An Artificial Neural Network-Assisted Hybrid Design Approach for Induction Motors in Vehicular Application

This article presents an artificial neural network (ANN)-based hybrid design methodology for motors used in electric vehicle applications. The proposed method uses ANN to achieve a semi-optimized motor geometry, followed by the drive cycle analysis for the desired vehicle. For this, a large pool of motor design data is used as a training set for the ANN. The semi-optimized motor geometry is further processed for power factor improvement, overall motor efficiency, and electromagnetic noise reduction. The proposed method reduces the overall complexity of the iterative motor design and optimization process. The implementation of the method is demonstrated with a case study wherein a 110 kW three-phase induction motor is designed for an electric bus using the NREL drive cycle. The performance of the motor is verified using a finite element analysis motor using Maxwell ANSYS. The work described in this article was motivated by the complexities of the iterative motor design process, which involves a high level of human resources engagement and time consumption. To address this, the presented work proposes a design approach that bypasses all the complex parts of the work by applying machine learning. The main feature of the approach is that it adopts an ANN-based method that provides a primitive set of motor design parameters for different structures/models of the motor. It eases the work of the motor designer, who has to select the best possible motor structure among these structures and revamp it for further improvement of motor performance. The application of the method is more prolific if the motor is designed for an electric vehicle that exhibits variable loading conditions. The assessment of the proposed model by designing a heavy-duty exhibit shows a significant reduction in the process complexities.