Increasing robustness of a datadriven permanent magnet synchronous motor by predicting torque using machine learning

Abstract

This research investigated the efficacy of two machine learning models in accurately forecasting torque in Permanent Magnet Synchronous Motors (PMSMs), which are vital components in the industrial and automotive industries. The models used a dataset including 50,000 data points and employed several approaches and algorithms. The Gradient Boosting methodology and the XGBoost algorithm were used to forecast torque by using input factors such as current, voltage, and rotational speed. Nevertheless, precise calibration is essential to avoid overfitting, and the use of this technique may be restricted due to computational requirements. The second model used an Artificial Neural Network (ANN) structure, making use of TensorFlow and Keras libraries. The XGBoost model demonstrated enhanced training velocity and interpretability, while the ANN model offered a more profound comprehension of data connections. Additional study is required to combine the advantages of both strategies and assess their practical effectiveness in real-life situations. This research emphasizes the capacity of machine learning in the field of electric motor engineering by using torque calculations, hence opening up possibilities for enhanced motor technologies that provide more efficiency, dependability, and sustainability.