1-09　The application of Machine Learning in coke strength prediction

Abstract

The coke strength (DI) depends on coal blend ratio and operation conditons which are coal moisture content, coal particle size, coke oven temperature and more. Consequently, it is important for the target quality to predict coke strength at the time to decide coal blend. The conventional prediction method of DI has been obtained by an inductive model from the average coal properties. However, this prediction has often larger error because there is no additivity in some coal properties. Therefore, we try to predict DI with AI which makes a deductive model from coal blending ratio and time series parameters of an objective variable. As a result, AI prediction is more accurate than the conventional one for consideration of coal compatibility and adopting operation conditions as explanatory variables.