Fuzzy Inference Systems (FIS) for classification and identification of plasma columns in Tokamak Reactors

This paper is mainly concerned with the application of a novel technique of data interpretation to the identification of plasma columns in Tokamak reactors fro nuclear fusion applications. theproposed method exploits several concepts derived from fuzzy system theory and particularly the idea of fuzzy inference. Additionally, a qualitative reasoning technique for easily classifying the type of plasma equilibrium by direct inspection of the data measurement reading has been developed. The proposed technique is used to analyze simulated databases of elongated multi-category plasma equilibria based on ASDEX-Upgrade geometry and poloidal field coils configuration. As well as demonstrating the successful recovery of sclr equilibrium parmeters, we show that the fuzzy inference technique can yield pratical advantages compared with earlier methods, also exploiting neural systems. An original technique of identification based on the extraction of a set of fuzzy rules directly from the available database has been proposed for the ASDEX-Upgrade configuration; however, the proposed methodologies can be applied to any inference model, an adaptive fuzzy-neural inference system based on adaptive network schemes is proposed. Finally, a technique for sensor selection and ranking based on fuzzy logic aiming to carry out dimensionality reduction is prsented.