Systematic error correction for experimentally modeled sensors by using ANNs

The paper deals with the compensation of the systematic uncertainty of sensors subject to nonlinear and combined influence parameters. The compensation is based on a second sensor and a digital artificial neural network (ANN). This heuristic fully a-posteriori approach allows the twofold problems of (i) the complex mathematical modeling of the influence on the measurement, and (ii) the effective solution of the nonlinear model to be simultaneously bypassed. Experimental results of the characterization of a variable-reluctance proximity transducer highlight the effectiveness of the proposed compensation scheme.