Methods of coke quality prediction: A review. (1st May 2018)

• Record Type: Journal Article
• Authors: North, Lauren; Blackmore, Karen; Nesbitt, Keith; Mahoney, Merrick R.
• Is Part Of: Fuel. Volume 219(2018)
• Page Start: 426
• Page End: 445
• Publication Date: 2018-05-01
• Subjects: Coking coal -- Cokemaking -- Coke quality -- Prediction -- Regression -- Neural network; Fuel -- Periodicals; Coal -- Periodicals; Coal; Fuel; Periodicals; 662.6
• DOI: 10.1016/j.fuel.2018.01.090
• Languages: English
• ISSNs: 0016-2361

Abstract

Highlights: Comprehensive review of methods of coke quality prediction completed. The most common methods of prediction were found to be regression based approaches. Data used to derive models limits their ability to generalize to other regions. Findings identify limitations in appropriateness and repeatability of methods. Abstract: The prediction of coke quality from global coal basins is critical to coke producers and steel makers for both the selection and effective utilisation of coals. This review analysed the methods described within published models for the prediction of coke quality. Of particular focus were methods that sought to predict coke strength after reaction (CSR) and the related coke reactivity index (CRI). Using the cross industry standard process for data mining (CRISP-DM) as an analysis framework, the models were compared in terms of their data treatment and use of analytical techniques. On reviewing these papers, our results indicate that it is difficult to apply models beyond the conditions under which they were derived, and that many models do not report enough detail to allow complete replication.