Structural and Material Identification of Historic Monuments through AI-based Photogrammetry and Ambient Vibration Testing

Abstract

In practice, assessing the structural integrity of existing structures with complex forms poses a challenge. Although photogrammetry has gained increasing recognition as a versatile and cost-effective technique for obtaining numerical three-dimensional models, it still presents some difficulties in converting the scanned models into exploitable finite element models. This paper specifically focuses on developing a suitable, reliable, and effective procedure for converting images into numerical three-dimensional models for use in the evaluation and assessment of structures. It presents a framework for the structural and material characterization of historic elements with a combined approach using ambient vibration testing. For this purpose, three elements of increasing complexity in a historic site are 3D scanned, converted to finite element models, and updated using experimentally determined natural frequencies. Relying on the shape accuracy obtained by the photogrammetry scans, a genetic algorithm is used to identify the elastic modulus and mass density of the material for the best match of the experimental and numerical frequencies. The integration of photogrammetry, AVT, and AI-based optimization genetic algorithms in this framework provides a robust and reliable approach for characterizing and investigating historic monuments.