应用ＢＰ神经网络，提出了一种液体火箭发动机故障实时检测系统。它采用非线性辨识技术，在建立发动机数学模型和输出包含故障信息的监视指标信号之后，用阈值线与监视指标比较，从而预报发动机故障。液体火箭发动机启动与稳态过程的试验数据检验表明：基于非线性辨识技术的故障检测系统性能优越。 Using BP neural network, a real-time fault detection system of liquid rocket engine is proposed. It uses non-linear identification technology, after establishing the mathematical model of the engine and outputting the monitoring index signal containing the fault information, the threshold value is compared with the monitoring index to predict the engine fault. Tests of liquid rocket engine start-up and steady-state process show that the fault detection system based on nonlinear identification technology has superior performance.