Abstract: The multi fault phenomena erists frequently in the turbo pump system of liquid rocket engine. The feature expression of common fault of turbo pump was studied for building up the standard fault pattern and satisfying the needs of neural network study and diagnosis. According to the problem of BP neural network, the method of multi fault diagnosis and classification using parallel BP neural network was studied. The results show that the parallel BP neural network can solve the difficult problem of multi fault diagnosis of turbo pump.

Key words: Liquid propellant rocket engine;Turbine pump;Artificial neural network;Fault diagnosis

摘要： 针对液体火箭发动机的涡轮泵系统中 ,常出现多故障同时发生的现象 ,分析了涡轮泵常见故障的特征表现 ,建立了涡轮泵系统的标准故障模式。在此基础上 ,提出了采用建立并行BP神经网络进行多故障诊断分类的方法。结果表明 ,并行BP神经网络结构简单 ,学习诊断速度快 ,对单一故障的诊断分类优于基本BP网络 ,且能对并发故障进行诊断分类

关键词: 液体推进剂火箭发动机;涡轮泵;人工神经元网;故障诊断