AI-Based Optimization Method of Motor Design Parameters for Enhanced NVH Performance in Electric Vehicles

Abstract

The high-frequency whining noise produced by motors in modern electric vehicles can cause a significant issue, which leads to passenger annoyance. This noise becomes even more noticeable due to the quiet nature of electric vehicles, which lack background noise sources to mask the high-frequency whining noise. To improve the noise caused by motors, it is essential to optimize various motor design parameters. However, this task requires expert knowledge and a considerable time investment. In this project, the application of artificial intelligence was applied to optimize the NVH performance of motors during the design phase. Firstly, three benchmark motor types were modelled using the Motor-CAD CAE tool. Machine learning models were trained using DoE methods to simulate batch runs of CAE inputs and outputs. By applying AI, a CatBoost-based regression model was developed to estimate motor performance, including NVH and torque, based on motor design parameters, achieving impressive R-squared values of 0.94 - 0.99. Additionally, further key design predictors were analysed through SHAP. Subsequently, various optimization algorithms were investigated, including particle swarm optimization, genetic algorithm, and reinforcement learning, to determine the optimal adjustments of motor design parameters for improved NVH performance. Throughout this process, improvements in NVH performance were achieved while applying constraints to maintain torque levels and motor cost. Finally, the AI model and optimization algorithms were integrated into a user interface dashboard, enabling motor design engineers to efficiently predict motor NVH performance by selecting input parameters, applying attribute balance constraints, and executing optimizations.