Predicting hydrodynamic parameters and volumetric gas–liquid mass transfer coefficient in an external-loop airlift reactor by support vector regression

Abstract For modeling, design and scale-up of the airlift reactors, it is crucial to estimate hydrodynamic parameters and volumetric gas–liquid mass transfer coefficient for different flow regimes. Prediction of these variables had begun by applying empirical power low correlations and later evolved in use of the artificial neural networks (ANN) as the best option available in the literature. The objective of this study was to present the support vector regression (SVR) model that predicts the gas holdup, downcomer liquid velocity and volumetric gas–liquid mass transfer coefficient values in the external-loop airlift reactor better than ANN. Furthermore, to demonstrate the applicability of the SVR model, it was used on the different literature data sets with wide-ranging databanks. The statistical error analysis revealed that the proposed generalized SVR model had more precisely prediction than ANN with an average absolute relative error (AARE) of 2.17%, 1.32% and 9.64% for gas holdup, downcomer liquid velocity and volumetric gas-liquid mass transfer coefficient values, respectively.