Transient response control of the ultrasonic motor with adaptive frequency sweep

Inhalt:
This study investigates the transient response of the traveling wave ultrasonic motor (TWUSM) with varying fre-quency sweep speeds to identify limitations in response times. Simplifying the system for analysis, the stator was iso-lated from the rotor and modeled using an equivalent circuit that incorporated nonlinear capacitance and resistance. This model effectively captured both static and dynamic behaviors of the TWUSM stator. To optimize the start-up response, the Soft Actor Critic (SAC) deep reinforcement learning algorithm was employed within the simulated environment defined by the equivalent circuit model. The SAC model demonstrated superior per-formance, achieving faster responses than traditional constant frequency sweeps during simulations. However, when applied in a real control system, the SAC model's performance was hindered by delays in inference calculation, leading to suboptimal results compared to the simulated outcomes. Future work will focus on minimizing the discrepancy between simulation and real-world application by optimizing real-time inference processes and expanding the model to include rotor-stator interactions. Additionally, comprehensive dynamic modeling of the entire motor system under varying operational conditions will be pursued to enhance the ro-bustness of frequency control using advanced reinforcement learning techniques.