作　　者： ; ;

机构地区： 中山大学数学与计算科学学院数学系

出　　处： 《数理统计与应用概率》 1994年第3期74-79,共6页

摘　　要： 本文阐述运用逐步Fisher判别法,通过对内燃机车机油光谱数据作分析,判断柴油机磨损状态是正常、异常或严重异常,其正判率达到90％。对于已判定处于异常或严重异常磨损状态的机车,运用成分向量的主成分分析及逐步聚类分析法,诊断磨损部件,准确率在75％以上。该项研究成果在应用上已取得成效。 Spectrometric oil analysis is one of the most important methods in Diesel engine wear conditions monitoring and failures diagnosis.Using the Spectrometer (such as the FAS-2C spectrometer, made in U.S.A.), we are able to determine the content of various metal element quickly and accurately. These spectrum data provide us the important message about the Diesel engine wear condition. But how to use these spectrum data to discriminate that the wear condition of Diesel engine is normal or not is a key and knotty problem.We have used multivariate statistical analysis methods to solve this problem. We used the stepwise Fisher's discriminating method to sieve the most important factors and to set up discriminating function. By means of Fisher's discriminating function, we are able to discriminate whether the wear condition of Diesel engine is normal, abnormal or unususally abnormal.In case of the wear condition of Diesel engine is abnormal, we will use principal component analysis of compositional vector and cluster analysis to diagnose the failure and unusually wear components.We have applied these methods to type DF4 internal combustion engine and have received excellent results. For the wear condition the rate of discriminating is about ninety percent (90% ) and for the failue diagnosis the rate of discriminating correctly is about seventy-seven percent (77% ).

关 键 词： 光谱数据 故障诊断 多元统计分析

领　　域： [理学] [理学]