Compensation of Measurement Errors for a Magnetoresistive Angular Sensor Array Using Artificial Neuronal Networks

Abstract

Sensing setups based on the magnetic field possess many benefits that make such sensors an important class in many application fields. Especially in industrial or automotive applications, such sensing concepts are very important parts in many essential system components. Magnetic field sensors are typically used to measure angular positions and rotational velocity as well as linear positions. Due to their high robustness and accuracy, MR-based sensors have a dominant role in the angular sensing domain. Nevertheless, novel vehicle concepts such as electrical driven or hybrid cars as well as increasing quality and reliability requirements demand improvements. However, those sensors face challenges due to new technologies, which are accompanied by new regulations. From this follows that manufactures can either adapt existing technologies or develop new sensing concepts. The goal of the following work is the evaluation of the capability of neuronal networks to enhance the measurement results of magnetic field sensors if these are disturbed by magnetic stray fields. Therefore, the system setup and possible error sources are analyzed before different neural networks are designed, tried and tested, whereby special attention is given to the network configuration in order to enable a deployment close to the individual sensors. Finally, the results are verified with an adapted measurement setup.

Keywords

• xMR
• magnetoresistive sensors
• sensor array
• artificial neural network
• convolutional neural network
• perceptron