Enhanced Fuzzy Systems for Type 2 Fuzzy and their Application in Dynamic System Identification

The paper proposes a novel fuzzy system structure to enhance the performance of fuzzy neural network systems. The structure of enhanced fuzzy system (EFS) is to decompose each fuzzy variable into fuzzy subsystems called component fuzzy systems to act as type 2 fuzzy, and each component fuzzy system is based on one traditional fuzzy set with one pair of symmetry fuzzy sets. In addition, in order to illustrate the performance of EFS, the paper utilizes the common back propagation learning algorithm for neural networks in the identification of dynamic systems. From simulation results, it is evident that the proposed EFS have much faster convergent speed in terms of epochs in the tracking model and better testing error than those of using other identification methods.