AI-Driven Early Evaluation Of Earthquake-Damaged Masonry Structures
Please login to view abstract download link
This work introduces a visual-analysis workflow driven by artificial intelligence, together with an initial methodology for assessing the resilience of masonry buildings following seismic events that have struck Italy since the 1980s. A collection of images gathered during official inspections by the Italian Civil Protection Department was processed to automatically recognize characteristic earthquake-related damage in three-dimensional masonry elements. These photographs, taken from both the interior and exterior of affected buildings and domes, were organized into training, validation, and testing subsets. The framework is conceived to strengthen the role of AI as a decision-support resource for post-event evaluations, with a particular emphasis on preserving and understanding heritage masonry [1]. On the mechanical side, the procedure assists in constructing discrete no-tension representations of walls and domes subjected to seismic actions, informed by the AI-generated damage predictions. Verification of structural capacity is performed through a recently developed strut-net method, which checks whether a pattern of compressed masonry struts can carry the vertical and lateral forces acting on the analyzed structures [2, 3].
