摘要
提出了基于特征融合和稀疏表示的机械故障诊断方法。用小波包分解齿轮振动信号,提取各频带的能量,组成初始特征向量。为了去除初始特征向量中的冗余信息,采用核主元分析进行特征融合,对初始特征向量进行降维,将降维之后的特征向量进行归一化,作为稀疏表示分类的输入。齿轮故障诊断实验表明,所提方法能较好地诊断出齿轮不同类型的故障。
A kind of the mechanical fault diagnosis method based on feature fusion and sparse representa- tion is proposed. The energy of every subband of gear vibration signal is extracted with wavelet packet decompo- sition, and the energy constitutes original feature vector. The feature fusion is implemented with KPCA, and the dimension reduction of original feature vector is used as the input of classification based on sparse representa- tion. The experiment demonstrates that the different fault type of gear is detected by the presented method.
出处
《机械传动》
CSCD
北大核心
2017年第1期54-58,共5页
Journal of Mechanical Transmission
基金
西安石油大学全日制硕士研究生优秀学位论文培育项目(2015YP140407)
关键词
故障诊断
特征融合
稀疏表示
齿轮
Fault diagnosis Feature fusion Sparse representation Gear
