Automated Optical Image Analysis of Coke Texture and Structure, and their Connection with Coke Porosity, Reactivity, Strength and Parent Coal Blend

Metallurgical coke is one of the most utilised forms of carbon used in ironmaking in blast furnace operation where it provides reducing gases, fuel for supporting the necessary temperatures and carbon for the hot metal to achieve specific properties. Strength is one of the most important characteristics of coke as in the blast furnace it also supports the burden to ensure good permeability for gasses and liquid phase drainage. To fulfil this task, the coke has to have a “large” mean size and narrow size distribution, and also must maintain a suitable size distribution throughout the furnace. It therefore requires high strength, good abrasion resistance, and resistance to chemical and thermal degradation.

CSIRO has developed the Mineral4/Recognition4 Optical Image Analysis (OIA) package, which performs identification and characterisation of different textures and structures in an image or a set of images (Donskoi, et al. 2017, 2013a, b, 2010, 2007). This imaging software allows structural characterisation which includes automated segmentation of porosity, Inert Maceral Derived Components (IMDC) and Reactive Maceral Derived Components (RMDC). Structural characterisation also performs segmentation of coke matrix into nodes and walls. Additionally, the software can automatically identify boundary/connections between IMDC and RMDC. All these approaches allow comprehensive characterisation of coke structure and porosity using traditional and newly developed structural descriptors within the Mineral4 software.

During this project textural coke characterisation was also developed. Microscope modifications and the development of an automated rotating polariser and the software controlling it, allowed for the collection of large images of coke under different polarisation angles and for the calculation of RMax, RMin and Bireflectance maps. Further software development during this project allowed identification of five major carbon types: isotropic and fused inerts, and three fused vitrinite types with high, medium and low bireflectance.

This project focussed on these newly developed techniques and the comprehensive characterisation of 28 cokes in order to identify statistically relevant relationships between coal blend properties, coke structure, coke texture and coke indices.  

The objectives of the project were to:

• Further develop structural coke characterisation which will include the identification of different types of IMDC and RMDC such as IMDC resulting from fusible and infusible inerts and RMDC resulting
• from high and low rank vitrinite.
• Develop novel automated image analysis methods for coke texture characterisation, including characterisation of the association of different carbon types and their structural characterisation (size distribution, wall thickness, porosity distribution corresponding to certain textural types etc).
• Understand the correlation between textural and structural characterisation of the same parts of coke. Apply a statistical approach to study the relationships between new and earlier developed coke structural characteristics as well as relationships between structural and textural attributes.
• To verify porosity figures measured optically. To understand the effect of porosity not measured optically (<1μm) and calculate presence of such porosity in different textures. To understand the dependence of coke CRI from porosity distribution, coke textural and structural composition.
• Extend the range of cokes studied in ACARP project C23051, especially to those produced from coal blends having rank from the larger rank range than previously used.
• To determine the most important parameters characterising coke structure and texture and achieve a more integrated understanding of relationships between the different characteristics of coke and the parent coal blend through examining relationships between coke micro/nano-porosity, different IMDC and RMDC structures, coke textural characteristics, reactivity and strength, and characteristics
• of initial coal blend.
• To apply models of coke strength indices developed during project C23051 to new cokes, to study their robustness and to develop new models (including new cokes) for a larger spectrum of parameters.

The project found that the analysis of data produced from a wide range of cokes should be done with special care. The data for low RoMax cokes can dramatically change the dependences for the overall group of cokes, eg make a certain dependence significant even though it was not observed for high RoMax cokes; or make a dependence insignificant or even change its sign when data for two types of cokes were clustered. This research showed that the structural and textural methodology for coke characterisation developed by CSIRO can comprehensively describe the coke texture and structure, and highlight the dependences between coke texture/structure, reactivity, strength and the parent coal blend. When it comes to an understanding of coke strength, the structural analysis was most informative, while textural characterisation was more helpful for coke reactivity. A number of important suggestions have been made to further improve the research in this area. In particular, it would be highly useful to extend the research to cover a larger coke database, aiming in three particular directions: increase the number of cokes made from low (RoMax below 1.00) and high (RoMax over 1.4) rank coals, and also apply this comprehensive characterisation to cokes produced in the Northern hemisphere. The latter would allow for better technical marketing of Australian coals, when compared to those from competing supply regions. It would also be important to adapt the developed techniques to the characterisation of cokes made from blends of different coals, therefore improving the reliability of coal blend optimisation and its ability to predict corresponding coke properties.