Intelligent fault tolerance control using long short-term memory for efficient system performance under fault...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence in Engineering
• Data-Driven Control

Background:

• Fault-tolerant control (FTC) is essential for system stability and performance under fault conditions.
• Traditional FTC methods often rely on complex diagnostic and process models.
• There is a need for data-driven approaches that simplify fault tolerance.

Purpose of the Study:

• To develop and validate a novel data-driven fault-tolerant control system.
• To demonstrate the efficacy of a single Long Short-Term Memory (LSTM) network for integrated fault diagnosis and control reconfiguration.
• To eliminate the requirement for explicit diagnostic and process models in FTC systems.

Main Methods:

• Implementation of a fault-tolerant control system based on a single Long Short-Term Memory (LSTM) network.
• Integration with MATLAB via a digital twin concept for simulation and validation.
• Application and testing on an assembler case study with faultless and faulty sensor scenarios.

Main Results:

• The LSTM-based system successfully managed control reconfiguration in the presence of faults.
• Simulations achieved a Root Mean Square Error (RMSE) of [Formula: see text] after 6553 iterations.
• Achieved system output accuracies of 92.81% in faultless conditions and 67.16% in the worst-case fault scenarios.

Conclusions:

• The proposed data-driven LSTM-FTC system is effective for enhancing control system operation during faults.
• The model-free nature of the LSTM-FTC system offers a significant advantage over traditional methods.
• The digital twin approach facilitated robust simulation and validation of the fault-tolerant control strategies.