Machine Learning-Assisted LIBS Identification of Epoxy Resins in CFRP for Recycling Processes.

Area of Science:

• Materials Science
• Analytical Chemistry
• Computer Science

Background:

• Efficient sorting of resin-based CFRP composites is crucial for sustainable recycling processes.
• Current methods often lack the precision needed for complex composite streams.
• Optimizing recycling streams requires accurate material identification and separation.

Purpose of the Study:

• To develop and validate an integrated methodology for precise material discrimination in CFRP composites.
• To leverage Laser-Induced Breakdown Spectroscopy (LIBS) and Machine Learning (ML) for automated sorting.
• To enhance the efficiency of composite recycling through advanced classification techniques.

Main Methods:

• Utilized LIBS to determine the chemical composition of composite materials, focusing on Epoxy resin (Bisphenol-A).
• Processed LIBS spectrograms, standardized feature dimensionality, and removed noise from datasets.
• Applied statistical analysis for feature selection, followed by Linear Discriminant Analysis (LDA) for dimensionality reduction.
• Employed a Support Vector Classification (SVC) model with 5-fold cross-validation for accurate classification.

Main Results:

• Achieved an average nested accuracy score of 0.8317 ± 0.0212 using the integrated LIBS-ML approach.
• Successfully grouped chemically similar materials based on their spectral fingerprints.
• Demonstrated the effectiveness of feature selection and LDA in preparing data for ML classification.

Conclusions:

• The integrated LIBS-ML methodology offers a robust solution for accurate material discrimination in CFRP composites.
• This approach shows significant potential for advancing automated sorting technologies in the composite recycling industry.
• The study highlights the synergy between spectroscopy and machine learning for environmental applications.