Wake Pattern Identification Using Graph Matching

Flow patterns generated in the wakes of immersed objects such as bluff bodies and swimming animals are related to the flow generation mechanisms (e.g., vortex shedding) and can provide a window to understanding hydrodynamic performance such as drag or propulsive efficiency. In this work the essential flow features are represented by critical points of the velocity field. The pattern is encoded in a graph constructed from the critical points, with node properties determined by the critical point character and graph edges (connections between critical points) carrying a scale-independent “weight” property associated with proximity of the connected points. Flows are compared by determining a best match between the weighted graphs of the two flows, accounting for constraints based on the character of matched critical points. Relative similarity is assessed from differences in weights of the matched edges. The approach is able to correctly match flow fields of the same type from potential flow representations of several bluff body and aquatic locomotion flows, even when distortions of up to 20% of the vortex spacing are applied to the flows.