Knowledge extraction of sonophotocatalytic treatment for acid blue 113 dye removal by artificial neural networks

Abstract

Removing decolorizing acid blue 113 (AB113) dye from textile wastewater is challenging due to its high stability and resistance to removal. In this study, we used an artificial neural network (ANN) model to estimate the effect of five different variables on AB113 dye removal in the sonophotocatalytic process. The five variables considered were reaction time (5-25 min), pH (3-11), ZnO dosage (0.2-1.0 g/L), ultrasonic power (100-300 W/L), and persulphate dosage (0.2-3 mmol/L). The most effective model had a 5-7-1 architecture, with an average deviation of 0.44 and R² of 0.99. A sensitivity analysis was used to analyze the impact of different process variables on removal efficiency and to identify the most effective variable settings for maximum dye removal. Then, an imaginary sonophotocatalytic system was created to measure the quantitative impact of other process parameters on AB113 dye removal. The optimum process parameters for maximum AB 113 removal were identified as 6.2 pH, 25 min reaction time, 300 W/L ultrasonic power, 1.0 g/L ZnO dosage, and 2.54 mmol/L persulfate dosage. The model created was able to identify trends in dye removal and can contribute to future experiments.

Keywords: Artificial neural networks; Index of relative importance; Prediction; Sensitivity analysis; Sonophotocatalytic; Textile wastewater.