Chemically Informed Targeted Multiblock Data Fusion for Coal Analysis Using Laser-Induced Breakdown Spectroscopy and...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Materials Science

Background:

• Accurate coal quality characterization is crucial for optimizing coal-fired power plants.
• Combining Laser-induced Breakdown Spectroscopy (LIBS) and Near-infrared Reflectance Spectroscopy (NIRS) offers complementary elemental and molecular data.
• Effective data fusion of LIBS and NIRS for coal analysis remains a significant challenge.

Purpose of the Study:

• To develop a novel targeted multiblock analysis framework for integrating LIBS and NIRS data.
• To improve the accuracy and interpretability of coal quality analysis.
• To establish a method for classifying coal types based on spectral features.

Main Methods:

• A supervised feature extraction method was used to create feature blocks from LIBS and NIRS spectra.
• Targeted multiblock models were constructed by combining primary prediction models weighted by predictive accuracy.
• Coal type classification was performed using spectral features to accommodate different coal ranks.

Main Results:

• The targeted multiblock model significantly reduced average absolute errors for calorific value, volatile matter, ash, and moisture content compared to baseline methods.
• Distance correlations between coal properties aligned with the multiblock weight distribution, enhancing chemical interpretability.
• The framework demonstrated improved accuracy and provided a clear route for integrating multispectral and chemical information.

Conclusions:

• The proposed targeted multiblock strategy enhances the accuracy and chemical interpretability of LIBS-NIRS coal analysis.
• This approach effectively integrates complementary spectral data for comprehensive coal quality assessment.
• The framework offers a pathway to optimize combustion efficiency and reduce emissions in coal-fired power plants.