Título: Forecasting Chaotic Time Series and Plasma Disruption Instabilities by Using Artificial Neural Networks
Autores: Oliveira, Kenya Andrésia de; Vannucci, Álvaro; Silva, Elton Cesar da
Resumo: Two-layer feedforward neural network was used to forecast chaotic time series and disruptive instabilities observed in the TEXT tokamak plasma discharges with very promising results. In both cases it was verified that a neural network with an architecture of the type m:2m:m:1, where m is the embedding dimension of the attractor of the dynamical system in consideration, is a very good initial guess for the process of finding the ideal architecture for the neural network, which is usually hard to achieve. A 15:30:15:1 (m = 15) neural network was capable, for example, to forecast the disruptive instabilities in X-rays signals up to 4 ms in advance, period of time about fourfold larger than the one obtained previously, when experimental magnetic signals from Mirnov coils were used. These very good forecasting results and those obtained by using chaotic temporal series like Lorenz system clearly suggest that there is an interplay between the architecture of a multilayer network and the embedding dimension m of the time series used. They are quite significant and opens up to the possibility of using neural networks for making predictions over the evolution of nonlinear systems, such as confined plasmas, for example.
Código DOI: 10.21528/CBRN2001-020
Artigo em pdf: 5cbrn_020.pdf
Arquivo BibTex: 5cbrn_020.bib