A New Modeling Method of the Rolling Load by the Function Synthesis Algorithm Using Genetic Programming

Abstract

This paper describes a rolling load modeling method that uses GP (Genetic Programming). It is important to predict the rolling load accurately for manufacturing high quality products in steel industry. Usually, the rolling load is predicted by using a statistical method based on a mathematical model. Even if the adaptive learning is applied to the conventional model, the prediction accuracy can not be improved for high quality manufacturing. In this paper, a new function structure of rolling load model is proposed and function components are determined by GP. This approach makes it possible, not only to achieve the high accuracy prediction, but also to reduce the calculation time for the real time pass scheduling and to apply the model to the rare rolling case with poor data base. It is observed that the new model reduces the standard deviation of the error by 17%, compared with the conventional method.