Machine learning models for predicting surfactant-enhanced oil removal from contaminated soil

Area of Science:

• Environmental Science
• Chemical Engineering
• Data Science

Background:

• Soil contamination by petroleum hydrocarbons poses significant environmental risks.
• Surfactant-enhanced remediation (SER) is effective but complex to model due to nonlinear interactions.

Purpose of the Study:

• To develop and validate machine learning models for predicting SER efficiency.
• To identify key factors influencing soil remediation outcomes.

Main Methods:

• Trained and validated six predictive models using a dataset of 2394 samples.
• Employed categorical boosting (CB), extreme gradient boosting, and decision tree models.
• Utilized cross-validation and Monte Carlo sensitivity analysis.

Main Results:

• The categorical boosting (CB) model achieved the highest performance (R² = 0.985, RMSE = 0.068).
• 96.4% of CB model predictions fell within the statistical applicability domain.
• Agitation speed, surfactant concentration, liquid-to-soil ratio, and washing time were identified as key factors.

Conclusions:

• AI-driven modeling, particularly CB, offers an efficient tool for optimizing SER designs.
• This approach can reduce operational risks and costs compared to conventional methods.
• Machine learning enhances the design and implementation of effective soil cleanup strategies.