The application of a semi-empirical modeling technique to real processes
Date
1998
Authors
Rietz, Christine Anne
Major Professor
Advisor
Rollins, Derrick K.
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Because of the increased demands on industry to maintain tighter quality specifications, to increase efficient energy utilization, and to conform to rigid safety and environmental standards, the need for improved process control and performance has become apparent. One solution to this demand has been the model predictive control (MPC) approach. MPC entails utilizing a process model to predict future behavior. Based on this prediction, a defined control objective is minimized in order to determine the optimal control moves of the manipulated variable. The accuracy of the process model is the foundation of any MPC strategy, and model identification is the most difficult and time-consuming element for practitioners to build. This work focuses on the application and evaluation of a predictive modeling technique to real open and closed-loop processes.
This technique is semi-empirical in nature and is referred to as the semi-empirical technique (SET). The approach of SET is to use an assumed model structure based upon knowledge of dynamic process behavior and available process information in an intelligent manner. This technique relies on a minimal amount of process data, where only two or three step changes are required to determine the model structure and its parameters. The greatest novelty of this method is the dynamic algorithm that modifies its prediction equation when input changes are detected or measurements are obtained for the process response. If a measurement of the output variable becomes available, the SET algorithm can advantageously use this information to correct discrepancies between the process and the prediction.
Thus, SET is a simplistic method of intelligently modeling a process, and SET exploits process information to enhance predictive performance. This work demonstrates that SET can be successfully applied to real processes with remarkable results. Since the prediction equation modifies when an input change is detected, it is shown that the accuracy of SET is dependent upon the detection of this change. This work also demonstrates the ability of SET to accurately predict the process response for a changing sampling frequency and a variety of sampling frequencies of the output variable.
Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Type
thesis