Machine learning-aided model for predicting oily sludge pyrolysis under various feedstock and operating conditions

Area of Science:

• Chemical Engineering
• Environmental Science
• Data Science

Background:

• Oily sludge pyrolysis offers resource recovery and safe residue disposal.
• Experimental optimization of pyrolysis is costly and time-consuming.
• Machine learning presents a viable alternative for process optimization.

Purpose of the Study:

• To develop and validate a machine learning model for predicting and optimizing oily sludge pyrolysis.
• To identify critical factors influencing pyrolysis performance.
• To reduce the time and cost associated with experimental optimization.

Main Methods:

• Six machine learning models were evaluated, with eXtreme Gradient Boosting (XGB) selected for its superior predictive accuracy.
• A multi-task XGB model was constructed using oily sludge ultimate/proximate composition and pyrolysis operating conditions as inputs.
• Model performance was verified, achieving an average R-squared value of 0.90.

Main Results:

• The XGB model accurately predicted oily sludge pyrolysis performance.
• Sludge ash and hydrogen content, along with pyrolysis temperature, were identified as the most influential factors.
• Ultimate composition (42.5%), proximate properties (35.8%), and operating conditions (21.7%) significantly impacted pyrolysis outcomes.

Conclusions:

• Machine learning, particularly XGB, provides an effective tool for understanding and optimizing oily sludge pyrolysis.
• This approach facilitates efficient resource recovery and waste management.
• The developed model offers valuable insights for industrial applications, reducing experimental burdens.