基于KPCA-HSMM设备退化状态识别方法的研究

为消除多通道观测信息冗余，压缩高维故障特征，提出了基于核主元分析( KPCA)多通道特征信息融合的隐半马尔可夫模型( HSMM)设备退化状态识别的新方法。首先，对采集的单通道振动信号进行小波相关滤波处理，构造单通道振动信号的小波相关特征尺度熵向量；然后，利用KPCA方法对多通道的小波相关特征尺度熵向量进行冗余消除和特征融合，得到多通道的融合小波相关特征尺度熵向量；并以此融合特征向量作为HSMM的输入进行训练，建立基于HSMM的设备运行状态分类器，从而卖现设备退化状态的识别。实验结果表明，该方法能有效的识别设备的退化状态，从而为多通道特征信息融合设备退化状态识别开辟新的途径。

The Study of Equipment Degradation State Recognition Method Based on KPCA-HSMM

In order to reduce redundancy among Multi-channel observations by sensors and compress high dimensional fault features, a new equipment degradation state recognition method based on kernel principal component analysis-hidden semi-Markov model ( KPCA-HSMM) was proposed. The single channel vibration signal was processed by the way of wavelet transform correlation filter and wavelet correlation feature scale entropy vector of single channel vibration signal was constructed; the wavelet correlation feature scale entropy vectors of multi-channel were fused by KPCA to obtain the fused eigenvector and eliminate redundancy; the fused eigenvector of multi-channel was inputted to the HSMM for training, and running states classification model of equipment based on HSMM was con?structed. to recognize the equipment degradation states. The experimental results show that the pro?posed method is very effective for equipment degradation state recognition and opens up a new way for recognizing equipment degradation state by multi-channel feature information fusion.