Speaker
Description
To advance autonomous operation at the Shanghai Synchrotron Radiation Facility (SSRF), we developed a diagnostic agent using a locally deployed Qwen3-30B model. Unlike black-box predictors, this agent employs physics-informed context engineering by langgraph to emulate expert reasoning: it iteratively excludes healthy subsystems, identifies parameter drifts, and validates hypotheses by referencing historical fault records. The system is designed to handle general accelerator anomalies, ranging from transient trips to long-term drifts, providing a secure, intelligent, and modular framework that facilitates continuous upgrades. Currently, a lightweight supervisory layer continuously monitors beam status and triggers the agent only upon anomaly detection, ensuring efficient, on-demand diagnosis.
| In which format do you inted to submit your paper? | LaTeX |
|---|