Robust dynamic sliding-mode control using adaptive RENN for magnetic levitation system.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Robotics

Background:

• Magnetic levitation systems require precise control despite inherent uncertainties.
• Traditional control methods like PID and Sliding Mode Control (SMC) face limitations with unknown dynamics and hardware constraints.
• Dynamic Sliding Mode Control (DSMC) aims to mitigate chattering but struggles with unknown uncertainty bounds.

Purpose of the Study:

• To propose a Robust Dynamic Sliding Mode Control system (RDSMC) for magnetic levitation systems.
• To enhance control performance and robustness by addressing uncertainties.
• To utilize a Recurrent Elman Neural Network (RENN) for online estimation of system uncertainties.

Main Methods:

• Derivation of the magnetic levitation system's dynamic model.
• Development of a PID-type DSMC to reduce chattering.
• Integration of an RENN estimator within the RDSMC to approximate nonlinear uncertainties online.
• Application of Lyapunov stability theorem for adaptive RENN parameter learning.
• Design of a robust compensator to handle approximation errors and higher-order terms.

Main Results:

• The proposed RDSMC effectively estimates and compensates for lumped uncertainties in the magnetic levitation system.
• Experimental results demonstrate successful tracking of various periodic trajectories.
• The RENN-based approach significantly improves control performance and robustness compared to conventional methods.
• The system exhibits enhanced stability and reduced sensitivity to unmodeled dynamics.

Conclusions:

• The developed RDSMC system offers a robust and effective solution for controlling magnetic levitation systems with unknown uncertainties.
• The combination of RENN and DSMC provides a powerful framework for adaptive and stable control.
• The findings validate the practical applicability of the proposed control strategy for magnetic levitation systems.