Methods comparison for neural network-based structural damage recognition and classification

Abstract

Machine learning has brought significant advancements to the field of structural health monitoring, providing flexible and efficient solutions for detecting both local and global damage in various infrastructures. Local damage detection focuses on identifying issues such as cracks and spalling in specific areas of concrete structures, including bridges, highways, and tunnels. Techniques like artificial neural networks (ANNs) and deep neural networks (DNNs) have proven effective in surface defect recognition, demonstrating their versatility across different structural environments. Additionally, cost-effective methods leveraging devices such as smartphones have been explored for rapid road integrity assessments, offering practical and affordable solutions. On a larger scale, global damage detection involves classifying structural collapse modes and types of damage, utilizing feature extraction and deep learning models to improve the accuracy of identifying large-scale failures. These advancements highlight the growing importance of machine learning and computer vision in enhancing the resilience and real-time monitoring of critical infrastructure systems.