Coal Blend Analysis by Automated Petrography

Leading coal laboratories and ironmaking plants use light microscopy methods to quantitatively estimate the composition of blends of product coking coals. Although various automatic imaging systems have been used by plants for routine blend analysis, the overwhelming majority of blend analysis performed by petrographic laboratories is by manual point counting. The automatic systems produce results only on a bulk basis (vitrinite reflectance distributions and coal blend proportions) and the methods employed are not consistent across the industry. The manual methods can produce more detailed information but are not suitable for routine plant analyses. The main aim of this project was to develop a more advanced blend analysis technique based on the MACE®3001 automated coal petrography system. This system was developed with ACARP assistance under Projects C6055 and C8056. A capability to analyse coal blends is a prerequisite for the system to become a viable alternative to manual petrographic methods for characterising coking coals and power station feeds. Automation is necessary to remove operator subjectivity, to permit more complex and extensive measurements than is practical with manual methods, and to make analyses more accessible to the coal industry.

The project was approached as an image analysis and statistical analysis problem to be solved within the design framework of the existing automated coal petrography system. The aim was to produce a workable version suitable for users of existing R&D systems. The project was successful and a software solution was found to determine vitrinite reflectance distributions and coal blend proportions on a grain-by-grain basis. To achieve this, a new technique was developed to accurately identify and measure individual grains of vitrinite and telo-vitrinite (a sub-group). The identification of vitrinite populations was assisted by accurate reflectance calibrations (improved reflectance standards are under development in project C12059). The performance of the blend analyser was assessed using samples provided by an international petrography standards body. The system was tested on a variety of shipment blends. It was not possible to fully automate the blend analyser at this time. The software was designed to work with manual microscopes and several interactive steps are required during the acquisition of data and during processing. In addition to standard coal blend analysis, several other potential applications were demonstrated. For example, in expert hands it is viable as a forensic tool. The high sensitivity of the technique may allow small secondary populations of vitrinite introduced by different mining patterns or contamination to be detected in single product coals. It was demonstrated that vitrinite populations separated by as little as 0.1 % reflectance could be accurately distinguished by using telo-vitrinite indicators. The provision of blend data on a grain-by-grain basis is an important development. For the first time it is possible to relate blend data to individual component grain properties such as maceral abundance, texture and density. It is now possible to develop more complex models to describe coal behaviour during processing and utilization.