Machine learning-assisted simulated annealing for deciphering multi-factor coupling mechanisms in complex wastewater...

Area of Science:

• Environmental Chemistry
• Adsorption Science
• Wastewater Treatment

Background:

• Phenolic and oil pollutants (POs) pose challenges in complex wastewater systems.
• Understanding adsorption interactions is crucial for effective pollutant removal.

Purpose of the Study:

• Develop a multi-factor-coupling response model to elucidate adsorption interactions.
• Identify critical factors and synergistic mechanisms governing POs adsorption.
• Optimize adsorption performance for enhanced wastewater treatment.

Main Methods:

• Bayesian-optimized Gaussian Process Regression (GPR) for predictive modeling.
• Shapley additive explanations (SHAP) for feature importance analysis.
• Simulated annealing algorithm for optimization and experimental validation.

Main Results:

• GPR model achieved R² of 0.99, significantly reducing RMSE and MAE.
• Initial concentration (71.54%) and specific surface area (22.01%) were identified as primary controlling factors.
• Optimized adsorption capacity (Q) reached 635.80 mg/g, a 15.40% improvement in coal chemical wastewater.

Conclusions:

• The developed framework integrates predictive modeling, intelligent optimization, and experimental validation.
• Specific surface area and hierarchical pore structures enhance adsorption capacity.
• Optimized environmental parameters and concentration gradients improve mass transfer for effective wastewater treatment.