Speaker
Description
Synchrotron light sources must deliver brilliant and stable photon beams to various user experiments, and their reliability depends on the cooperative operation of many subsystems. As light sources are becoming more complex, accelerator operation is getting harder for human operators to perform stably. Therefore, we believe that artificial intelligence (AI) and machine learning (ML) offer promising solutions toward more reliable and sustainable operation. This presentation focuses on the following three directions for AI-assisted reliable operations. First, AI/ML-based intelligent monitoring of the operating condition for each subsystem enables abnormal behavior to be recognized before causing critical trouble. Second, automation of routine tasks helps us to mitigate the growing difficulty of securing and training sufficient operators. Third, an AI agent can make it easier to realize advanced operating conditions that would be difficult for humans and require time-consuming adjustments. We present the ongoing projects and future plans to implement these functionalities in SPring-8/SACLA.