Preliminary Design of Permanent Magnet Motor Using Machine Learning Algorithm and Analytical Method

The global attention toward electric vehicles is growing tremendously, mainly because of environmental issues in recent years. There has been a significant increase in the development of hybrid and pure electric vehicles as they are considered as an effective solution for reducing the carbon footprint. There is a lot of research happening, especially in the design of high-performance e-motors for electric powertrain applications. In this paper, the focus is on the permanent magnet synchronous motors (PMSM) due to its higher efficiency and more advantageous torque characteristics compared to other types of motors. This paper presents a procedure for determining the initial design parameters using analytical calculation method for a PMSM, followed by developing machine learning algorithms (XGBoost, random forest, and artificial neural networks) with the available benchmarking data and compare their performance to determine the motor design parameters. A comparison study with the results obtained from analytical calculation and machine learning algorithm is carried out in determining the initial sizing parameters, and we have obtained an accuracy of 80%. We believe that this machine learning algorithm design approach will help in saving the time needed for theoretical design, and with an optimum design solution, can reduce the time and iterations of FEA required while designing an e-motor.