Using Machine Learning to Reduce Design Time for Permanent Magnet Volume Minimization in IPMSMs for Automotive Applications

Abstract

Interior permanent magnet synchronous motors (IPMSMs) have been widely used as traction motors in electric vehicles. Finite element analysis is commonly used to design IPMSMs but is highly time-intensive. To shorten the design period for IPMSMs, various surrogate models have been constructed to predict relevant characteristics, and they have been used in the optimization of IPMSM geometry. However, to date, no surrogate models have been able to accurately predict the characteristics over the wide speed range required for automotive applications. Herein, we propose a method for accurately predicting the speed-torque characteristics of an IPMSM by using machine learning techniques. To improve the prediction accuracy, we set the motor parameters as the prediction target of the machine learning methods. We then used the trained surrogate model and a real-coded genetic algorithm to minimize the volume of the permanent magnet and showed that the design time can be significantly reduced compared with the case where only finite element analysis is used.

Automatic Design System With Generative Adversarial Network and Convolutional Neural Network for Optimization Design of Interior Permanent Magnet Synchronous Motor

“…For example, when training a prediction model to replace FEA, the training data are generally generated by FEA. Assuming that the FEA for generating training data takes 12.4 minutes to analyze the speed-torque characteristics of an IPMSM [15], an FEA of 100,000 datasets would take more than two years.…”

“…First, for data acquisition, we use semi-supervised learning where the training data are generated using machine learning. In [15], the authors proposed a method for training a prediction model that can accurately predict the speed-torque characteristics of double-layered IPMSMs from a small number of design parameters and FEA results using machine learning. This prediction model can be used to calculate the operating characteristics of various double-layered IPMSMs from their design parameters to generate a large dataset in a short time.…”

“…First, the relationship between rotor shape and speed-torque characteristics is learned using the method proposed in [15]. For the prediction model in [15], the input variables are the design parameters, such as the PM thickness, and current conditions.…”

“…First, the relationship between rotor shape and speed-torque characteristics is learned using the method proposed in [15]. For the prediction model in [15], the input variables are the design parameters, such as the PM thickness, and current conditions. Although the objective is to predict the speedtorque characteristics, the prediction models do not directly predict the speed-torque characteristics; instead, we set three motor parameters (PM flux linkage a and d-and q-axis inductances Ld, Lq) as the prediction targets to improve accuracy.…”

“…The objective of the prediction models is to accurately learn the function f in the following equation. See [15] for more details.…”