Optimization of industrial Fenton process of phenolic effluent using artificial neural network and design of experiments

Abstract

This work aimed to optimize the homogeneous Fenton process used to remove phenol in a chemical industry located in Mauá (São Paulo, Brazil). Phenolic effluent exhibits significant variability in its chemical composition, which poses a great challenge in finding the optimal reagents ratio for conducting Fenton treatment. Often, there is a need for repetitions to treat the same batch, leading to increased reagent consumption and treatment time. Thus, to optimize the industrial process, the present work applied modern optimization techniques on real data, such as design of experiments (DoE) and artificial neural network (ANN). ANN was used to perform simulations in the company’s process, thereby predicting the quantities of ferrous sulfate heptahydrate and hydrogen peroxide. The following optimal conditions have been determined by DoE (phenol concentration ≤ 4 ppm): pH from 2.7 to 4 and temperature from 28 to 40 °C. Before the implementation of the optimized parameters, the efficiency for a single treatment was 45%. After, the efficiency was 82% and the company saved, on average, 51.6% of reagent costs. The main contribution of the present work is the development and industrial testing of data science methods that aims to optimize the industrial Fenton process.