采用振动信号二维特征向量聚类的配电开关机械状态识别新方法
A novel mechanical state identification method for distribution switch based on vibration signal 2-D feature vector with clustering algorithm

配电开关动作产生的振动信号具有非线性非平稳特性，蕴含机械状态信息. 提出一种基于振动信号二维特征向量和模糊K均值聚类的配电开关机械状态识别新方法. 利用HHT带通滤波对配电开关振动信号进行时频分解，分别求取各子频带信号的能量值和重心频率，得到振动信号的二维特征向量作为反映配电开关的机械状态的特征量. 提取配电开关在正常、 底座螺丝松动、 机械结构卡涩及卸掉A相触头绝缘拉杆等4种典型状态实测振动信号的二维特征向量做模糊K均值聚类，结果表明，所提取的特征向量能有效地表征配电开关的机械状态.
Vibration signals of distribution switches that contain mechanical information are characterized by nonlinearity and nonstationarity. Thus，A novel mechanical state identification method for distribution switch based on vibration signal 2-D feature vector with fuzzy K-mean clustering algorithm was proposed in this paper. Taking advantage of HHT band-pass filter，vibration signals would be decomposed in time and frequency domain in order to obtain each sub-band reconstructed signal’s energy and center frequency as the 2-D feature vector，which could represent the mechanical state for distribution switch. FKM clustering was applied to these 2-D feature vectors of observed vibration signals in four typical conditions including normal states，screw loosing states，mechanical structure clamping stagnation states and relieved insulated pull rod of phase A contacts states. Results show that the feature quantity can represent the mechanical state of distribution switch accurately and effectively