Development of an Intelligent Fault Classification System for BEPCII-U Superconducting Cavities

Background The superconducting radio-frequency (SRF) system is a core component of large-scale accelerators like the Beijing Electron-Positron Collider Upgrade project (BEPCII-U), and its stable operation is critical. Traditional manual fault diagnosis for SRF cavities, relying on operators’ waveform interpretation, is inefficient and subjective.

Purpose This study aims to develop an intelligent fault classification system for BEPCII-U SRF cavities, establish a "machine classification + manual verification" mechanism, and improve fault diagnosis accuracy and efficiency.

Methods Historical BEPCII SRF cavity fault data were collected and screened to build a dataset. A four-layer fault classification system (FTP synchronization, image preprocessing, deep learning classification, notification) was designed, and its performance was verified by experiments.

Results The system stably completes fault data processing and classification, significantly reducing diagnosis time and improving accuracy compared with manual diagnosis. Its four-layer architecture and dual mechanism ensure reliability and avoid misdiagnosis.

Conclusions The system addresses the limitations of traditional manual diagnosis, improves BEPCII-U SRF cavity fault diagnosis level, and provides a technical solution for intelligent operation and maintenance of large-scale accelerator SRF systems.