摘要
通过特征提取最大限度地反应出不同故障种类的差别对故障的准确诊断具有重要意义。故采用EEMD样本熵用于汽轮机振动故障的特征提取。将振动信号利用EEMD分解得到IMF分量,计算IMF的样本熵作为多为特征向量,大大提高了不同故障之间的区分程度,并通过计算多维空间下各类故障之间的形心距体现出不同故障种类的区分程度。最后,将IMF样本熵作为SVM的特征向量,对故障进行诊断。选取汽轮机转子正常、质量不平衡、油膜涡动、动静碰摩、转子不对中5种样本进行诊断,并与能量特征提取方法进行对比,结果显示该方法具有更高的诊断准确性。
The obvious distinction of different fault types reflected by eigenvectors has great significance for the accuracy of fanlt diagnosis. So the EEMD sample entropy is used to achieve the eigenvectors of turbine vibration faults. The vibration signals are decomposed by EEMD to achieve IMF components, the sample entropy of IMF is calculated as multi-dimensional vector, which greatly improves the degree of distinction between different faults. And the core of distance between different faults calculated in hyperspace can reflect the degree of distinction between different faults. Finally, the sample entropy of IMF treated as eigenvectors of SVM is used to fault diagnosis. Five kinds of fault typical:normal rotor state, rotor misalignment, mass unbalance, rubbing and oil whirl are chose to diagnose. Meanwhile, compared with the EEMD energy, the result demonstrates the more accuracy of this method.
出处
《汽轮机技术》
北大核心
2015年第6期457-460,共4页
Turbine Technology