ASTM clustering for improving coal analysis by near-infrared spectroscopy

Area of Science:

• Analytical Chemistry
• Materials Science
• Geochemistry

Background:

• Near-infrared (NIR) spectroscopy is a valuable tool for analyzing coal properties.
• Accurate prediction of coal characteristics is crucial for quality control and utilization.
• Existing methods may lack precision when applied to diverse coal sample sets.

Purpose of the Study:

• To enhance the accuracy of predicting nine key coal properties using NIR spectroscopy.
• To investigate the effectiveness of pre-classification clustering on coal samples.
• To evaluate multivariate techniques for coal sample discrimination and classification.

Main Methods:

• Application of multivariate analysis to NIR spectra of coal samples.
• Clustering of coal samples into homogeneous groups based on ASTM standards.
• Utilizing Soft Independent Modelling of Class Analogy (SIMCA) and Linear Discriminant Analysis (LDA) for classification.
• Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS) in the NIR range.

Main Results:

• Clustering coal samples into homogeneous groups improved prediction error determination compared to analyzing the entire set.
• Calibrations for some clusters approached ASTM/ISO quality standards.
• SIMCA models showed overlapping characteristics, hindering unique classification.
• LDA improved sample classification but was not universally satisfactory across all groups.

Conclusions:

• Pre-classification clustering of coal samples enhances NIR spectral analysis accuracy for property prediction.
• LDA shows potential for coal classification, though further refinement is needed.
• Accurate prediction for new samples requires correct group assignment, highlighting the importance of classification methods.