Coal Classification Based on Reflection Spectroscopy and the IAT-TELM Algorithm.

Area of Science:

• Materials Science
• Computational Chemistry
• Machine Learning Applications

Background:

• Coal is a primary global energy source, with its quality significantly impacting industrial processes.
• Traditional coal classification methods (manual examination, chemical assays) are inefficient and lack consistent accuracy.
• Accurate coal type identification is crucial for optimizing industrial applications and resource management.

Purpose of the Study:

• To develop a rapid, accurate, and cost-effective algorithm for coal classification using spectral data.
• To address the limitations of traditional coal classification methods through advanced machine learning.
• To enhance coal classification performance by improving existing machine learning models.

Main Methods:

• Collected and preprocessed reflectance spectra of bituminous coal, anthracite, and lignite.
• Developed an affine transformation function extreme learning machine (AT-TELM) model.
• Optimized the AT-TELM model by refining weight matrices and biases, resulting in the improved AT-TELM (IAT-TELM).

Main Results:

• The IAT-TELM model achieved a high coal classification accuracy of 97.8%.
• The optimized model effectively handled skewed data distributions caused by random weight and bias assignments.
• Spectral analysis combined with machine learning proved superior to traditional classification methods.

Conclusions:

• The IAT-TELM algorithm offers a significant advancement in coal classification technology.
• This spectral analysis and machine learning approach provides a cost-effective, rapid, and precise solution for industrial needs.
• The developed method demonstrates high potential for practical application in the coal industry.