Data mining – new perspectives on predicting coke quality in recovery stamp charged coke making process

Abstract

Data mining technique has been used to develop the coke quality prediction model for stamp charge coke making. The developed model is being used to optimize the coke strength after reaction based on its coal blend properties and operating parameters. The study shows the actual plant coke CSR and predicted coke CSR of same coal blend were much closer proximity. Actual plant data have been used for developing and validation of the model. The technique was successfully used to produce superior coke from the same coke oven batteries without damaging the oven health.