Exact analytical inversion of TSK fuzzy systems with singleton and linear consequents


Ulu C., Güzelkaya M., Eksin İ.

INTERNATIONAL JOURNAL OF APPROXIMATE REASONING, cilt.55, sa.6, ss.1357-1368, 2014 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 55 Sayı: 6
  • Basım Tarihi: 2014
  • Doi Numarası: 10.1016/j.ijar.2014.01.007
  • Dergi Adı: INTERNATIONAL JOURNAL OF APPROXIMATE REASONING
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Sayfa Sayıları: ss.1357-1368
  • Anahtar Kelimeler: TSK fuzzy systems, Inversion, Linear consequents, Piecewise linear membership functions, INTERNAL-MODEL CONTROL, UNIVERSAL APPROXIMATION, CONTROLLER-DESIGN
  • Yıldız Teknik Üniversitesi Adresli: Hayır

Özet

In literature, exact inversion methods for TSK fuzzy systems exist only for the systems with singleton consequents. These methods have binding limitations such as strong triangular partitioning, monotonic rule bases and/or invertibility check. These extra limitations lessen the modeling capabilities of the TSK fuzzy systems. In this study, an exact analytical inversion method for TSK fuzzy systems with singleton and linear consequents is presented. The only limitation of the proposed method is that the inversion variable should be represented by piecewise linear membership functions (PWL-MFs). In this case, the universe of discourse of the inversion variable is divided into specific regions in which only one linear piece exists for each PWL-MF at most. In the proposed method, the analytical formulation of TSK fuzzy system is expressed in terms of the inversion variable by using linear equations of PWL-MFs. Thus, the fuzzy system output in any region can be obtained by using the appropriate parameters of the linear equations of PWL-MFs defined within the related region. This expression provides a way to obtain linear and quadratic equations in terms of the inversion variable for TSK fuzzy systems with singleton and linear consequents, respectively. So, it becomes very easy to find exact inverse solutions for each region by using explicit analytical solutions for linear or quadratic equations. The proposed inversion method has been illustrated through simulation examples. (c) 2014 Elsevier Inc. All rights reserved.