Applications of Feed-Forward Neural Network to Study Irregular-Shape Particle Effects on Hydrodynamics Behavior in a Liquid–Solid Circulating Fluidized Bed Riser

Feed-forward neural network (FFNN) modeling techniques are applied to study the flow behavior of different-size irregular-shape particles in a pilot scale liquid–solid circulating fluidized bed (LSCFB) riser. The adequacy of the developed model is examined by comparing the model predictions with experimental data obtained from the LSCFB using lava rocks (dmean 500 and 920 µm) and water as solids and liquid phases, respectively. Axial and radial solid holdup profiles are measured in the riser at four axial locations (H 1, 2, 3 and 3.8 m above the distributor) above the liquid distributor for different operating liquids. In the model training, the effects of various auxiliary and primary liquid velocities, superficial liquid velocities and superficial solid velocities on radial phase distribution at different axial positions are considered. For model validation along with other experimental parameters, dimensionless normalized superficial liquid velocities and net superficial liquid velocities are also introduced. The correlation coefficient values of the predicted output and the experimental data are found to be 0.95 and 0.94 for LR-500 and LR-920 particles, respectively which reflects the competency of the developed FFNN model.