Comparison of response surface method and artificial neural networks in predicting formaldehyde and methanol removal using moving bed sequential batch reactor (MBSBR) and Fixed bed sequential batch reactor (FBSBR): Process optimization and kinetic study

Abstract

Formaldehyde (FA) is a carcinogenic pollutant in industrial wastewater that requires removal prior to environmental discharge, often alongside biodegradable methanol (MeOH). The present study investigates the removal efficiency of FA and MeOH using innovative sequencing batch reactors (SBR), specifically the moving bed (MBSBR) and the fixed bed (FBSBR) systems, which acclimated petrochemical sludge. Analytical methods included colorimetric measurements and gas chromatography, while Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) were used for experimental design and modeling. The FBSBR achieved superior removal efficiencies of 99 % for FA, 99.5 % for MeOH, and 98.7 % for COD within nine days, compared to 15 days for MBSBR. The research showed that lower pollutant concentrations improved removal efficiencies, with ANOVA confirming the reliability of RSM model. The high F values (ranging from 68.95 to 229.93) and the very low p-value (<0.0001) of the quadratic equations showed that the proposed RSM model was highly reliable for FA, MeOH, and COD removal. The modified Stover-Kincannon model showed that the maximum specific growth rate (U_max) and half-saturation constant (K B) for FA biodegradation were 70.9 g/L·d and 71 g/L·d in the MBSBR, and 76.9 g/L·d and 76.8 g/L·d in the FBSBR, respectively. Given the high efficiency of these bioreactors, it is recommended to use them to remove FA and other xenobiotic pollutants. The ANN model outperformed RSM in predictive accuracy, suggesting its use in real-time monitoring to enhance wastewater treatment efficiency.