Investigation of artificial neural network methodology for modeling of a liquid–solid circulating fluidized bed riser

An artificial neural network (ANN) approach is investigated to model and study the phase holdup distributions of a liquid–solid circulating fluidized bed (LSCFB) system. The ANN model is developed based on different operating parameters of the LSCFB including primary and auxiliary liquid velocities, and superficial solids velocity. The competency of the model is examined by comparing the model predicted and the experimental phase holdup of the LSCFB riser reactor. It is also found that the ANN model successfully predicted the radial non-uniformity of phase holdup that is observed in the experimental runs of the riser. When compared, the model predicted output and trend of radial flow structure for solids holdup are in well agreement with the experiments. The mean absolute percentage error is around 6% and the correlation coefficient value of the predicted output and the experimental data is 0.992.

The ANN model is developed based on different operating parameters of the LSCFB including primary and auxiliary liquid velocities, and superficial solid velocity. When compared, the model predicted output and trend of radial flow structure for solid holdup are in good agreement with the experiments.Figure optionsDownload as PowerPoint slideHighlights
► Model is validated by comparing model predicted and a pilot scale LSCFB data.
► ANN model predicted and experimental data are in agreement.
► Statistical performance measures of the ANN model are quite competitive.
► Mean absolute percentage error (MAPE) is only around 6%.
► Correlation coefficient of the predicted output and the experimental data is 0.992.