Cd adsorption prediction of Fe mono/composite modified biochar based on machine learning: Application for...

Area of Science:

• Environmental Science
• Materials Science
• Computational Chemistry

Background:

• Cadmium contamination in water poses significant environmental and health risks.
• Biochar, a sustainable material, shows potential for heavy metal remediation.
• Tailoring biochar properties for enhanced adsorption requires efficient predictive tools.

Purpose of the Study:

• To develop and compare machine learning models (ANN and RF) for predicting cadmium adsorption on Fe-modified biochar.
• To identify key structural and chemical factors influencing biochar adsorption capacity.
• To guide the targeted synthesis of high-performance Fe-modified biochar for cadmium removal.

Main Methods:

• Construction and evaluation of Artificial Neural Network (ANN) and Random Forest (RF) models.
• SHAP (SHapley Additive exPlanations) analysis to determine feature importance.
• Experimental synthesis and characterization of Fe-Ca modified biochar (FeCa-BC) from various feedstocks.
• Validation of ML model predictions using experimental data and a GUI toolbox.
• Analysis of adsorption mechanisms using FTIR, XRD, and XPS.

Main Results:

• The RF model demonstrated superior predictive performance (R² = 0.98) compared to ANN.
• Biochar structural characteristics were the most significant factor (55.44%) influencing adsorption.
• Optimal adsorption conditions were identified based on elemental composition (C, Fe, O, H, N) and pH.
• Experimental FeCa-BC validated the RF model's predictive accuracy (R² = 0.89).
• The developed GUI toolbox achieved prediction errors below 5%.

Conclusions:

• Machine learning, particularly RF, is a powerful tool for predicting and optimizing biochar properties for heavy metal adsorption.
• Targeted synthesis of Fe-modified biochar based on ML insights can significantly enhance cadmium removal efficiency.
• Surface complexation, precipitation, and ion exchange are key mechanisms for cadmium adsorption on FeCa-BC.
• This study provides a framework for data-driven design of functional biochar for environmental remediation.