Neural-network-based semi-empirical turbulent particle transport modelling founded on gyrokinetic analyses of JT-60U plasmas. (22nd August 2019)

• Record Type: Journal Article
• Authors: Narita, E.; Honda, M.; Nakata, M.; Yoshida, M.; Hayashi, N.; Takenaga, H.
• Is Part Of: Nuclear fusion. Volume 59:Number 10(2019:Oct.)
• Publication Date: 2019-08-22
• Subjects: turbulent transport -- gyrokinetic modelling -- neural network -- density profile prediction; Nuclear fusion -- Periodicals; 621.48405
• DOI: 10.1088/1741-4326/ab2f43
• Languages: English
• ISSNs: 0029-5515

Abstract

Abstract: Novel turbulent particle transport modelling has been proposed following joint analyses with gyrokinetic calculations and JT-60U experimental data. Here the diagonal (diffusion) and off-diagonal (pinch) transport mechanisms are treated individually. Besides the decomposition, realistic particle sources from neutral-beam fuelling, which have not been discussed in earlier gyrokinetic studies on particle transport, are taken into account. Taking advantage of the features offered by the modelling, the contribution from each transport mechanism to the turbulent particle flux has been quantitatively clarified. Furthermore, a framework has been developed to calculate the turbulent particle flux driven by each transport mechanism accurately and quickly, taking a neural-network-based approach. The framework can be used for fast prediction of density profiles and for investigating the effects of the transport mechanisms on density profile formation.