Nonlinear inverse modeling of sensor characteristics based on compensatory neurofuzzy systems

To correct the nonlinearity error of output response of the sensor, a new approach for sensor inverse modeling based on compensatory neurofuzzy systems is proposed in this paper. Such a compensatory fuzzy logical system is proved to be a universal approximator. The compensatory fuzzy neural networks not only adaptively adjust fuzzy membership functions but also dynamic optimize the adaptive fuzzy reasoning by using a compensatory learning algorithm. The proposed neurofuzzy system is then applied to construct input-output characteristic inverse modeling of pressure sensor. Experimental result has shown that the proposed inverse modeling approach automatically compensates the effect of the associated nonlinearity to estimate the applied pressure. Hence, the performance of the pressure sensor is highly improved. As compared with other neural networks modeling methods, the proposed approach has the advantages of simplicity, flexibility, and high accuracy