Abstract: A system’s architecture for condition monitoring of propulsion system using BP ART hybrid neural networks was presented The topology of this hybrid architecture was:the first processing unit consisted of BP(Back Propagation)neural networks,one BP network per subassembly,the second unit was made up of ART(Adaptive Resonance Theory)neural network which had an autoassociative architecture,only one network for the whole propulsion system Each output of ART network represented a“health state”of propulsion system The hybrid architecture was made full uses of advantages of every neural network An application example of this system wasA system’s architecture for condition monitoring of propulsion system using BP ART hybrid neural networks was presented The topology of this hybrid architecture was:the first processing unit consisted of BP(Back Propagation)neural networks,one BP network per subassembly,the second unit was made up of ART(Adaptive Resonance Theory)neural network which had an autoassociative architecture,only one network for the whole propulsion system Each output of ART network represented a“health state”of propulsion system The hybrid architecture was made full uses of advantages of every neural network An application example of this system was illustrated,which demonstrated that this hybrid neural networks’ system could be implemented effectively in developing an advanced real time system for condition monitoring and fault diagnosis of propulsion system

Key words: Liquid propellant rocket engine;Engine failure;Fault diagnosis;Artificial neural network;Real time display

摘要： 应用ＢＰ－ＡＲＴ混合神经网络提出了一种供推进系统状态监控实时使用的系统，其拓扑结构为： 第一层处理单元由ＢＰ神经网络组成， 每个ＢＰ网络代表一个相应的推进系统组件； 第二层处理单元为一个ＡＲＴ神经网络， 网络的每一个输出代表推进系统的一种“健康状态”， 据此可对其故障进行“诊断”。该混合结构充分发挥了两类网络的优点， 给出的具体应用实例也显示出在推进系统实时状态监控与故障诊断应用中的有效性

关键词: 液体推进剂火箭发动机;发动机故障;故障诊断;人工神经元网络;实时显示