Data-Driven Distributed Fault Detection and Fault-Tolerant Control for Large-Scale Systems: A Subspace...

Area of Science:

• Control Engineering
• Systems Science
• Data-Driven Modeling

Background:

• Centralized fault detection and control (FD/FTC) methods are inadequate for large-scale systems due to subsystem interconnections.
• Existing approaches struggle with distributed architectures and the need for global system information.

Purpose of the Study:

• To propose a data-driven framework for integrated fault detection and fault-tolerant control (FD/FTC) in large-scale systems.
• To enable distributed design by utilizing only local and neighboring subsystem data.

Main Methods:

• A subspace predictor-assisted framework organically combines FD and FTC.
• Each subsystem's predictor uses only local and neighboring input/output (I/O) data, facilitating distributed design.
• The integrated architecture is entirely data-driven, requiring no large-scale system mechanism information.

Main Results:

• The proposed framework enables a distributed design for FD and FTC.
• The method successfully integrates fault detection and fault-tolerant control without global data.
• Feasibility and effectiveness were verified through numerical simulations and a cascaded CSTR case study.

Conclusions:

• The subspace predictor-assisted framework provides an effective data-driven solution for distributed FD and FTC in large-scale systems.
• This approach overcomes the limitations of centralized methods and demonstrates practical applicability.