Comparative and exploratory study of ATR and diffuse reflectance mid-infrared spectroscopy for coal property analysis

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Geoscience

Background:

• Accurate and rapid coal quality assessment is crucial for the energy industry.
• Mid-infrared (MIR) spectroscopy offers potential for fast, non-destructive analysis.
• Different sampling geometries, like attenuated total reflectance (ATR) and diffuse reflectance FTIR (DRF), may influence analytical performance.

Purpose of the Study:

• To systematically compare ATR and DRF sampling geometries for mid-infrared Fourier transform infrared (FTIR) spectroscopy in coal analysis.
• To evaluate the benefits of multimodal data fusion at different levels (low, mid, high) for improving prediction accuracy.
• To establish a unified framework for benchmarking sampling geometries and fusion strategies in coal quality assessment.

Main Methods:

• Analysis of 200 coal samples using ATR and DRF FTIR spectroscopy.
• Preprocessing included extended multiplicative scatter correction (EMSC), wavelength selection, and partial least squares (PLS) regression.
• Systematic benchmarking of sampling geometries and three levels of data fusion (low, mid, high).

Main Results:

• Diffuse reflectance FTIR (DRF) provided superior ash prediction, while attenuated total reflectance (ATR) excelled in volatile matter and moisture prediction.
• Calorific value and fixed carbon predictions were comparable between ATR and DRF.
• Sulfur prediction remained challenging for both methods.
• High-level data fusion consistently improved prediction accuracy for five out of six properties, whereas low- and mid-level fusion showed no significant benefit.

Conclusions:

• The choice between ATR and DRF depends on the specific coal property being analyzed.
• High-level data fusion of ATR and DRF spectra can enhance the predictive power of FTIR spectroscopy for coal quality assessment.
• This study provides practical guidance for selecting optimal FTIR modalities and fusion strategies for routine coal analysis.