论文部分内容阅读
研究了用于液体火箭发动机（ＬＲＥ）故障仿真与故障检测的神经网络ＢＰ（ＢａｃｋＰｒｏｐａｎａｔｉｏｎ）算法。在ＢＰ算法中采用了加噪声等技术来避免系统误差陷入局部极小，训练出精度高（误差小于０．０２）的神经网络。试验表明：神经网络ＢＰ算法成功地用于故障仿真与故障检测。
A neural network BP (BackPropanation) algorithm for fault simulation and fault detection of Liquid Rocket Engine (LRE) is studied. BP algorithm is used to increase the noise and other techniques to avoid the system error into a local minimum, training a high precision (error less than 0.02) neural network. Experiments show that BP neural network algorithm is successfully used in fault simulation and fault detection.