Investigation of Artificial Neural Network Techniques for Thermistor Linearization
Naveen Kumar Vaegae, Venkata Lakshmi Narayana Komanapalli · DOAJ (DOAJ: Directory of Open Access Journals) · 2015
This paper presents the development of an artificial neural network (ANN)-based linearizer for thermistor connected in operational amplifier based inverting amplifier circuit. The op-amp based thermistor signal conditioning circuit exhibits a stable temperature-voltage relation over a temperature range of 0°C - 100°C with moderate linearity. The transformation technique is used for the analysis and design of the thermistor circuit. An ANN-based direct modeling technique is used for nonlinearity estimation of thermistor circuit to improve the linearity. In this paper, four neural network algorithms Levenberg-Marquardt, Broyden-Fletcher-Goldfarb-Shanno, scaled conjugate gradient and extreme learning machine algorithms are used for training and learning of ANN. The merits and de- merits of the algorithms are investigated for thermistor linearization. An embedded unit is used to implement the learning and training of the ANN techniques and the performance of the developed techniques is experimentally verified on a prototype unit. A notable feature of the developed technique is the non-linearity error remains very low over the entire dynamic range of the sensor.