Identifiable Conditions Of T-S Fuzzy Model And Its Application

Complex and uncertain systems are often poorly modeled with conventional approaches that attempt to find a global function or analytical structure for a nonlinear system. Fuzzy modeling is an effective approach that can describe complex, ill-defined nonlinear or uncertain system. This dissertation closely surrounds fuzzy modeling and identification methods for nonlinear systems to discuss and to research.First of all, the essential definitions and principles related to fuzzy system's identification have been introduced. Based on the above explaining, the classification and the approximation abilities of fuzzy system are presented.In the next place, the identifiability of fuzzy system methods is discussed. First, the input space of Mamdani fuzzy model is confirmed by choosing membership functions and confirming the number of rules and the rules. Then, parameter identification is done by recursive least squares algorithm (RLS). The performance of the identification method is analyzed. The persistent excitation conditions are established, under which the parameters of the fuzzy system model converge to their true values. The validity of the persistent excitation conditions is identified by the second-order nonlinear moving average system. Then the mean square error is given. Second, this part investigates the identifiability of nonlinear system modeling based on T-S fuzzy model. After confirming the input space, fuzzy modeling identifiable conditions is studied and discrimination is given based on the principal components of the QR decomposition method. Finally, the effectiveness of this method is verified by the simulation experiment of the nonlinear system modeling.In the third part, the structure of fuzzy model is determined by comparing the fitting degree of objective functions. Because there are some uncertain factors, including noise, interferences and the coupling among variables in practical engineering applications, the similar linear correlation among columns may exist in fuzzy relation matrix. In this chapter, the new algorithm by using objective function is proposed to confirm the structure and parameters of fuzzy model for nonlinear systems. First of all, the structure of fuzzy model is confirmed by using fuzzy clustering. Thus, the objective function and parameter estimation are recursively calculated to simply the structure of model. The U-D matrix decomposition is used to reduce the amount of computation in the determination process of fuzzy model. The simulation results demonstrate the effectiveness of the proposed method. Finally, fuzzy identifying methods are applied into practice engineering about the cooling system of turbogenerator with seal oil in power plant simulator. The applicability of fuzzy modeling proposed in this article is demonstrated by the simulated results.