Screening Biochar and Optimizing Experimental Conditions for Arsenic Sorption by Machine Learning
作者
Han Su,Kai Zhang,Fei Wang,Zhi Zheng,Meng Niu,Xingyu Liu,Fei Lian,Bo Li
出处
期刊:ACS ES&T water [American Chemical Society] 日期:2025-09-10卷期号:5 (10): 5808-5819被引量:1
标识
DOI:10.1021/acsestwater.5c00339
摘要
To understand the structure–activity relationship between biochar properties and their sorption capabilities, as well as to assess the effect of sorption conditions, predictive models were developed based on a compiled dataset with 477 records. The ensemble model (Random Forest-Gradient Boosting Decision Tree) achieved better arsenic sorption prediction than individual models with biochar properties and experimental conditions as inputs. Model interpretation suggested that the initial concentration of arsenic had the greatest impact on its sorption onto biochar. The importance of the characteristics of biochar on arsenic sorption was ranked as O content > O/C ratio > Fe content > specific surface area > C content > H content > H/C ratio > (O+N)/C ratio. For sorption conditions, the contribution of each factor was in the order of adsorbent amount > pH > reaction temperature. Biochar with higher O content (>1.4%), higher O/C ratio, about 10–20% Fe content, about 1.8 g/L adsorbent dosage, higher surface area, lower H content, and lower H/C ratio exhibited stronger sorption capacity for arsenic. This information is valuable for guiding the development of high-performance biochars for arsenic removal and demonstrates the transformative potential of machine learning models in decoding complex nonlinearity relationships.