Curvature-based energy spectra from neural representations of turbulent Rayleigh-Bénard convection

Kurzfassung

Physics-informed neural networks are used to analyze turbulent Rayleigh-Bénard convection. Taking advantage of automatic differentiation, the availability of derivatives allows the local computation of various turbulent parameters. In the work presented here, we determine the curvature of the velocity vector field as an indicator of the prevalent length scale, to compute an energy spectrum based on its combined histogram with the kinetic energy.