Automated coal petrography using random forest

Abstract Coal is the backbone of the steel industry because of its manifold use in coke making, pulverised coal injection and electricity generation. Coal behaviour is a factor of its phase fractions (organic and inorganic) along with its maturity level. Coal petrography is an important tool for maceral determination and rank measurement based on its optical properties. Manual calculation of phase fractions is time-consuming and depends on operator's expertise and efficiency. To add value to plant operations, a faster, accurate and repetitive data is required. As a result, an attempt has been made to develop a machine learning based method for the automatic calculation of different phases present in coal. A random forest based model is developed to classify different phases of coal macerals (organic constituents) and minerals (inorganic constituents). The efficacy of the proposal is improved after introducing a hierarchical classification approach wherein random forest based classifier is used to segment macerals ignoring background. Features related to micro-structures of the coal microscopic images are extracted and utilized in random forest based classification. Methodology developed provides a better and quick alternative for manual petrographic analysis. A comparative analysis suggest that the final output shows better than 90% classification accuracy compared to ground truth. Its industrial application will save time, money and labour with the increase in efficiency level.

[1]  Edward Lester,et al.  A novel automated image analysis method for maceral analysis , 2002 .

[2]  Colin R. Ward,et al.  A scanning electron microscope method for automated, quantitative analysis of mineral matter in coal , 1996 .

[3]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[4]  Kevin Bouchard,et al.  Mineral grains recognition using computer vision and machine learning , 2019, Comput. Geosci..

[5]  Marta Skiba,et al.  The application of artificial intelligence for the identification of the maceral groups and mineral components of coal , 2017, Comput. Geosci..

[6]  D. Mukherjee,et al.  Coal petrography: a pattern recognition approach , 1994 .

[7]  Dipti Prasad Mukherjee,et al.  Improved Random Forest for Classification , 2018, IEEE Transactions on Image Processing.

[8]  Martin Allen,et al.  Repeatability of maceral analysis using image analysis systems , 1995 .

[9]  Yaokun Wu,et al.  Intelligent Image Segmentation for Organic-Rich Shales Using Random Forest, Wavelet Transform, and Hessian Matrix , 2020, IEEE Geoscience and Remote Sensing Letters.

[10]  R. Bustin.  Quantifying macerals: Some statistical and practical considerations , 1991 .

[11]  G. O'brien,et al.  Improving coke strength prediction using automated coal petrography , 2012 .

[12]  Nick J. Miles,et al.  Automated maceral analysis using fluorescence microscopy and image analysis , 1995 .

[13]  Dipti Prasad Mukherjee,et al.  Calculation of phase fraction in steel microstructure images using random forest classifier , 2018, IET Image Process..

[14]  Siddharth Misra,et al.  Machine learning for locating organic matter and pores in scanning electron microscopy images of organic-rich shales , 2019, Fuel.

[15]  Wolfgang Riepe,et al.  Characterization of coal and coal blends by automatic image analysis: Use of the Leitz texture analysis system , 1984 .

[16]  N. Otsu.  A threshold selection method from gray level histograms , 1979 .

[17]  J. Esterle,et al.  Coal characterisation by automated coal petrography , 2003 .

[18]  Furkan Elmaz,et al.  Classification of solid fuels with machine learning , 2020 .