In this letter a deep learning-based model is proposed for online inspection of silicone rubber outdoor insulators. The inclined plane tracking and erosion test is used as per ASTM D2303 in order to simulate standard erosion on silicone rubber insulation composites. Photos taken for the tested composites are used as training and testing inputs for a convolutional neural network topology in the proposed deep learning model, thereby classifying the degree of erosion damage into light, moderate and severe. The remarkable classification accuracy obtained shows the potential of utilizing the proposed framework for online monitoring of outdoor silicone rubber insulators in the transmission and distribution grid.

中文翻译：

---

应用迁移学习深度学习评估硅橡胶材料表面侵蚀


在这封信中，提出了一种基于深度学习的模型，用于硅橡胶户外绝缘子的在线检测。根据 ASTM D2303 使用斜面漏电起痕和侵蚀测试，以模拟硅橡胶绝缘复合材料的标准侵蚀。在所提出的深度学习模型中，将测试复合材料拍摄的照片用作卷积神经网络拓扑的训练和测试输入，从而将侵蚀损坏程度分为轻度、中度和严重。所获得的显着分类精度显示了利用所提出的框架对输配电网中的户外硅橡胶绝缘子进行在线监测的潜力。