A Novel Classification Method for Predicting the Casting Behaviour in the Steelmaking Practice

M. Vannucci and V. Colla (Italy)


Neural Networks, Model–based Reasoning, Intelligent Manufacturing


The paper presents a novel classification method, the Model–based algorithm, which is based on the principle of classifying the input patterns on the basis of their resem blance to some predefined models. The proposed algorithm has been applied, together with other traditional algorithms, in order to process data coming from a real industrial context. The overall purpose of the work is the prediction of the occurrence of a criti cal situation during continuous casting in common steel making practice and this task is pursued by dividing the data in two classes corresponding to good and bad casting behaviour respectively. The classifiers’ performances are heavily affected by the fact that one of the classes to be recognised is poorly represented in the available database. The performance obtained by the different techniques on the experimental data are compared and discussed.

Important Links:

Go Back