Control Relevant Identification of Sheet and Film Processes

Sheet and film processes, which are characterized by a large number of inputs and outputs and strong interactions, illustrate the difficult and interesting control relevant identification problems existing in large scale processes. Existing techniques for model identification require substantially more input-output data than are typically available for these processes, and do not inform the control engineer on how to conduct experiments to provide the most useful process information for control purposes. Theoretical results dealing with the model identification requirements for controller performance and stability are developed. Next, the level of accuracy in the identified model is quantified using rigorous statistical analysis techniques. A direct relationship is derived between the model accuracy and controller design. In order to improve model accuracy, methods are developed to address the problem of input design, where input moves are selected to provide data rich in the process information relevant for closed loop control. This is followed by controller simulation results which demonstrate how the quality of the model affects the closed loop performance resulting from two controller design methods. A proposed combined identification and control procedure provides more accurate estimated models and higher quality cross-directional film profiles than industrially accepted identification and control procedures.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Chemical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Featherstone, Andrew Paul

Contributors dc:contributor

• Braatz, Richard D.

• Engineering, General

Language dc:language

eng

Identifier

(MiAaPQ)AAI9737101

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/82436