John C. Campbell
Ph.D., University of Wisconsin-Madison, 1997
The work presented in this thesis advocates a constrained infinite horizon controller. The infinite horizon controller uses a process model and an initial state estimate, thus the modelling and estimation issues are of critical importance. A model structure is proposed that captures the main characteristics of the process including the scanning sensor and large time-delay. Model parameters are identified from process data, and symmetry is enforced to improve the parameter estimates.
The thickness of the film is estimated from sparse measurements using a periodic Kalman filter. The covariances of the estimates are used to evaluate scanning patterns. Since constant disturbance models can be used by the controller to achieve offset free control, a general disturbance model and a measured input disturbance model are presented. Lifted systems are constructed for both disturbance models. The lifted systems are useful for discussing the issues of observability and detectability. Conditions of detectability are established for both of the disturbance models.
Once the state of the film can be adequately reconstructed from the sensor data, a control strategy can be implemented to improve film properties. One important control contribution can be made by having the controller handle hard constraints on the actuators without employing clipping of the controller output signal. Determining targets and dealing with with regulator infeasibility are discussed as well. Finally, simulations bring modelling, estimation, and regulation together in an effort to better understand the periodic filter and disturbance models that are key to improved gage control.
|[ Home | People | Projects | Recent Presentations | Publications | Tech Reports | Contact Us ]|
University of Wisconsin
Department of Chemical Engineering
Madison WI 53706