Unsupervised deep learning for super-resolution reconstruction of turbulence

Recent attempts to use deep learning for super-resolution reconstruction of turbulent flows have used supervised learning, which requires paired data training. This limitation hinders more practical applications reconstruction. Therefore, we present an unsupervised model that adopts a cycle-consistent generative adversarial network can be trained with unpaired turbulence Our is validated using three examples: (i) recovering the original flow field from filtered direct numerical simulation (DNS) homogeneous isotropic turbulence; (ii) reconstructing full-resolution fields partially measured DNS channel flows; and (iii) generating DNS-resolution large eddy (LES) flows. In examples (ii), are available our demonstrates qualitatively quantitatively similar performance as best supervised-learning model. More importantly, in example (iii), where impossible, successfully reconstructs high-resolution statistical quality LES data. indeed possible, opening new door wide application fields.