Machine learning-aided biochar design for the adsorptive removal of emerging inorganic pollutants in water

The escalating presence of emerging inorganic pollutants (EIPs) including vanadium (V), antimony (Sb), thallium (Tl), mercury (Hg), fluoride (F−), and rare earth elements (REEs) in aquatic environments poses a significant threat to water quality and human health. Therefore, remediation of EIPs conta...

Full description

Saved in:
Bibliographic Details
Published in:Separation and purification technology Vol. 362; p. 131421
Main Authors: Ullah, Habib, Khan, Sangar, Zhu, Xiaoying, Chen, Baoliang, Rao, Zepeng, Wu, Naicheng, Idris, Abubakr M
Format: Journal Article
Language:English
Published: Elsevier B.V 30.07.2025
Subjects:
Adsorption
Biochar
Inorganic water pollution
Machine learning (ML)
XGBoost algorithm
Machine learning (ML)
Inorganic water pollution
XGBoost algorithm
Adsorption
Biochar
ISSN:1383-5866
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The escalating presence of emerging inorganic pollutants (EIPs) including vanadium (V), antimony (Sb), thallium (Tl), mercury (Hg), fluoride (F−), and rare earth elements (REEs) in aquatic environments poses a significant threat to water quality and human health. Therefore, remediation of EIPs contaminated water is of pressing concern. Biochar adsorption offers a promising, environmentally benign, and cost-effective approach for EIP removal. However, inconsistent experimental methodologies and varying research objectives in previous studies hinder the selection of optimal biochar for specific EIP. Developing biochar materials with high adsorption capacity is crucial for effectively removing EIPs from water. However, the optimization of biochar designing using advanced artificial intelligence (AI) methodologies has not been thoroughly reviewed. This study employed a dataset of 528 data points from 61 biochar samples, collected from adsorption experiments conducted between 2014 and 2024, encompassing 24 variables related to various EIPs. To predict adsorption capacity and elucidate adsorption mechanisms, Random Forest (RF), Support Vector Regression (SVR), XGBoost, and CatBoost machine learning algorithms were applied. The XGBoost model outperformed the others, achieving a coefficient of determination (R2) of 0.96 and a lower root mean squared error (RMSE) of 0.4. Feature importance and SHAP value analysis identified reaction pH, initial concentration and pyrolysis temperature as key predictors of adsorption efficiency. Future predictions from the XGBoost model indicate that reaction pH, initial concentration pyrolysis temperature and biochar pH, are critical factors influencing EIP adsorption. This research offers novel insights into EIPs adsorption and establishes a framework for designing sustainable biochar-based adsorbents for wastewater treatment.
ISSN:1383-5866
DOI:10.1016/j.seppur.2025.131421