Efficacy of ultrasonic treatment for diazinon removal in aqueous solution: optimization by RSM and modeling by artificial neural networks (ANN) and support vector machine (SVM)

Abstract Diazinon (DZ) is a widely utilized agricultural pesticide posing significant risks to human and environmental health through water contamination, highlighting the need for effective pollution control strategies. This study aims to examine the efficacy of ultrasonic treatment for DZ removal as an eco-friendly method. To this end, optimal conditions were determined through statistical response surface methodology (RSM) based on central composite design (CCD) by appraising the impact of influential parameters such as initial DZ concentration, ultrasonic power, treatment time, solution volume, pH value as well as their interactions. The results revealed high DZ concentrations and larger solution volumes lowered degradation efficiency, while higher applied power improved it. Under optimal conditions – an initial DZ concentration of 4.41 mg/l, ultrasonic power of 302.88 W, pH 6.18, a reaction time of 56.1 min, and a solution volume of 105.6 ml – 49.6 % degradation was achieved. Further, artificial neural networks (ANN) and support vector machine (SVM) were applied for modeling DZ removal. The optimum ANN model was attained with 7 neurons in the hidden layer. The Gaussian RBF kernel with C = 51.69 and ε = 0.09 was identified as the optimal kernel for the support vector machine. The SVM model indicated superior accuracy ( R 2 = 0.994) compared with the ANN and RSM model ( R 2 = 0.98), confirming its stronger predictive capability for system behavior. The results confirm that ultrasonic treatment is an effective advanced oxidation processes (AOP) technique for removal of DZ from aquatic environments.