基于人工噪声神经网络BP算法的火箭发动机故障仿真与检测  

FAILURE SIMULATION AND DETECTION OF A LIQUID ROCKET ENGINE BASED ON ARTIFICIAL NOISE BP NEURAL NETWORK

作　　者：黄敏超[1] 张育林[1] 冯心 

机构地区：[1]国防科技大学航天技术系

出　　处：《推进技术》1994年第2期1-6,33,共7页Journal of Propulsion Technology

基　　金：国家自然科学基金

摘　　要：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。BP (Back Propagation) neural network used in liquid rocket engine(LRE)failure simulation and detection was described. In order that the system errordid not fall into partial minimum, additional noise was recommended in BP algorithm, and an accurate neural network was trained (with error below 0. 02). It wasshown that the BP neural network could be successfully employed in simulation anddetection of LRE failures.

关 键 词：液体推进剂 火箭发动机 故障诊断 

分 类 号：V448.15[航空宇航科学与技术—飞行器设计]