A Method for Real-Time Fault Detection of Liquid Rocket Engine Based on Adaptive Genetic Algorithm Optimizing Back...

Area of Science:

• Aerospace Engineering
• Artificial Intelligence
• Mechanical Engineering

Background:

• Liquid rocket engines (LREs) require robust real-time fault diagnosis for operational safety and reliability.
• Traditional fault detection methods may lack the sensitivity and adaptability needed for complex LRE systems.
• Optimizing neural network parameters is crucial for improving diagnostic accuracy.

Purpose of the Study:

• To develop and validate a real-time fault diagnosis method for LREs.
• To enhance the performance of back propagation (BP) neural networks using an adaptive genetic algorithm (GA).
• To improve system sensitivity and robustness in fault detection.

Main Methods:

• An adaptive genetic algorithm was employed to optimize a BP neural network.
• The optimized BP network was used for real-time prediction of sensor data.
• A threshold judgment mechanism was implemented to detect engine malfunctions based on prediction errors.
• The method was simulated and verified using operational data from a liquid hydrogen and liquid oxygen rocket engine.

Main Results:

• The proposed method demonstrated effective real-time fault diagnosis for the LRE.
• The adaptive GA-optimized BP network showed higher system sensitivity and robustness compared to a single BP model.
• Performance was superior to a BP model optimized by a traditional GA.

Conclusions:

• The adaptive GA-optimized BP neural network offers a valuable approach for real-time LRE fault diagnosis.
• The method exhibits significant improvements in sensitivity and robustness, indicating strong engineering application potential.
• This technique enhances the safety and reliability of liquid rocket engine operations.