Artificial neural networks modeling of ozone bubble columns: Mass transfer coefficient, gas hold-up, and bubble size

This study aims at applying artificial neural network (ANN) modeling approach in designing ozone bubble columns. Three multi-layer perceptron (MLP) ANN models were developed to predict the overall mass transfer coefficient (kLa, s-1), the gas hold-up (εG, dimensionless), and the Sauter mean bubble diameter (dS, m) in different ozone bubble columns using simple inputs such as bubble column's geometry and operating conditions. The obtained results showed excellent prediction of kLa, εG, and dS values as the coefficient of multiple determination (R2) values for all ANN models exceeded 0.98. The ANN models were then used to determine the local mass transfer coefficient (kL, m.s-1). A very good agreement between the modeled and the measured kL values was observed (R2 = 0.85).