Monitoring Large-Scale Radio-Detection Arrays with Machine Learning

In recent years, radio-detection techniques, such as those employed in the GRAND experiment, have emerged as a promising method for detecting ultra-high-energy cosmic rays (UHECRs). One of the key advantages of radio detection is its cost-effectiveness, allowing for the deployment of large arrays that can cover vast areas necessary for measuring the low fluxes of UHECRs. However, this comes with the challenge of monitoring the functionality of a massive number of antennas (up to tens of thousands) over a vast area (tens of thousands of km2). In this talk, we will present an approach to addressing this challenge using a combination of dimensionality reduction (UMAP) and clustering (DBSCAN) algorithms applied to periodically triggered monitoring data of a GRAND prototype setup. Our method aims to identify malfunctions and periods of poor operation, enabling efficient maintenance and optimization of the radio-detection system.