On‐Site Application of Artificial Intelligence to Sludge Dewatering for the Optimization of Flocculant Dosage

Abstract

Sludge dewatering is an essential process in sludge disposal to reduce its volume and improve the efficiency of final disposal. Artificial intelligence (AI) models for the dewatering of night‐soil sludge and digested sludge were developed in this study using laboratory experimental data with belt filter presses, and all models showed excellent prediction accuracy on both test dataset and new dataset with a correlation coefficient more than 0.9. Model interpretation by Shapley additive explanations indicated that the dosage method of poly‐ferric sulfate was the most important factor in all models, and fiber‐like content and dosage of poly‐ferric sulfate ranked second for night‐soil sludge and digested sludge, respectively. On‐site case study with screw presses demonstrated that the optimized polymer dosages determined by AI models were in good accordance with the observed ones, which highlight the potential applicability of developed AI models to different types of sludge dewatering machines for the optimization of flocculant dosage. It was supposed that the optimized polymer dosage could be primarily determined by sludge characteristics rather than the type of dewatering machines used. However, further investigation is needed to assess the applicability of these models to other types of sludge dewatering machines, such as centrifuges.