Rapid characterization of MSW and RDF feedstocks for waste-to-energy process using LIBS and ML techniques

Area of Science:

• Waste Management & Valorization
• Analytical Chemistry
• Computational Science

Background:

• Municipal solid waste (MSW) heterogeneity complicates biofuel and bioproduct generation.
• Accurate characterization of MSW-derived refuse-derived fuels (RDF) is crucial for efficient processing.
• Traditional waste analysis methods are labor-intensive and time-consuming.

Purpose of the Study:

• To develop a rapid and accurate characterization technique for RDF.
• To enhance the efficiency of waste analysis for biofuel production.
• To integrate Laser-Induced Breakdown Spectroscopy (LIBS) with machine learning (ML) for predicting RDF properties.

Main Methods:

• Utilized LIBS to obtain spectral data from RDF samples.
• Applied data pre-processing techniques to LIBS spectra.
• Developed and trained ML models using domain and theory-based spectral features.
• Validated model performance in predicting key process parameters.

Main Results:

• Achieved high accuracy in predicting High Heating Value (HHV), carbon content, and volatile matter.
• Demonstrated an average relative root-mean-square error (RRMSE) of 2.13%.
• Attained a coefficient of determination (R²) of 0.98 or higher for all predicted parameters on testing data.

Conclusions:

• The combined LIBS-ML approach offers a fast and accurate alternative to traditional RDF characterization.
• This method significantly improves waste sorting and processing efficiency.
• The approach supports enhanced environmental compliance in waste management.