Application of Machine Learning to Evaluate Insulator Surface Erosion

This article proposes a new automated inspection system that can estimate erosion in silicone rubber (SIR) samples using a computer vision-based method. In this work, we used SIR samples that were damaged under laboratory conditions. The proposed work is expected to classify SIR samples into one of three classes based on the degree of erosion following the IEC-60587 standard in defining failed samples. We use various preprocessing and feature extraction methods and classify using the artificial neural network (ANN) and deep convolutional neural network (CNN). We compare their performance and find that the best results were achieved using a deep CNN architecture. This work serves as a proof of concept and can be further extended to outdoor on-field test cases.