动态神经网络在液体火箭发动机故障检测与分离中的应用

LIQUID ROCKET ENGINE FAULT DETECTION AND ISOLATION WITH DYNAMIC NEURAL NETWORK

应用动态神经网络在线辨识方法，提出了一种液体火箭发动机故障实时检测与分离基本系统。检测逻辑通过度量包含发动机故障信息的辨识残差信号实现火箭发动机故障检测，故障分离通过分析辨识误差相关函数的不同空间特征来实现。仿真研究表明动态神经网络可成功地应用于泵压式液体火箭发动机故障检测与分离。

A real-time algorithm of fault detection and isolation for a liquid rocket engine is proposed based on an on-line identification method of a dynamic neural network.The detection of the rocket engine fault is achieved by evaluating the identified residual signals indicating the engine fault through the detection logic.The isolation of the fault type can be obtained from analysis of key correlation functions of the identified errors.Simulation research shows that the dynamic neural network has great success in the fault detection and isolation of the liquid rocket engine.