Parameter identification of Takagi-Sugeno Fuzzy model of surge tank system

Author

Aftab, Muhammad Saleheen ; Kadri, Muhammad Bilal

Author_Institution

Pakistan Navy Eng. Coll., Nat. Univ. of Sci. & Technol., Islamabad, Pakistan

fYear

2013

fDate

25-26 Sept. 2013

Firstpage

1

Lastpage

4

Abstract

This paper presents Takagi-Sugeno (TS) fuzzy modeling of a dynamic system. TS fuzzy model of a highly nonlinear system of Surge Tank has been proposed which has fixed structure that is linear in parameters. The rule base for the fuzzy model is obtained from the input-output data set, extracted from the measurements performed on the actual system. Number of rules has been determined utilizing the fuzzy clustering approach that incorporates fuzzy c-means (FCM) algorithm. After the formation of the rules, parameters of the Takagi-Sugeno Systems have been identified. Two different approaches are presented for the training of the fuzzy system: Batch Least Squares (BLS) algorithm, and the Gradient method. The trained fuzzy systems from the two approaches are tested with different forms of inputs and their performance is analyzed and validated by comparing with the real system.

Keywords

fuzzy control; fuzzy set theory; gradient methods; least squares approximations; nonlinear control systems; parameter estimation; pattern clustering; BLS algorithm; FCM algorithm; TS fuzzy modeling; Takagi-Sugeno fuzzy modeling; Takagi-Sugeno systems; batch least squares algorithm; dynamic system; fuzzy c-means algorithm; fuzzy clustering; fuzzy system; gradient method; input-output data set; nonlinear system; parameter identification; rule base; surge tank system; Clustering algorithms; Data models; Fuzzy systems; Gradient methods; Liquids; Surges; Training data; Batch Least Squares; Fuzzy C-Means; Fuzzy Clustering; Gradient Method; System Identification; Takagi-Sugeno System;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer,Control & Communication (IC4), 2013 3rd International Conference on

Conference_Location

Karachi

Print_ISBN

978-1-4673-6011-1

Type

conf

DOI

10.1109/IC4.2013.6653774

Filename

6653774