Machine learning modeling of thermally assisted biodrying process for municipal sludge

Area of Science:

• Environmental Science
• Chemical Engineering
• Data Science

Background:

• Municipal sludge (MS) utilization is crucial for resource recovery, with activated carbon production being a key application.
• Drying is an essential pretreatment step in converting MS into activated carbon.
• Thermally assisted biodrying offers a promising method for MS treatment.

Purpose of the Study:

• To develop accurate machine learning (ML) models for predicting moisture ratio (MR) and composting temperature (CT) during MS biodrying.
• To optimize ML model hyperparameters using Bayesian optimization.
• To evaluate feature importance for model interpretability.

Main Methods:

• Six ML models were employed to predict MR and CT.
• Bayesian optimization was used for hyperparameter tuning.
• SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDPs) were utilized for feature analysis.

Main Results:

• Gaussian Process Regression (GPR) emerged as the best-performing model for both MR (R²=0.9967) and CT (R²=0.9958) prediction.
• Optimized GPR models achieved high prediction accuracy with low RMSE values.
• A user-friendly graphical user interface (GUI) was developed for the GPR model.

Conclusions:

• The study successfully established accurate ML-based prediction models for MS biodrying parameters.
• GPR provides a robust tool for real-time monitoring and optimization of the drying process.
• The developed GUI facilitates practical application of the models for researchers and engineers, aiding in efficient MS management.