MODELLING OF ROLL FORCE IN COLD ROLLING MILL

Abstract

In iron and steelmaking, as in other industrial activities, increasingly, stringent customer requirements for product quality and regularity lead to tighter control of manufacturing processes. Today Iron and steel makers need to be increasingly innovative and must develop new techniques, especially in the field of artificial intelligence. Indeed, artificial intellegnce (Al) has greatly helped the steel industry to face its evolutionary challenges, typically with expert systems and fuzzy logic. Nowadays, artificial neural networks (ANNs) play an increasingly*important role in this field. Sollac's (Iron and steel industry of France) 20 years of modeling experience with ANN put it in an ideal position to deeply evaluate the contribution of ANN techniques to this process modeling. ANNs have a potential to improve the accuracy of the computation by substituting or correcting the mathematical model. Accurate prediction of roll force is of significant importance in the presetting stage in a cold rolling mill (CRM). Currently, mathematical models are employed for this purpose in CRMs. The possibility of using ANN model for prediction of more accurate roll force as compared to that predicted by mathematical model is investigated; For this problem, the data from TATA STEEL's cold rolling mill is used. MATLAB's Neural network Toolbox has been used to design and develop the required ANN model. To overcome the drawbacks of the mathematical model, an alternative model is required to predict more accurate roll force in cold rolling mill. It, is proposed to use ANN model for this purpose because the neural networks have more general functional forms than the well developed statistical methods can effectively deal with, it is a data driven model and it can take into account the factors whose exact physical relations are not known.