Machine-learning-aided predicting and engineering of hydrochar nitrogen-containing functional groups.

Area of Science:

• Materials Science
• Chemical Engineering
• Data Science

Background:

• Hydrochar, a carbonaceous material from hydrothermal treatment (HTT), possesses properties influenced by nitrogen-containing functional groups (NCFGs).
• Controlling NCFGs is crucial for tailoring hydrochar properties for specific applications.
• Predictive modeling offers a pathway to optimize NCFG content.

Purpose of the Study:

• To apply machine learning (ML) for predicting and regulating four key NCFGs in hydrochar: amine-N (N-A), pyrrolic-N (N-5), pyridinic-N (N-6), and quaternary-N (N-Q).
• To develop a robust ML model capable of guiding the targeted production of hydrochar with desired NCFG profiles.
• To create an accessible online system for hydrochar production optimization.

Main Methods:

• Development of four single-target random forest (RF) models to predict individual NCFGs.
• Construction of a multi-target RF model integrating predictions for all four NCFGs.
• External validation of the ML model using an independent set of 33 hydrochar samples.
• Incorporation of validation data to refine the multi-target model's performance.
• Development of an online prediction and optimization system.

Main Results:

• Single-target RF models achieved high predictive accuracy, with test R² values ranging from 0.904 to 0.943.
• The initial multi-target RF model demonstrated strong performance (average test R² = 0.882).
• External validation confirmed model robustness (R² = 0.877).
• Refinement with validation data further improved the multi-target model (R² = 0.894).

Conclusions:

• Machine learning provides a powerful tool for predicting and controlling NCFGs in hydrochar.
• The developed ML models and online system can guide the targeted synthesis of hydrochar.
• This approach facilitates the production of tailored hydrochar materials for advanced applications.