A nonlinearized multivariate dominant factor-based partial least squares (PLS) model for coal analysis by using...

Area of Science:

• Analytical Chemistry
• Geochemistry
• Spectroscopy

Background:

• Accurate elemental concentration measurement in coal is crucial for resource evaluation and combustion control.
• Conventional methods often struggle with complex spectral interferences and nonlinearities inherent in coal analysis.
• Laser-induced breakdown spectroscopy (LIBS) offers rapid elemental analysis but requires sophisticated chemometric models for accurate quantification.

Purpose of the Study:

• To develop an advanced chemometric model for improved elemental concentration measurement in coal using LIBS.
• To enhance the accuracy and robustness of carbon concentration determination in bituminous coal.
• To address nonlinear self-absorption and inter-element interference in LIBS spectral data.

Main Methods:

• Application of a nonlinearized multivariate dominant factor-based partial least-squares (PLS) model.
• Construction of a comprehensive dominant factor using characteristic line intensities of multiple elements.
• Integration of nonlinear transformations of line intensities to model self-absorption and inter-element interference.
• Utilizing leave-one-out cross-validation and L-curve methods for model validation.

Main Results:

• The proposed model significantly improved prediction performance compared to conventional PLS.
• Reduced the root mean square error of prediction (RMSEP) for carbon concentration from 4.47% to 3.77%.
• Demonstrated superior robustness and performance across different calibration and prediction sample sets.

Conclusions:

• The nonlinearized dominant factor-based PLS model offers a more effective approach for coal elemental analysis using LIBS.
• The model accurately quantifies carbon concentration by addressing spectral nonlinearities and interferences.
• The developed method provides a robust and accurate tool for coal elemental analysis.