Sugeno Inference System for Estimating Non-linear Characteristics in Smart Sensor Applications

Abstract

This work exploits Sugeno inference system in estimating the non-linear characteristics values closer to the physical quantities sensors measure. In general, the non-linearity of the sensor is modelled by fixing the input of the sensor to the membership function of the inference system and output response of the sensor to output constant of the inference system. In addition, the input membership function of the inference system is organized in a way that the middle arm of the first membership is vertically aligned to the left arm of the second membership and so for all of the membership functions. While the relationship of the input and output of the inference system thus the sensor is governed by the if-then-rules. The root-mean-square error of the overall developed system is relatively too small to be ignored in sensitive applications, signifying the closer fit to the actual physical quantity data. The minimal multiplication and addition to compute n number of membership functions is found to be n and 2×n respectively. The promising results offer the method to be applied in programmable smart sensor applications.