Sensing Wettability Condition of Insulation Surface Employing Convolutional Neural Network

Wettability of polymeric insulators is a prime indicator of the insulator surface condition. New polymeric insulator surfaces are hydrophobic in nature, where discrete water droplets are formed on the insulator surface. But as the insulation becomes aged, the surface loses its hydrophobicity, leading to the formation of continuous water channels, which subsequently leads to dry band arcing and even flashover, thereby affecting the long-term performance of the insulators. Therefore, accurate sensing of wettability of polymeric insulators is important for reliable insulator diagnostics. Considering the above fact, in this letter, we propose a deep learning framework for accurate sensing of wettability class (WC) of insulators. We captured the images of water droplets with varying WCs on the surface of an 11 kV silicone rubber suspension insulator. After suitable preprocessing, the captured images were fed to a pretrained deep convolutional neural network model AlexNet for the purpose of wettability classification. We also observed that our proposed method is capable of sensing different WC with a high degree of accuracy, which can be practically implemented for real-life monitoring of insulators.