Machine Learning Simulating and Predicting the Adsorption Performance of Activated Carbons for Removing Methylene Blue from Wastewater
Zhu-Ling Guo, Lei Huang, Jia Yan, Yonghui Liu, Yufang Guo, Zhenxing Wang, Qian Li, Meng Li, Xiujing Xing, Limin Li, Hongguo Zhang
Abstract
Methylene blue (MB) contamination in aqueous environments has emerged as a pressing environmental challenge, thereby necessitating the development of effective remediation strategies. Activated carbons (ACs), as highly promising adsorbent materials, have garnered considerable research attention worldwide for MB removal. This study proposes a machine learning (ML) approach to simulate and predict the performance of ACs in removing MB from aqueous solutions. We compiled a database from 282 literature sources, containing 301 data points encompassing variables across two dimensions: ACs characteristics and operational conditions. Following data preprocessing and logarithmic transformation of the prediction target, a random forest (RF) algorithm was fine-tuned to establish the MB adsorption capacity prediction model. Experimental results demonstrate that the optimized RF model exhibits high predictive accuracy, with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^2$</tex> of 0.9998 and RMSE of 2.446 for the validation set. Among the factors, the specific surface area of ACs, initial MB concentration in water, and pore volume of ACs were identified as the primary influencing factors. Furthermore, partial dependence analysis was employed to investigate the impact of individual variables on adsorption capacity, providing crucial insights for adsorbent design and process optimization. This research develops a comprehensive framework for applying machine learning (ML) to address environmental problems, providing a practical tool to facilitate the design and implementation of ACs-based water treatment systems.