Disturbance pattern classification and neuro-adaptive control

Author

Cooper, D.J. ; Megan, L. ; Hinde, R.F., Jr.

Author_Institution

Dept. of Chem. Eng., Connecticut Univ., Storrs, CT, USA

Volume

12

Issue

2

fYear

1992

fDate

4/1/1992 12:00:00 AM

Firstpage

42

Lastpage

48

Abstract

An adaptation strategy is given that is based on an analysis of patterns exhibited in the recent history of the controller error and manipulated process input variable. The focus of the strategy is on adaptation requirements due to load disturbances. The manipulated input pattern associated with a disturbance is analyzed to determine whether the disturbance has the potential for changing process character. If so, the corresponding controller error pattern is analyzed to determine the appropriate adaptation to the controller´s internal model. A vector quantizing neural network is studied as a pattern analysis tool for implementing the method. The strategy is limited to a single parameter adaptation where the gain of the controller´s internal model is the adjustable parameter. Details and a demonstration of the method are presented using a simulated process constructed to challenge the strategy. A model-based PI algorithm is employed.<>

Keywords

adaptive control; neural nets; pattern recognition; controller error; controller error pattern; load disturbance pattern classification; manipulated process input variable; model-based PI algorithm; neuro-adaptive control; pattern analysis tool; vector quantizing neural network; Adaptation model; Adaptive control; Artificial neural networks; Error analysis; Error correction; History; Pattern analysis; Pattern classification; Predictive models; Programmable control;

fLanguage

English

Journal_Title

Control Systems, IEEE

Publisher

ieee

ISSN

1066-033X

Type

jour

DOI

10.1109/37.126852

Filename

126852