Polarized Light Microscopy and Image Processing as a Near Real-time Monitoring Tool for Coal Mine Dust Monitoring

Abstract

Airborne dust in coal mining environments presents a persistent occupational health hazard, with complex and heterogeneous particle compositions that complicate exposure assessment. Despite regulatory advancements, the reemergence of severe lung diseases has underscored the limitations of conventional monitoring approaches. This dissertation addresses the critical need for real-time, source-informative dust characterization by developing and evaluating a novel methodology that integrates portable optical microscopy with advanced image processing techniques.

The first study introduces a monitoring concept aimed at distinguishing among coal, carbonate, and silicate particles—the three primary sources of airborne dust in underground coal mines. A particle tracking algorithm was developed to enable direct analysis of composite samples by capturing sequential images of the same substrate area following staged depositions of known materials. This approach eliminated the need for post hoc SEM-EDX analysis by assigning particle identities based on deposition order. Classification models were constructed by establishing grayscale intensity thresholds under plane- and cross-polarized lighting conditions. The results demonstrated the method's potential to support dust source apportionment, thereby informing exposure mitigation strategies and operational decision-making. Particle size and loading density were identified as key factors influencing classification performance.

The second study extends the methodology to further subclassify the silicate fraction into silica and other silicates. This work employed machine learning—specifically, random forest classifiers—trained on grayscale intensity features extracted from polarized light images. The objective was to evaluate the feasibility of expanding the classification scheme to four particle classes: coal, carbonates, other silicates, and silica. While the approach showed promise, challenges related to particle agglomeration, optical similarity between silica and aluminosilicates, and particle size were observed. Ultimately, the study concluded that reliable silica classification was not feasible using optical methods alone.

The third study applies the developed methodology to dust samples collected from multiple locations within an active underground coal mine. A simplified classification scheme, based on the thresholding approach from the first study, was adopted to facilitate practical implementation. Comparative analysis with scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDX) was conducted to evaluate performance under field conditions. The study highlighted practical limitations, including particle agglomeration, mixed optical signals, and particle intensity variability, as well as the need for site-specific calibration.

Collectively, these studies advance the state of knowledge in dust exposure monitoring by demonstrating the potential of integrating portable microscopy and image analysis for near-real-time characterization of airborne particles in mining environments. The findings inform the development of practical monitoring tools and support ongoing efforts to reduce dust-related health risks in the mining industry.