Failure type detection and predictive maintenance for the next generation of imaging atmospheric Cherenkov telescopes

TL;DR

This paper proposes a machine learning-based approach for failure detection and predictive maintenance in the next generation of atmospheric Cherenkov telescopes, leveraging auxiliary sensor data to improve operational reliability.

Contribution

It introduces a novel predictive maintenance model tailored for large-scale astrophysical telescope arrays, integrating supervised and reinforcement learning techniques.

Findings

Successful development of failure prediction models

Enhanced maintenance scheduling efficiency

Potential reduction in telescope downtime

Abstract

The next generation of imaging atmospheric Cherenkov telescopes will be composed of hundreds of telescopes working together to attempt to unveil some fundamental physics of the high-energy Universe. Along with the scientific data, a large volume of housekeeping and auxiliary data coming from weather stations, instrumental sensors, logging files, etc., will be collected as well. Driven by supervised and reinforcement learning algorithms, such data can be exploited for applying predictive maintenance and failure type detection to these astrophysical facilities. In this paper, we present the project aiming to trigger the development of a model that will be able to predict, just in time, forthcoming component failures along with their kind and severity