Coal identification based on a deep network and reflectance spectroscopy.

Area of Science:

• Geoscience
• Spectroscopy
• Artificial Intelligence

Background:

• Rapid field identification of coal types is crucial for resource management and exploration.
• Traditional methods can be time-consuming and require specialized laboratory equipment.
• Developing efficient, on-site coal characterization techniques is a significant challenge.

Purpose of the Study:

• To develop and validate a rapid, field-deployable method for coal type identification.
• To integrate spectroscopy with advanced deep learning algorithms for enhanced accuracy.
• To provide a cost-effective and convenient solution for on-site coal analysis.

Main Methods:

• Collection and preprocessing of field spectral data from various coal samples.
• Development of a deep learning model combining convolutional neural networks (CNNs) for feature extraction and extreme learning machines (ELMs) for classification.
• Optimization of the CNN-ELM model parameters using the whale optimization algorithm (WOA).

Main Results:

• The proposed spectroscopy-based deep learning method achieved high accuracy in identifying coal types in field conditions.
• The whale optimization algorithm effectively improved the performance and robustness of the coal identification model.
• Experimental validation demonstrated the method's capability for rapid and precise coal classification.

Conclusions:

• The integrated spectroscopy and deep learning approach offers a low-cost, convenient, and effective solution for rapid field identification of coal types.
• This method significantly advances the capabilities for on-site geological and mining assessments.
• The optimized CNN-ELM model provides a reliable tool for real-time coal characterization.