Coal Classification Method Based on Improved Local Receptive Field-Based Extreme Learning Machine Algorithm and...

Area of Science:

• Geosciences
• Spectroscopy
• Artificial Intelligence

Background:

• Traditional coal classification relies on experienced technicians, leading to inefficiencies and automation challenges.
• Accurate coal identification impacts production efficiency, environmental pollution, and economic outcomes.
• Spectral data from coal often exhibits high dimensionality, strong correlation, and redundancy.

Purpose of the Study:

• To develop a fast and high-precision method for coal identification and analysis.
• To apply visible-infrared spectroscopy and machine learning for coal mine identification.
• To provide guidance for coal mining and production through advanced analytical techniques.

Main Methods:

• Utilizing visible-infrared spectroscopy for spectral data acquisition.
• Employing the local receptive field (LRF) for advanced feature extraction from spectral data.
• Combining LRF with an extreme learning machine (ELM) optimized by an improved coyote optimization algorithm (I-COA).

Main Results:

• The developed coal classification model effectively identifies coal types using spectral data.
• The local receptive field (LRF) demonstrates superior spectral characteristic extraction compared to Convolutional Neural Networks (CNN) and Principal Component Analysis (PCA).
• The improved extreme learning machine with local receptive field (ELM-LRF) achieved high precision in coal classification.

Conclusions:

• Visible-infrared spectroscopy combined with machine learning offers a powerful approach for coal identification.
• The proposed ELM-LRF method, optimized with I-COA, provides an effective and efficient solution for coal classification.
• This technique enhances coal mining and production by enabling accurate and automated coal analysis.