Data-Driven Mode Loss Detection for the Emitted Radio Frequency Beam of ITER Gyrotrons

Gyrotrons are the sources of the millimeter-length coherent electromagnetic (EM) wave injected into magnetically confined plasma by the electron cyclotron heating system. In ITER’s re-baseline, the electron cyclotron heating system foresees employing an increasing number of gyrotrons to provide on-demand power for extended pulses. Their reliability relies on maintaining the correct mode for stable EM beam emission throughout the pulse. When a wrong mode is generated, the radio frequency (RF) wave does not couple with the gyrotron output window, and the power remains inside the gyrotron body. The consequence is twofold: the plasma does not receive the expected power, and the energy left inside the gyrotron leads to a potentially dangerous overheating. To ensure correct gyrotron operations and prompt action in the event of wrong mode generation, a dependable mode loss detection system is of crucial importance. Since RF measurements are not available in the current baseline due to the layout arrangement of the transmission lines, alternative detection methods must be explored. In this work, the effect of mode loss during operation on the acquired gyrotron data at different time scales is studied, and the phenomenon is characterized. Moreover, two data-driven methods based on the collector calorimetry and the electron beam current trend prediction have been developed as mode loss detections, and their performances are evaluated against dedicated datasets.