Prediction of drop size distribution and mean drop size in an L-shaped pulsed packed column using artificial neural network (ANN) model and semi-empirical correlation

Abstract

Abstract This study proposes the use of artificial neural network (ANN) and semi-empirical models for predicting mean drop size and drop size distribution in an L-shaped pulsed packed extraction column under non-mass-transfer conditions, employing toluene-water (T/W) and n-butyl acetate-water (B/W) systems. The ANN model was trained using the Levenberg–Marquardt algorithm and demonstrated excellent predictive performance, achieving R2 values of 0.981 and 0.986 for drop size distribution and mean drop size, respectively. Furthermore, the ANN model demonstrated superior accuracy in predicting mean drop size, with an average absolute relative error (AARE) of only 2%, significantly outperforming the semi-empirical model’s AARE of 6%. Moreover, the ANN model significantly reduced the maximum prediction error in drop size distribution, particularly under conditions where the semi-empirical model showed poor performance. The findings underscore the critical role of pulsation intensity and interfacial tension in determining drop size. Notably, higher pulsation intensities were found to significantly reduce the influence of interfacial tension. Finally, new semi-empirical correlations derived from the experimental data were established to predict both mean drop size and drop size distribution.