Fault Detection and Diagnosis for Liquid Rocket Engines Based on Long Short-Term Memory and Generative Adversarial Networks

The development of health monitoring technology for liquid rocket engines (LREs) can effectively improve the safety and reliability launch vehicles, which has important theoretical engineering significance. Therefore, we propose a fault detection diagnosis (FDD) method large LOX/kerosene engine based on long short-term memory (LSTM) generative adversarial networks (GANs). Specifically, first modeled using MATLAB/Simulink simulated engine’s normal operation states involving various startup steady-state stages utilizing injection. Second, created an LSTM-GAN model trained with operating data LSTM as generator multilayer perceptron (MLP) discriminator. Third, test were input into discriminator to obtain discrimination results realize detection. Finally, predicted samples calculate absolute error between real value each parameter. Then index, standardized (SAE), was constructed. SAE analyzed diagnosis. highlight that proposed detects faults in processes, diagnoses process without missing alarm or being affected by false alarms. Compared conventional redline cut-off system (RCS), adaptive threshold algorithm (ATA), support vector machine (SVM), is more concise timely.