基于XGBoost算法的瓷支柱绝缘子振动声学检测信号缺陷识别方法研究

Research on Defect Recognition Technology of Vibro⁃Acoustic Detection Signal for Porcelain Pillar Insulators Based on XGBoost Algorithm

针对瓷支柱绝缘子振动声学检测信号频谱分析存在误判的问题,提出基于XGBoost算法的瓷支柱绝缘子振动声学检测信号缺陷识别方法。从28个时域特征、功率密度谱特征和小波域特征中按照重要性提取了14个特征作为缺陷识别的依据,训练了瓷支柱绝缘子振动声学检测信号缺陷识别模型。结果表明,通过模型对瓷支柱绝缘子振动声学检测信号缺陷进行分类识别,准确率达到95.83%,取得了较好的缺陷识别效果。将XGBoost算法应用于现场检测信号识别,正确率达到96.6%,能够满足工程应用需要。

In order to solve the problem of misjudgment in the frequency spectrum analysis of vibro⁃acoustic detection signals for porcelain pillar insulators,this paper proposes a defect recognition method for vibro⁃acoustic detection signals for porcelain pillar insulators based on XGBoost algorithm.From twenty⁃eight time⁃domain features,power density spectrum features and wavelet domain features,the fourteen features are extracted according to their importance as the basis for defect recognition,and the defect recognition model of porcelain pillar insulator vibro⁃acoustic detection signals is trained.The results show that the accuracy rate is 95.83%when the model is used to classify and identify the defects of vibro⁃acoustic detection signals of porcelain pillar insulators,achieving a good defect recognition effect.The XGBoost algorithm is applied to on⁃site signal recognition,with an accuracy rate of 96.6%,which can meet the needs of the engineering application.