A modified OPTICS clustering algorithm for analyzing flow characteristics

Current clustering-based methods still need improvement for the flow features identifying and analysis in the field of wind engineering. Therefore, this paper proposed a modified OPTICS (Ordering Points to Identify the Clustering Structure) algorithm to extract flow features. It is a density-based clustering method with the Euclidean distance replaced by the correlation distance. It is employed to study the streamwise vortex shedding and the spanwise distribution of the A-mode in a low-Reynolds-number circular cylinder wake flow obtained by Large Eddy Simulation. Its performance is further compared with the k-means, and the original OPTICS method. The results indicated that the OPTICS algorithm based on the correlation distance can identify the streamwise vortex shedding and the spanwise distribution of the A-mode effectively with reasonable initial parameters. Compared with the k-means, this method is insensitive to the initialization parameters and its results are consequently more stable and well-determined.