Advanced Characterization of Industrial Smoke: Particle Composition and Size Analysis with Single Particle Aerosol...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Data Science

Background:

• Industrialization increases exposure to harmful smoke particles with complex compositions and sizes.
• Mass spectrometry is key for aerosol analysis but struggles with complex data for particle size-composition relationships.
• Existing methods lack precise prediction of particle size from mass spectrometry data.

Purpose of the Study:

• To develop a novel method for precise prediction of industrial smoke particle size using mass spectrometry data.
• To investigate the relationship between lead isotope abundance, particle size, and soldering temperature.
• To enhance machine learning model generalization for analyzing complex aerosol data.

Main Methods:

• Combined single particle aerosol mass spectrometry (SPAMS) with optimized machine learning algorithms.
• Utilized kernel principal component analysis (KPCA) for nonlinear dimensionality reduction of mass spectrometry data.
• Developed a systematic stratified random sampling algorithm (SSRSA) to handle imbalanced sample distributions.

Main Results:

• Achieved a prediction accuracy (R^2) of 0.843 for particle size using SPAMS data and random forest (RF).
• Demonstrated a significant correlation between lead (Pb) isotope abundance and soldering temperature.
• Observed an increase in smaller particle sizes with rising soldering temperatures.

Conclusions:

• This study presents an innovative approach for accurate analysis of industrial smoke particle composition and size.
• The findings offer crucial insights for health risk assessment and developing effective pollution control strategies.
• The developed machine learning model improves the analysis of complex aerosol mass spectrometry data.