基于信息融合的涡轮发动机故障诊断方法研究

对近年来信息融合方法发展进行总结，着重介绍贝叶斯信息融合法、神经网络信息融合法、基于特征的信息融合法和D-S证据理论信息融合法，分析各种方法的原理和特点。以CFM56发动机故障诊断为例，用BP神经网络的输出结果为输入，构建D-S证据融合的识别框架，进行故障诊断。结果表明，采用D-S证据理论的方法，缩短了故障诊断的时间并提高了故障诊断的精确度。

Turbine Aero-engine Fault Diagnosis Methods Based on Information Fusion

This paper summarizes the recent development of information fusion. The methods, including bayesian, neural network, feature and D-S evidential reasoning information fusion, are demonstrated. Moreover, the principle andcharacteristic of methods mentioned are analyzed. The fault recognition frameworks of D-S evidential reasoning, which use output results of BP neural network as input ones are constructed to diagnose faults by taking CFM56 aero enginefault diagnosis as examples. The result indicates that the fault diagnosis time is shortened and fault diagnosis definition is improved.