Predicting Strip Tearing in Cold Rolling Tandem Mill using Neural Network

Strip tearing during cold rolling process has always been considered among the main concerns for steel companies, while several works have been done so far regarding the examination of this issue. In this paper, experimental data from cold rolling tandem mill is used for detecting strip tearing. Sensors are placed across the cold rolling tandem mill, to collect information on parameters (such as angular velocity of the rolls, voltage and the electrical current of electrical motors driving rolls, roll gap, and strip tension force between rolls) directly from the cold rolling tandem mill. The information includes two modes: perfect rolling and ruptured rolling. A neural network is designed by means of MATLAB software and, then, trained using the information from the related data files. Finally, the neural network is examined by new data. It is concluded that the neural network has good accuracy in distinguishing between perfect and defected rolling.