Research data supporting "Data-driven classification of sheared stratified turbulence from experimental shadowgraphs"

For a more detailed description, refer to READ_ME.pdf This dataset corresponds to an automated analysis of stratified turbulent flow visualisations in an inclined duct laboratory experiment. The goal is to reduce the complexity of 113 long shadowgraph movies previously shared on another repository doi.org/10.17863/CAM.104471. Our analysis takes each of the 50,155 frames and automatically extracts statistics corresponding to the morphology of density interfaces embeddd within the flow, before clustering these statistics in a low-dimensional space. The resulting clusters are physically interpretable and allow us to classify the flow instantaneously into a distinct type of turbulence. The main dataset is the Matlab array pca_and_clustering_data.mat, providing the interface morphology 10-D vector and the ensuing clustering of each of the 50,155 frames processed in the associated article (see the methodology and variable names in Section III of the article). In addition, we provide: - the archive movies.zip contains six .mp4 movies showing the temporal evolution of all experiments at theta=1, 2, 3, 4, 5 and 6 degrees (including the principal components and clusters). Note that clusters initially named A, B, C, D, E in the movies correspond to L, B, O, G, U respectively in the paper. - metadata_spreadsheet.xlsx shows all the experimental parameters and information about temporal spacing between frames. - clustering_code.zip contains the Matlab code used to cluster the data (including the implementation of the OPTICS algorithm).