Bumpless Transfer Fault-Tolerant Control for Continuous-Time Switched Systems via Learning-Based Fault Reconstruction

Area of Science:

• Control Systems Engineering
• Fault-Tolerant Control
• Switched Systems

Background:

• Switched linear systems are susceptible to actuator faults and external disturbances, compromising operational reliability.
• Existing fault-tolerant control methods often require fault differentiability or struggle with decoupling disturbances.
• Bumpless transfer is crucial for maintaining system stability during control mode switching.

Purpose of the Study:

• To develop a novel learning-based robust unknown input observer (UIO) for fault reconstruction and state estimation in switched linear systems.
• To design an efficient bumpless transfer fault-tolerant (FT) controller that mitigates fault effects and minimizes control bumps.
• To ensure the bumpless transfer constraint and improve the system's anti-disturbance capability.

Main Methods:

• A new learning-based robust UIO is proposed, which does not require fault differentiability and completely decouples disturbances.
• Fault reconstruction is achieved through iterative learning on the system's timeline, utilizing historical information.
• An FT controller is designed using the estimated fault information, incorporating a novel inequality transformation for bumpless transfer and improved disturbance rejection.

Main Results:

• The developed UIO effectively reconstructs actuator faults and estimates system states while being robust to magnitude-bounded disturbances.
• The proposed FT controller successfully counteracts fault effects, suppresses control bumps, and enhances the system's anti-disturbance capability.
• Solvability conditions for the UIO and controller are established under average dwell-time switching, demonstrating theoretical robustness.

Conclusions:

• The integrated approach of learning-based UIO and bumpless transfer FT control offers a robust solution for switched linear systems facing actuator faults and disturbances.
• The method's effectiveness and applicability are validated through simulations on an inverted pendulum system controlled by a direct current motor.
• This work contributes to advancing fault-tolerant control strategies for complex dynamic systems.