基于卷积神经网络的气体绝缘组合开关盆式 绝缘子螺栓松动检测方法*

盆式绝缘子是GIS的关键绝缘器件,它与两侧气室法兰通过螺栓进行紧固连接,当螺栓松动时会导致盆式绝缘子应力分布不均,严重时会引起绝缘子破裂,从而影响GIS运行的安全性和可靠性。文章搭建了盆式绝缘子螺栓松动超声波检测系统,以获取不同螺栓不同工况下的超声信号,基于卷积神经网络对超声信号进行特征提取,并且与BP神经网络的训练结果进行对比分析。实验结果表明,卷积神经网络可以自动提取GIS盆式绝缘子螺栓松动特征量,对十种螺栓松动工况的识别准确率达到100%,相比于BP神经网络具有较高的识别准确率,该方法可以直接用于盆式绝缘子螺栓松动检测。

Looseness detection method of gas insulated switchgear basin insulator bolts based on convolutional neural network

Basin-type insulator is the key insulation device of GIS. It is fastened and connected with the flanges of the gas chambers on both sides by bolts. When the bolts are loose, the stress distribution of the basin insulator will be uneven, and in severe cases, the insulator will be cracked, which will affect the safety and stability of GIS operation. The article builds an ultrasonic detection system for the looseness of flange bolts of basin insulators to obtain ultrasonic signals of different bolts under different working conditions, The features of the ultrasonic signals are extracted based on the convolutional neural network and compared with BP neural network. The experimental results show that the convolutional neural network can automatically extract the GIS basin insulator bolt loosening feature quantity, and the recognition accuracy of ten bolt loosening conditions reaches 100%, which has certain advantages compared with the BP neural network. The method is applied to a basin-type insulation bolt loosening detection system, and the detection is accurate and the effect is obvious.