A Thermocouple Model Based on Neural Networks

Thermocouples are temperature sensors of common use in industrial applications. Classical mathematical models for these sensors consist of a set of two to four polynomial expressions reproducing their behaviour in different temperature ranges. In this work we propose a new 'one stage' model for these sensors covering the whole sensing range. The modelization has been carried out with an artificial neural network, which reproduces the sensor behaviour in the whole span, providing even better results than the standard piecewise model.