Getting fast and reliable predictions of turbulent transport properties is an important challenge in magnetic fusion. Previous research (Heinonen and Diamond 2020 Phys. Rev. E101 061201) proposed a data-driven approach using neural networks to predict the particle flux and Reynolds stress in a minimal model of drift-wave turbulence. The present work extends this approach to the interchange instability driven by the magnetic curvature. This study highlights the importance of assessing the reliability of data-driven models, especially in view of their application to more complex high-fidelity simulations. In particular, a figure of merit is introduced to identify regions of the input space where the model’s predictions cannot be trusted. The data-driven model predictions are used to gain insight into the vorticity gradient’s contribution to the turbulent flux and the antiviscous nature of the Reynolds stress.
This paper presents a data-driven method to predict turbulent transport properties in magnetic fusion, which is crucial for understanding and controlling fusion plasmas. The approach uses neural networks to model the particle flux and Reynolds stress, and introduces a reliability metric to identify regions where the model's predictions are uncertain. The results provide insights into the complex interactions between turbulence and transport.