Hydrodesulfurization of Dibenzothiophene: A Machine Learning Approach

Area of Science:

• Catalysis
• Chemical Engineering
• Materials Science

Background:

• Hydrodesulfurization (HDS) is crucial for removing sulfur from fuels.
• Dibenzothiophene (DBT) and its derivatives are difficult sulfur compounds to eliminate.
• Developing efficient HDS catalysts remains an industrial challenge.

Purpose of the Study:

• To investigate key factors influencing catalyst efficiency in DBT hydrodesulfurization.
• To apply machine learning (ML) algorithms for predicting HDS performance.
• To identify critical catalyst structural parameters affecting DBT conversion and selectivity.

Main Methods:

• Utilized machine learning regression techniques: Lasso, Ridge, and Random Forest.
• Estimated DBT conversion and selectivity based on catalyst properties.
• Analyzed regression coefficients to determine the importance of structural parameters.

Main Results:

• Random Forest and Lasso regression provided adequate predictions for DBT conversion.
• Regularized regression models showed similar, suitable outcomes for selectivity estimation.
• Catalyst pore size and slab length were identified as essential predictors for selectivity.

Conclusions:

• ML models effectively predict HDS catalyst performance for DBT removal.
• Catalyst structural properties significantly influence selectivity.
• Findings align with existing experimental data, validating the ML approach.