基于频率响应信号PCA的陶瓷绝缘子缺陷检测

为了对陶瓷绝缘子缺陷进行检测，利用频率响应分析法(FRA)识别陶瓷绝缘子主要固有模态的峰值范围，运用时间频域特征提取的主成分分析(PCA)对陶瓷绝缘子进行缺陷检测。通过选取安装在500kV输电塔上材质为方石英和氧化铝的67个陶瓷绝缘子样本，研究了瓷体、瓷帽和内部三种典型的陶瓷绝缘子缺陷类型。通过PCA对时间数据和频率响应数据提取特征进行分析，实现了对完好、瓷体缺陷、瓷帽缺陷、内部缺陷和材质的区分问题。实验结果表明：内部缺陷表现为固有模态的消失或新模态的产生，基于时间数据的PCA对数据方差贡献最大的向量达到99.74%，可以区分瓷体缺陷和材质；基于频率响应数据的PCA对数据方差贡献最大的向量达到96.70%，可实现陶瓷绝缘子内部缺陷检测。

Defect Detection of Ceramic Insulator based on PCA of Frequency Response Signal

In order to detect the defects of ceramic insulators, frequency response analysis (FRA) is used to identify the peak range of the main natural modes of ceramic insulators, and principal component analysis (PCA) based on time-frequency feature extraction is used to detect the defects of ceramic insulators. By selecting 67 samples of ceramic insulators made of cristobalite and alumina installed on 500 kV transmission towers, three typical types of ceramic insulator defects, namely porcelain body, porcelain cap and interior, were studied. PCA was used to analyze the extracted features of time data and frequency response data, and realized the distinction of integrity, porcelain body defects, porcelain cap defects, internal defects and materials. The experimental results show that the internal defects are manifested by the disappearance of natural modes or the generation of new modes. The vector with the largest contribution of PCA based on time data to the data variance reaches 99.74%, which can distinguish the defects and materials of porcelain body; PCA based on frequency response data has the largest contribution to the data variance vector of 96.70%, which can realize the internal defect detection of ceramic insulators.