Using Machine Learning to Locate Three-Dimensional Magnetic Reconnection within PHASMA

The PHASMA (PHAse Space MApping) facility at WVU uses pulsed plasma guns to investigate magnetic reconnection through the interaction of merging magnetic flux ropes. This study, encompassing approximately 650 shots of helium double flux rope, utilizes parameters such as the flux function and fast photodiodes to identify and locate magnetic reconnection within PHASMA through the application of machine learning techniques. Shots are clustered via having similar bias and arc currents via unsupervised machine learning. Magnetic flux evolution movies are then constructed from these grouped shots. Photodiode measurements are also correlated using unsupervised machine learning to identify similarities that exist between shots. Based on these parameters, we attempt to predict magnetic reconnection at a distant location based on a predictive neural network that uses nonlocal (edge) magnetic measurements and line-integrated fast photodiode measurements in PHASMA. This analysis will enable new studies of reconnection in highly turbulent and irreproducible systems by providing a means of localizing the time and location of reconnection to better synchronize triggered measurements, such as Thomson scattering measurements of the electron distribution function.