A CROSS-DOMAIN FRAMEWORK FOR EXPLAINABLE STRUCTURAL INSPECTION

Abstract: The digital transformation of structural inspection has led to the widespread use of image-based artificial intelligence for automated damage detection. Deep learning models, particularly convolutional neural networks, have demonstrated strong capability in identifying surface-level defects such as cracks from photographic data. However, although these systems provide pixel-level predictions, they lack mechanisms for semantic interpretation, crossdomain data integration, and structured knowledge representation. For deployment in civil engineering workflows, AI-based inspection must move beyond detection toward contextual reasoning and interoperable knowledge capturing.