Eccentricity and Tilt Modeling and Calibration Method for Ring Magnetic Sensor Array-Based Current Measurement

This paper introduces an innovative current measurement method leveraging deep learning to reduce errors caused by the eccentricity and tilt of current-carrying wires in a tunnel magnetoresistance (TMR) ring array configuration. The study first examines the measurement principles of the 4-TMR ring array, followed by developing a mathematical model for TMR sensor arrangement. This model enables the generation of sample data reflecting arbitrary eccentricities and tilts of the current-carrying wire. The sample data are then used to train a deep neural network (DNN) model, which accurately calculates the current in the wire. Experimental results show that the current measurement error using this method remains within 0.5%, outperforming traditional Ampere’s law-based measurement techniques. This approach demonstrates significant potential for practical application in high-precision current sensing within engineering contexts.