Improvement of filamentary plasma identification via neural networks

摘要

The identification of plasma shape and position for control purposes in thermonuclear fusion devices is usually performed via external magnetic measurements. The result should be available in a few ms to guarantee an effective control action. Fast solution of such inverse problem can be achieved via Equivalent Currents (EC) method or Artificial Neural Networks (ANN). Anyway, in the case of EC, their location become critical during the dynamical evolution of the fusion event. ANN may be beneficial in optimally locating, in real time, the EC for various plasma configurations.