8th International Congress on Science and Technology of Ironmaking - ICSTI 2018

Despite the extensive research in the prediction of coke properties from the parent coals, at present, no single model exists to adequately predict coke quality from all coal basins. Whilst heavily used as a feature in early coke quality prediction models, the binding behaviour of the maceral components, termed fusibility, has had limited success in being implemented into a global prediction model. Part of the limitation of existing models is the arbitrarily assumed fixed proportion of fusing macerals within each coal, despite evidence suggesting that the behaviour of macerals varies by coal basin. Current methods of determining coal fusibility rely on manual point count methods on both the coal and coke structures, which is a costly process with significant measurement variability. This study developed a method of predicting the fusibility by using a data mining approach, applied to commonly measured coal quality parameters. The proposed method has implications for determining and interpreting the contributing factors to coal fusibility. Further, as implemented within a coke quality prediction model, this method contributes to the knowledge of the relationship between fusibility and coke quality parameters.