Shape recognition performance analysis and improvement in Sendzimir rolling mills

Twenty-high Sendzimir rolling mills (ZRMs) typically use small diameter work rolls to provide massive rolling force. Because of the small diameter of the work rolls, a rolled steel strip has a complex shape mixed with quarter, edge, and center waves. When the strip shape is controlled automatically, actuator saturation occurs in the shape actuator such as AS-U roll. These problems affect productivity and the quality of products made from the rolled material. We analyzed problems with the automatic shape control of ZRMs. The shape recognition performance was analyzed by comparing the measured and recognized shapes by multi-layer perceptron (MLP) method. In addition, neural networks were developed using the radial basis function (RBF) method, and are proposed to improve the shape recognition performance of the automatic shape control system in a ZRM. Through simulation results, we found that shape recognition performance can be improved by the proposed method based on RBF neural networks.