Adapted Methodology for Aerospace Sealant Inspection Using Neural Networks

This study presents a novel deep learning-based method for inspecting aerospace sealants, utilizing a modified Mask Region-Based Convolutional Neural Network (Mask RCNN) for defect detection and segmentation. Inspired by medical image analysis techniques, the methodology involves training the modified Mask RCNN model to detect and classify defects in aerospace sealants, such as bubbles, cracks, and irregular application patterns. Several images of sealant applied to various surfaces are used for training, with data augmentation techniques enhancing the dataset to ensure robust performance under diverse conditions. Once trained, the model automatically generates detailed reports that highlighting the professional roles involved. The proposed method aims to improve maintenance efficiency, reduce human error, and ensure the quality of aerospace sealants, ultimately contributing to the overall safety and performance of aerospace components.