Machine learning modeling of microplastics removal by coagulation in water and wastewater treatment

Abstract

Microplastics (MPs) pose a global concern due to their persistence and potential toxicity. Coagulation is the common treatment technology for removing particles including MPs in water and wastewater. This research aims to address this challenge by developing machine learning models, including Artificial Neural Network (ANN), Least Square Support Vector Machine (LSSVM), Particle Swarm Optimization-Adaptive Neuro-Fuzzy Inference System (PSO-ANFIS), and Radial Basis Function (RBF) to predict the removal efficiency of MPs by coagulation under different conditions. Various input parameters, such as MP and coagulant concentration, solution pH and temperature were considered in these models. Through statistical analyses, the RBF model exhibited the highest accuracy with an R 2 value of 0.96 and R 2 value for ANN, PSO-ANFIS and RBF was 0.91, 0.83 and 0.79, respectively. Sensitivity analysis revealed that water temperature had the most significant negative effect, while coagulant aid showed the most positive effect on the coagulation performance for MP removal. The modeling approach and its findings provide valuable insights for improving the efficiency of MP removal in dynamic water and wastewater treatment processes. • MP removal by coagulation was analyzed and modeled using ML procedures. • ANN, PSO-ANFIS, LSSVM and RBF were developed to model MP removal performance. • RBF model showed the highest accuracy with R 2 value of 0.9841. • Water temperature showed significant negative effect on MP removal. • Type and dose of coagulant aid showed significant positive effect on MP removal.