Deep learning-Guided optimization of cobalt catalysts for antibiotic degradation

Area of Science:

• Environmental Science
• Materials Science
• Chemical Engineering

Background:

• Antibiotic pollution in aquatic environments is a growing global concern.
• Advanced oxidation processes (AOPs) show promise for antibiotic removal but require optimization.
• Development of efficient inorganic catalysts is crucial for effective AOPs.

Purpose of the Study:

• To integrate machine learning and optimization algorithms for improved antibiotic removal in AOPs.
• To accelerate the development of novel inorganic catalysts for environmental remediation.
• To identify key factors influencing antibiotic degradation mechanisms.

Main Methods:

• Data preprocessing and exploratory data analysis from 207 research papers.
• Application of the TabNet deep learning model for classification and regression tasks.
• Utilization of the Sparrow Search Algorithm (SSA) for optimizing experimental conditions.
• Synthesis and testing of cobalt-based catalysts (Co-CuO, Co3O4, CoFe2O4).
• Interpretation of model predictions using SHapley Additive exPlanations (SHAP) analysis.

Main Results:

• TabNet model achieved 82.02% classification accuracy and an R² of 0.96 for regression.
• Sparrow Search Algorithm identified optimal experimental conditions for antibiotic degradation.
• Model predictions for catalyst degradation rates were within a 2% error margin.
• SHAP analysis effectively distinguished between free and non-free radical degradation mechanisms.

Conclusions:

• The integrated machine learning and optimization approach significantly enhances antibiotic removal efficiency in AOPs.
• The developed model accurately predicts catalyst performance and aids in designing new materials.
• This research offers a novel theoretical framework and practical solutions for environmental engineering challenges related to antibiotic pollution.