Applications of system identification to paper machine model development and controller design

The evolution of the computer-based data collection and analysis systems has lead to significant advancements in the application of modern system identification techniques to paper machine model development. Process data obtained from paper machines excited with pseudo-random binary sequences (PRBS) can be used to determine process dynamics, isolate multivariable process interactions and develop advanced computer models to evaluate existing and new control strategies. This paper examines some of the traditional parametric identification techniques such as least squares, maximum likelihood or instrumental variable methods applied to data collected from a paper machine. The resulting paper machine model includes the process dynamics, the shape of the responses in a cross machine direction sense and defines the interactions between the machine direction and cross machine direction control elements. Finally, the paper illustrates the application of the models to minimum variance methods of coordinating multiple actuators to reduce overall finished paper variability