The mathematical modeling of the tandem mills involves non lineal expressions of difficult analytic and numeric solutions. If the model proposed by Alexander (1972), considered the most complete (with larger effort computational) it be considered to describe the process in each stand, this hinders the application of that model in control and supervision systems "on-line". For resolve such difficulty, in this work a representation of the sensibility of the process by means of Neural networks is presented. The new representation allows to obtain the value of the output thickness, for each stand, in the form predictive, eliminating the need of the sensor of thickness, usually ray-X, that introduce delays in the loops control. Other aspect of this work is the application of predictive control for tandem mills, based in neural networks. Simulation results are presented.
Automobile Industry; Tandem mills; Automation; Predictive Control; Neural networks
