Deep-learning Sonic-RAY tomography for architectural heritage digital reconstruction and structural diagnosis. S-RAY

S-RAY aims to develop novel non-invasive technology for the in situ inspection and precise 3D reconstruction of the interior morphology of historical masonry elements. The project addresses one of the main challenges in conservation of heritage structures, which is dealing with the many unknowns and uncertainties about the existing structure (construction materials, geometry, existing damage, etc.). The project includes: (i) the design and fabrication of new equipment to perform the automatic inspection; and (ii) the development of novel processing algorithms based on deep learning (DL) able to provide the accurate 3D reconstruction of masonry elements from raw sonic tomography data.

The main research motivation is to overcome common limitations encountered in current non-invasive systems able to provide information about the interior of structures by: (a) reducing uncertainties at the level of interpretation and improving the processing of tomographic data for historic construction materials; (b) reducing the time consumption, at operational and data processing level; (c) enhancing practicability and applicability of inspection systems, with a focus on increasing the scalability of the inspection from local measurements to full buildings. The main research hypothesis is that these limitations can be addressed by means of three different disciplines that have gained great importance in the recent years: (a) Artificial Intelligence (AI) and DL algorithms; (b) automation and robotics; and (c) unmanned aerial vehicles (UAV) and long-range systems.

The project is based on the idea that a non-invasive ready-to-use tool (apparatus and method) that provides an accurate internal 3D reconstruction of masonry structural elements can be of great help for several professionals in the field of conservation of the built heritage: from architects and engineers to assist in their calculations to archaeologist and specialist in the virtualization of the built heritage.