论文部分内容阅读
本文描述一个用来诊断SSME(航天飞机主发动机)突发性性能衰退和失效模式的系统。该系统将SSME看作一批组件的集合,并利用装在每个组件中的本地敏感器所获得的几组参数推导出时间特征数据构成要分析的故障模式。这里使用了一种混合结构:第一个处理层由基于自适应谐振原理(ART-2)的神经网络构成,每个组件配备1个ART网络;第2层由按内容访问的存储器(CAM)网络构成,每个组件配备1个;最后一层是一个反推神经网络,它处理来自所有CAM网络的数据。文中介绍了仅含一个高压燃料涡轮泵(HFTP)的原型系统。本项工作的长期目标是通过建立集运行参数控制部件、组件、敏感器和神经网络系统为一体的反遗回路,构造一个使用上述结构的系统以确保SSME总是处于正常工作状态。
This article describes a system used to diagnose sudden performance degradation and failure modes in SSME (Space Shuttle Main Engine). The system treats SSME as a collection of components and derives time characteristic data from the sets of parameters obtained by local sensors installed in each component to compose the failure mode to be analyzed. A hybrid architecture is used here: The first processing layer consists of a neural network based on the adaptive resonance principle (ART-2), each with one ART network; the second layer consists of content-accessible memory (CAM) The network consists of 1 per component and the last layer is a backpropagation neural network that processes data from all CAM networks. This article describes a prototype system that contains only one high-pressure fuel turbine pump (HFTP). The long-term goal of this work is to construct a system that uses the above structure to ensure that SSME is always in a normal working condition by establishing an anti-aliasing circuit that integrates the operating parameters of control components, components, sensors and neural network systems.