The Hydrophobicity Class Identification of Silicone-Rubber Samples using Deep Learning Algorithms

Abstract

This paper presents an approach to classify the hydrophobicity characteristic of silicone rubber (SiR) samples using deep learning algorithms. By deforming the hydrophobicity property of SiR samples using corona discharges, images of water droplets placed on the sample surface were acquired. From the images, the contact angles of the droplets were determined to find the hydrophobicity classes. The generated water droplet image dataset was trained, validated, and tested utilizing AlexNet, VGGNet, and ResNet. The result shows that the modified AlexNet model with an accuracy of 99.36% is a reliable diagnostic method to identify the hydrophobicity qualification of the SiR samples.