Cornerstones are the key stones: Using interpretable machine learning to probe the clogging process in 2D granular hoppers

This dataset contains the information recorded for approximately 50,000 hopper flows under varying conditions. Each file represents a single flow event, from beginning of flow to final clog forming. The files are matlab structures, containing the positions, radii, frame number, tracked particle ID's and velocities for every grain in the camera field of view throughout the flow. Other values, such as the position of the outlet and the total mass ejected during the flow are also included.

Methods
Data are the tracked positions of grains throughout individual hopper flows (beginning of flow to final clog). A camera records images of grains near the outlet at 130 frames per second, which was then analyzed using matlab code. The grain centers were located and species (size) identified; this information was then fed into a tracking algorithm to uniquely identify grains throughout flow. From these unique identifiers, velocity and ejected mass information was calculated.