The noise immunization capability for two-termed fuzzy logic control systems

We investigate how the percentage of overlap in symmetric membership functions affects the performance of a linear time invariant system. A noise immunization analysis based on a simplified fuzzy noise additive model was made. The model basically realizes its fuzzy inference and proceeds its fuzzification and defuzzification as usual, except the measured output is noise-corrupted. From the analysis and computer simulations, it was found that the percentage of overlap affects the noise immunization capability once the signal/noise ratio is large enough. For first order systems used in these studies, the SNR ratio should have at least 20 dB to notice the substantial improvement caused by the percentage of overlap on the system performance.