Development and Discussion of a Virtual Sensor for Stationary Torque Prediction in Electric Machines

The automotive industry is undergoing a significant transformation, driven by the rise of electric mobility and autonomous driving, which places new demands on the knowledge of designers, especially regarding the behavior of electric motors. Accurate torque prediction is essential for accurate efficiency prediction and range determination. This article presents a novel methodology for torque prediction. By using data-driven AI models and comparing them with physical models, this study explores optimal strategies for torque prediction in electric motors. Experimental validation was conducted using two identical 440 kW induction motors (IMs) coupled via a cardan shaft and equipped with speed and torque sensors. The results show a high accuracy of the physical model with a mean absolute error (MAE) of 20 Nm. The AI-based data-driven model provides an accuracy of 14 Nm MAE. This research is closed with a discussion about the advantages and limits of the approaches.<bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><i>Index Terms</i>— Electromechanical sensors, intelligent sensors, machine learning (ML), mechanical sensors, predictive models, sensor systems and applications.</b>