Figure 5
From: A scalable convolutional neural network approach to fluid flow prediction in complex environments
Local Orientational Vector Field Entropy (LOVE). (A) Channel flow around a circular obstacle. The inlet flow enters from the left of the channel with a uniform velocity parallel to the abscissa. To quantify how the obstacle’s impact on the flow diminishes with downstream distance, \(\Delta x\), we compute the orientational entropy of the flow within a sequence of obstacle-shaped neighborhoods centered at successively further downstream locations. One can see from the plotted flow field snapshots (colored circles) that the induced vorticity immediately dowstream from the obstacle gradually dampens out, resulting in the flow field returning to its uniform inlet profile sufficiently far downstream. (B) The local orientational entropy of the flow field as a function of displacement from the downstream edge of the obstacle for laminar flow (blue) and time-averaged vortical flow (red). (C) The same entropy, but now comparing the time-averaged vortical flow (red) to the time-averaged entropy (black) over five different temporal snapshots of the dynamic flow field. For the latter curve, standard error bars are plotted alongside the data.
