Do particle-related parameters influence circulating fluidized bed (CFB) riser flux and elutriation?

In fluidized bed systems, particle properties (e.g., particle diameter, Reynolds number) are acknowledged to be important in dictating various fluidization phenomena, and are thereby commonly included in empirical correlations and physical models. The goal of this study was to harness machine learning tools to determine the relative dominance of the parameters in riser flux and elutriation to provide insights in advancing such predictive tools. The 1320 dataset involved monodisperse, binary, and polydisperse Geldart Group B particles investigated in a pilot-scale circulating fluidized bed riser. Regarding both riser flux and elutriation, (i) random forest ranking and self-organizing map (SOM) weight planes indicate that the pressure at the riser bottom was by far the most dominant relative to the particle-related parameters; and (ii) the neural network model based on the pressure at the riser bottom alone gave an le of almost 1, affirming the lack of influence of particle-related properties. (C) 2020 Elsevier Ltd. All rights reserved.