Strip Snap Analytics in Cold Rolling Process Using Machine Learning

Abstract

Strip snap, also known as strip breakage or belt tearing, is an undesirable quality incident which results in yield loss and reduced work speed in the cold rolling process of strip products. Therefore, it is necessary to reveal a functional relationship between certain selected variables and strip snap event for the aim of quality improvement. In this study, the probability of strip snap occurrence was quantified by a selected measured variable. Several machine learning algorithms were adopted to predict this target probability. To validate this approach, a case study was conducted based on real-world data collected from an electrical steel reversing mill. The excessively good performance indicates several variables which are strongly correlated with the target.