Evaluating the performance of cross-directional control systems

Methods have been developed for monitoring the performance of SISO control systems, which compare the variance of the output achieved from the plant to an estimate of the variance that could be achieved by minimum variance control. The methods require knowledge of only the inherent delay within the process. This paper extends these ideas to analyze the performance of cross-directional control systems for sheet processes such as paper machines, which are large multivariable systems. Because such systems are inherently ill-conditioned, the performance is compared against that of a generalized minimum variance controller designed to avoid controlling poorly controllable spatial and dynamic modes. The results of applying the algorithm to data from a paper machine are presented.