A temperature-compensated system for magnetic field measurements based on artificial neural networks

Abstract

This paper presents a personal-computer-controlled system, assembled mainly with IEEE 488 general purpose instruments and aimed at the measurement of magnetic field intensity. The main sensor element in the system is a magnetoresistive field sensor which includes four permalloy strips connected in a Wheatstone bridge configuration. The temperature dependence of this sensor can vary almost 25% in the experimental operating temperature range (20/spl deg/C-100/spl deg/C). In order to overcome this problem, a two-terminal integrated circuit temperature transducer is connected to the system, and its temperature information is used to correct the temperature drift error of the magnetic sensor. Artificial neural networks are used for data reduction and final results show an improvement in the system's accuracy from 20% to 2%.