An artificial neural network for prediction of gas holdup in bubble columns with oily solutions

Gas holdup in a bubble column reactor filled with oil-based liquids was estimated by an artificial neural network (ANN). The ANN was trained using experimental data from the literature with various sparger pore diameters and a bubbly flow regime. The trained ANN was able to predict that the gas holdup of data did not seen during the training period over the studied range of physical properties, operating conditions, and sparger pore diameter with average normalized square error < 0.05. Comparisons of the neural network predictions to correlations obtained from experimental data show that the neural network was properly designed and could powerfully estimate gas holdup in bubble column with oily solutions.