Ensuring the accuracy of the approximation of R/T-characteristics of NTC-thermistor based on neural network modeling

Abstract

The research is aimed at improving the accuracy of the approximation of characteristics of the semiconductor thermoresistive temperature conductor on the example of an NTC-type thermistor (B57703M series) using neural network techniques for intelligent processing of measurement information.
The objective of the study is to develop feed forward neural network models with Back Propagation and Resilient Propagation learning algorithms in order to ensure the accuracy of approximation of R/T-characteristics of NTC-thermistors in the working temperature range. It is shown that the use of the developed neural network models can provide higher accuracy of the approximation in comparison with the known Steinhart-Hart polynomial model.
Statistical estimation has shown that for the purpose of solving the problem of neural network approximation of R/T-characteristics of NTC-thermistors, the Back Propagation algorithm is preferable to the Resilient Propagation algorithm. The practical use of the developed models improves the accuracy of individual calibration of NTC-thermistor’s temperature range 218.15 – 428.15 K.