A new method for model predictive control design and its application to an industrial process /
The presence of natural constraints in many controlled chemical processes introduces nonlineaxities in control actions. The activation of those constraints decreases the degrees of freedom of the system and may lead to instabilities, especially when the number of input variables is less than or equ...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1995.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=742746001&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | The presence of natural constraints in many controlled chemical processes introduces nonlineaxities in control actions. The activation of those constraints decreases the degrees of freedom of the system and may lead to instabilities, especially when the number of input variables is less than or equal to the number of output variables However, in certain chemical processes more input than output variables may be available. In a control scheme this offers more degrees of freedom and less possibility for instabilities, even if some of the constraints are active. Moreover, use of more input than output variables can reduce the cost of controlling a process, by selecting the most economic combination of input moves. Thus, the issues of robust stability and performance of non-square constrained model predictive controllers naturally arises. In this work we obtain robust stability and performance conditions for multivariable non-square Dynamic Matrix Controllers with End-condition (EDMC). In the derivation of these conditions we consider the presence of input and input move constraints and soft output constraints. Modeling error is -also considered as an upper and lower bound on each one of the model coefficients. The robust stability conditions we derive are sufficient and impose a lower bound on the move suppression coefficients. It is shown that use of more input variables to control a certain number of output variables improves the closed-loop performance of the system. The method is easily applied, since for any non- square system we need to solve a simple off-line optimization problem. This method was implemented to ail industrial process with three input and two output variables. Several tests proved the efficiency of the method. Some of the results are presented and the real responses are compared to the simulated ones. The performance of this control scheme is also compared to the ones of the corresponding square systems. In some cases and for the same level of uncertainty, the square systems do not have feasible solutions. In all other cases the robust performance of the non-square system is superior to the performances of the square systems. |
|---|---|
| Item Description: | Vita. "Major Subject: Chemical Engineering". |
| Physical Description: | xii, 158 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references. |