Clusters identification and meso-scale structures in a circulating fluidized bed based on image processing

To understand the behaviors of particle clusters in the circulating fluidized bed (CFB), experiments were conducted with glass beads to acquire the image sequences of gas-solid flow on a CFB riser with a 100 mm × 25 mm cross-section and 3.2 m in length by adopting high-speed photography. An image multilevel thresholding approach using k-means algorithm was applied to perform image segmentation to identify clusters as well as core clusters in the riser, automatically. Cluster characteristics, such as the density and the number of clusters were obtained subsequently. The results show the image segmentation method based on k-means algorithm has made some improvement in terms of precision and systematicness for cluster identification. In addition, the internal structure of the cluster was analysed. Collectively, the cluster always consists of a dense core with highest solids holdup surrounded by a relatively dilute cloud with no clear boundary. High solids holdup enhances the cluster formation. On the contrary, the core cluster disappears at low solids holdup condition, indicating the cluster is only composed of cluster cloud in this case. Furthermore, based on the present experimental data, the correlations between the cluster density and the local time-mean solids holdup are presented.