Explainable AI Model for Predicting Heavy Metal Adsorption Capacity of Engineered Biochar Materials

Abstract

The development of engineered biochar materials for heavy metal removal has gained significant attention due to their high adsorption potential, low cost, and environmental sustainability in wastewater treatment applications. Accurate prediction of adsorption capacity is essential for optimizing biochar design and improving contaminant removal efficiency under varying treatment conditions. This study presents an explainable artificial intelligence (XAI) model for predicting the heavy metal adsorption capacity of engineered biochar materials using data-driven analytical techniques. The proposed framework integrates machine learning algorithms with explainability methods to establish transparent relationships between biochar physicochemical properties and adsorption performance. Key input parameters, including surface area, pore volume, pH, functional group composition, pyrolysis temperature, feedstock type, and heavy metal concentration, were utilized to train and validate predictive models for adsorption capacity estimation. Advanced explainable AI techniques were incorporated to identify the relative importance of influential variables and provide interpretable insights into adsorption mechanisms and process behavior. Performance evaluation demonstrated that the proposed XAI model achieved high predictive accuracy and improved generalization capability compared to conventional empirical modeling approaches. The explainability framework also enabled enhanced understanding of feature interactions and decision-making processes within the machine learning model, thereby improving model transparency and scientific reliability. Comparative analysis revealed significant advantages in predictive consistency, process optimization capability, and material selection support for wastewater treatment applications.