Hazard function exploration of tokamak tearing mode stability boundaries

Abstract

The technical feasibility of hazard function learning and exploration on multi-scenario tokamak plasma time-series datasets is demonstrated. In particular, the onset intensity (events per time) of born-rotating tearing mode instabilities in DIII-D tokamak plasmas is modeled as a function (the hazard) of gross plasma equilibrium covariates. It is shown that the hazard function is calibrated in a precise probabilistic sense. Partial dependence visualizations of this function suggest that the approach can be developed into powerful tokamak database query software geared towards physics discovery.