Environmental Risk Assessment of Potential Toxic Elements in Co-Pyrolysis of Sludges and Plastics Based on Machine...

Area of Science:

• Environmental Science
• Chemical Engineering
• Data Science

Background:

• Co-pyrolysis of sludge and plastics is a key strategy for waste management and resource recovery.
• Assessing environmental risks associated with co-pyrolysis requires understanding the behavior of potential toxic elements (PTEs).

Purpose of the Study:

• To develop high-precision, interpretable machine learning models for predicting environmental risks in sludge and plastic co-pyrolysis.
• To quantify the contributions of operational parameters and PTE chemical speciation to environmental risks.

Main Methods:

• Construction of six machine learning models using experimental datasets (2015-2025).
• Inclusion of operational parameters and eight PTEs with four chemical speciation fractions (F1-F4).
• Application of feature importance and Shapley Additive Explanation (SHAP) analysis for risk factor identification and model interpretation.

Main Results:

• XGBoost, Random Forest, and CatBoost models demonstrated high performance with accuracies up to 0.94.
• Cd-F4 (residual fraction of Cadmium), As-F1 (acid-soluble/exchangeable fraction of Arsenic), and Cu-F4 (residual fraction of Copper) were identified as significant risk factors.
• SHAP analysis confirmed Cd-F4 as the primary risk discriminant, with F1 and F4 fractions being crucial for risk level differentiation.

Conclusions:

• An interpretable machine learning framework was proposed for sludge and plastic co-pyrolysis.
• The study provides a theoretical foundation for optimizing co-pyrolysis processes and assessing environmental risks.
• Understanding PTE speciation (F1 and F4) is critical for effective risk management in waste co-pyrolysis.