Fast Identification Problems in 3-D Iron Core Fusion Devices

In nuclear fusion devices, numerical algorithms to model the plasma behavior and all of the other field sources are needed for the control system, especially when plasma instabilities are present. The control measures have to be taken in very strict time intervals (real-time problems); therefore, the involved codes have to be very fast. These days, suitable identification procedures are mainly based on information coming from the magnetic measurements, typically in the axisymmetric frame. Unfortunately, the presence of iron materials strongly increases the complexity of the identification, for both the lack of linear behavior and 2-D hypothesis. This paper proposes fast and effective iron modeling useful for practical real-time applications; the iron is modeled by means of an integral approach, whereas an inverse problem that relates the available magnetic measurements with the unknown currents and magnetizations is solved in a fast way by means of modern graphical-processing unit architectures. © 2017 IEEE.