Identificación de la temperatura electrónica de un plasma frío por medio de redes neuronales artificiales

Abstract

In plasma nuclear fusion devices, the performance of the experiments is related to their diagnostic capability, that is, the capability to effectively obtain the plasma parameters involved in each experiment. In the control area, for plasma magnetic confinement devices as Tokamaks or Stellerators, create, maintain and finish a plasma discharge are tasks linked to plasma parameters themselves. Conventional techniques for extracting plasma parameters, such as the Langmuir probe, involve an indirect measure and usually require off-line processing to reach a certain parameter. This offline nature restricts any plasma control task in real time. This thesis aims to address the problem of parametric plasma extraction, specifically the electron temperature, through techniques of rapid online recovery, focused towards computer science and artificial intelligence, as such as artificial neural networks.