摘要
将粗糙集理论的属性约简与核Fisher鉴别分析方法结合起来,提出一种基于粗糙核Fisher鉴别分析的故障特征提取方法.首先采用粗糙集理论的属性约简删除与分类无关或关系不大的特征,降低输入特征维数,排除干扰特征的影响,减小了特征提取计算量;再采用核Fisher鉴别分析方法进一步提取非线性特征;最后将该方法应用于航空发动机滑油系统故障特征提取及故障识别中.结果表明:该方法获取的特征在提高分类正确率的同时,还有效地降低了输入特征维数,提高了分类效率,并且对分类器具有较强的适应性和鲁棒性.
A new approach based on rough kernel fisher discriminant analysis （RKFDA） was proposed for aeroengine fault feature extraction, which combined rough set and kernel Fisher discriminant analysis. Firstly, rough set was used to exclude the features irrelevant to the fault; reduce the dimension of features, remove the effect of disturbance characteristics and cut down the cost of computation. Secondly, kernel Fisher discriminant analysis was employed on the obtained subset of features to extract the nonlinear features. Finally, fault extraction and recognition experiments in aeroengine lubricating oil system were carried out to test the performance of this method. The results show that the extracted features based on the proposed method could improve the recognition for aeroengine fault, and reduce efficiently the dimension of features with strong adaptability and robustness for various classifiers.
出处
《航空动力学报》
EI
CAS
CSCD
北大核心
2008年第7期1346-1352,共7页
Journal of Aerospace Power
基金
国家自然科学基金(60672179)
军队重点科研基金(2003KJ01705)
