DECAF cross-machine comparison of born-rotating mode locking forecaster developed for real-time implementation on KSTAR

Operation of reactor scale tokamaks with high thermal and magnetic energy density will require low occurrence of plasma disruptions in which the quenching of these energies can compromise critical plasma facing components or vessel integrity. The presence of rotating MHD modal instabilities can deteriorate the plasma performance in tokamaks and their locking to the wall leading to plasma disruptions. Prediction, observation, and avoidance of these modes is therefore essential in the operation of tokamaks. A mode locking forecaster based on a torque balance model has been developed across devices of ranging aspect ratio and error fields. Analysis of the KSTAR, NSTX, MAST-U, and DIII-D databases has been conducted and compared. Based on the success of database analysis, real-time forecasting and identification modules were written and installed on the KSTAR superconducting tokamak. Over 50 dedicated plasma, of different scenarios, were run experimentally to test this system (real-time DECAF [1]). The results show a 100% success rate in identifying true positives in this collection of nearly equal disrupted / non-disrupted plasmas. Real-time results are shown, with comparisons to the analogous offline analysis, and lessons learned in this process are discussed. [1] S.A. Sabbagh, et al., Phys. Plasmas 30, 032506 (2023); https://doi.org/10.1063/5.0133825