An adaptive approach for linearization of temperature sensor characteristics

This paper discusses an adaptive approach, based on iterative construction of a piecewise linear approximant, for calibration of industrial sensors. Given the non-linear characteristics of the sensor, the approach consists of two steps: (i) the inverted characteristic is computed in pointwise form on the basis of an iterative Newton-Raphson scheme; (ii) an iterative algorithm is employed in order to construct a continuously improving set of piecewise linear approximations of the computationally inverted sensor characteristic. The main feature of the presented approach is the irregular partitioning of the measurement range, induced from the error minimization procedure performed in (ii). On this basis, the lookup table necessary for calibration of the sensor is constructed in the "cheapest" possible way, allowing on the one hand optimal linear approximation in every particular measurement sub-range and on the other an effective reduction of the necessary storage capacity.