Description
An Array of Cherenkov Telescopes is equipped with a multitude of sensors spread all over the instrumentation and collects a large volume of housekeeping and auxiliary data coming from telescopes, weather stations and other devices in the array site. In this poster we will present how we intend to exploit the sensor’s information to perform predictive maintenance (PdM) using with the most advanced artificial intelligence algorithms. This technique will be useful to detect in advance the remaining useful life of the array components, and to estimate the correct timing for performing their maintenance. The application of PdM will allow to minimize the array downtime, to increase the telescopes sub-components longevity, and to reduce the costs due to unforeseen maintenance. Our model used a time series data coming from several different sensors (temperature, current, torque, etc.) dedicated to monitoring several mechanical components of the telescopes (engines, cameras, encoders, etc.). The adopted unsupervised machine learning approach will allow us to perform the correct trade-off between preventive and corrective maintenance.
