Real-time Torque Estimation of the Ultrasonic Motor Based on the Machine Learning Model

Abstract

This study focuses on the real-time estimation of ultrasonic motor output torque using machine learning models. By collecting sample data under various operating conditions, a neural network is employed to uncover the hidden relationships between the output torque and the electrical, mechanical parameters of the ultrasonic motor. The neural network, trained on this diverse dataset, enables precise torque estimation without the need for direct torque measurement sensors. This approach significantly simplifies the experimental setup, reducing the dependency on extensive sensor arrays. Consequently, this simplification paves the way for future sensorless control of ultrasonic motors. The successful implementation of this model demonstrates the feasibility of using data-driven techniques in complex motor control applications. By leveraging the power of neural networks, the study showcases how advanced computational methods can effectively address the challenges of real-time torque estimation in ultrasonic motors, offering a promising solution for enhancing the performance and efficiency of motor systems.