一种应用机理模型和SOM神经网络的液体火箭推进系统实时故障诊断方法  

A Real-time Fault Diagnosis Approach to Liquid Rocket Propulsion Systems Using Mechanism Model and SOM Neural Networks

作　　者：杨尔辅[1] 徐用懋[1] 张振鹏[2] 

机构地区：[1]清华大学,北京100084 [2]北京航空航天大学,北京100083

出　　处：《上海海运学院学报》2001年第3期280-284,290,共6页Journal of Shanghai Maritime University

基　　金：中国博士后科学基金 (中博基 2 0 0 0 -2 3) ;国家自然科学基金 ( 5 9486 0 0 5 )

摘　　要：为了解决液体火箭推进系统实时故障诊断的问题 ,提出了一种应用故障机理模型和SOM(Self-organizingMapping)神经网络的实时故障诊断方法。其基本过程是先建立所研究对象的故障机理模型 ,通过计算机仿真的办法获得液体火箭推进系统可能的故障模式及故障数据库 ,然后利用SOM神经网络的自组织特征映射功能建立非线性的故障模式识别器 ,完成对系统的实时故障诊断。从而解决了单纯依靠故障机理模型进行诊断时所遇到的实时性问题和单纯依靠SOM神经网络诊断时所遇到的故障样本获取问题。In order to real time diagnose the failures of liquid rocket propulsion systems, a real time fault diagnosis approach using mechanism model and SOM(Self organizing Mapping) neural networks is put forward. Firstly, a mechanism model of liquid rocket propulsion systems is set up. Secondly, the possible fault modes and database of liquid rocket propulsion systems are acquired via computer simulations. Finally, a SOM neural network is used to real time complete failure pattern recognition of liquid rocket propulsion systems. So the real time problem ,which is often met when the mechanism model based diagnosis method is simply applied, is resolved. In addition, the difficulty in acquiring enough failure samples when the neural networks based diagnosis approach is independently used is also overcome. The diagnosis results given in this investigation demonstrate that the real time fault diagnosis approach is effective.

关 键 词：液体火箭 推进系统 故障诊断 过程建模 神经网络 模式识别 计算机仿真 机理模型 

分 类 号：V430[航空宇航科学与技术—航空宇航推进理论与工程] TP206.3[自动化与计算机技术—检测技术与自动化装置]