摘要
针对航空发动机磨损故障诊断自动化及智能化程度不高的问题,提出一种基于油液数据挖掘的航空发动机磨损故障诊断知识获取方法。该方法利用自组织神经网络对原始多维特征数据进行特征融合,得到融合值;利用Parzen窗法制定融合值的界限值,将样本划分为正常、警告和异常3种状态;利用Weka软件对油液数据进行规则提取。该方法能够从油液光谱数据中识别出不同磨损状态信息,并提取出知识规则用于构建航空发动机磨损诊断系统的知识库,实现了基于润滑油光谱磨损数据的航空发动机故障诊断的自动化与智能化。应用某型飞机发动机实际油液光谱数据对提出的磨损故障诊断知识获取方法进行验证,结果表明:经特征融合得到的融合值能够准确反映航空发动机的劣化趋势;利用融合值的界限值划分样本状态,再进行规则提取时具有很高的识别率。
Aimed at the problem of low automation and intellectualization of aeroengine wear fault diagnosis,a knowledge acquisition method of aeroengine wear fault diagnosis based on oil data mining was proposed.This method uses self-organizing neural network to fuse the original multi-dimensional feature data and get the fusion value,uses Parzen window method to establish the limit value of fusion value and divide the sample into normal state,warning state and abnormal state,and uses the software of Weka to extract knowledge rules from oil analysis data.This method can identify different wear state information from oil spectrum data,and extract knowledge rules to build knowledge base of aeroengine wear diagnosis system,which realizes the automation and intellectualization of aeroengine fault diagnosis based on oil spectral wear datalubricant spectrum.The proposed knowledge acquisition method for wear fault diagnosis was validated by using the actual oil spectrum data of an aircraft engine.The results show that the fusion value obtained by data fusion can accurately reflect the deterioration state of aero-engine,it has a high recognition rate by using the boundary value of the fusion value to divide the sample state and extract the knowledge rules.
作者
张全德
陈果
郑宏光
陈明衡
王培文
王洪伟
李华
ZHANG Quande;CHEN Guo;ZHENG Hongguang;CHEN Mingheng;WANG Peiwen;WANG Hongwei;LI Hua(Quality Safety Department,China's People's Liberation Army 5720 Factory,Wuhu Anhui 241000,China;College of Civil Aviation,Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 210016,China;The Sixth Research Office,Beijing Aeronautical Technology Research Center,Beijing 100076,China;Mechanical and Electronic Department,China's Peopl e's Liberation Army 5720 Factory,Wuhu Anhui 241000,China)
出处
《润滑与密封》
CAS
CSCD
北大核心
2019年第3期128-134,共7页
Lubrication Engineering
基金
国家自然科学基金项目(51675263)
关键词
故障诊断
数据挖掘
特征融合
界限值
规则提取
fault diagnosis
data mining
self-organizing map
feature fusion
threshold
rule extraction
