Classification of e-waste using machine learning-assisted laser-induced breakdown spectroscopy.

Area of Science:

• Materials Science
• Environmental Science
• Computer Science

Background:

• Sustainable waste management is vital for economic growth and resource conservation.
• Electronic waste (e-waste) management requires efficient metal classification.
• Artificial intelligence (AI) offers potential for rapid, contactless classification in e-waste.

Purpose of the Study:

• To evaluate an AI-based laser-induced breakdown spectroscopy (LIBS) system for classifying aluminum alloys in e-waste.
• To compare the effectiveness of different machine learning (ML) models for this classification task.

Main Methods:

• Utilized laser-induced breakdown spectroscopy (LIBS) to obtain spectral data from five types of aluminum alloys.
• Employed machine learning (ML) classifiers, including Principal Component Analysis (PCA) and K-nearest neighbor (kNN) variants.
• Trained supervised ML models with 10-fold cross-validation on 80% of spectral data and tested on 20%.

Main Results:

• Principal Component Analysis (PCA) was ineffective in differentiating alloy spectra.
• Standard K-nearest neighbor (kNN) models achieved less than 30% accuracy.
• Ensembling kNN with the random subspace method significantly improved classification accuracy to 98%.

Conclusions:

• An AI-based LIBS system provides effective, non-contact classification of e-waste aluminum alloys.
• This technology can be integrated with robotic systems to minimize manual labor in recycling.
• The study demonstrates a viable approach for enhancing e-waste management and resource recovery.