Fig. 4: Unconditional generation of non-equilibrium turbulence over periodic hills.
From: Conditional neural field latent diffusion model for generating spatiotemporal turbulence
a Instantaneous velocity magnitude of the DNS flow data. b Instantaneous velocity magnitudes of three randomly generated realizations by CoNFiLD. c Turbulence statistics at selected locations from CoNFiLD (red lines) and DNS (blue dots), including the mean streamwise velocity \(\bar{u}\) (upper left), mean vertical velocity \(\bar{v}\) (upper right), Reynolds shear stress \(\overline{u^{\prime} v^{\prime} }\) (lower left), and the total turbulence kinetic energy (\(\bar{k}\)) (lower right). The spatial coordinates are normalized by the hill height h, and the statistical quantities are normalized by the bulk velocity Ub. See Supplementary Video 2 for animation.
