Quantum machine learning – A novel approach for hydrodynamics analysis and modeling of liquid-solid circulating fluidized bed risers

The Liquid-Solid Circulating Fluidized Bed (LSCFB) system is commonly employed in chemical and biochemical applications, serving as a reactor where solids act as catalysts or biofilm carriers. Mass and heat transfer heavily rely on the distribution of solids holdups, which are influenced by various operating conditions, including superficial liquid velocities and auxiliary liquid velocities. These operating conditions play a key role in controlling the hydrodynamics of solid distribution. Quantum machine learning (QML)-based modeling techniques are employed to investigate the solids holdup distribution in a Liquid-Solid Circulating Fluidized Bed (LSCFB) system. A novel application of quantum machine learning for the hydrodynamics analysis and modeling of the LSCFB riser. This work aims to lay the groundwork for investigating the applicability of QML for understanding and predicting the hydrodynamic behavior of CFB risers by leveraging the unique properties of quantum systems during the present Noisy Intermediate-Scale Quantum (NISQ) era. The unavailability of full-scale quantum infrastructure limits proper exploitation of different quantum computing opportunities including embedding. In this study glass beads particles with 500 µm used as the solid phase and water as liquid phase. At various axial locations (H = 1.0, 2.0, 3.0, and 3.8 m above the distributor) and with varying superficial liquid velocities, radial nonuniformity of solids holdup is observed. The derived models account for the impact of operating parameters, such as auxiliary and primary liquid velocities and superficial particulate velocity, on the radial phase distribution at various axial positions in the riser. By comparing the predicted solids holdup data with experimental data obtained from a pilot-scale LSCFB reactor, the accuracy of the models is determined. The models exhibit reasonable correlations with the experimental data and consistent phase distribution trends. The correlation coefficients between predicted outputs and experimental data for sequential and nonsequential QML models are 0.95 and 0.96, respectively.