基于粒计算的旋转机械信号特征提取技术

针对现有数据挖掘技术未能有效提取旋转机械信号中的敏感特征,本文提出了基于粒计算的特征提取技术。即在邻域粗糙集中,分别于一定包含度下,选择不同粒度层的敏感特征,利用敏感特征构建核属性集。将新技术应用于实验室齿轮信号的分析,结果表明本文提出的特征提取技术更好地完成了数据挖掘的作用,建立了核属性集,为下一步旋转机械的故障诊断打下了基础。

Regarding the fact that current date mining techniques are unable to extract sensitive features from rotary mechanical signals effectively, this essay has brought up the Features Extraction Technique based on Granular Computing. That is selecting sensitive features from different granularity layers respectively in a certain inclusion degree between Neighborhood rough sets. Then, using the sensitive features to build up a core feature set. Yt clearly demonstrates that the Features Extraction Technique brought up by this essay has better accomplished data mining after its application to the analysis of laboratory gear signal. Features Extraction Technique not only establishes core feature sets, but also lays a foundation for fault diagnosis of rotary machines.