Approbation of Intelligent Models for Predicting Technological Disruptions in the Aluminum Production

This paper presents the outcomes of approbation of the developed intelligent models and techniques for early prediction of one of the most serious technological disruptions - anode spikes - based on daily monitored parameter values using a combination of mathematical statistics, data mining, and machine learning methods. The first part of the paper provides a brief description of the problem statement and initial monitoring parameters, and then discusses the proposed predictive models based on ensemble learning methods and association rules. The second part of the paper presents the results of the implementation and approbation of the predictive models on real data from pilot sites of the aluminum production complex. The approbation results showed that the accuracy of the developed predictive models is more than 70 % and makes it possible to detect disruptions at the early stages of their development, when deviations in the technological process are detected only by indirect features - a certain combination of values of controlled parameters.