基于振动信号MTF变换的并联电容器故障诊断

Fault Diagnosis of Shunt Capacitor Based on Markov Transfer Field Transformation of Vibration Signal

针对并联电容器运行过程中常出现内部缺陷、接触缺陷、重叠缺陷、油质缺陷等4类典型缺陷,提出了一种基于振动信号马尔可夫转移场变换的并联电容器故障诊断方法。首先,根据现场工况制作了4种典型缺陷样品,测试了不同缺陷下试样的机械振动信号并搭建振动数据集;然后,基于马尔可夫转移场变换将一维振动信号转变为二维图像以提高信号可视能力;最后,通过卷积神经网络对马尔可夫密度图进行特征自提取和分类,讨论了不同分位数下信号的可视化能力,并与常见的诊断方法进行了比较。结果表明:将振动信号进行马尔可夫转移场变换提高了信号的可视化能力,使深度学习算法能够更全面的提取信号特征,文中方法的平均识别准确率达到了98%左右,能较好地实现并联电容器的故障诊断,并优于其他常见方法。

Aiming at the four typical defects of shunt capacitor,such as internal defect,contact defect,overlapping defect and oil defect,a fault diagnosis method of shunt capacitor based on Markov transfer field transformation of vibration signal is proposed.Firstly,four typical defect samples are made according to the field conditions,the mechanical vibration signals of the samples under different defects are tested,and the vibration data set is built.Then,the one-dimensional vibration signal is transformed into two-dimensional image based on Markov transfer field transform to improve the visibility of the signal.Finally,the feature Self Extraction and classification of Markov density map are carried out by convolution neural network.The visualization ability of signals under different quantiles is discussed and compared with common diagnosis methods.The results show that the Markov transfer field transformation of the vibration signal improves the visualization ability of the signal,and the deep learning algorithm can extract the signal features more comprehensively.The average recognition accuracy of the proposed method is about 98%,which can better realize the fault diagnosis of shunt capacitor,and is better than other common methods.