改进ART1神经网络在航空发动机故障诊断中的应用  

A Application in Aero-Engine Fault Diagnosis Based on Reformative ART1 Neural Network

作　　者：李果[1] 李学仁[1] 何秀然[1] 

机构地区：[1]陕西西安空军工程大学工程学院

出　　处：《微计算机信息》2005年第09S期156-158,共3页Control & Automation

摘　　要：根据航空发动机故障诊断原理和飞参数据,提出了一种改进ART1神经网络故障诊断方法。该法对数据预处理后,运用四阈值法和二次判断建立故障代码消除“伪故障”,并对ART1神经网络的输入层进行了改进,在诊断突发性故障的基础上,增加了对因为部件原因而引起的渐变性故障的诊断。实验结果证明,该方法精确度高,噪声抑制力强,而且诊断准确,大大降低了虚警率。According to principles of aero-engine fault diagnosis and flight data,this paper introduces a fault diagnosis method based on reformative ART1 neural network.After data pretreatment,it uses four -circumscription and quadraticestimation to build fault codes and eliminate inveracious fault, and also ameliorate input layer of ART1 network.So this network can diagnoses not only gusty fault,but also gradual fault.It is proved that this method has high precision,strong restrain against noisos,and low rate of inveracious alarms.

关 键 词：航空发动机 自适应共振理论(ART)网络 故障诊断 四阈值法 二次判断 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]