Abstract
The current study presents an integrated methodology for improving the performance of conventional fluidized bed reactors using swirling flow via a blade distributor and conical shape. The primary objective of this study is to investigate the combined effect of swirling flow and conical geometry on bed flow and heat transfer, which has not been sufficiently studied, especially through experimental techniques supported by machine learning modeling able to predict performance. A blade distributor was employed in the SCFBR, while the bed heat transfer coefficient and surface particle temperature were estimated and measured at different inlet velocities ranging from 0.993 to 2.5 m/s. The study results emphasized that the SCFBR clearly outperforms the CFBR based on lower bed pressure drop, more uniform particle distribution, and higher heat transfer coefficient. The heat transfer coefficient of the SCFBR increased by 40% compared to that of the CFBR, corresponding to a more homogenized distribution of particle surface temperature. The machine learning findings elucidate the superiority of the Extra Trees model in modeling the bed pressure drop and heat transfer coefficient, achieving with an R² of 0.973, an RMSE of 52.11, and a MAE of 35.11 for the heat transfer coefficient and an R² of 0.965, an RMSE of 8.76, and a MAE of 2.27 for the bed pressure drop. The good agreement between the experimental and ML results demonstrated the reliability of the proposed methodology. This study emphasizes the importance of integrating swirling flow with conical geometry, supported ML modeling, and represents a promising method for developing thermally efficient fluidized bed reactors with low energy consumption for advanced industrial applications.