An unsupervised clustering approach for yield stress prediction during flat rolling

In flat metal production complex automation technologies are often exploited in order to improve quality according to growing market demands and to optimize the costs associated with possible wastes. In this paper an unsupervised clustering technique is proposed in order to alleviate plant conduction problems normally rising when new materials are treated in a metal flat rolling mill. The case of steel cold flat rolling is deeply treated as an example where these data-mining technologies can be successfully applied so as to classify the produced materials without the use of special sensors.