Peak-Based Machine Learning for Plastic Type Classification in Time-of-Flight Secondary Ion Mass Spectrometry

Area of Science:

• Analytical Chemistry
• Materials Science
• Data Science

Background:

• Accurate plastic identification is crucial for recycling and material management.
• Traditional methods for plastic classification can be time-consuming and may lack specificity.
• Advanced analytical techniques combined with machine learning offer potential for improved classification.

Purpose of the Study:

• To develop and validate a machine learning model for classifying six distinct plastic types using ToF-SIMS data.
• To investigate the feature importance and chemical properties of different plastics based on spectral data.
• To enhance the explainability of plastic classification models.

Main Methods:

• Utilized Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) for material characterization.
• Applied data preprocessing techniques, including examination of local maxima.
• Implemented dimensionality reduction (Principal Component Analysis) for data visualization.
• Conducted ensemble analysis using Decision Tree, Random Forest, Gradient Boosting, and LIGHTGBM algorithms.

Main Results:

• Achieved successful classification of six different plastic types.
• Identified key spectral features indicative of each plastic type.
• Demonstrated the capability to infer chemical properties from ToF-SIMS data.
• The ensemble model provided high accuracy in plastic differentiation.

Conclusions:

• ToF-SIMS data, when analyzed with machine learning, provides an effective method for plastic classification.
• The developed approach enhances the understanding of chemical properties related to plastic types.
• This methodology offers a powerful tool for material identification, with implications for waste management and polymer science.