摘要
提出一种故障特征选择方法 ,该方法采用类内类间距离作为特征评价准则 ,并利用遗传算法良好的寻优能力 ,解决特征的优选问题。轴承诊断实例证明 ,该方法有较好的寻优特征子集的能力 ,能够提高BIT系统的诊断精度 ,降低系统的虚警率 ,因而在机电BIT故障特征选择中有较好的应用前景。
A method of feature selection was proposed. A between-calss and within-class distance measurement criterion and genetic algorithm (GA) for optimal feature selection were presented. According to the results of bearing diagnosis example,it is proved that this method possesses excellent optimization property,and can enhance the diagnostic correctness,decrease false alarm rate. The method has good prospects in the BIT(built-in test) fault feature selection.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2004年第12期1048-1050,1062,共4页
China Mechanical Engineering
基金
国防预研基金资助项目
关键词
机电BIT
特征选择
距离准则
遗传算法
mechantronics BIT (MEBIT)
feature selection
distance measurement
genetic algorithm
