Smart computing approach for design and scale-up of conical spouted beds with open-sided draft tubes

Open-sided draft tubes provide an optimal gas distribution through a cross flow pattern between the spout and the annulus in conical spouted beds. The design, optimization, control, and scale-up of the spouted beds require precise information on operating and peak pressure drops. In this study, a multilayer perceptron (MLP) neural network was employed for accurate prediction of these hydrodynamic characteristics. A relatively huge number of experiments were accomplished and the most influential dimensionless groups were extracted using the Buckingham-pi theorem. Then, the dimensionless groups were used for developing the MLP model for simultaneous estimation of operating and peak pressure drops. The iterative constructive technique confirmed that 4-14-2 is the best structure for the MLP model in terms of absolute average relative deviation (AARD%), mean square error (MSE), and regression coefficient (R-2). The developed MLP approach has an excellent capacity to predict the transformed operating (MSE = 0.00039, AARD% = 1.30, and R-2 = 0.76099) and peak (MSE = 0.22933, AARD% = 11.88, and R-2 = 0.89867) pressure drops. (C) 2020 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.

Automated summary

An MLP neural network was employed to accurately predict hydrodynamic characteristics in conical spouted beds, using a large number of experiments and dimensionless groups extracted via the Buckingham-pi theorem. The developed MLP model has excellent predictive capacity for operating and peak pressure drops, providing precise information for design and optimization of spouted beds.