A method for identifying weak and transient circulation cells in river bends, and application to LES simulation data

River bend flow is characterized by secondary circulation that can occur as one or more rotating cells. Distinguishing between one-cell and multiple-cell flow is often done visually with mean flow streamlines. However, there are instances where one cell dominates the flow and thus weaker cells are not distinguishable in the mean flow alone. Nonetheless, even a transient or weak cell can affect momentum and pollutant transport. Thus, a more robust method for identifying cells, capable of distinguishing weak or transient cells, is desirable. The proposed method requires a time series of velocity data on a plane, as may be collected via Particle Image Velocimetry or similar methods. Here, Large Eddy Simulation data are used for illustration. Vortices in every time step snapshot are identified using the method from Graftieaux et al., 2001 and their centre, circulation strength, and area are recorded. Applying MATLABs density-based spatial clustering of applications with noise (DBSCAN) to the vortex centre locations divides the vortices based on which cell they represent. Vortices that do not clearly lie within any cluster are excluded automatically. In the included figure, five cells are identified using instantaneous vortex clustering. Each cluster can be assigned a location based on the circulation-weighted mean instantaneous vortex locations. These cluster centres approximate the centres of cells identifiable in the mean flow. The mean circulation of a cluster can be calculated as the circulation of all contributing vortices, averaged over time. This allows for easy comparison of the circulation contributions of each cell under varying conditions, even when the cell in question is not strong enough to appear in mean flow streamlines. This method is still somewhat subjective, as there are parameters that must be tuned. The following best practices are proposed: Within reason, err on the side of dividing into more clusters, as clusters can be combined but cannot be separated after analysis; and cluster positively and negatively rotating cells separately. Graftieaux, L., Michard, M., & Nathalie, G. (2001). Combining PIV, POD and vortex identification algorithms for the study of unsteady turbulent swirling flows. Measurement Science and Technology, 12(9), 14221429. https://doi.org/10.1088/0957-0233/12/9/307