.png){kind=icon}
.png){kind=icon}
Return
High-precision displacement control of giant magnetostrictive actuator based on LSTM-enhanced feedforward compensation inverse model
.png){kind=icon}
PRE
AI
Y
Y
S
张
D
DOI：10.1088/1361-665X/ae2b17.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Giant magnetostrictive actuators (GMAs) are widely utilized in fast actuation systems owing to their superior performance characteristics. However, the presence of material hysteresis and structural nonlinearities results in low tracking accuracy and limited response speed during actuation, significantly impeding the advancement of its practical applications. To address this issue, a high-precision displacement control strategy based on long short-term memory (LSTM)-enhanced feedforward inverse compensation model is novelly proposed. Specifically, the nonlinear mathematical model of the GMA is established based on the Jiles–Atherton (J–A) hysteresis. On this basis, an accurate feedforward inverse model is incorporated to the baseline controller of GMA, in order to compensate for system nonlinearities. Benefitting from its online learning and adaptive adjustment capabilities, the LSTM network is employed to dynamically update the feedforward model parameters, thereby reducing tracking errors and suppressing disturbances. To further enhance the response speed, the LSTM network is improved by incorporating Kalman filter (KF) -based rapid prediction, enabling precise displacement tracking control of the GMA. The effectiveness of the proposed method is validated through simulation analysis and experimental results. The results demonstrate that when the control signal is a sinusoidal or harmonic signal, the dynamic tracking error is maintained within 1 μm, with a maximum root mean square error of 0.4727 μm. The error remains within an acceptable range, thereby confirming the effectiveness of the proposed LSTM-enhanced feedforward compensation inverse model control algorithm.
Journal
IF:
3.8
Papers:
8.5K
Citations:
2.5W
Organization
Cited Papers