Artificial neural network modeling on the prediction of mass transfer coefficient for ozone absorption in RPB

It has been proved that Higee technology can intensify the processes involving the multiphase mass transfer, and be applied to the ozone-based advanced oxidation processes. Modeling and prediction of mass transfer coefficient are rare in this field. A modeling approach based on artificial neural network (ANN) was developed in this work to predict mass transfer coefficient of ozone absorption process in rotating packed bed (RPB). Serial experiments were conducted to obtain data for the establishment of ANN model, which was then employed to predict the overall mass transfer coefficient (K(L)a) using dimensionless quantities such as Reynolds number of gas and liquid, Froude number and Weber number, calculated in terms of the geometry of RPB and operating conditions. To optimize the model structure and performance, random grid search for hyperparameters was adopted in this work. The final model exhibits a prediction ability with R-2 of 0.9896 and 0.9877, RMSE of 0.01801 and 0.03085, and MAE of 0.01265 and 0.02219 on the training set and the test set, respectively. (C) 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.