Deep Learning–Driven Torque Ripple Suppression in Permanent Magnet Motor Drives for High-Precision Applications

Abstract

The Permanent Magnet Motor Drives, especially the Permanent Magnet Synchronous Motors (PMSMs), is in common use in high precision control like robotics, electric vehicles, aerospace systems, and computer numerical control (CNC) machine where their high efficiency, small structure and better dynamic response is required. In spite of these benefits, torque ripple has been a significant cost to overcome the drive performances due to the generation of vibration, acoustic noise, as well as poor motion accuracy. Traditional methods of reducing torque ripple, such as, harmonic current injection and model-based control methods can tend to demand precise modelling of the motor and can be vulnerable to changes in parameters and external influence. This paper will detail strategies to overcome these constraints by offering a deep learning-based strategy of torque ripple suppression in Permanent Magnet Motor Drives that will enhance precise and stable power delivery. The deep neural network model is trained to reflects the nonlinear correlations between motor working variables (current components, rotor position, and speed) and makes it possible to predict and compensate the torque ripple in real-time. The suggested approach is tested with a simulated Permanent Magnet Synchronous Motor drive system and conditions of various load and speed are applied. The outcomes of the simulation show that the deep learning-based controller leads to a large decrease in the torque ripple when it exists relative to the conventional control methods without destabilising the torque output and speed response. The suggested solution increases the accuracy, stability and performance of the motor drive systems in state of the art industry.