含有噪声的、正常和稳定的传感器数据训练 ART2神经元网络 ,用于液体火箭发动机( L RE)故障检测。每个传感器连续窗的功率谱输入 ART2神经元网络进行学习 ,试验学习好的神经网络 ,验证其能否有效地检测出发动机故障以及故障发生时间。传感器数据来自某变推力液体火箭发动机地面试车 RS61。试验结果表明 ,神经元网络显示的故障发生时间与试车后专家分析的故障开始时间相符 ART2 neural network trained with noisy, normal and stable sensor data for fault detection of Liquid Rocket Motor (L RE). The power spectra of the continuous windows of each sensor are input into the ART2 neural network to learn the good neural network and verify whether it can effectively detect the engine failure and the time when the fault occurs. Sensor data from a variable thrust liquid rocket engine ground test RS61. The experimental results show that the failure time displayed by the neural network coincides with the failure time started by the expert after the test