Anomaly Detection System for Accelerator Magnet Power Supplies Based on Deep Autoencoders

Introduction / Main Floor: Anomaly Detection System for Accelerator Magnet Power Supplies Based on Deep Autoencoders

This article introduces an intelligent anomaly detection project deployed at the Indus-2 synchrotron of India's RRCAT laboratory. The project uses a deep autoencoder neural network to real-time monitor the operating status of 117 magnet power supply units, achieving a detection accuracy of 95.2%, and provides an innovative AI solution for the safe operation of large-scale scientific facilities.