Skip to Main Content
A metal rolling process is examined, and shown to be an implicit, discrete, multistage, control process with admissible control set dependent only upon the system state. By means of numerical techniques, dynamic programming is applied to this process to generate a closed-loop roll setting policy which is optimal in the sense that it achieves a specified final state at minimum dollar cost. Digital simulation and Monte Carlo techniques are used to compare performance of the optimally controlled mill with that of the the same mill, operating in accord with current rolling practices. Both models are examined using random initial conditions and the process sensors are assumed to contribute noise to the measurement of the system state. An averaging technique is used to decrease the effects of this noise on system performance.