Motor Control System Implementation and Speed Estimation Method Using ANN with Offline Training and Online Inference

This paper proposes a motor control system implementation and speed estimation method using an artificial neural network (ANN) with offline training and online inference. Recent studies have increasingly explored the application of ANN to motor drive systems. However, these studies have focused on improving control performance, and discussion on how to implement and apply ANNs to motor drive systems remains relatively limited. To address this limitation, this paper proposes the procedures required to apply an ANN to a motor drive system, including data collection under various operating conditions, ANN model design and training, and online inference. The proposed method reduces the computational burden during training and achieves fast inference in the driving environment. In addition, an ANN model suitable for the motor drive system is designed through estimation performance comparison. The validity of the proposed system is verified through simulation and experimental results.