data. The current lack of real-time data analysis impedes the direct feedback and the possibility for fine*tuning in time-critical beam*time experiments, and data collection, storage, data management, and curation of the data become more and more challenging. This contribution is focusing on real-time analysis methods using deep learning and evaluation of large data volumes generated at the super-radiant terahertz facility (TELBE) at Helmholtz Zentrum Dresden Rossendorf (HZDR). Here, the pulse-resolved data acquisition at a 100 kHz repetition rate enables femtosecond timing accuracy and high dynamic range but creates datasets at a rate of GB per minute, which are challenging to handle. We will also introduce our online ultrafast DAQ system that uses a GPU platform for real-time image processing, and a custom high-performance FPGA board for interfacing the image sensors and provide a continuous data transfer.