A Cognitive Neural Linearization Model Design for Temperature Measurement System based on Optimization Algorithm

Abstract

Abstract – The main core of this paper is to design an experimental method for estimating of the nonlinearity, calibrating and testing of the different types of thermocouples temperature sensors (J, K, T, S and R) using multi-layer perceptron (MLP) neural network based on slice genetic (SG) optimization learning algorithm. Temperature sensor has a nonlinearity behavior nature in its output response but it requires a linear behavior output with accepts approximation in accuracy level, noise and measurement errors. Therefore, neural network topology is proposed with five main steps algorithm to reduce the effected noise and minimize the measured errors. Matlab simulation results and laboratory work (LabVIEW) validate the preciously of the proposed cognitive neural linearization algorithm in terms of calculating the temperature from the different types of thermocouples temperature sensors and minimizing the error between the actual temperature output and neural linearization temperature output as well as overcoming the problem of the over learning in the linearization model with the minimum number of fitness evaluation for the learning algorithm..