ANN-based error reduction for experimentally modeled sensors

A method for correcting the effects of multiple error sources in differential transducers is proposed. The difference in actual characteristics of the sensing elements of the differential scheme, and an easily controllable auxiliary quantity (e.g. supply voltage of conditioning circuit) provide independent information for the correction. This is carried out by a nonlinear multidimensional inverse model of the transducer based on an artificial neural network. Experimental results of the correction of a variable-reluctance displacement transducer, subject to the combined interference of structural and geometrical parameters, highlight the effectiveness of the proposed method.