A New Intelligent Scheme for Simultaneous Cold Junction Compensation and Linearization of Thermocouples

The paper presents a novel neural network based system for processing thermocouple signals, performing the tasks of linearization of the sensor transfer curve and cold junction compensation in the same stride. Keeping an eye on the viability of the complete system as a commercial product, a thermistor has been considered for sensing the reference junction temperature. Cost of the remaining hardware is also minimal. With the use of versatile Differential Evolution (DE) algorithm during the training of a multilayered Artificial Neural Network (ANN) to initialize the weight and biases of the ANN, it has been possible to achieve remarkably low measurement errors, as revealed by the simulation studies.