Processing calibration results for measuring transducers with an integrated sensor

Abstract

The proper production process control is possible only if there is an information-measuring system equipped with a large set of measuring transducers of physical quantities. Currently, measuring transducers with semiconductor sensors are widely used. Along with high sensitivity to the measured value and repeatability of parameters, these transducers are sensible to stray quantities, which leads to additional errors. One way to reduce the additional errors of measuring transducers with an integrated sensor is to apply algorithmic correction, which involves a multifactorial calibration experiment, the processing of the results of which allows obtaining the conversion function. Algorithmic correction solves the issue of linearizing the conversion function and minimizes the effect of influencing factors on the measurement results. The conversion function quality and, consequently, the accuracy of further measurements largely depend on the correctness of the calibration experiment results. Random outliers (spurious errors) in the calibration results inevitably cause a distortion of the conversion function and the further erroneous measurement results. This paper proposes a theoretical justification for the possibility of generating the conversion function of a measuring transducer with an integrated sensor by processing the results of a multifactorial calibration experiment including random parasitic errors. Experimental results are provided to illustrate the effectiveness of the proposed calibration results processing technique, which allows obtaining the correct conversion function in the conditions of distorted initial data.

Neural Network-Based Sensor Calibration for Micro and Nanoelectronics Applications

“…Но существуют погрешности при измерении из-за нелинейности и влияния помех, необходимость учета которых приводит к поиску способов коррекции характеристик измерительных преобразователей с оптическими датчиками. Среди них эффективными являются методы, улучшающие конструкции оптических датчиков за счет новых используемых в них элементов (в частности, световодов), и алгоритмические методы, позволяющие формировать путем калибровки функцию преобразования с учетом нелинейности датчика по отношению к измеряемому значению, а также к паразитным воздействиям [1].…”