Artificial neural network based predictive negative hydrogen ion helicon plasma source for fusion grade large sized ion source

Abstract

Ion source operation and control is a challenging task, and the manual search of the optimized input parameter by the operator becomes a time consuming and cumbersome process. HELicon Experiment for Negative Ion Source (HELEN-I) with single driver developed at the Institute for Plasma Research (IPR) for the production negative hydrogen ion involves a multitude of interacting systems (such as Radio Frequency (RF) power, gas feed pressure, magnetic fields). This paper presents an Artificial Intelligence (AI) driven multiple-input and multiple-output model for HELEN-I to get optimized values of ion saturation current (mA) and high plasma density of order 1018/m3. The developed Artificial Neural Network (ANN) based model predicts the outputs of the HELEN-I and is further utilizes a heuristic global optimization Particle Swarm Optimization (PSO) algorithm to select optimal source parameters to predict the desired outputs without conducting real-time experiments. The experimental studies further validate the results obtained by the proposed approach.