A non-neural fuzzy rule base approach for modeling complex systems

Studies the modeling of complex systems using fuzzy rules for the description of the underlying plant. The current research is a modification and an extension of a bisection and homogeneity algorithm that generates rules directly from the data. This modified approach examines correlation-product inference as a method for approximating nonlinear 2D systems and 3D chaotic systems. The iterative process of trajectory reconstruction combines fuzzy inference and a variable-time-step Euler method. Experimental results are presented for the Van der Pol oscillator and for the Lorenz system.