Sonocatalytic degradation of caffeine using CeO2 nanorods: Modeling by artificial neural network

Abstract

This study has investigated the utilization of CeO2 nanorods (NRs) in the sonocatalytic degradation of caffeine. The degradation performance was determined by examining the influence of three parametric conditions, namely, the initial pH of the solution (3.5 – 7.5), initial concentration of caffeine (5 – 30 mg/L), and dosage of CeO2 NRs (0.5 – 2.0 g/L). All experiments were conducted in an ultrasonic bath (37 kHz, 150 W) that served as a sonocatalytic reactor. The mathematical modeling of the process with the catalyst was developed using Feedforward artificial neural networks (FFNN). The FFNN was employed to develop suitable modeling for determining the performance of the sonocatalytic degradation of caffeine (%) using CeO2 NRs. A three-layer FFNN with [4−10-1] topology was successfully developed to predict the sonocatalytic degradation of caffeine using CeO2 NRs. The FFNN was able to offer highly accurate predictions with the overall R2 and MSE validation values of 0.991 and 0.00225, respectively. The ANN model has also provided excellent predictive performance by achieving the highest R2 value. Thus, these results showed the promising finding of the sonocatalysis degradation of caffeine using CeO2 NRs via experiments and the ANN model.